intel/llvm
1
0
Fork 0
mirror of https://github.com/intel/llvm.git synced 2026-01-20 10:58:11 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
7be1ac7e1d49b45550c74bf507de86e3bd0813de
llvm/lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp

7134 lines
318 KiB
C++
Raw Normal View History

Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
//===-- SymbolFileDWARF.cpp ------------------------------------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#include "SymbolFileDWARF.h"
// Other libraries and framework includes
#include "clang/AST/ASTConsumer.h"
#include "clang/AST/ASTContext.h"
#include "clang/AST/Decl.h"
#include "clang/AST/DeclGroup.h"
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
#include "clang/AST/DeclObjC.h"
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
#include "clang/AST/DeclTemplate.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "clang/Basic/Builtins.h"
#include "clang/Basic/IdentifierTable.h"
#include "clang/Basic/LangOptions.h"
#include "clang/Basic/SourceManager.h"
#include "clang/Basic/TargetInfo.h"
#include "clang/Basic/Specifiers.h"
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
#include "clang/Sema/DeclSpec.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
This change brings in the latest LLVM/Clang, and completes the support in the LLDB expression parser for incomplete types. Clang now imports types lazily, and we complete those types as necessary. Changes include: - ClangASTSource now supports three APIs which it passes to ClangExpressionDeclMap. CompleteType completes a TagDecl or an ObjCInterfaceDecl when needed; FindExternalVisibleDecls finds named entities that are visible in the expression's scope; and FindExternalLexicalDecls performs a (potentially restricted) search for entities inside a lexical scope like a namespace. These changes mean that entities in namespaces should work normally. - The SymbolFileDWARF code for searching a context for a specific name is now more general, and can search arbitrary contexts. - We are continuing to adapt our calls into LLVM from interfaces that take start and end iterators when accepting multiple items to interfaces that use ArrayRef. - I have cleaned up some code, especially our use of namespaces. This change is neutral for our testsuite and greatly improves correctness for large programs (like Clang) with complicated type systems. It should also lay the groundwork for improving the expression parser's performance as we are lazier and lazier about providing type information. llvm-svn: 136555
2011-07-30 02:42:06 +00:00
#include "llvm/Support/Casting.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "lldb/Core/Module.h"
#include "lldb/Core/PluginManager.h"
#include "lldb/Core/RegularExpression.h"
#include "lldb/Core/Scalar.h"
#include "lldb/Core/Section.h"
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
#include "lldb/Core/StreamFile.h"
Move DIE location reporting into the DWARFDebugInfo class, use it from there in SymbolFileDWARF::ParseType (and eventually in other interesting places as well.) llvm-svn: 138644
2011-08-26 19:44:13 +00:00
#include "lldb/Core/StreamString.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "lldb/Core/Timer.h"
#include "lldb/Core/Value.h"
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
#include "lldb/Host/Host.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "lldb/Symbol/Block.h"
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
#include "lldb/Symbol/ClangExternalASTSourceCallbacks.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "lldb/Symbol/CompileUnit.h"
#include "lldb/Symbol/LineTable.h"
#include "lldb/Symbol/ObjectFile.h"
#include "lldb/Symbol/SymbolVendor.h"
#include "lldb/Symbol/VariableList.h"
Move the SourceManager from the Debugger to the Target. That way it can store the per-Target default Source File & Line. Set the default Source File & line to main (if it can be found.) at startup. Selecting the current thread & or frame resets the current source file & line, and "source list" as well as the breakpoint command "break set -l <NUM>" will use the current source file. llvm-svn: 139323
2011-09-08 22:13:49 +00:00
#include "lldb/Target/ObjCLanguageRuntime.h"
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
#include "lldb/Target/CPPLanguageRuntime.h"
Move the SourceManager from the Debugger to the Target. That way it can store the per-Target default Source File & Line. Set the default Source File & line to main (if it can be found.) at startup. Selecting the current thread & or frame resets the current source file & line, and "source list" as well as the breakpoint command "break set -l <NUM>" will use the current source file. llvm-svn: 139323
2011-09-08 22:13:49 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "DWARFCompileUnit.h"
#include "DWARFDebugAbbrev.h"
#include "DWARFDebugAranges.h"
#include "DWARFDebugInfo.h"
#include "DWARFDebugInfoEntry.h"
#include "DWARFDebugLine.h"
#include "DWARFDebugPubnames.h"
#include "DWARFDebugRanges.h"
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
#include "DWARFDeclContext.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include "DWARFDIECollection.h"
#include "DWARFFormValue.h"
#include "DWARFLocationList.h"
#include "LogChannelDWARF.h"
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
#include "SymbolFileDWARFDebugMap.h"
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
#include <map>
Disable the debug logging I accidentally left enabled. llvm-svn: 118758
2010-11-11 01:09:45 +00:00
//#define ENABLE_DEBUG_PRINTF // COMMENT OUT THIS LINE PRIOR TO CHECKIN
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
#ifdef ENABLE_DEBUG_PRINTF
#include <stdio.h>
#define DEBUG_PRINTF(fmt, ...) printf(fmt, ## __VA_ARGS__)
#else
#define DEBUG_PRINTF(fmt, ...)
#endif
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
#define DIE_IS_BEING_PARSED ((lldb_private::Type*)1)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
using namespace lldb;
using namespace lldb_private;
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
//static inline bool
//child_requires_parent_class_union_or_struct_to_be_completed (dw_tag_t tag)
//{
// switch (tag)
// {
// default:
// break;
// case DW_TAG_subprogram:
// case DW_TAG_inlined_subroutine:
// case DW_TAG_class_type:
// case DW_TAG_structure_type:
// case DW_TAG_union_type:
// return true;
// }
// return false;
//}
//
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
static AccessType
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
DW_ACCESS_to_AccessType (uint32_t dwarf_accessibility)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
switch (dwarf_accessibility)
{
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
case DW_ACCESS_public: return eAccessPublic;
case DW_ACCESS_private: return eAccessPrivate;
case DW_ACCESS_protected: return eAccessProtected;
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
default: break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
return eAccessNone;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Added a logging helper class for SymbolFileDWARF::ParseType() that will enable us to track the depth of parsing and what is being parsed. This helps when trying to track down difficult type parsing issues and is only enabled in non-production builds. llvm-svn: 150203
2012-02-09 20:03:49 +00:00
#if defined(LLDB_CONFIGURATION_DEBUG) or defined(LLDB_CONFIGURATION_RELEASE)
class DIEStack
{
public:
void Push (DWARFCompileUnit *cu, const DWARFDebugInfoEntry *die)
{
m_dies.push_back (DIEInfo(cu, die));
}
void LogDIEs (Log *log, SymbolFileDWARF *dwarf)
{
StreamString log_strm;
const size_t n = m_dies.size();
log_strm.Printf("DIEStack[%zu]:\n", n);
for (size_t i=0; i<n; i++)
{
DWARFCompileUnit *cu = m_dies[i].cu;
const DWARFDebugInfoEntry *die = m_dies[i].die;
std::string qualified_name;
die->GetQualifiedName(dwarf, cu, qualified_name);
log_strm.Printf ("[%zu] 0x%8.8x: %s name='%s'\n",
i,
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
qualified_name.c_str());
}
log->PutCString(log_strm.GetData());
}
void Pop ()
{
m_dies.pop_back();
}
class ScopedPopper
{
public:
ScopedPopper (DIEStack &die_stack) :
m_die_stack (die_stack),
m_valid (false)
{
}
void
Push (DWARFCompileUnit *cu, const DWARFDebugInfoEntry *die)
{
m_valid = true;
m_die_stack.Push (cu, die);
}
~ScopedPopper ()
{
if (m_valid)
m_die_stack.Pop();
}
protected:
DIEStack &m_die_stack;
bool m_valid;
};
protected:
struct DIEInfo {
DIEInfo (DWARFCompileUnit *c, const DWARFDebugInfoEntry *d) :
cu(c),
die(d)
{
}
DWARFCompileUnit *cu;
const DWARFDebugInfoEntry *die;
};
typedef std::vector<DIEInfo> Stack;
Stack m_dies;
};
#endif
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
void
SymbolFileDWARF::Initialize()
{
LogChannelDWARF::Initialize();
PluginManager::RegisterPlugin (GetPluginNameStatic(),
GetPluginDescriptionStatic(),
CreateInstance);
}
void
SymbolFileDWARF::Terminate()
{
PluginManager::UnregisterPlugin (CreateInstance);
LogChannelDWARF::Initialize();
}
const char *
SymbolFileDWARF::GetPluginNameStatic()
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
return "dwarf";
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const char *
SymbolFileDWARF::GetPluginDescriptionStatic()
{
return "DWARF and DWARF3 debug symbol file reader.";
}
SymbolFile*
SymbolFileDWARF::CreateInstance (ObjectFile* obj_file)
{
return new SymbolFileDWARF(obj_file);
}
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
TypeList *
SymbolFileDWARF::GetTypeList ()
{
if (m_debug_map_symfile)
return m_debug_map_symfile->GetTypeList();
return m_obj_file->GetModule()->GetTypeList();
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
//----------------------------------------------------------------------
// Gets the first parent that is a lexical block, function or inlined
// subroutine, or compile unit.
//----------------------------------------------------------------------
static const DWARFDebugInfoEntry *
GetParentSymbolContextDIE(const DWARFDebugInfoEntry *child_die)
{
const DWARFDebugInfoEntry *die;
for (die = child_die->GetParent(); die != NULL; die = die->GetParent())
{
dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_compile_unit:
case DW_TAG_subprogram:
case DW_TAG_inlined_subroutine:
case DW_TAG_lexical_block:
return die;
}
}
return NULL;
}
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
SymbolFileDWARF::SymbolFileDWARF(ObjectFile* objfile) :
SymbolFile (objfile),
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
UserID (0), // Used by SymbolFileDWARFDebugMap to when this class parses .o files to contain the .o file index/ID
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
m_debug_map_symfile (NULL),
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
m_clang_tu_decl (NULL),
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
m_flags(),
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
m_data_debug_abbrev (),
m_data_debug_aranges (),
m_data_debug_frame (),
m_data_debug_info (),
m_data_debug_line (),
m_data_debug_loc (),
m_data_debug_ranges (),
m_data_debug_str (),
Convert over to the latest and greatest on disc accelerator hash tables. Renamed the DWARF sections to ".apple_names" and ".apple_types" until we get more buy in from other vendors. llvm-svn: 140702
2011-09-28 17:06:40 +00:00
m_data_apple_names (),
m_data_apple_types (),
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_data_apple_namespaces (),
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
m_abbr(),
m_info(),
m_line(),
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_apple_names_ap (),
m_apple_types_ap (),
m_apple_namespaces_ap (),
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
m_apple_objc_ap (),
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
m_function_basename_index(),
m_function_fullname_index(),
m_function_method_index(),
m_function_selector_index(),
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
m_objc_class_selectors_index(),
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
m_global_index(),
Separated the DWARF index for types from that the index of the namespaces since we can't parse DW_TAG_namespace DIEs as types. They are only decls in clang. All of the types we handle right now have both clang "XXXType" classes to go with the "XXXDecl" classes which means they can be used within the lldb_private::Type class. I need to check to see which other decls that don't have associated type objects need to float around the debugger and possibly make a lldb_private::Decl class to manage them. llvm-svn: 116558
2010-10-15 02:03:22 +00:00
m_type_index(),
m_namespace_index(),
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
m_indexed (false),
m_is_external_ast_source (false),
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
m_using_apple_tables (false),
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
m_supports_DW_AT_APPLE_objc_complete_type (eLazyBoolCalculate),
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
m_ranges(),
m_unique_ast_type_map ()
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
}
SymbolFileDWARF::~SymbolFileDWARF()
{
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
if (m_is_external_ast_source)
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
{
ModuleSP module_sp (m_obj_file->GetModule());
if (module_sp)
module_sp->GetClangASTContext().RemoveExternalSource ();
}
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
}
static const ConstString &
GetDWARFMachOSegmentName ()
{
static ConstString g_dwarf_section_name ("__DWARF");
return g_dwarf_section_name;
}
All UnwindPlan objects are now passed around as shared pointers. ArchDefaultUnwindPlan plug-in interfaces are now cached per architecture instead of being leaked for every frame. Split the ArchDefaultUnwindPlan_x86 into ArchDefaultUnwindPlan_x86_64 and ArchDefaultUnwindPlan_i386 interfaces. There were sporadic crashes that were due to something leaking or being destroyed when doing stack crawls. This patch should clear up these issues. llvm-svn: 125541
2011-02-15 00:19:15 +00:00
UniqueDWARFASTTypeMap &
SymbolFileDWARF::GetUniqueDWARFASTTypeMap ()
{
if (m_debug_map_symfile)
return m_debug_map_symfile->GetUniqueDWARFASTTypeMap ();
return m_unique_ast_type_map;
}
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
ClangASTContext &
SymbolFileDWARF::GetClangASTContext ()
{
if (m_debug_map_symfile)
return m_debug_map_symfile->GetClangASTContext ();
ClangASTContext &ast = m_obj_file->GetModule()->GetClangASTContext();
if (!m_is_external_ast_source)
{
m_is_external_ast_source = true;
llvm::OwningPtr<clang::ExternalASTSource> ast_source_ap (
new ClangExternalASTSourceCallbacks (SymbolFileDWARF::CompleteTagDecl,
SymbolFileDWARF::CompleteObjCInterfaceDecl,
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
SymbolFileDWARF::FindExternalVisibleDeclsByName,
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
SymbolFileDWARF::LayoutRecordType,
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
this));
ast.SetExternalSource (ast_source_ap);
}
return ast;
}
void
SymbolFileDWARF::InitializeObject()
{
// Install our external AST source callbacks so we can complete Clang types.
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
ModuleSP module_sp (m_obj_file->GetModule());
if (module_sp)
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
{
const SectionList *section_list = m_obj_file->GetSectionList();
const Section* section = section_list->FindSectionByName(GetDWARFMachOSegmentName ()).get();
// Memory map the DWARF mach-o segment so we have everything mmap'ed
// to keep our heap memory usage down.
if (section)
<rdar://problem/10560053> Fixed "target modules list" (aliased to "image list") to output more information by default. Modified the "target modules list" to have a few new options: "--header" or "-h" => show the image header address "--offset" or "-o" => show the image header address offset from the address in the file (the slide applied to the shared library) Removed the "--symfile-basename" or "-S" option, and repurposed it to "--symfile-unique" "-S" which will show the symbol file if it differs from the executable file. ObjectFile's can now be loaded from memory for cases where we don't have the files cached locally in an SDK or net mounted root. ObjectFileMachO can now read mach files from memory. Moved the section data reading code into the ObjectFile so that the object file can get the section data from Process memory if the file is only in memory. lldb_private::Module can now load its object file in a target with a rigid slide (very common operation for most dynamic linkers) by using: bool Module::SetLoadAddress (Target &target, lldb::addr_t offset, bool &changed) lldb::SBModule() now has a new constructor in the public interface: SBModule::SBModule (lldb::SBProcess &process, lldb::addr_t header_addr); This will find an appropriate ObjectFile plug-in to load an image from memory where the object file header is at "header_addr". llvm-svn: 149804
2012-02-05 02:38:54 +00:00
m_obj_file->MemoryMapSectionData(section, m_dwarf_data);
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
}
If the new .apple_names and .apple_types DWARF accelerator tables are available, we currently will still index the DWARF ourselves and assert if the name lookups differ. This will help us transition to the new accelerator tables and make sure they are workng before we switch over entirely. llvm-svn: 140788
2011-09-29 16:58:15 +00:00
get_apple_names_data();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (m_data_apple_names.GetByteSize() > 0)
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
m_apple_names_ap.reset (new DWARFMappedHash::MemoryTable (m_data_apple_names, get_debug_str_data(), ".apple_names"));
if (m_apple_names_ap->IsValid())
m_using_apple_tables = true;
else
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_apple_names_ap.reset();
}
If the new .apple_names and .apple_types DWARF accelerator tables are available, we currently will still index the DWARF ourselves and assert if the name lookups differ. This will help us transition to the new accelerator tables and make sure they are workng before we switch over entirely. llvm-svn: 140788
2011-09-29 16:58:15 +00:00
get_apple_types_data();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (m_data_apple_types.GetByteSize() > 0)
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
m_apple_types_ap.reset (new DWARFMappedHash::MemoryTable (m_data_apple_types, get_debug_str_data(), ".apple_types"));
if (m_apple_types_ap->IsValid())
m_using_apple_tables = true;
else
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_apple_types_ap.reset();
}
get_apple_namespaces_data();
if (m_data_apple_namespaces.GetByteSize() > 0)
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
m_apple_namespaces_ap.reset (new DWARFMappedHash::MemoryTable (m_data_apple_namespaces, get_debug_str_data(), ".apple_namespaces"));
if (m_apple_namespaces_ap->IsValid())
m_using_apple_tables = true;
else
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_apple_namespaces_ap.reset();
}
If the new .apple_names and .apple_types DWARF accelerator tables are available, we currently will still index the DWARF ourselves and assert if the name lookups differ. This will help us transition to the new accelerator tables and make sure they are workng before we switch over entirely. llvm-svn: 140788
2011-09-29 16:58:15 +00:00
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
get_apple_objc_data();
if (m_data_apple_objc.GetByteSize() > 0)
{
m_apple_objc_ap.reset (new DWARFMappedHash::MemoryTable (m_data_apple_objc, get_debug_str_data(), ".apple_objc"));
if (m_apple_objc_ap->IsValid())
m_using_apple_tables = true;
else
m_apple_objc_ap.reset();
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
bool
SymbolFileDWARF::SupportedVersion(uint16_t version)
{
return version == 2 || version == 3;
}
uint32_t
Testcase fixes with the new symbol lookup code for Objective-C, making symbol lookups for various raw Objective-C symbols work correctly. The IR interpreter makes these lookups because Clang has emitted raw symbol references for ivars and classes. Also improved performance in SymbolFiles, caching the result of asking for SymbolFile abilities. llvm-svn: 145758
2011-12-03 04:38:43 +00:00
SymbolFileDWARF::CalculateAbilities ()
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
uint32_t abilities = 0;
if (m_obj_file != NULL)
{
const Section* section = NULL;
const SectionList *section_list = m_obj_file->GetSectionList();
if (section_list == NULL)
return 0;
uint64_t debug_abbrev_file_size = 0;
uint64_t debug_aranges_file_size = 0;
uint64_t debug_frame_file_size = 0;
uint64_t debug_info_file_size = 0;
uint64_t debug_line_file_size = 0;
uint64_t debug_loc_file_size = 0;
uint64_t debug_macinfo_file_size = 0;
uint64_t debug_pubnames_file_size = 0;
uint64_t debug_pubtypes_file_size = 0;
uint64_t debug_ranges_file_size = 0;
uint64_t debug_str_file_size = 0;
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
section = section_list->FindSectionByName(GetDWARFMachOSegmentName ()).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section_list = &section->GetChildren ();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugInfo, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section != NULL)
{
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_info_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugAbbrev, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_abbrev_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugAbbrevData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugAranges, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_aranges_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugArangesData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugFrame, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_frame_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugFrameData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugLine, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_line_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugLineData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugLoc, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_loc_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugLocData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugMacInfo, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_macinfo_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugMacInfoData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugPubNames, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_pubnames_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugPubNamesData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugPubTypes, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_pubtypes_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugPubTypesData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugRanges, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_ranges_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugRangesData);
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
section = section_list->FindSectionByType (eSectionTypeDWARFDebugStr, true).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (section)
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
debug_str_file_size = section->GetFileSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
m_flags.Set (flagsGotDebugStrData);
}
if (debug_abbrev_file_size > 0 && debug_info_file_size > 0)
abilities |= CompileUnits | Functions | Blocks | GlobalVariables | LocalVariables | VariableTypes;
if (debug_line_file_size > 0)
abilities |= LineTables;
}
return abilities;
}
const DataExtractor&
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
SymbolFileDWARF::GetCachedSectionData (uint32_t got_flag, SectionType sect_type, DataExtractor &data)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (m_flags.IsClear (got_flag))
{
m_flags.Set (got_flag);
const SectionList *section_list = m_obj_file->GetSectionList();
if (section_list)
{
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
SectionSP section_sp (section_list->FindSectionByType(sect_type, true));
if (section_sp)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// See if we memory mapped the DWARF segment?
if (m_dwarf_data.GetByteSize())
{
Fixed a few things in the ELF object file: 1 - sections only get a valid VM size if they have SHF_ALLOC in the section flags 2 - symbol names are marked as mangled if they start with "_Z" Also fixed the DWARF parser to correctly use the section file size when extracting the DWARF. llvm-svn: 153496
2012-03-27 02:40:46 +00:00
data.SetData(m_dwarf_data, section_sp->GetOffset (), section_sp->GetFileSize());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
else
{
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
if (m_obj_file->ReadSectionData (section_sp.get(), data) == 0)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
data.Clear();
}
}
}
}
return data;
}
const DataExtractor&
SymbolFileDWARF::get_debug_abbrev_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugAbbrevData, eSectionTypeDWARFDebugAbbrev, m_data_debug_abbrev);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const DataExtractor&
SymbolFileDWARF::get_debug_aranges_data()
{
return GetCachedSectionData (flagsGotDebugArangesData, eSectionTypeDWARFDebugAranges, m_data_debug_aranges);
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DataExtractor&
SymbolFileDWARF::get_debug_frame_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugFrameData, eSectionTypeDWARFDebugFrame, m_data_debug_frame);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const DataExtractor&
SymbolFileDWARF::get_debug_info_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugInfoData, eSectionTypeDWARFDebugInfo, m_data_debug_info);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const DataExtractor&
SymbolFileDWARF::get_debug_line_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugLineData, eSectionTypeDWARFDebugLine, m_data_debug_line);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const DataExtractor&
SymbolFileDWARF::get_debug_loc_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugLocData, eSectionTypeDWARFDebugLoc, m_data_debug_loc);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const DataExtractor&
SymbolFileDWARF::get_debug_ranges_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugRangesData, eSectionTypeDWARFDebugRanges, m_data_debug_ranges);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
const DataExtractor&
SymbolFileDWARF::get_debug_str_data()
{
Modified both the ObjectFileMachO and ObjectFileELF to correctly set the SectionType for Section objects for DWARF. Modified the DWARF plug-in to get the DWARF sections by SectionType so we can safely abstract the LLDB core from section names for the various object file formats. Modified the SectionType definitions for .debug_pubnames and .debug_pubtypes to use the correct case. llvm-svn: 109054
2010-07-21 22:54:26 +00:00
return GetCachedSectionData (flagsGotDebugStrData, eSectionTypeDWARFDebugStr, m_data_debug_str);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
const DataExtractor&
Convert over to the latest and greatest on disc accelerator hash tables. Renamed the DWARF sections to ".apple_names" and ".apple_types" until we get more buy in from other vendors. llvm-svn: 140702
2011-09-28 17:06:40 +00:00
SymbolFileDWARF::get_apple_names_data()
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
{
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
return GetCachedSectionData (flagsGotAppleNamesData, eSectionTypeDWARFAppleNames, m_data_apple_names);
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
}
const DataExtractor&
Convert over to the latest and greatest on disc accelerator hash tables. Renamed the DWARF sections to ".apple_names" and ".apple_types" until we get more buy in from other vendors. llvm-svn: 140702
2011-09-28 17:06:40 +00:00
SymbolFileDWARF::get_apple_types_data()
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
{
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
return GetCachedSectionData (flagsGotAppleTypesData, eSectionTypeDWARFAppleTypes, m_data_apple_types);
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const DataExtractor&
SymbolFileDWARF::get_apple_namespaces_data()
{
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
return GetCachedSectionData (flagsGotAppleNamespacesData, eSectionTypeDWARFAppleNamespaces, m_data_apple_namespaces);
}
const DataExtractor&
SymbolFileDWARF::get_apple_objc_data()
{
return GetCachedSectionData (flagsGotAppleObjCData, eSectionTypeDWARFAppleObjC, m_data_apple_objc);
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
DWARFDebugAbbrev*
SymbolFileDWARF::DebugAbbrev()
{
if (m_abbr.get() == NULL)
{
const DataExtractor &debug_abbrev_data = get_debug_abbrev_data();
if (debug_abbrev_data.GetByteSize() > 0)
{
m_abbr.reset(new DWARFDebugAbbrev());
if (m_abbr.get())
m_abbr->Parse(debug_abbrev_data);
}
}
return m_abbr.get();
}
const DWARFDebugAbbrev*
SymbolFileDWARF::DebugAbbrev() const
{
return m_abbr.get();
}
DWARFDebugInfo*
SymbolFileDWARF::DebugInfo()
{
if (m_info.get() == NULL)
{
Timer scoped_timer(__PRETTY_FUNCTION__, "%s this = %p", __PRETTY_FUNCTION__, this);
if (get_debug_info_data().GetByteSize() > 0)
{
m_info.reset(new DWARFDebugInfo());
if (m_info.get())
{
m_info->SetDwarfData(this);
}
}
}
return m_info.get();
}
const DWARFDebugInfo*
SymbolFileDWARF::DebugInfo() const
{
return m_info.get();
}
DWARFCompileUnit*
SymbolFileDWARF::GetDWARFCompileUnitForUID(lldb::user_id_t cu_uid)
{
DWARFDebugInfo* info = DebugInfo();
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
if (info && UserIDMatches(cu_uid))
return info->GetCompileUnit((dw_offset_t)cu_uid).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
return NULL;
}
DWARFDebugRanges*
SymbolFileDWARF::DebugRanges()
{
if (m_ranges.get() == NULL)
{
Timer scoped_timer(__PRETTY_FUNCTION__, "%s this = %p", __PRETTY_FUNCTION__, this);
if (get_debug_ranges_data().GetByteSize() > 0)
{
m_ranges.reset(new DWARFDebugRanges());
if (m_ranges.get())
m_ranges->Extract(this);
}
}
return m_ranges.get();
}
const DWARFDebugRanges*
SymbolFileDWARF::DebugRanges() const
{
return m_ranges.get();
}
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
lldb::CompUnitSP
SymbolFileDWARF::ParseCompileUnit (DWARFCompileUnit* dwarf_cu, uint32_t cu_idx)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompUnitSP cu_sp;
if (dwarf_cu)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompileUnit *comp_unit = (CompileUnit*)dwarf_cu->GetUserData();
if (comp_unit)
{
// We already parsed this compile unit, had out a shared pointer to it
cu_sp = comp_unit->shared_from_this();
}
else
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
ModuleSP module_sp (m_obj_file->GetModule());
if (module_sp)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
const DWARFDebugInfoEntry * cu_die = dwarf_cu->GetCompileUnitDIEOnly ();
if (cu_die)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
const char * cu_die_name = cu_die->GetName(this, dwarf_cu);
const char * cu_comp_dir = cu_die->GetAttributeValueAsString(this, dwarf_cu, DW_AT_comp_dir, NULL);
LanguageType cu_language = (LanguageType)cu_die->GetAttributeValueAsUnsigned(this, dwarf_cu, DW_AT_language, 0);
if (cu_die_name)
{
std::string ramapped_file;
FileSpec cu_file_spec;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
if (cu_die_name[0] == '/' || cu_comp_dir == NULL || cu_comp_dir[0] == '\0')
{
// If we have a full path to the compile unit, we don't need to resolve
// the file. This can be expensive e.g. when the source files are NFS mounted.
if (module_sp->RemapSourceFile(cu_die_name, ramapped_file))
cu_file_spec.SetFile (ramapped_file.c_str(), false);
else
cu_file_spec.SetFile (cu_die_name, false);
}
else
{
std::string fullpath(cu_comp_dir);
if (*fullpath.rbegin() != '/')
fullpath += '/';
fullpath += cu_die_name;
if (module_sp->RemapSourceFile (fullpath.c_str(), ramapped_file))
cu_file_spec.SetFile (ramapped_file.c_str(), false);
else
cu_file_spec.SetFile (fullpath.c_str(), false);
}
cu_sp.reset(new CompileUnit (module_sp,
dwarf_cu,
cu_file_spec,
MakeUserID(dwarf_cu->GetOffset()),
cu_language));
if (cu_sp)
{
dwarf_cu->SetUserData(cu_sp.get());
if (m_debug_map_symfile)
{
// Let the symbol file register the compile unit with
// the symbol vendor using its compile unit index
// when we are doing DWARF in .o files + debug map
m_debug_map_symfile->SetCompileUnit(this, cu_sp);
}
else
{
// Figure out the compile unit index if we weren't given one
if (cu_idx == UINT32_MAX)
DebugInfo()->GetCompileUnit(dwarf_cu->GetOffset(), &cu_idx);
m_obj_file->GetModule()->GetSymbolVendor()->SetCompileUnitAtIndex(cu_idx, cu_sp);
}
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
}
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
return cu_sp;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
uint32_t
SymbolFileDWARF::GetNumCompileUnits()
{
DWARFDebugInfo* info = DebugInfo();
if (info)
return info->GetNumCompileUnits();
return 0;
}
CompUnitSP
SymbolFileDWARF::ParseCompileUnitAtIndex(uint32_t cu_idx)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompUnitSP cu_sp;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
DWARFDebugInfo* info = DebugInfo();
if (info)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = info->GetCompileUnitAtIndex(cu_idx);
if (dwarf_cu)
cu_sp = ParseCompileUnit(dwarf_cu, cu_idx);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
return cu_sp;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
static void
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
AddRangesToBlock (Block& block,
DWARFDebugRanges::RangeList& ranges,
addr_t block_base_addr)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
const size_t num_ranges = ranges.GetSize();
for (size_t i = 0; i<num_ranges; ++i)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
const DWARFDebugRanges::Range &range = ranges.GetEntryRef (i);
const addr_t range_base = range.GetRangeBase();
assert (range_base >= block_base_addr);
block.AddRange(Block::Range (range_base - block_base_addr, range.GetByteSize()));;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
block.FinalizeRanges ();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Function *
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
SymbolFileDWARF::ParseCompileUnitFunction (const SymbolContext& sc, DWARFCompileUnit* dwarf_cu, const DWARFDebugInfoEntry *die)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
DWARFDebugRanges::RangeList func_ranges;
const char *name = NULL;
const char *mangled = NULL;
int decl_file = 0;
int decl_line = 0;
int decl_column = 0;
int call_file = 0;
int call_line = 0;
int call_column = 0;
DWARFExpression frame_base;
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
assert (die->Tag() == DW_TAG_subprogram);
if (die->Tag() != DW_TAG_subprogram)
return NULL;
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
if (die->GetDIENamesAndRanges (this,
dwarf_cu,
name,
mangled,
func_ranges,
decl_file,
decl_line,
decl_column,
call_file,
call_line,
call_column,
&frame_base))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// Union of all ranges in the function DIE (if the function is discontiguous)
AddressRange func_range;
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
lldb::addr_t lowest_func_addr = func_ranges.GetMinRangeBase (0);
lldb::addr_t highest_func_addr = func_ranges.GetMaxRangeEnd (0);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (lowest_func_addr != LLDB_INVALID_ADDRESS && lowest_func_addr <= highest_func_addr)
{
func_range.GetBaseAddress().ResolveAddressUsingFileSections (lowest_func_addr, m_obj_file->GetSectionList());
if (func_range.GetBaseAddress().IsValid())
func_range.SetByteSize(highest_func_addr - lowest_func_addr);
}
if (func_range.GetBaseAddress().IsValid())
{
Mangled func_name;
if (mangled)
func_name.SetValue(mangled, true);
else if (name)
func_name.SetValue(name, false);
FunctionSP func_sp;
std::auto_ptr<Declaration> decl_ap;
if (decl_file != 0 || decl_line != 0 || decl_column != 0)
Fixed a crasher (an assert was firing in the DWARF parser) when setting breakpoints on inlined functions by name. This involved fixing the DWARF parser to correctly back up and parse the concrete function when we find inlined functions by name, then grabbing any appropriate inlined blocks and returning symbol contexts with the block filled in. After this was fixed, the breakpoint by name resolver needed to correctly deal with symbol contexts that had the inlined block filled in in the symbol contexts. llvm-svn: 119017
2010-11-14 00:22:48 +00:00
decl_ap.reset(new Declaration (sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(decl_file),
decl_line,
decl_column));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Removed some commented out code from the DWARF parser. Also reduce the size of the lldb_private::Symbol objects by removing the lldb_private::Function pointer that was in each symbol. Running Instruments has shown that when debugging large applications with DWARF in .o files that lldb_private::Symbol objects are one of the highest users of memory. No one was using the Symbol::GetFunction() call anyway. llvm-svn: 140881
2011-09-30 20:52:25 +00:00
// Supply the type _only_ if it has already been parsed
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
Type *func_type = m_die_to_type.lookup (die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
assert(func_type == NULL || func_type != DIE_IS_BEING_PARSED);
func_range.GetBaseAddress().ResolveLinkedAddress();
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
const user_id_t func_user_id = MakeUserID(die->GetOffset());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
func_sp.reset(new Function (sc.comp_unit,
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
func_user_id, // UserID is the DIE offset
func_user_id,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
func_name,
func_type,
func_range)); // first address range
if (func_sp.get() != NULL)
{
Removed some commented out code from the DWARF parser. Also reduce the size of the lldb_private::Symbol objects by removing the lldb_private::Function pointer that was in each symbol. Running Instruments has shown that when debugging large applications with DWARF in .o files that lldb_private::Symbol objects are one of the highest users of memory. No one was using the Symbol::GetFunction() call anyway. llvm-svn: 140881
2011-09-30 20:52:25 +00:00
if (frame_base.IsValid())
func_sp->GetFrameBaseExpression() = frame_base;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
sc.comp_unit->AddFunction(func_sp);
return func_sp.get();
}
}
}
return NULL;
}
size_t
SymbolFileDWARF::ParseCompileUnitFunctions(const SymbolContext &sc)
{
assert (sc.comp_unit);
size_t functions_added = 0;
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu = GetDWARFCompileUnitForUID(sc.comp_unit->GetID());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (dwarf_cu)
{
DWARFDIECollection function_dies;
const size_t num_funtions = dwarf_cu->AppendDIEsWithTag (DW_TAG_subprogram, function_dies);
size_t func_idx;
for (func_idx = 0; func_idx < num_funtions; ++func_idx)
{
const DWARFDebugInfoEntry *die = function_dies.GetDIEPtrAtIndex(func_idx);
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
if (sc.comp_unit->FindFunctionByUID (MakeUserID(die->GetOffset())).get() == NULL)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (ParseCompileUnitFunction(sc, dwarf_cu, die))
++functions_added;
}
}
//FixupTypes();
}
return functions_added;
}
bool
SymbolFileDWARF::ParseCompileUnitSupportFiles (const SymbolContext& sc, FileSpecList& support_files)
{
assert (sc.comp_unit);
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = GetDWARFCompileUnitForUID(sc.comp_unit->GetID());
assert (dwarf_cu);
const DWARFDebugInfoEntry * cu_die = dwarf_cu->GetCompileUnitDIEOnly();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (cu_die)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
const char * cu_comp_dir = cu_die->GetAttributeValueAsString(this, dwarf_cu, DW_AT_comp_dir, NULL);
dw_offset_t stmt_list = cu_die->GetAttributeValueAsUnsigned(this, dwarf_cu, DW_AT_stmt_list, DW_INVALID_OFFSET);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// All file indexes in DWARF are one based and a file of index zero is
// supposed to be the compile unit itself.
support_files.Append (*sc.comp_unit);
<rdar://problem/8196933> Use the metadata in the dSYM bundle Info.plist to remap source paths when they keys are available. llvm-svn: 152836
2012-03-15 21:01:31 +00:00
return DWARFDebugLine::ParseSupportFiles(sc.comp_unit->GetModule(), get_debug_line_data(), cu_comp_dir, stmt_list, support_files);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
return false;
}
struct ParseDWARFLineTableCallbackInfo
{
LineTable* line_table;
const SectionList *section_list;
lldb::addr_t prev_sect_file_base_addr;
lldb::addr_t curr_sect_file_base_addr;
bool is_oso_for_debug_map;
bool prev_in_final_executable;
DWARFDebugLine::Row prev_row;
SectionSP prev_section_sp;
SectionSP curr_section_sp;
};
//----------------------------------------------------------------------
// ParseStatementTableCallback
//----------------------------------------------------------------------
static void
ParseDWARFLineTableCallback(dw_offset_t offset, const DWARFDebugLine::State& state, void* userData)
{
LineTable* line_table = ((ParseDWARFLineTableCallbackInfo*)userData)->line_table;
if (state.row == DWARFDebugLine::State::StartParsingLineTable)
{
// Just started parsing the line table
}
else if (state.row == DWARFDebugLine::State::DoneParsingLineTable)
{
// Done parsing line table, nothing to do for the cleanup
}
else
{
ParseDWARFLineTableCallbackInfo* info = (ParseDWARFLineTableCallbackInfo*)userData;
// We have a new row, lets append it
if (info->curr_section_sp.get() == NULL || info->curr_section_sp->ContainsFileAddress(state.address) == false)
{
info->prev_section_sp = info->curr_section_sp;
info->prev_sect_file_base_addr = info->curr_sect_file_base_addr;
// If this is an end sequence entry, then we subtract one from the
// address to make sure we get an address that is not the end of
// a section.
if (state.end_sequence && state.address != 0)
info->curr_section_sp = info->section_list->FindSectionContainingFileAddress (state.address - 1);
else
info->curr_section_sp = info->section_list->FindSectionContainingFileAddress (state.address);
if (info->curr_section_sp.get())
info->curr_sect_file_base_addr = info->curr_section_sp->GetFileAddress ();
else
info->curr_sect_file_base_addr = 0;
}
if (info->curr_section_sp.get())
{
lldb::addr_t curr_line_section_offset = state.address - info->curr_sect_file_base_addr;
// Check for the fancy section magic to determine if we
if (info->is_oso_for_debug_map)
{
// When this is a debug map object file that contains DWARF
// (referenced from an N_OSO debug map nlist entry) we will have
// a file address in the file range for our section from the
// original .o file, and a load address in the executable that
// contains the debug map.
//
// If the sections for the file range and load range are
// different, we have a remapped section for the function and
// this address is resolved. If they are the same, then the
// function for this address didn't make it into the final
// executable.
bool curr_in_final_executable = info->curr_section_sp->GetLinkedSection () != NULL;
// If we are doing DWARF with debug map, then we need to carefully
// add each line table entry as there may be gaps as functions
// get moved around or removed.
if (!info->prev_row.end_sequence && info->prev_section_sp.get())
{
if (info->prev_in_final_executable)
{
bool terminate_previous_entry = false;
if (!curr_in_final_executable)
{
// Check for the case where the previous line entry
// in a function made it into the final executable,
// yet the current line entry falls in a function
// that didn't. The line table used to be contiguous
// through this address range but now it isn't. We
// need to terminate the previous line entry so
// that we can reconstruct the line range correctly
// for it and to keep the line table correct.
terminate_previous_entry = true;
}
else if (info->curr_section_sp.get() != info->prev_section_sp.get())
{
// Check for cases where the line entries used to be
// contiguous address ranges, but now they aren't.
// This can happen when order files specify the
// ordering of the functions.
lldb::addr_t prev_line_section_offset = info->prev_row.address - info->prev_sect_file_base_addr;
Section *curr_sect = info->curr_section_sp.get();
Section *prev_sect = info->prev_section_sp.get();
assert (curr_sect->GetLinkedSection());
assert (prev_sect->GetLinkedSection());
lldb::addr_t object_file_addr_delta = state.address - info->prev_row.address;
lldb::addr_t curr_linked_file_addr = curr_sect->GetLinkedFileAddress() + curr_line_section_offset;
lldb::addr_t prev_linked_file_addr = prev_sect->GetLinkedFileAddress() + prev_line_section_offset;
lldb::addr_t linked_file_addr_delta = curr_linked_file_addr - prev_linked_file_addr;
if (object_file_addr_delta != linked_file_addr_delta)
terminate_previous_entry = true;
}
if (terminate_previous_entry)
{
line_table->InsertLineEntry (info->prev_section_sp,
state.address - info->prev_sect_file_base_addr,
info->prev_row.line,
info->prev_row.column,
info->prev_row.file,
false, // is_stmt
false, // basic_block
false, // state.prologue_end
false, // state.epilogue_begin
true); // end_sequence);
}
}
}
if (curr_in_final_executable)
{
line_table->InsertLineEntry (info->curr_section_sp,
curr_line_section_offset,
state.line,
state.column,
state.file,
state.is_stmt,
state.basic_block,
state.prologue_end,
state.epilogue_begin,
state.end_sequence);
info->prev_section_sp = info->curr_section_sp;
}
else
{
// If the current address didn't make it into the final
// executable, the current section will be the __text
// segment in the .o file, so we need to clear this so
// we can catch the next function that did make it into
// the final executable.
info->prev_section_sp.reset();
info->curr_section_sp.reset();
}
info->prev_in_final_executable = curr_in_final_executable;
}
else
{
// We are not in an object file that contains DWARF for an
// N_OSO, this is just a normal DWARF file. The DWARF spec
// guarantees that the addresses will be in increasing order
// so, since we store line tables in file address order, we
// can always just append the line entry without needing to
// search for the correct insertion point (we don't need to
// use LineEntry::InsertLineEntry()).
line_table->AppendLineEntry (info->curr_section_sp,
curr_line_section_offset,
state.line,
state.column,
state.file,
state.is_stmt,
state.basic_block,
state.prologue_end,
state.epilogue_begin,
state.end_sequence);
}
}
info->prev_row = state;
}
}
bool
SymbolFileDWARF::ParseCompileUnitLineTable (const SymbolContext &sc)
{
assert (sc.comp_unit);
if (sc.comp_unit->GetLineTable() != NULL)
return true;
DWARFCompileUnit* dwarf_cu = GetDWARFCompileUnitForUID(sc.comp_unit->GetID());
if (dwarf_cu)
{
const DWARFDebugInfoEntry *dwarf_cu_die = dwarf_cu->GetCompileUnitDIEOnly();
<rdar://problem/10413589> Fixed a potential crasher where we weren't checking we got a valid DIE in a compile unit. llvm-svn: 145226
2011-11-28 03:29:03 +00:00
if (dwarf_cu_die)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/10413589> Fixed a potential crasher where we weren't checking we got a valid DIE in a compile unit. llvm-svn: 145226
2011-11-28 03:29:03 +00:00
const dw_offset_t cu_line_offset = dwarf_cu_die->GetAttributeValueAsUnsigned(this, dwarf_cu, DW_AT_stmt_list, DW_INVALID_OFFSET);
if (cu_line_offset != DW_INVALID_OFFSET)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/10413589> Fixed a potential crasher where we weren't checking we got a valid DIE in a compile unit. llvm-svn: 145226
2011-11-28 03:29:03 +00:00
std::auto_ptr<LineTable> line_table_ap(new LineTable(sc.comp_unit));
if (line_table_ap.get())
{
ParseDWARFLineTableCallbackInfo info = {
line_table_ap.get(),
m_obj_file->GetSectionList(),
0,
0,
m_debug_map_symfile != NULL,
false,
DWARFDebugLine::Row(),
SectionSP(),
SectionSP()
};
uint32_t offset = cu_line_offset;
DWARFDebugLine::ParseStatementTable(get_debug_line_data(), &offset, ParseDWARFLineTableCallback, &info);
sc.comp_unit->SetLineTable(line_table_ap.release());
return true;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
return false;
}
size_t
SymbolFileDWARF::ParseFunctionBlocks
(
const SymbolContext& sc,
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
Block *parent_block,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *die,
addr_t subprogram_low_pc,
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
uint32_t depth
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
)
{
size_t blocks_added = 0;
while (die != NULL)
{
dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_inlined_subroutine:
Backout previous fix that was designed to allow functions within functions as this function is what parses the blocks for a function... llvm-svn: 137461
2011-08-12 16:22:48 +00:00
case DW_TAG_subprogram:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_TAG_lexical_block:
{
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
Block *block = NULL;
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
if (tag == DW_TAG_subprogram)
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
{
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
// Skip any DW_TAG_subprogram DIEs that are inside
// of a normal or inlined functions. These will be
// parsed on their own as separate entities.
if (depth > 0)
break;
block = parent_block;
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
}
else
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
BlockSP block_sp(new Block (MakeUserID(die->GetOffset())));
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
parent_block->AddChild(block_sp);
block = block_sp.get();
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
}
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
DWARFDebugRanges::RangeList ranges;
const char *name = NULL;
const char *mangled_name = NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
int decl_file = 0;
int decl_line = 0;
int decl_column = 0;
int call_file = 0;
int call_line = 0;
int call_column = 0;
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
if (die->GetDIENamesAndRanges (this,
dwarf_cu,
name,
mangled_name,
ranges,
decl_file, decl_line, decl_column,
call_file, call_line, call_column))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (tag == DW_TAG_subprogram)
{
assert (subprogram_low_pc == LLDB_INVALID_ADDRESS);
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
subprogram_low_pc = ranges.GetMinRangeBase(0);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Few little fixes to reading in inlined functions. Also added a test case with some inlining. llvm-svn: 110892
2010-08-12 01:20:14 +00:00
else if (tag == DW_TAG_inlined_subroutine)
{
// We get called here for inlined subroutines in two ways.
// The first time is when we are making the Function object
// for this inlined concrete instance. Since we're creating a top level block at
// here, the subprogram_low_pc will be LLDB_INVALID_ADDRESS. So we need to
// adjust the containing address.
// The second time is when we are parsing the blocks inside the function that contains
// the inlined concrete instance. Since these will be blocks inside the containing "real"
// function the offset will be for that function.
if (subprogram_low_pc == LLDB_INVALID_ADDRESS)
{
Added more functionality to Range template classes in RangeMap.h and converted remaining DWARF areas that were using ranges over to this class. Also converted lldb_private::Block to use it. llvm-svn: 141460
2011-10-08 00:49:15 +00:00
subprogram_low_pc = ranges.GetMinRangeBase(0);
Few little fixes to reading in inlined functions. Also added a test case with some inlining. llvm-svn: 110892
2010-08-12 01:20:14 +00:00
}
}
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
AddRangesToBlock (*block, ranges, subprogram_low_pc);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (tag != DW_TAG_subprogram && (name != NULL || mangled_name != NULL))
{
std::auto_ptr<Declaration> decl_ap;
if (decl_file != 0 || decl_line != 0 || decl_column != 0)
Few little fixes to reading in inlined functions. Also added a test case with some inlining. llvm-svn: 110892
2010-08-12 01:20:14 +00:00
decl_ap.reset(new Declaration(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(decl_file),
decl_line, decl_column));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
std::auto_ptr<Declaration> call_ap;
if (call_file != 0 || call_line != 0 || call_column != 0)
Few little fixes to reading in inlined functions. Also added a test case with some inlining. llvm-svn: 110892
2010-08-12 01:20:14 +00:00
call_ap.reset(new Declaration(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(call_file),
call_line, call_column));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
block->SetInlinedFunctionInfo (name, mangled_name, decl_ap.get(), call_ap.get());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
++blocks_added;
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
if (die->HasChildren())
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
blocks_added += ParseFunctionBlocks (sc,
block,
dwarf_cu,
die->GetFirstChild(),
subprogram_low_pc,
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
depth + 1);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
break;
default:
break;
}
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
// Only parse siblings of the block if we are not at depth zero. A depth
// of zero indicates we are currently parsing the top level
// DW_TAG_subprogram DIE
if (depth == 0)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
die = NULL;
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
else
die = die->GetSibling();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
return blocks_added;
}
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
bool
SymbolFileDWARF::ParseTemplateDIE (DWARFCompileUnit* dwarf_cu,
const DWARFDebugInfoEntry *die,
ClangASTContext::TemplateParameterInfos &template_param_infos)
{
const dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_template_type_parameter:
case DW_TAG_template_value_parameter:
{
const uint8_t *fixed_form_sizes = DWARFFormValue::GetFixedFormSizesForAddressSize (dwarf_cu->GetAddressByteSize());
DWARFDebugInfoEntry::Attributes attributes;
const size_t num_attributes = die->GetAttributes (this,
dwarf_cu,
fixed_form_sizes,
attributes);
const char *name = NULL;
Type *lldb_type = NULL;
clang_type_t clang_type = NULL;
uint64_t uval64 = 0;
bool uval64_valid = false;
if (num_attributes > 0)
{
DWARFFormValue form_value;
for (size_t i=0; i<num_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
switch (attr)
{
case DW_AT_name:
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
name = form_value.AsCString(&get_debug_str_data());
break;
case DW_AT_type:
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
const dw_offset_t type_die_offset = form_value.Reference(dwarf_cu);
lldb_type = ResolveTypeUID(type_die_offset);
if (lldb_type)
clang_type = lldb_type->GetClangForwardType();
}
break;
case DW_AT_const_value:
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
uval64_valid = true;
uval64 = form_value.Unsigned();
}
break;
default:
break;
}
}
if (name && lldb_type && clang_type)
{
bool is_signed = false;
template_param_infos.names.push_back(name);
clang::QualType clang_qual_type (clang::QualType::getFromOpaquePtr (clang_type));
if (tag == DW_TAG_template_value_parameter && ClangASTContext::IsIntegerType (clang_type, is_signed) && uval64_valid)
{
llvm::APInt apint (lldb_type->GetByteSize() * 8, uval64, is_signed);
template_param_infos.args.push_back (clang::TemplateArgument (llvm::APSInt(apint), clang_qual_type));
}
else
{
template_param_infos.args.push_back (clang::TemplateArgument (clang_qual_type));
}
}
else
{
return false;
}
}
}
return true;
default:
break;
}
return false;
}
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
bool
SymbolFileDWARF::ParseTemplateParameterInfos (DWARFCompileUnit* dwarf_cu,
const DWARFDebugInfoEntry *parent_die,
ClangASTContext::TemplateParameterInfos &template_param_infos)
{
if (parent_die == NULL)
return NULL;
Args template_parameter_names;
for (const DWARFDebugInfoEntry *die = parent_die->GetFirstChild();
die != NULL;
die = die->GetSibling())
{
const dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_template_type_parameter:
case DW_TAG_template_value_parameter:
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
ParseTemplateDIE (dwarf_cu, die, template_param_infos);
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
break;
default:
break;
}
}
if (template_param_infos.args.empty())
return false;
return template_param_infos.args.size() == template_param_infos.names.size();
}
clang::ClassTemplateDecl *
SymbolFileDWARF::ParseClassTemplateDecl (clang::DeclContext *decl_ctx,
Fixed an issue where a class that resides inside another class wasn't getting an access specifier set on it, causing an assertion to fire when building with a Debug+Asserts build of clang. llvm-svn: 143010
2011-10-26 03:31:36 +00:00
lldb::AccessType access_type,
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
const char *parent_name,
int tag_decl_kind,
const ClangASTContext::TemplateParameterInfos &template_param_infos)
{
if (template_param_infos.IsValid())
{
std::string template_basename(parent_name);
template_basename.erase (template_basename.find('<'));
ClangASTContext &ast = GetClangASTContext();
return ast.CreateClassTemplateDecl (decl_ctx,
Fixed an issue where a class that resides inside another class wasn't getting an access specifier set on it, causing an assertion to fire when building with a Debug+Asserts build of clang. llvm-svn: 143010
2011-10-26 03:31:36 +00:00
access_type,
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
template_basename.c_str(),
tag_decl_kind,
template_param_infos);
}
return NULL;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
size_t
SymbolFileDWARF::ParseChildMembers
(
const SymbolContext& sc,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *parent_die,
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
clang_type_t class_clang_type,
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
const LanguageType class_language,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
std::vector<clang::CXXBaseSpecifier *>& base_classes,
std::vector<int>& member_accessibilities,
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
DWARFDIECollection& member_function_dies,
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
AccessType& default_accessibility,
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
bool &is_a_class,
LayoutInfo &layout_info
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
)
{
if (parent_die == NULL)
return 0;
size_t count = 0;
const DWARFDebugInfoEntry *die;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const uint8_t *fixed_form_sizes = DWARFFormValue::GetFixedFormSizesForAddressSize (dwarf_cu->GetAddressByteSize());
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
uint32_t member_idx = 0;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
for (die = parent_die->GetFirstChild(); die != NULL; die = die->GetSibling())
{
dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_member:
Possibly too soon for this commit. llvm-svn: 153946
2012-04-03 08:45:33 +00:00
case DW_TAG_APPLE_Property:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
DWARFDebugInfoEntry::Attributes attributes;
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
const size_t num_attributes = die->GetAttributes (this,
dwarf_cu,
fixed_form_sizes,
attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
Declaration decl;
Updated the lldb_private::Flags class to have better method names and made all of the calls inlined in the header file for better performance. Fixed the summary for C string types (array of chars (with any combo if modifiers), and pointers to chars) work in all cases. Fixed an issue where a forward declaration to a clang type could cause itself to resolve itself more than once if, during the resolving of the type itself it caused something to try and resolve itself again. We now remove the clang type from the forward declaration map in the DWARF parser when we start to resolve it and avoid this additional call. This should stop any duplicate members from appearing and throwing all the alignment of structs, unions and classes. llvm-svn: 117437
2010-10-27 03:32:59 +00:00
//DWARFExpression location;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const char *name = NULL;
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
const char *prop_name = NULL;
const char *prop_getter_name = NULL;
const char *prop_setter_name = NULL;
uint32_t prop_attributes = 0;
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
bool is_artificial = false;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
lldb::user_id_t encoding_uid = LLDB_INVALID_UID;
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
AccessType accessibility = eAccessNone;
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
uint32_t member_byte_offset = UINT32_MAX;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
size_t byte_size = 0;
size_t bit_offset = 0;
size_t bit_size = 0;
uint32_t i;
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
for (i=0; i<num_attributes && !is_artificial; ++i)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name: name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_type: encoding_uid = form_value.Reference(dwarf_cu); break;
case DW_AT_bit_offset: bit_offset = form_value.Unsigned(); break;
case DW_AT_bit_size: bit_size = form_value.Unsigned(); break;
case DW_AT_byte_size: byte_size = form_value.Unsigned(); break;
case DW_AT_data_member_location:
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
if (form_value.BlockData())
{
Value initialValue(0);
Value memberOffset(0);
const DataExtractor& debug_info_data = get_debug_info_data();
uint32_t block_length = form_value.Unsigned();
uint32_t block_offset = form_value.BlockData() - debug_info_data.GetDataStart();
if (DWARFExpression::Evaluate(NULL, // ExecutionContext *
NULL, // clang::ASTContext *
NULL, // ClangExpressionVariableList *
NULL, // ClangExpressionDeclMap *
NULL, // RegisterContext *
debug_info_data,
block_offset,
block_length,
eRegisterKindDWARF,
&initialValue,
memberOffset,
NULL))
{
member_byte_offset = memberOffset.ResolveValue(NULL, NULL).UInt();
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
case DW_AT_accessibility: accessibility = DW_ACCESS_to_AccessType (form_value.Unsigned()); break;
<rdar://problem/11049371> http://llvm.org/bugs/show_bug.cgi?id=12232 Fixed a case where a missing "break" in a switch statement could cause an assertion to fire and kill the debug session. The fix was derived from the findings of Andrea Bigagli, thanks Andrea. llvm-svn: 152741
2012-03-14 21:00:47 +00:00
case DW_AT_artificial: is_artificial = form_value.Unsigned() != 0; break;
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
case DW_AT_APPLE_property_name: prop_name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_APPLE_property_getter: prop_getter_name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_APPLE_property_setter: prop_setter_name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_APPLE_property_attribute: prop_attributes = form_value.Unsigned(); break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
default:
<rdar://problem/11049371> http://llvm.org/bugs/show_bug.cgi?id=12232 Fixed a case where a missing "break" in a switch statement could cause an assertion to fire and kill the debug session. The fix was derived from the findings of Andrea Bigagli, thanks Andrea. llvm-svn: 152741
2012-03-14 21:00:47 +00:00
case DW_AT_declaration:
case DW_AT_description:
case DW_AT_mutable:
case DW_AT_visibility:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_sibling:
break;
}
}
}
Fixed a problem where we did not read properties correctly if the setter/getter were not present in the debug information. The fixes are as follows: - We not only look for the method by its full name, but also look for automatically-generated methods when searching for a selector in an Objective-C interface. This is necessary to find accessors. - Extract the getter and setter name from the DW_TAG_APPLE_Property declaration in the DWARF if they are present; generate them if not. llvm-svn: 154067
2012-04-05 00:12:52 +00:00
if (prop_name)
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
{
Fixed a problem where we did not read properties correctly if the setter/getter were not present in the debug information. The fixes are as follows: - We not only look for the method by its full name, but also look for automatically-generated methods when searching for a selector in an Objective-C interface. This is necessary to find accessors. - Extract the getter and setter name from the DW_TAG_APPLE_Property declaration in the DWARF if they are present; generate them if not. llvm-svn: 154067
2012-04-05 00:12:52 +00:00
ConstString fixed_getter;
ConstString fixed_setter;
// Check if the property getter/setter were provided as full
// names. We want basenames, so we extract them.
if (prop_getter_name && prop_getter_name[0] == '-')
{
ObjCLanguageRuntime::ParseMethodName (prop_getter_name,
NULL,
&fixed_getter,
NULL,
NULL);
prop_getter_name = fixed_getter.GetCString();
}
if (prop_setter_name && prop_setter_name[0] == '-')
{
ObjCLanguageRuntime::ParseMethodName (prop_setter_name,
NULL,
&fixed_setter,
NULL,
NULL);
prop_setter_name = fixed_setter.GetCString();
}
// If the names haven't been provided, they need to be
// filled in.
if (!prop_getter_name)
{
prop_getter_name = prop_name;
}
if (!prop_setter_name && prop_name[0] && !(prop_attributes & DW_APPLE_PROPERTY_readonly))
{
StreamString ss;
ss.Printf("set%c%s:",
toupper(prop_name[0]),
&prop_name[1]);
fixed_setter.SetCString(ss.GetData());
prop_setter_name = fixed_setter.GetCString();
}
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
}
// Clang has a DWARF generation bug where sometimes it
<rdar://problem/10335756> Fixed an issue where bad DWARF from clang would get recycled from DWARF back into types and cause clang to assert and die, killing the lldb binary, when it tried to used the type in an expression. llvm-svn: 142897
2011-10-25 01:25:35 +00:00
// represents fields that are references with bad byte size
// and bit size/offset information such as:
//
// DW_AT_byte_size( 0x00 )
// DW_AT_bit_size( 0x40 )
// DW_AT_bit_offset( 0xffffffffffffffc0 )
//
// So check the bit offset to make sure it is sane, and if
// the values are not sane, remove them. If we don't do this
// then we will end up with a crash if we try to use this
// type in an expression when clang becomes unhappy with its
// recycled debug info.
Added a hack that allows expressions to ignore Objective-C member accessibility restrictions. Ultimately Clang should have a flag that ignores accessibility for Objective-C like it has one for C++. llvm-svn: 117768
2010-10-30 01:56:10 +00:00
<rdar://problem/10335756> Fixed an issue where bad DWARF from clang would get recycled from DWARF back into types and cause clang to assert and die, killing the lldb binary, when it tried to used the type in an expression. llvm-svn: 142897
2011-10-25 01:25:35 +00:00
if (bit_offset > 128)
{
bit_size = 0;
bit_offset = 0;
}
// FIXME: Make Clang ignore Objective-C accessibility for expressions
Added a hack that allows expressions to ignore Objective-C member accessibility restrictions. Ultimately Clang should have a flag that ignores accessibility for Objective-C like it has one for C++. llvm-svn: 117768
2010-10-30 01:56:10 +00:00
if (class_language == eLanguageTypeObjC ||
class_language == eLanguageTypeObjC_plus_plus)
accessibility = eAccessNone;
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
if (member_idx == 0 && !is_artificial && name && (strstr (name, "_vptr$") == name))
{
// Not all compilers will mark the vtable pointer
// member as artificial (llvm-gcc). We can't have
// the virtual members in our classes otherwise it
// throws off all child offsets since we end up
// having and extra pointer sized member in our
// class layouts.
is_artificial = true;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (is_artificial == false)
{
Type *member_type = ResolveTypeUID(encoding_uid);
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
clang::FieldDecl *field_decl = NULL;
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
if (tag == DW_TAG_member)
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
{
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
if (member_type)
{
if (accessibility == eAccessNone)
accessibility = default_accessibility;
member_accessibilities.push_back(accessibility);
field_decl = GetClangASTContext().AddFieldToRecordType (class_clang_type,
name,
member_type->GetClangLayoutType(),
accessibility,
bit_size);
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)field_decl, MakeUserID(die->GetOffset()));
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
}
else
{
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
if (name)
GetObjectFile()->GetModule()->ReportError ("0x%8.8llx: DW_TAG_member '%s' refers to type 0x%8.8llx which was unable to be parsed",
MakeUserID(die->GetOffset()),
name,
encoding_uid);
else
GetObjectFile()->GetModule()->ReportError ("0x%8.8llx: DW_TAG_member refers to type 0x%8.8llx which was unable to be parsed",
MakeUserID(die->GetOffset()),
encoding_uid);
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
}
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
if (member_byte_offset != UINT32_MAX || bit_size != 0)
{
/////////////////////////////////////////////////////////////
// How to locate a field given the DWARF debug information
//
// AT_byte_size indicates the size of the word in which the
// bit offset must be interpreted.
//
// AT_data_member_location indicates the byte offset of the
// word from the base address of the structure.
//
// AT_bit_offset indicates how many bits into the word
// (according to the host endianness) the low-order bit of
// the field starts. AT_bit_offset can be negative.
//
// AT_bit_size indicates the size of the field in bits.
/////////////////////////////////////////////////////////////
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
ByteOrder object_endian = GetObjectFile()->GetModule()->GetArchitecture().GetDefaultEndian();
uint64_t total_bit_offset = 0;
total_bit_offset += (member_byte_offset == UINT32_MAX ? 0 : (member_byte_offset * 8));
if (object_endian == eByteOrderLittle)
{
total_bit_offset += byte_size * 8;
total_bit_offset -= (bit_offset + bit_size);
}
else
{
total_bit_offset += bit_offset;
}
layout_info.field_offsets.insert(std::make_pair(field_decl, total_bit_offset));
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
}
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
if (prop_name != NULL)
{
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
clang::ObjCIvarDecl *ivar_decl = NULL;
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
if (field_decl)
{
ivar_decl = clang::dyn_cast<clang::ObjCIvarDecl>(field_decl);
assert (ivar_decl != NULL);
}
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
GetClangASTContext().AddObjCClassProperty (class_clang_type,
prop_name,
Added support for the DW_AT_APPLE_Property tag for unbacked properties. We support two variants: one in which the getter/setter are provided by selector ("mySetter:") and one in which the getter/setter are provided by signature ("-[MyClass mySetter:]"). llvm-svn: 153675
2012-03-29 19:07:06 +00:00
member_type->GetClangLayoutType(),
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
ivar_decl,
prop_setter_name,
prop_getter_name,
We now record metadata for Objective-C interfaces, Objective-C methods, and Objective-C properties. llvm-svn: 154972
2012-04-18 01:06:17 +00:00
prop_attributes,
MakeUserID(die->GetOffset()));
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
We now record metadata for Objective-C interfaces, Objective-C methods, and Objective-C properties. llvm-svn: 154972
2012-04-18 01:06:17 +00:00
if (ivar_decl)
GetClangASTContext().SetMetadata((uintptr_t)ivar_decl, MakeUserID(die->GetOffset()));
Add code that reads the APPLE_property debug info, and makes up properties from them. llvm-svn: 144440
2011-11-12 01:36:43 +00:00
}
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
++member_idx;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
break;
case DW_TAG_subprogram:
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
// Let the type parsing code handle this one for us.
member_function_dies.Append (die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
case DW_TAG_inheritance:
{
is_a_class = true;
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
if (default_accessibility == eAccessNone)
default_accessibility = eAccessPrivate;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// TODO: implement DW_TAG_inheritance type parsing
DWARFDebugInfoEntry::Attributes attributes;
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
const size_t num_attributes = die->GetAttributes (this,
dwarf_cu,
fixed_form_sizes,
attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
Declaration decl;
DWARFExpression location;
lldb::user_id_t encoding_uid = LLDB_INVALID_UID;
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
AccessType accessibility = default_accessibility;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
bool is_virtual = false;
bool is_base_of_class = true;
off_t member_offset = 0;
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_type: encoding_uid = form_value.Reference(dwarf_cu); break;
case DW_AT_data_member_location:
if (form_value.BlockData())
{
Value initialValue(0);
Value memberOffset(0);
const DataExtractor& debug_info_data = get_debug_info_data();
uint32_t block_length = form_value.Unsigned();
uint32_t block_offset = form_value.BlockData() - debug_info_data.GetDataStart();
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
if (DWARFExpression::Evaluate (NULL,
NULL,
NULL,
NULL,
Change the DWARFExpression::Evaluate methods to take an optional RegisterContext* - normally this is retrieved from the ExecutionContext's StackFrame but when we need to evaluate an expression while creating the stack frame list this can be a little tricky. Add DW_OP_deref_size, needed for the _sigtramp FDE expression. Add support for processing DWARF expressions in RegisterContextLLDB. Update callers to DWARFExpression::Evaluate. llvm-svn: 119885
2010-11-20 01:28:30 +00:00
NULL,
Enabled extra warnings and fixed a bunch of small issues. llvm-svn: 124250
2011-01-25 23:55:37 +00:00
debug_info_data,
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
block_offset,
block_length,
eRegisterKindDWARF,
&initialValue,
memberOffset,
NULL))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
member_offset = memberOffset.ResolveValue(NULL, NULL).UInt();
}
}
break;
case DW_AT_accessibility:
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
case DW_AT_virtuality: is_virtual = form_value.Unsigned() != 0; break;
default:
case DW_AT_sibling:
break;
}
}
}
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Type *base_class_type = ResolveTypeUID(encoding_uid);
assert(base_class_type);
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
Clean up a bit of the type getting code where lldb_private:Type now has clang_type_t GetClangFullType(); // Get a completely defined clang type clang_type_t GetClangLayoutType(); // Get a clang type that can be used for type layout clang_type_t GetClangForwardType(); // A type that can be completed if needed, but is more efficient. llvm-svn: 125691
2011-02-16 23:00:21 +00:00
clang_type_t base_class_clang_type = base_class_type->GetClangFullType();
assert (base_class_clang_type);
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
if (class_language == eLanguageTypeObjC)
{
Clean up a bit of the type getting code where lldb_private:Type now has clang_type_t GetClangFullType(); // Get a completely defined clang type clang_type_t GetClangLayoutType(); // Get a clang type that can be used for type layout clang_type_t GetClangForwardType(); // A type that can be completed if needed, but is more efficient. llvm-svn: 125691
2011-02-16 23:00:21 +00:00
GetClangASTContext().SetObjCSuperClass(class_clang_type, base_class_clang_type);
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
}
else
{
Clean up a bit of the type getting code where lldb_private:Type now has clang_type_t GetClangFullType(); // Get a completely defined clang type clang_type_t GetClangLayoutType(); // Get a clang type that can be used for type layout clang_type_t GetClangForwardType(); // A type that can be completed if needed, but is more efficient. llvm-svn: 125691
2011-02-16 23:00:21 +00:00
base_classes.push_back (GetClangASTContext().CreateBaseClassSpecifier (base_class_clang_type,
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
accessibility,
is_virtual,
is_base_of_class));
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
break;
default:
break;
}
}
return count;
}
clang::DeclContext*
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
SymbolFileDWARF::GetClangDeclContextContainingTypeUID (lldb::user_id_t type_uid)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
DWARFDebugInfo* debug_info = DebugInfo();
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
if (debug_info && UserIDMatches(type_uid))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
DWARFCompileUnitSP cu_sp;
const DWARFDebugInfoEntry* die = debug_info->GetDIEPtr(type_uid, &cu_sp);
if (die)
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
return GetClangDeclContextContainingDIE (cu_sp.get(), die, NULL);
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
}
return NULL;
}
clang::DeclContext*
SymbolFileDWARF::GetClangDeclContextForTypeUID (const lldb_private::SymbolContext &sc, lldb::user_id_t type_uid)
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
if (UserIDMatches(type_uid))
return GetClangDeclContextForDIEOffset (sc, type_uid);
return NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Type*
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
SymbolFileDWARF::ResolveTypeUID (lldb::user_id_t type_uid)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
if (UserIDMatches(type_uid))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
if (debug_info)
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DWARFCompileUnitSP cu_sp;
const DWARFDebugInfoEntry* type_die = debug_info->GetDIEPtr(type_uid, &cu_sp);
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
const bool assert_not_being_parsed = true;
return ResolveTypeUID (cu_sp.get(), type_die, assert_not_being_parsed);
}
}
return NULL;
}
Type*
SymbolFileDWARF::ResolveTypeUID (DWARFCompileUnit* cu, const DWARFDebugInfoEntry* die, bool assert_not_being_parsed)
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
{
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
if (die != NULL)
{
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::ResolveTypeUID (die = 0x%8.8x) %s '%s'",
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
die->GetName(this, cu));
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
// We might be coming in in the middle of a type tree (a class
// withing a class, an enum within a class), so parse any needed
// parent DIEs before we get to this one...
const DWARFDebugInfoEntry *decl_ctx_die = GetDeclContextDIEContainingDIE (cu, die);
switch (decl_ctx_die->Tag())
{
case DW_TAG_structure_type:
case DW_TAG_union_type:
case DW_TAG_class_type:
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
{
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
// Get the type, which could be a forward declaration
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::ResolveTypeUID (die = 0x%8.8x) %s '%s' resolve parent forward type for 0x%8.8x",
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
die->GetName(this, cu),
decl_ctx_die->GetOffset());
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
//
// Type *parent_type = ResolveTypeUID (cu, decl_ctx_die, assert_not_being_parsed);
// if (child_requires_parent_class_union_or_struct_to_be_completed(die->Tag()))
// {
// if (log)
// GetObjectFile()->GetModule()->LogMessage (log.get(),
// "SymbolFileDWARF::ResolveTypeUID (die = 0x%8.8x) %s '%s' resolve parent full type for 0x%8.8x since die is a function",
// die->GetOffset(),
// DW_TAG_value_to_name(die->Tag()),
// die->GetName(this, cu),
// decl_ctx_die->GetOffset());
// // Ask the type to complete itself if it already hasn't since if we
// // want a function (method or static) from a class, the class must
// // create itself and add it's own methods and class functions.
// if (parent_type)
// parent_type->GetClangFullType();
// }
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
}
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
break;
default:
break;
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
}
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
return ResolveType (cu, die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
return NULL;
}
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
// This function is used when SymbolFileDWARFDebugMap owns a bunch of
// SymbolFileDWARF objects to detect if this DWARF file is the one that
// can resolve a clang_type.
bool
SymbolFileDWARF::HasForwardDeclForClangType (lldb::clang_type_t clang_type)
{
clang_type_t clang_type_no_qualifiers = ClangASTType::RemoveFastQualifiers(clang_type);
const DWARFDebugInfoEntry* die = m_forward_decl_clang_type_to_die.lookup (clang_type_no_qualifiers);
return die != NULL;
}
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
lldb::clang_type_t
SymbolFileDWARF::ResolveClangOpaqueTypeDefinition (lldb::clang_type_t clang_type)
{
// We have a struct/union/class/enum that needs to be fully resolved.
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
clang_type_t clang_type_no_qualifiers = ClangASTType::RemoveFastQualifiers(clang_type);
const DWARFDebugInfoEntry* die = m_forward_decl_clang_type_to_die.lookup (clang_type_no_qualifiers);
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
if (die == NULL)
Updated the lldb_private::Flags class to have better method names and made all of the calls inlined in the header file for better performance. Fixed the summary for C string types (array of chars (with any combo if modifiers), and pointers to chars) work in all cases. Fixed an issue where a forward declaration to a clang type could cause itself to resolve itself more than once if, during the resolving of the type itself it caused something to try and resolve itself again. We now remove the clang type from the forward declaration map in the DWARF parser when we start to resolve it and avoid this additional call. This should stop any duplicate members from appearing and throwing all the alignment of structs, unions and classes. llvm-svn: 117437
2010-10-27 03:32:59 +00:00
{
// We have already resolved this type...
return clang_type;
}
// Once we start resolving this type, remove it from the forward declaration
// map in case anyone child members or other types require this type to get resolved.
// The type will get resolved when all of the calls to SymbolFileDWARF::ResolveClangOpaqueTypeDefinition
// are done.
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
m_forward_decl_clang_type_to_die.erase (clang_type_no_qualifiers);
Updated the lldb_private::Flags class to have better method names and made all of the calls inlined in the header file for better performance. Fixed the summary for C string types (array of chars (with any combo if modifiers), and pointers to chars) work in all cases. Fixed an issue where a forward declaration to a clang type could cause itself to resolve itself more than once if, during the resolving of the type itself it caused something to try and resolve itself again. We now remove the clang type from the forward declaration map in the DWARF parser when we start to resolve it and avoid this additional call. This should stop any duplicate members from appearing and throwing all the alignment of structs, unions and classes. llvm-svn: 117437
2010-10-27 03:32:59 +00:00
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
Changed lldb_private::Type over to use the intrusive ref counted pointers so we don't have to lookup types in a type list by ID. Changed the DWARF parser to remove the "can externally complete myself" bits from the type when we are in the process of completing the type itself to avoid an onslaught of external visible decl requests from the clang::ExternalASTSource. llvm-svn: 142461
2011-10-18 23:36:41 +00:00
// Disable external storage for this type so we don't get anymore
// clang::ExternalASTSource queries for this type.
ClangASTContext::SetHasExternalStorage (clang_type, false);
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit *dwarf_cu = debug_info->GetCompileUnitContainingDIE (die->GetOffset()).get();
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
Type *type = m_die_to_type.lookup (die);
const dw_tag_t tag = die->Tag();
Added the ability to log when things get completed in DWARF. This can be enabled using: (lldb) log enable --verbose lldb completion This will print out backtraces for all type completion calls which will help us verify that we don't ever complete a type when we don't need to. llvm-svn: 153787
2012-03-30 23:50:54 +00:00
LogSP log (LogChannelDWARF::GetLogIfAny(DWARF_LOG_DEBUG_INFO|DWARF_LOG_TYPE_COMPLETION));
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
if (log)
Added the ability to log when things get completed in DWARF. This can be enabled using: (lldb) log enable --verbose lldb completion This will print out backtraces for all type completion calls which will help us verify that we don't ever complete a type when we don't need to. llvm-svn: 153787
2012-03-30 23:50:54 +00:00
{
Added the ability to log a message with a backtrace when verbose logging is enabled to the Module class. Used this new function in the DWARF parser. llvm-svn: 155404
2012-04-23 22:55:20 +00:00
GetObjectFile()->GetModule()->LogMessageVerboseBacktrace (log.get(),
"0x%8.8llx: %s '%s' resolving forward declaration...",
MakeUserID(die->GetOffset()),
DW_TAG_value_to_name(tag),
type->GetName().AsCString());
Added the ability to log when things get completed in DWARF. This can be enabled using: (lldb) log enable --verbose lldb completion This will print out backtraces for all type completion calls which will help us verify that we don't ever complete a type when we don't need to. llvm-svn: 153787
2012-03-30 23:50:54 +00:00
}
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
assert (clang_type);
DWARFDebugInfoEntry::Attributes attributes;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ClangASTContext &ast = GetClangASTContext();
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
switch (tag)
{
case DW_TAG_structure_type:
case DW_TAG_union_type:
case DW_TAG_class_type:
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
{
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
LayoutInfo layout_info;
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
if (die->HasChildren())
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
LanguageType class_language = eLanguageTypeUnknown;
bool is_objc_class = ClangASTContext::IsObjCClassType (clang_type);
if (is_objc_class)
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
class_language = eLanguageTypeObjC;
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
// For objective C we don't start the definition when
// the class is created.
ast.StartTagDeclarationDefinition (clang_type);
}
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
int tag_decl_kind = -1;
AccessType default_accessibility = eAccessNone;
if (tag == DW_TAG_structure_type)
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
tag_decl_kind = clang::TTK_Struct;
default_accessibility = eAccessPublic;
Added support for the new ".apple_objc" accelerator tables. These tables are in the same hashed format as the ".apple_names", but they map objective C class names to all of the methods and class functions. We need to do this because in the DWARF the methods for Objective C are never contained in the class definition, they are scattered about at the translation unit level and they don't even have attributes that say the are contained within the class itself. Added 3 new formats which can be used to display data: eFormatAddressInfo eFormatHexFloat eFormatInstruction eFormatAddressInfo describes an address such as function+offset and file+line, or symbol + offset, or constant data (c string, 2, 4, 8, or 16 byte constants). The format character for this is "A", the long format is "address". eFormatHexFloat will print out the hex float format that compilers tend to use. The format character for this is "X", the long format is "hex float". eFormatInstruction will print out disassembly with bytes and it will use the current target's architecture. The format character for this is "i" (which used to be being used for the integer format, but the integer format also has "d", so we gave the "i" format to disassembly), the long format is "instruction". Mate the lldb::FormatterChoiceCriterion enumeration private as it should have been from the start. It is very specialized and doesn't belong in the public API. llvm-svn: 143114
2011-10-27 17:55:14 +00:00
}
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
else if (tag == DW_TAG_union_type)
{
tag_decl_kind = clang::TTK_Union;
default_accessibility = eAccessPublic;
}
else if (tag == DW_TAG_class_type)
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
tag_decl_kind = clang::TTK_Class;
default_accessibility = eAccessPrivate;
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolContext sc(GetCompUnitForDWARFCompUnit(dwarf_cu));
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
std::vector<clang::CXXBaseSpecifier *> base_classes;
std::vector<int> member_accessibilities;
bool is_a_class = false;
// Parse members and base classes first
DWARFDIECollection member_function_dies;
ParseChildMembers (sc,
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
dwarf_cu,
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
die,
clang_type,
class_language,
base_classes,
member_accessibilities,
member_function_dies,
default_accessibility,
is_a_class,
layout_info);
// Now parse any methods if there were any...
size_t num_functions = member_function_dies.Size();
if (num_functions > 0)
{
for (size_t i=0; i<num_functions; ++i)
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
ResolveType(dwarf_cu, member_function_dies.GetDIEPtrAtIndex(i));
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
}
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
}
if (class_language == eLanguageTypeObjC)
{
<rdar://problem/11113279> Fixed type lookups to "do the right thing". Prior to this fix, looking up a type using "foo::bar" would result in a type list that contains all types that had "bar" as a basename unless the symbol file was able to match fully qualified names (which our DWARF parser does not). This fix will allow type matches to be made based on the basename and then have the types that don't match filtered out. Types by name can be fully qualified, or partially qualified with the new "bool exact_match" parameter to the Module::FindTypes() method. This fixes some issue that we discovered with dynamic type resolution as well as improves the overall type lookups in LLDB. llvm-svn: 153482
2012-03-26 23:03:23 +00:00
std::string class_str (ClangASTType::GetTypeNameForOpaqueQualType(ast.getASTContext(), clang_type));
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
if (!class_str.empty())
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
DIEArray method_die_offsets;
if (m_using_apple_tables)
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
if (m_apple_objc_ap.get())
m_apple_objc_ap->FindByName(class_str.c_str(), method_die_offsets);
}
else
{
if (!m_indexed)
Index ();
ConstString class_name (class_str.c_str());
m_objc_class_selectors_index.Find (class_name, method_die_offsets);
}
if (!method_die_offsets.empty())
{
DWARFDebugInfo* debug_info = DebugInfo();
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
DWARFCompileUnit* method_cu = NULL;
const size_t num_matches = method_die_offsets.size();
for (size_t i=0; i<num_matches; ++i)
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
const dw_offset_t die_offset = method_die_offsets[i];
DWARFDebugInfoEntry *method_die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &method_cu);
if (method_die)
ResolveType (method_cu, method_die);
else
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
{
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
if (m_using_apple_tables)
{
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_objc accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, class_str.c_str());
}
}
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
}
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
}
}
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
// If we have a DW_TAG_structure_type instead of a DW_TAG_class_type we
// need to tell the clang type it is actually a class.
if (class_language != eLanguageTypeObjC)
{
if (is_a_class && tag_decl_kind != clang::TTK_Class)
ast.SetTagTypeKind (clang_type, clang::TTK_Class);
}
// Since DW_TAG_structure_type gets used for both classes
// and structures, we may need to set any DW_TAG_member
// fields to have a "private" access if none was specified.
// When we parsed the child members we tracked that actual
// accessibility value for each DW_TAG_member in the
// "member_accessibilities" array. If the value for the
// member is zero, then it was set to the "default_accessibility"
// which for structs was "public". Below we correct this
// by setting any fields to "private" that weren't correctly
// set.
if (is_a_class && !member_accessibilities.empty())
{
// This is a class and all members that didn't have
// their access specified are private.
ast.SetDefaultAccessForRecordFields (clang_type,
eAccessPrivate,
&member_accessibilities.front(),
member_accessibilities.size());
}
if (!base_classes.empty())
{
ast.SetBaseClassesForClassType (clang_type,
&base_classes.front(),
base_classes.size());
// Clang will copy each CXXBaseSpecifier in "base_classes"
// so we have to free them all.
ClangASTContext::DeleteBaseClassSpecifiers (&base_classes.front(),
base_classes.size());
}
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
}
Improved the type's handling of anonymous structs, so that the expression parser can look up members of anonymous structs correctly. This meant creating all the proper IndirectFieldDecls in each Record after it has been completely populated with members. llvm-svn: 151868
2012-03-02 01:03:45 +00:00
ast.BuildIndirectFields (clang_type);
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
ast.CompleteTagDeclarationDefinition (clang_type);
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
if (!layout_info.field_offsets.empty())
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
{
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
if (type)
layout_info.bit_size = type->GetByteSize() * 8;
if (layout_info.bit_size == 0)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
layout_info.bit_size = die->GetAttributeValueAsUnsigned(this, dwarf_cu, DW_AT_byte_size, 0) * 8;
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
clang::QualType qual_type(clang::QualType::getFromOpaquePtr(clang_type));
const clang::RecordType *record_type = clang::dyn_cast<clang::RecordType>(qual_type.getTypePtr());
if (record_type)
{
const clang::RecordDecl *record_decl = record_type->getDecl();
if (log)
Added logging so we can see the field names and offsets of any structures for when we enable the assisted layout. llvm-svn: 149167
2012-01-28 02:22:27 +00:00
{
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
Added logging so we can see the field names and offsets of any structures for when we enable the assisted layout. llvm-svn: 149167
2012-01-28 02:22:27 +00:00
"SymbolFileDWARF::ResolveClangOpaqueTypeDefinition (clang_type = %p) caching layout info for record_decl = %p, bit_size = %llu, alignment = %llu, field_offsets[%u], base_offsets[0], vbase_offsets[0])",
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
clang_type,
record_decl,
layout_info.bit_size,
layout_info.alignment,
(uint32_t)layout_info.field_offsets.size());
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
Added logging so we can see the field names and offsets of any structures for when we enable the assisted layout. llvm-svn: 149167
2012-01-28 02:22:27 +00:00
llvm::DenseMap <const clang::FieldDecl *, uint64_t>::const_iterator pos, end = layout_info.field_offsets.end();
for (pos = layout_info.field_offsets.begin(); pos != end; ++pos)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::ResolveClangOpaqueTypeDefinition (clang_type = %p) field = { bit_offset=%u, name='%s' }",
clang_type,
(uint32_t)pos->second,
pos->first->getNameAsString().c_str());
}
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
m_record_decl_to_layout_map.insert(std::make_pair(record_decl, layout_info));
}
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
}
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
}
Moved byte-size computation out of a startDefinition() ... endDefinition() block, preventing crashes where the byte size of a not-yet-complete type was being computed. llvm-svn: 151546
2012-02-27 20:07:01 +00:00
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
return clang_type;
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
case DW_TAG_enumeration_type:
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ast.StartTagDeclarationDefinition (clang_type);
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
if (die->HasChildren())
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolContext sc(GetCompUnitForDWARFCompUnit(dwarf_cu));
ParseChildEnumerators(sc, clang_type, type->GetByteSize(), dwarf_cu, die);
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
}
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ast.CompleteTagDeclarationDefinition (clang_type);
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
return clang_type;
default:
assert(false && "not a forward clang type decl!");
break;
}
return NULL;
}
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
Type*
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolFileDWARF::ResolveType (DWARFCompileUnit* dwarf_cu, const DWARFDebugInfoEntry* type_die, bool assert_not_being_parsed)
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
{
if (type_die != NULL)
{
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
Type *type = m_die_to_type.lookup (type_die);
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
if (type == NULL)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
type = GetTypeForDIE (dwarf_cu, type_die).get();
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (assert_not_being_parsed)
Don't assert but report & return a NULL type if we end up parsing a type we are in the middle of parsing. llvm-svn: 147914
2012-01-11 02:21:12 +00:00
{
if (type != DIE_IS_BEING_PARSED)
return type;
GetObjectFile()->GetModule()->ReportError ("Parsing a die that is being parsed die: 0x%8.8x: %s %s",
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
type_die->GetOffset(),
DW_TAG_value_to_name(type_die->Tag()),
type_die->GetName(this, dwarf_cu));
Don't assert but report & return a NULL type if we end up parsing a type we are in the middle of parsing. llvm-svn: 147914
2012-01-11 02:21:12 +00:00
}
else
return type;
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
}
return NULL;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
CompileUnit*
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolFileDWARF::GetCompUnitForDWARFCompUnit (DWARFCompileUnit* dwarf_cu, uint32_t cu_idx)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// Check if the symbol vendor already knows about this compile unit?
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
if (dwarf_cu->GetUserData() == NULL)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// The symbol vendor doesn't know about this compile unit, we
// need to parse and add it to the symbol vendor object.
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
return ParseCompileUnit(dwarf_cu, cu_idx).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
return (CompileUnit*)dwarf_cu->GetUserData();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
bool
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolFileDWARF::GetFunction (DWARFCompileUnit* dwarf_cu, const DWARFDebugInfoEntry* func_die, SymbolContext& sc)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
sc.Clear();
// Check if the symbol vendor already knows about this compile unit?
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, UINT32_MAX);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.function = sc.comp_unit->FindFunctionByUID (MakeUserID(func_die->GetOffset())).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (sc.function == NULL)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.function = ParseCompileUnitFunction(sc, dwarf_cu, func_die);
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
if (sc.function)
{
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
sc.module_sp = sc.function->CalculateSymbolContextModule();
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
return true;
}
return false;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
uint32_t
SymbolFileDWARF::ResolveSymbolContext (const Address& so_addr, uint32_t resolve_scope, SymbolContext& sc)
{
Timer scoped_timer(__PRETTY_FUNCTION__,
"SymbolFileDWARF::ResolveSymbolContext (so_addr = { section = %p, offset = 0x%llx }, resolve_scope = 0x%8.8x)",
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
so_addr.GetSection().get(),
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
so_addr.GetOffset(),
resolve_scope);
uint32_t resolved = 0;
if (resolve_scope & ( eSymbolContextCompUnit |
eSymbolContextFunction |
eSymbolContextBlock |
eSymbolContextLineEntry))
{
lldb::addr_t file_vm_addr = so_addr.GetFileAddress();
DWARFDebugInfo* debug_info = DebugInfo();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
if (debug_info)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
dw_offset_t cu_offset = debug_info->GetCompileUnitAranges().FindAddress(file_vm_addr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (cu_offset != DW_INVALID_OFFSET)
{
uint32_t cu_idx;
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = debug_info->GetCompileUnit(cu_offset, &cu_idx).get();
if (dwarf_cu)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, cu_idx);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
assert(sc.comp_unit != NULL);
resolved |= eSymbolContextCompUnit;
if (resolve_scope & eSymbolContextLineEntry)
{
LineTable *line_table = sc.comp_unit->GetLineTable();
if (line_table != NULL)
{
if (so_addr.IsLinkedAddress())
{
Address linked_addr (so_addr);
linked_addr.ResolveLinkedAddress();
if (line_table->FindLineEntryByAddress (linked_addr, sc.line_entry))
{
resolved |= eSymbolContextLineEntry;
}
}
else if (line_table->FindLineEntryByAddress (so_addr, sc.line_entry))
{
resolved |= eSymbolContextLineEntry;
}
}
}
if (resolve_scope & (eSymbolContextFunction | eSymbolContextBlock))
{
DWARFDebugInfoEntry *function_die = NULL;
DWARFDebugInfoEntry *block_die = NULL;
if (resolve_scope & eSymbolContextBlock)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
dwarf_cu->LookupAddress(file_vm_addr, &function_die, &block_die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
else
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
dwarf_cu->LookupAddress(file_vm_addr, &function_die, NULL);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
if (function_die != NULL)
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.function = sc.comp_unit->FindFunctionByUID (MakeUserID(function_die->GetOffset())).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (sc.function == NULL)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.function = ParseCompileUnitFunction(sc, dwarf_cu, function_die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/10103468> Symbol files (dSYM files on darwin) can now be specified during program execution: (lldb) target symbols add /path/to/symfile/a.out.dSYM/Contents/Resources/DWARF/a.out This command can be used when you have a debug session in progress and want to add symbols to get better debug info fidelity. llvm-svn: 153693
2012-03-29 21:43:25 +00:00
else
{
// We might have had a compile unit that had discontiguous
// address ranges where the gaps are symbols that don't have
// any debug info. Discontiguous compile unit address ranges
// should only happen when there aren't other functions from
// other compile units in these gaps. This helps keep the size
// of the aranges down.
sc.comp_unit = NULL;
resolved &= ~eSymbolContextCompUnit;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (sc.function != NULL)
{
resolved |= eSymbolContextFunction;
if (resolve_scope & eSymbolContextBlock)
{
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
Block& block = sc.function->GetBlock (true);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (block_die != NULL)
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.block = block.FindBlockByID (MakeUserID(block_die->GetOffset()));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.block = block.FindBlockByID (MakeUserID(function_die->GetOffset()));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (sc.block)
resolved |= eSymbolContextBlock;
}
}
}
}
}
}
}
return resolved;
}
uint32_t
SymbolFileDWARF::ResolveSymbolContext(const FileSpec& file_spec, uint32_t line, bool check_inlines, uint32_t resolve_scope, SymbolContextList& sc_list)
{
const uint32_t prev_size = sc_list.GetSize();
if (resolve_scope & eSymbolContextCompUnit)
{
DWARFDebugInfo* debug_info = DebugInfo();
if (debug_info)
{
uint32_t cu_idx;
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
for (cu_idx = 0; (dwarf_cu = debug_info->GetCompileUnitAtIndex(cu_idx)) != NULL; ++cu_idx)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompileUnit *dc_cu = GetCompUnitForDWARFCompUnit(dwarf_cu, cu_idx);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
bool file_spec_matches_cu_file_spec = dc_cu != NULL && FileSpec::Compare(file_spec, *dc_cu, false) == 0;
if (check_inlines || file_spec_matches_cu_file_spec)
{
SymbolContext sc (m_obj_file->GetModule());
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, cu_idx);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
assert(sc.comp_unit != NULL);
uint32_t file_idx = UINT32_MAX;
// If we are looking for inline functions only and we don't
// find it in the support files, we are done.
if (check_inlines)
{
Added the ability to restrict breakpoints by function name, function regexp, selector etc to specific source files. Added SB API's to specify these source files & also more than one module. Added an "exact" option to CompileUnit's FindLineEntry API. llvm-svn: 140362
2011-09-23 00:54:11 +00:00
file_idx = sc.comp_unit->GetSupportFiles().FindFileIndex (1, file_spec, true);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (file_idx == UINT32_MAX)
continue;
}
if (line != 0)
{
LineTable *line_table = sc.comp_unit->GetLineTable();
if (line_table != NULL && line != 0)
{
// We will have already looked up the file index if
// we are searching for inline entries.
if (!check_inlines)
Added the ability to restrict breakpoints by function name, function regexp, selector etc to specific source files. Added SB API's to specify these source files & also more than one module. Added an "exact" option to CompileUnit's FindLineEntry API. llvm-svn: 140362
2011-09-23 00:54:11 +00:00
file_idx = sc.comp_unit->GetSupportFiles().FindFileIndex (1, file_spec, true);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (file_idx != UINT32_MAX)
{
uint32_t found_line;
uint32_t line_idx = line_table->FindLineEntryIndexByFileIndex (0, file_idx, line, false, &sc.line_entry);
found_line = sc.line_entry.line;
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 02:20:48 +00:00
while (line_idx != UINT32_MAX)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
sc.function = NULL;
sc.block = NULL;
if (resolve_scope & (eSymbolContextFunction | eSymbolContextBlock))
{
const lldb::addr_t file_vm_addr = sc.line_entry.range.GetBaseAddress().GetFileAddress();
if (file_vm_addr != LLDB_INVALID_ADDRESS)
{
DWARFDebugInfoEntry *function_die = NULL;
DWARFDebugInfoEntry *block_die = NULL;
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
dwarf_cu->LookupAddress(file_vm_addr, &function_die, resolve_scope & eSymbolContextBlock ? &block_die : NULL);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (function_die != NULL)
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.function = sc.comp_unit->FindFunctionByUID (MakeUserID(function_die->GetOffset())).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (sc.function == NULL)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
sc.function = ParseCompileUnitFunction(sc, dwarf_cu, function_die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
if (sc.function != NULL)
{
Modified the host process monitor callback function Host::StartMonitoringChildProcess to spawn a thread for each process that is being monitored. Previously LLDB would spawn a single thread that would wait for any child process which isn't ok to do as a shared library (LLDB.framework on Mac OSX, or lldb.so on linux). The old single thread used to call wait4() with a pid of -1 which could cause it to reap child processes that it shouldn't have. Re-wrote the way Function blocks are handles. Previously I attempted to keep all blocks in a single memory allocation (in a std::vector). This made the code somewhat efficient, but hard to work with. I got rid of the old BlockList class, and went to a straight parent with children relationship. This new approach will allow for partial parsing of the blocks within a function. llvm-svn: 111706
2010-08-21 02:22:51 +00:00
Block& block = sc.function->GetBlock (true);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (block_die != NULL)
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.block = block.FindBlockByID (MakeUserID(block_die->GetOffset()));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.block = block.FindBlockByID (MakeUserID(function_die->GetOffset()));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
sc_list.Append(sc);
line_idx = line_table->FindLineEntryIndexByFileIndex (line_idx + 1, file_idx, found_line, true, &sc.line_entry);
}
}
}
else if (file_spec_matches_cu_file_spec && !check_inlines)
{
// only append the context if we aren't looking for inline call sites
// by file and line and if the file spec matches that of the compile unit
sc_list.Append(sc);
}
}
else if (file_spec_matches_cu_file_spec && !check_inlines)
{
// only append the context if we aren't looking for inline call sites
// by file and line and if the file spec matches that of the compile unit
sc_list.Append(sc);
}
if (!check_inlines)
break;
}
}
}
}
return sc_list.GetSize() - prev_size;
}
void
SymbolFileDWARF::Index ()
{
if (m_indexed)
return;
m_indexed = true;
Timer scoped_timer (__PRETTY_FUNCTION__,
"SymbolFileDWARF::Index (%s)",
GetObjectFile()->GetFileSpec().GetFilename().AsCString());
DWARFDebugInfo* debug_info = DebugInfo();
if (debug_info)
{
uint32_t cu_idx = 0;
const uint32_t num_compile_units = GetNumCompileUnits();
for (cu_idx = 0; cu_idx < num_compile_units; ++cu_idx)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = debug_info->GetCompileUnitAtIndex(cu_idx);
bool clear_dies = dwarf_cu->ExtractDIEsIfNeeded (false) > 1;
dwarf_cu->Index (cu_idx,
m_function_basename_index,
m_function_fullname_index,
m_function_method_index,
m_function_selector_index,
m_objc_class_selectors_index,
m_global_index,
m_type_index,
m_namespace_index);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// Keep memory down by clearing DIEs if this generate function
// caused them to be parsed
if (clear_dies)
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
dwarf_cu->ClearDIEs (true);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
m_function_basename_index.Finalize();
m_function_fullname_index.Finalize();
m_function_method_index.Finalize();
m_function_selector_index.Finalize();
m_objc_class_selectors_index.Finalize();
m_global_index.Finalize();
m_type_index.Finalize();
m_namespace_index.Finalize();
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
#if defined (ENABLE_DEBUG_PRINTF)
Modified version of a patch from Warren Paul that takes care of issues with indirect forms, deals with empty DW_AT_comp_dir attributes, and fixups for handling other signed integer types. llvm-svn: 125240
2011-02-09 23:39:34 +00:00
StreamFile s(stdout, false);
Added support for accessing and loading our new .debug_names and .debug_types DWARF accelerator table sections to the DWARF parser. These sections are similar to the .debug_pubnames and .debug_pubtypes, but they are designed to be hash tables that are saved to disc in a way that the sections can just be loaded into memory and used without any work on the debugger side. The .debug_pubnames and .debug_pubtypes sections are not ordered, contain a copy of the name in the section itself which makes these sections quite large, they only include publicly exported names (so no static functions, no types defined inside functions), many compilers put different information in them making them very unreliable so most debugger ignore these sections and parse the DWARF on their own. The tables must also be parsed and sorted in order to be used effectively. The new sections can be quickly loaded and very efficiently be used to do name to DIE lookups with very little up front work. The format of these new sections will be changing while we work out the bugs, but we hope to have really fast name to DIE lookups soon. llvm-svn: 138979
2011-09-01 23:16:13 +00:00
s.Printf ("DWARF index for '%s/%s':",
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
GetObjectFile()->GetFileSpec().GetDirectory().AsCString(),
GetObjectFile()->GetFileSpec().GetFilename().AsCString());
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
s.Printf("\nFunction basenames:\n"); m_function_basename_index.Dump (&s);
s.Printf("\nFunction fullnames:\n"); m_function_fullname_index.Dump (&s);
s.Printf("\nFunction methods:\n"); m_function_method_index.Dump (&s);
s.Printf("\nFunction selectors:\n"); m_function_selector_index.Dump (&s);
s.Printf("\nObjective C class selectors:\n"); m_objc_class_selectors_index.Dump (&s);
s.Printf("\nGlobals and statics:\n"); m_global_index.Dump (&s);
Separated the DWARF index for types from that the index of the namespaces since we can't parse DW_TAG_namespace DIEs as types. They are only decls in clang. All of the types we handle right now have both clang "XXXType" classes to go with the "XXXDecl" classes which means they can be used within the lldb_private::Type class. I need to check to see which other decls that don't have associated type objects need to float around the debugger and possibly make a lldb_private::Decl class to manage them. llvm-svn: 116558
2010-10-15 02:03:22 +00:00
s.Printf("\nTypes:\n"); m_type_index.Dump (&s);
Fixed an issue where we might not find global variables by name when we have a debug map with DWARF in the .o files due to the attemted shortcut that was being taken where the global variables were being searched for by looking in the symbol table. The problem with the symbols in the symbol table is we don't break apart the symbol names for symbols when they are mangled into basename and the fully mangled name since this would take a lot of CPU time to chop up the mangled names and try and find the basenames. The DWARF info typically has this broken up for us where the basename of the variable is in a the DW_AT_name attribute, and the mangled name is in the DW_AT_MIPS_linkage_name attribute. Now we correctly find globals by searching all OSO's for the information so we can take advantage of this split information. llvm-svn: 119012
2010-11-13 22:57:37 +00:00
s.Printf("\nNamepaces:\n"); m_namespace_index.Dump (&s);
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
#endif
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Added a function to test if a ClangNamespaceDecl matches the current symbol file. This will help us to minimize lookups that can't possibly match anything in the current symbol file. llvm-svn: 141838
2011-10-13 00:00:53 +00:00
bool
SymbolFileDWARF::NamespaceDeclMatchesThisSymbolFile (const ClangNamespaceDecl *namespace_decl)
{
if (namespace_decl == NULL)
{
// Invalid namespace decl which means we aren't matching only things
// in this symbol file, so return true to indicate it matches this
// symbol file.
return true;
}
clang::ASTContext *namespace_ast = namespace_decl->GetASTContext();
if (namespace_ast == NULL)
return true; // No AST in the "namespace_decl", return true since it
// could then match any symbol file, including this one
if (namespace_ast == GetClangASTContext().getASTContext())
return true; // The ASTs match, return true
// The namespace AST was valid, and it does not match...
Moved the list of found namespaces into the search context object. Having it populated and registered within a single FindExternalVisibleDecls call worked fine when there was only one call (i.e., when we were just looking in the global namespace). However, now FindExternalVisibleDecls is called for nested namespaces as well, which means that it is called not once but many times (once per module in which the parent namespace appears). This means that the namespace mapping is built up across many calls to the inferior FindExternalVisibleDecls, so I moved it into a data structure (the search context) that is shared by all calls. I also added some logging to make it easier to see what is happening during a namespace search, and cleaned up some existing logging. llvm-svn: 141888
2011-10-13 21:08:11 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage(log.get(), "Valid namespace does not match symbol file");
Moved the list of found namespaces into the search context object. Having it populated and registered within a single FindExternalVisibleDecls call worked fine when there was only one call (i.e., when we were just looking in the global namespace). However, now FindExternalVisibleDecls is called for nested namespaces as well, which means that it is called not once but many times (once per module in which the parent namespace appears). This means that the namespace mapping is built up across many calls to the inferior FindExternalVisibleDecls, so I moved it into a data structure (the search context) that is shared by all calls. I also added some logging to make it easier to see what is happening during a namespace search, and cleaned up some existing logging. llvm-svn: 141888
2011-10-13 21:08:11 +00:00
Added a function to test if a ClangNamespaceDecl matches the current symbol file. This will help us to minimize lookups that can't possibly match anything in the current symbol file. llvm-svn: 141838
2011-10-13 00:00:53 +00:00
return false;
}
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
bool
SymbolFileDWARF::DIEIsInNamespace (const ClangNamespaceDecl *namespace_decl,
DWARFCompileUnit* cu,
const DWARFDebugInfoEntry* die)
{
// No namespace specified, so the answesr i
if (namespace_decl == NULL)
return true;
Enabled the namespace-specific search functionality, which had previously been commented out while I tested it. It's not fully working yet, but it doesn't break our testsuite and it's an important piece of functionality. Also added some logging to SymbolFileDWARF to help diagnose entities that are found in a symbol file, but do not reside in the expected namespace. llvm-svn: 141894
2011-10-13 21:50:33 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
Added a function to test if a ClangNamespaceDecl matches the current symbol file. This will help us to minimize lookups that can't possibly match anything in the current symbol file. llvm-svn: 141838
2011-10-13 00:00:53 +00:00
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const DWARFDebugInfoEntry *decl_ctx_die = NULL;
clang::DeclContext *die_clang_decl_ctx = GetClangDeclContextContainingDIE (cu, die, &decl_ctx_die);
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
if (decl_ctx_die)
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
{
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
clang::NamespaceDecl *clang_namespace_decl = namespace_decl->GetNamespaceDecl();
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
if (clang_namespace_decl)
{
if (decl_ctx_die->Tag() != DW_TAG_namespace)
Enabled the namespace-specific search functionality, which had previously been commented out while I tested it. It's not fully working yet, but it doesn't break our testsuite and it's an important piece of functionality. Also added some logging to SymbolFileDWARF to help diagnose entities that are found in a symbol file, but do not reside in the expected namespace. llvm-svn: 141894
2011-10-13 21:50:33 +00:00
{
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage(log.get(), "Found a match, but its parent is not a namespace");
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
return false;
Enabled the namespace-specific search functionality, which had previously been commented out while I tested it. It's not fully working yet, but it doesn't break our testsuite and it's an important piece of functionality. Also added some logging to SymbolFileDWARF to help diagnose entities that are found in a symbol file, but do not reside in the expected namespace. llvm-svn: 141894
2011-10-13 21:50:33 +00:00
}
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
if (clang_namespace_decl == die_clang_decl_ctx)
return true;
else
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
return false;
}
else
{
// We have a namespace_decl that was not NULL but it contained
// a NULL "clang::NamespaceDecl", so this means the global namespace
// So as long the the contained decl context DIE isn't a namespace
// we should be ok.
if (decl_ctx_die->Tag() != DW_TAG_namespace)
return true;
}
}
Enabled the namespace-specific search functionality, which had previously been commented out while I tested it. It's not fully working yet, but it doesn't break our testsuite and it's an important piece of functionality. Also added some logging to SymbolFileDWARF to help diagnose entities that are found in a symbol file, but do not reside in the expected namespace. llvm-svn: 141894
2011-10-13 21:50:33 +00:00
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage(log.get(), "Found a match, but its parent doesn't exist");
Enabled the namespace-specific search functionality, which had previously been commented out while I tested it. It's not fully working yet, but it doesn't break our testsuite and it's an important piece of functionality. Also added some logging to SymbolFileDWARF to help diagnose entities that are found in a symbol file, but do not reside in the expected namespace. llvm-svn: 141894
2011-10-13 21:50:33 +00:00
Added the ability to see if a DIE is contained in a namespace. llvm-svn: 141832
2011-10-12 23:34:26 +00:00
return false;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
SymbolFileDWARF::FindGlobalVariables (const ConstString &name, const lldb_private::ClangNamespaceDecl *namespace_decl, bool append, uint32_t max_matches, VariableList& variables)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindGlobalVariables (name=\"%s\", namespace_decl=%p, append=%u, max_matches=%u, variables)",
name.GetCString(),
namespace_decl,
append,
max_matches);
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (!NamespaceDeclMatchesThisSymbolFile(namespace_decl))
return 0;
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
DWARFDebugInfo* info = DebugInfo();
if (info == NULL)
return 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// If we aren't appending the results to this list, then clear the list
if (!append)
variables.Clear();
// Remember how many variables are in the list before we search in case
// we are appending the results to a variable list.
const uint32_t original_size = variables.GetSize();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DIEArray die_offsets;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_names_ap.get())
{
const char *name_cstr = name.GetCString();
const char *base_name_start;
const char *base_name_end = NULL;
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (!CPPLanguageRuntime::StripNamespacesFromVariableName(name_cstr, base_name_start, base_name_end))
base_name_start = name_cstr;
m_apple_names_ap->FindByName (base_name_start, die_offsets);
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
// Index the DWARF if we haven't already
if (!m_indexed)
Index ();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_global_index.Find (name, die_offsets);
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const size_t num_die_matches = die_offsets.size();
if (num_die_matches)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
SymbolContext sc;
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
sc.module_sp = m_obj_file->GetModule();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
assert (sc.module_sp);
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DWARFCompileUnit* dwarf_cu = NULL;
const DWARFDebugInfoEntry* die = NULL;
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
bool done = false;
for (size_t i=0; i<num_die_matches && !done; ++i)
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
{
const dw_offset_t die_offset = die_offsets[i];
die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (die)
{
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
switch (die->Tag())
{
default:
case DW_TAG_subprogram:
case DW_TAG_inlined_subroutine:
case DW_TAG_try_block:
case DW_TAG_catch_block:
break;
case DW_TAG_variable:
{
sc.comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, UINT32_MAX);
assert(sc.comp_unit != NULL);
if (namespace_decl && !DIEIsInNamespace (namespace_decl, dwarf_cu, die))
continue;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
ParseVariables(sc, dwarf_cu, LLDB_INVALID_ADDRESS, die, false, false, &variables);
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
if (variables.GetSize() - original_size >= max_matches)
done = true;
}
break;
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, name.GetCString());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
// Return the number of variable that were appended to the list
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const uint32_t num_matches = variables.GetSize() - original_size;
if (log && num_matches > 0)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindGlobalVariables (name=\"%s\", namespace_decl=%p, append=%u, max_matches=%u, variables) => %u",
name.GetCString(),
namespace_decl,
append,
max_matches,
num_matches);
}
return num_matches;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
uint32_t
SymbolFileDWARF::FindGlobalVariables(const RegularExpression& regex, bool append, uint32_t max_matches, VariableList& variables)
{
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindGlobalVariables (regex=\"%s\", append=%u, max_matches=%u, variables)",
regex.GetText(),
append,
max_matches);
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
DWARFDebugInfo* info = DebugInfo();
if (info == NULL)
return 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// If we aren't appending the results to this list, then clear the list
if (!append)
variables.Clear();
// Remember how many variables are in the list before we search in case
// we are appending the results to a variable list.
const uint32_t original_size = variables.GetSize();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DIEArray die_offsets;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_names_ap.get())
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
{
DWARFMappedHash::DIEInfoArray hash_data_array;
if (m_apple_names_ap->AppendAllDIEsThatMatchingRegex (regex, hash_data_array))
DWARFMappedHash::ExtractDIEArray (hash_data_array, die_offsets);
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
// Index the DWARF if we haven't already
if (!m_indexed)
Index ();
m_global_index.Find (regex, die_offsets);
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
SymbolContext sc;
<rdar://problem/10103468> I started work on being able to add symbol files after a debug session had started with a new "target symfile add" command and quickly ran into problems with stale Address objects in breakpoint locations that had lldb_private::Section pointers into modules that had been removed or replaced. This also let to grabbing stale modules from those sections. So I needed to thread harded the Address, Section and related objects. To do this I modified the ModuleChild class to now require a ModuleSP on initialization so that a weak reference can created. I also changed all places that were handing out "Section *" to have them hand out SectionSP. All ObjectFile, SymbolFile and SymbolVendors were inheriting from ModuleChild so all of the find plug-in, static creation function and constructors now require ModuleSP references instead of Module *. Address objects now have weak references to their sections which can safely go stale when a module gets destructed. This checkin doesn't complete the "target symfile add" command, but it does get us a lot clioser to being able to do such things without a high risk of crashing or memory corruption. llvm-svn: 151336
2012-02-24 01:59:29 +00:00
sc.module_sp = m_obj_file->GetModule();
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
assert (sc.module_sp);
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFCompileUnit* dwarf_cu = NULL;
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
const DWARFDebugInfoEntry* die = NULL;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const size_t num_matches = die_offsets.size();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
if (num_matches)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
for (size_t i=0; i<num_matches; ++i)
{
const dw_offset_t die_offset = die_offsets[i];
die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (die)
{
sc.comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, UINT32_MAX);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
ParseVariables(sc, dwarf_cu, LLDB_INVALID_ADDRESS, die, false, false, &variables);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (variables.GetSize() - original_size >= max_matches)
break;
}
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x for regex '%s')\n",
die_offset, regex.GetText());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
// Return the number of variable that were appended to the list
return variables.GetSize() - original_size;
}
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
bool
SymbolFileDWARF::ResolveFunction (dw_offset_t die_offset,
DWARFCompileUnit *&dwarf_cu,
SymbolContextList& sc_list)
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
{
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
const DWARFDebugInfoEntry *die = DebugInfo()->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
return ResolveFunction (dwarf_cu, die, sc_list);
}
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
bool
SymbolFileDWARF::ResolveFunction (DWARFCompileUnit *cu,
const DWARFDebugInfoEntry *die,
SymbolContextList& sc_list)
{
SymbolContext sc;
if (die == NULL)
return false;
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
// If we were passed a die that is not a function, just return false...
if (die->Tag() != DW_TAG_subprogram && die->Tag() != DW_TAG_inlined_subroutine)
return false;
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
const DWARFDebugInfoEntry* inlined_die = NULL;
if (die->Tag() == DW_TAG_inlined_subroutine)
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
{
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
inlined_die = die;
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
while ((die = die->GetParent()) != NULL)
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
{
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
if (die->Tag() == DW_TAG_subprogram)
break;
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
assert (die->Tag() == DW_TAG_subprogram);
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
if (GetFunction (cu, die, sc))
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Address addr;
// Parse all blocks if needed
if (inlined_die)
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
sc.block = sc.function->GetBlock (true).FindBlockByID (MakeUserID(inlined_die->GetOffset()));
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
assert (sc.block != NULL);
if (sc.block->GetStartAddress (addr) == false)
addr.Clear();
}
else
{
sc.block = NULL;
addr = sc.function->GetAddressRange().GetBaseAddress();
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
if (addr.IsValid())
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
sc_list.Append(sc);
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
return true;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
return false;
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
}
void
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
SymbolFileDWARF::FindFunctions (const ConstString &name,
const NameToDIE &name_to_die,
SymbolContextList& sc_list)
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DIEArray die_offsets;
if (name_to_die.Find (name, die_offsets))
{
ParseFunctions (die_offsets, sc_list);
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
void
SymbolFileDWARF::FindFunctions (const RegularExpression &regex,
const NameToDIE &name_to_die,
SymbolContextList& sc_list)
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DIEArray die_offsets;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (name_to_die.Find (regex, die_offsets))
{
ParseFunctions (die_offsets, sc_list);
}
}
void
SymbolFileDWARF::FindFunctions (const RegularExpression &regex,
const DWARFMappedHash::MemoryTable &memory_table,
SymbolContextList& sc_list)
{
DIEArray die_offsets;
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
DWARFMappedHash::DIEInfoArray hash_data_array;
if (memory_table.AppendAllDIEsThatMatchingRegex (regex, hash_data_array))
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
DWARFMappedHash::ExtractDIEArray (hash_data_array, die_offsets);
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
ParseFunctions (die_offsets, sc_list);
}
}
void
SymbolFileDWARF::ParseFunctions (const DIEArray &die_offsets,
SymbolContextList& sc_list)
{
const size_t num_matches = die_offsets.size();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
if (num_matches)
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
SymbolContext sc;
DWARFCompileUnit* dwarf_cu = NULL;
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
for (size_t i=0; i<num_matches; ++i)
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const dw_offset_t die_offset = die_offsets[i];
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
ResolveFunction (die_offset, dwarf_cu, sc_list);
}
}
}
bool
SymbolFileDWARF::FunctionDieMatchesPartialName (const DWARFDebugInfoEntry* die,
const DWARFCompileUnit *dwarf_cu,
uint32_t name_type_mask,
const char *partial_name,
const char *base_name_start,
const char *base_name_end)
{
// If we are looking only for methods, throw away all the ones that aren't in C++ classes:
if (name_type_mask == eFunctionNameTypeMethod
|| name_type_mask == eFunctionNameTypeBase)
{
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
clang::DeclContext *containing_decl_ctx = GetClangDeclContextContainingDIEOffset(die->GetOffset());
if (!containing_decl_ctx)
return false;
bool is_cxx_method = DeclKindIsCXXClass(containing_decl_ctx->getDeclKind());
if (!is_cxx_method && name_type_mask == eFunctionNameTypeMethod)
return false;
if (is_cxx_method && name_type_mask == eFunctionNameTypeBase)
return false;
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
// Now we need to check whether the name we got back for this type matches the extra specifications
// that were in the name we're looking up:
if (base_name_start != partial_name || *base_name_end != '\0')
{
// First see if the stuff to the left matches the full name. To do that let's see if
// we can pull out the mips linkage name attribute:
Mangled best_name;
DWARFDebugInfoEntry::Attributes attributes;
die->GetAttributes(this, dwarf_cu, NULL, attributes);
uint32_t idx = attributes.FindAttributeIndex(DW_AT_MIPS_linkage_name);
if (idx != UINT32_MAX)
{
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, idx, form_value))
Fixed an issue that would cause an assertion to fire when an inlined function was found during a regex function find call. llvm-svn: 120814
2010-12-03 17:49:14 +00:00
{
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
const char *name = form_value.AsCString(&get_debug_str_data());
best_name.SetValue (name, true);
}
}
if (best_name)
{
const char *demangled = best_name.GetDemangledName().GetCString();
if (demangled)
{
std::string name_no_parens(partial_name, base_name_end - partial_name);
Make sure breakpoint partial name matches occur on namespace boundaries. <rdar://problem/10720345> "break set -n" name matching should only match at namespace boundaries llvm-svn: 151876
2012-03-02 02:24:42 +00:00
const char *partial_in_demangled = strstr (demangled, name_no_parens.c_str());
if (partial_in_demangled == NULL)
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
return false;
Make sure breakpoint partial name matches occur on namespace boundaries. <rdar://problem/10720345> "break set -n" name matching should only match at namespace boundaries llvm-svn: 151876
2012-03-02 02:24:42 +00:00
else
{
// Sort out the case where our name is something like "Process::Destroy" and the match is
// "SBProcess::Destroy" - that shouldn't be a match. We should really always match on
// namespace boundaries...
if (partial_name[0] == ':' && partial_name[1] == ':')
{
// The partial name was already on a namespace boundary so all matches are good.
return true;
}
else if (partial_in_demangled == demangled)
{
// They both start the same, so this is an good match.
return true;
}
else
{
if (partial_in_demangled - demangled == 1)
{
// Only one character difference, can't be a namespace boundary...
return false;
}
else if (*(partial_in_demangled - 1) == ':' && *(partial_in_demangled - 2) == ':')
{
// We are on a namespace boundary, so this is also good.
return true;
}
else
return false;
}
}
Fixed an issue that would cause an assertion to fire when an inlined function was found during a regex function find call. llvm-svn: 120814
2010-12-03 17:49:14 +00:00
}
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
}
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
return true;
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
}
uint32_t
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
SymbolFileDWARF::FindFunctions (const ConstString &name,
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
const lldb_private::ClangNamespaceDecl *namespace_decl,
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
uint32_t name_type_mask,
bool include_inlines,
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
bool append,
SymbolContextList& sc_list)
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
{
Timer scoped_timer (__PRETTY_FUNCTION__,
"SymbolFileDWARF::FindFunctions (name = '%s')",
name.AsCString());
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindFunctions (name=\"%s\", name_type_mask=0x%x, append=%u, sc_list)",
name.GetCString(),
name_type_mask,
append);
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
// If we aren't appending the results to this list, then clear the list
if (!append)
sc_list.Clear();
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (!NamespaceDeclMatchesThisSymbolFile(namespace_decl))
return 0;
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
// If name is empty then we won't find anything.
if (name.IsEmpty())
return 0;
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
// Remember how many sc_list are in the list before we search in case
// we are appending the results to a variable list.
Added alpha .debug_names and .debug_types support in the DWARF parser. llvm-svn: 138996
2011-09-02 04:03:59 +00:00
const uint32_t original_size = sc_list.GetSize();
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
const char *name_cstr = name.GetCString();
uint32_t effective_name_type_mask = eFunctionNameTypeNone;
const char *base_name_start = name_cstr;
const char *base_name_end = name_cstr + strlen(name_cstr);
if (name_type_mask & eFunctionNameTypeAuto)
{
if (CPPLanguageRuntime::IsCPPMangledName (name_cstr))
effective_name_type_mask = eFunctionNameTypeFull;
else if (ObjCLanguageRuntime::IsPossibleObjCMethodName (name_cstr))
effective_name_type_mask = eFunctionNameTypeFull;
else
{
if (ObjCLanguageRuntime::IsPossibleObjCSelector(name_cstr))
effective_name_type_mask |= eFunctionNameTypeSelector;
if (CPPLanguageRuntime::IsPossibleCPPCall(name_cstr, base_name_start, base_name_end))
effective_name_type_mask |= (eFunctionNameTypeMethod | eFunctionNameTypeBase);
}
}
else
{
effective_name_type_mask = name_type_mask;
if (effective_name_type_mask & eFunctionNameTypeMethod || name_type_mask & eFunctionNameTypeBase)
{
// If they've asked for a CPP method or function name and it can't be that, we don't
// even need to search for CPP methods or names.
if (!CPPLanguageRuntime::IsPossibleCPPCall(name_cstr, base_name_start, base_name_end))
{
effective_name_type_mask &= ~(eFunctionNameTypeMethod | eFunctionNameTypeBase);
if (effective_name_type_mask == eFunctionNameTypeNone)
return 0;
}
}
if (effective_name_type_mask & eFunctionNameTypeSelector)
{
if (!ObjCLanguageRuntime::IsPossibleObjCSelector(name_cstr))
{
effective_name_type_mask &= ~(eFunctionNameTypeSelector);
if (effective_name_type_mask == eFunctionNameTypeNone)
return 0;
}
}
}
DWARFDebugInfo* info = DebugInfo();
if (info == NULL)
return 0;
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
DWARFCompileUnit *dwarf_cu = NULL;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_names_ap.get())
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
DIEArray die_offsets;
uint32_t num_matches = 0;
if (effective_name_type_mask & eFunctionNameTypeFull)
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
// If they asked for the full name, match what they typed. At some point we may
// want to canonicalize this (strip double spaces, etc. For now, we just add all the
// dies that we find by exact match.
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
num_matches = m_apple_names_ap->FindByName (name_cstr, die_offsets);
for (uint32_t i = 0; i < num_matches; i++)
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
const dw_offset_t die_offset = die_offsets[i];
const DWARFDebugInfoEntry *die = info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
if (die)
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (namespace_decl && !DIEIsInNamespace (namespace_decl, dwarf_cu, die))
continue;
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
if (!include_inlines && die->Tag() == DW_TAG_inlined_subroutine)
continue;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
ResolveFunction (dwarf_cu, die, sc_list);
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x for '%s')",
die_offset, name_cstr);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
}
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
else
{
if (effective_name_type_mask & eFunctionNameTypeSelector)
{
if (namespace_decl && *namespace_decl)
return 0; // no selectors in namespaces
num_matches = m_apple_names_ap->FindByName (name_cstr, die_offsets);
// Now make sure these are actually ObjC methods. In this case we can simply look up the name,
// and if it is an ObjC method name, we're good.
for (uint32_t i = 0; i < num_matches; i++)
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
const dw_offset_t die_offset = die_offsets[i];
const DWARFDebugInfoEntry* die = info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (die)
{
const char *die_name = die->GetName(this, dwarf_cu);
if (ObjCLanguageRuntime::IsPossibleObjCMethodName(die_name))
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
{
if (!include_inlines && die->Tag() == DW_TAG_inlined_subroutine)
continue;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
ResolveFunction (dwarf_cu, die, sc_list);
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
}
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportError ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x for '%s')",
die_offset, name_cstr);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
}
die_offsets.clear();
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (effective_name_type_mask & eFunctionNameTypeMethod
|| effective_name_type_mask & eFunctionNameTypeBase)
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if ((effective_name_type_mask & eFunctionNameTypeMethod) &&
(namespace_decl && *namespace_decl))
return 0; // no methods in namespaces
// The apple_names table stores just the "base name" of C++ methods in the table. So we have to
// extract the base name, look that up, and if there is any other information in the name we were
// passed in we have to post-filter based on that.
// FIXME: Arrange the logic above so that we don't calculate the base name twice:
std::string base_name(base_name_start, base_name_end - base_name_start);
num_matches = m_apple_names_ap->FindByName (base_name.c_str(), die_offsets);
for (uint32_t i = 0; i < num_matches; i++)
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
const dw_offset_t die_offset = die_offsets[i];
const DWARFDebugInfoEntry* die = info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (die)
{
if (namespace_decl && !DIEIsInNamespace (namespace_decl, dwarf_cu, die))
continue;
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (!FunctionDieMatchesPartialName(die,
dwarf_cu,
effective_name_type_mask,
name_cstr,
base_name_start,
base_name_end))
continue;
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
if (!include_inlines && die->Tag() == DW_TAG_inlined_subroutine)
continue;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
// If we get to here, the die is good, and we should add it:
ResolveFunction (dwarf_cu, die, sc_list);
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x for '%s')",
die_offset, name_cstr);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
}
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
die_offsets.clear();
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
}
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
// Index the DWARF if we haven't already
if (!m_indexed)
Index ();
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (name_type_mask & eFunctionNameTypeFull)
FindFunctions (name, m_function_fullname_index, sc_list);
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
std::string base_name(base_name_start, base_name_end - base_name_start);
ConstString base_name_const(base_name.c_str());
DIEArray die_offsets;
DWARFCompileUnit *dwarf_cu = NULL;
if (effective_name_type_mask & eFunctionNameTypeBase)
{
uint32_t num_base = m_function_basename_index.Find(base_name_const, die_offsets);
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
for (uint32_t i = 0; i < num_base; i++)
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
const DWARFDebugInfoEntry* die = info->GetDIEPtrWithCompileUnitHint (die_offsets[i], &dwarf_cu);
if (die)
{
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (namespace_decl && !DIEIsInNamespace (namespace_decl, dwarf_cu, die))
continue;
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
if (!FunctionDieMatchesPartialName(die,
dwarf_cu,
effective_name_type_mask,
name_cstr,
base_name_start,
base_name_end))
continue;
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
if (!include_inlines && die->Tag() == DW_TAG_inlined_subroutine)
continue;
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
// If we get to here, the die is good, and we should add it:
ResolveFunction (dwarf_cu, die, sc_list);
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
die_offsets.clear();
}
Fix a couple of places in FindFunctions where I was using the passed in name_type_mask rather than the computed effective_name_type_mask. llvm-svn: 141624
2011-10-11 01:18:11 +00:00
if (effective_name_type_mask & eFunctionNameTypeMethod)
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (namespace_decl && *namespace_decl)
return 0; // no methods in namespaces
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
uint32_t num_base = m_function_method_index.Find(base_name_const, die_offsets);
{
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
for (uint32_t i = 0; i < num_base; i++)
{
const DWARFDebugInfoEntry* die = info->GetDIEPtrWithCompileUnitHint (die_offsets[i], &dwarf_cu);
if (die)
{
if (!FunctionDieMatchesPartialName(die,
dwarf_cu,
effective_name_type_mask,
name_cstr,
base_name_start,
base_name_end))
continue;
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
if (!include_inlines && die->Tag() == DW_TAG_inlined_subroutine)
continue;
Add a version of ResolveFunction that takes a "const DWARFDebugInfoEntry *" and a "DWARFCompileUnit *" to avoid doing a DIE lookup twice and to prepare for using namespaces in the lookups. llvm-svn: 141843
2011-10-13 00:59:38 +00:00
// If we get to here, the die is good, and we should add it:
ResolveFunction (dwarf_cu, die, sc_list);
}
}
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
die_offsets.clear();
}
Added function name types to allow us to set breakpoints by name more intelligently. The four name types we currently have are: eFunctionNameTypeFull = (1 << 1), // The function name. // For C this is the same as just the name of the function // For C++ this is the demangled version of the mangled name. // For ObjC this is the full function signature with the + or // - and the square brackets and the class and selector eFunctionNameTypeBase = (1 << 2), // The function name only, no namespaces or arguments and no class // methods or selectors will be searched. eFunctionNameTypeMethod = (1 << 3), // Find function by method name (C++) with no namespace or arguments eFunctionNameTypeSelector = (1 << 4) // Find function by selector name (ObjC) names this allows much more flexibility when setting breakoints: (lldb) breakpoint set --name main --basename (lldb) breakpoint set --name main --fullname (lldb) breakpoint set --name main --method (lldb) breakpoint set --name main --selector The default: (lldb) breakpoint set --name main will inspect the name "main" and look for any parens, or if the name starts with "-[" or "+[" and if any are found then a full name search will happen. Else a basename search will be the default. Fixed some command option structures so not all options are required when they shouldn't be. Cleaned up the breakpoint output summary. Made the "image lookup --address <addr>" output much more verbose so it shows all the important symbol context results. Added a GetDescription method to many of the SymbolContext objects for the more verbose output. llvm-svn: 107075
2010-06-28 21:30:43 +00:00
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if ((effective_name_type_mask & eFunctionNameTypeSelector) && (!namespace_decl || !*namespace_decl))
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
FindFunctions (name, m_function_selector_index, sc_list);
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
}
If the new .apple_names and .apple_types DWARF accelerator tables are available, we currently will still index the DWARF ourselves and assert if the name lookups differ. This will help us transition to the new accelerator tables and make sure they are workng before we switch over entirely. llvm-svn: 140788
2011-09-29 16:58:15 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// Return the number of variable that were appended to the list
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const uint32_t num_matches = sc_list.GetSize() - original_size;
if (log && num_matches > 0)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindFunctions (name=\"%s\", name_type_mask=0x%x, append=%u, sc_list) => %u",
name.GetCString(),
name_type_mask,
append,
num_matches);
}
return num_matches;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
uint32_t
Extended function lookup to allow the user to indicate whether inline functions are desired. This allows the expression parser, for instance, to filter out inlined functions when looking for functions it can call. llvm-svn: 150279
2012-02-10 22:52:19 +00:00
SymbolFileDWARF::FindFunctions(const RegularExpression& regex, bool include_inlines, bool append, SymbolContextList& sc_list)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Timer scoped_timer (__PRETTY_FUNCTION__,
"SymbolFileDWARF::FindFunctions (regex = '%s')",
regex.GetText());
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindFunctions (regex=\"%s\", append=%u, sc_list)",
regex.GetText(),
append);
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// If we aren't appending the results to this list, then clear the list
if (!append)
sc_list.Clear();
// Remember how many sc_list are in the list before we search in case
// we are appending the results to a variable list.
uint32_t original_size = sc_list.GetSize();
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_names_ap.get())
FindFunctions (regex, *m_apple_names_ap, sc_list);
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
Move the responsibility for translating the various eFunctionNameType lookups to the SymbolFIle (it was done mostly in the BreakpointResolverName resolver before.) Then tailor our searches to the way the indexed maps are laid out. This removes a bunch of test case failures using indexed dSYM's. llvm-svn: 141428
2011-10-07 22:23:45 +00:00
// Index the DWARF if we haven't already
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (!m_indexed)
Index ();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
FindFunctions (regex, m_function_basename_index, sc_list);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
FindFunctions (regex, m_function_fullname_index, sc_list);
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// Return the number of variable that were appended to the list
return sc_list.GetSize() - original_size;
}
Move DIE location reporting into the DWARFDebugInfo class, use it from there in SymbolFileDWARF::ParseType (and eventually in other interesting places as well.) llvm-svn: 138644
2011-08-26 19:44:13 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
SymbolFileDWARF::FindTypes (const SymbolContext& sc,
const ConstString &name,
const lldb_private::ClangNamespaceDecl *namespace_decl,
bool append,
uint32_t max_matches,
TypeList& types)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
So we can't use .debug_pubtypes as it, as designed, does not tell us about all types in all compile units. I added a new kind of accelerator table to the DWARF that allows us to index the DWARF compile units and DIEs in a way that doesn't require the data to stay loaded. Currently when indexing the DWARF we check if the compile unit had parsed its DIEs and if it hasn't we index the data and free all of the DIEs so we can reparse later when we need to after using one of our complete accelerator tables to determine we need to reparse some DWARF. If the DIEs had already been parsed we leave them loaded. The new accelerator table uses the "const char *" pointers from our ConstString class as the keys, and NameToDIE::Info as the value. This info contains the compile unit index and the DIE index which means we are pointed right to the DIE we need unlike the other DWARF accelerator tables that often just point us to the compile unit we would find our answer in. llvm-svn: 113933
2010-09-15 04:15:46 +00:00
DWARFDebugInfo* info = DebugInfo();
if (info == NULL)
return 0;
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
if (namespace_decl)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindTypes (sc, name=\"%s\", clang::NamespaceDecl(%p) \"%s\", append=%u, max_matches=%u, type_list)",
name.GetCString(),
namespace_decl->GetNamespaceDecl(),
namespace_decl->GetQualifiedName().c_str(),
append,
max_matches);
}
else
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindTypes (sc, name=\"%s\", clang::NamespaceDecl(NULL), append=%u, max_matches=%u, type_list)",
name.GetCString(),
append,
max_matches);
}
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// If we aren't appending the results to this list, then clear the list
if (!append)
types.Clear();
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (!NamespaceDeclMatchesThisSymbolFile(namespace_decl))
return 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DIEArray die_offsets;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_types_ap.get())
{
const char *name_cstr = name.GetCString();
m_apple_types_ap->FindByName (name_cstr, die_offsets);
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
if (!m_indexed)
Index ();
m_type_index.Find (name, die_offsets);
}
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const size_t num_die_matches = die_offsets.size();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
if (num_die_matches)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const uint32_t initial_types_size = types.GetSize();
DWARFCompileUnit* dwarf_cu = NULL;
const DWARFDebugInfoEntry* die = NULL;
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
for (size_t i=0; i<num_die_matches; ++i)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const dw_offset_t die_offset = die_offsets[i];
die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (die)
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (namespace_decl && !DIEIsInNamespace (namespace_decl, dwarf_cu, die))
continue;
Type *matching_type = ResolveType (dwarf_cu, die);
if (matching_type)
{
// We found a type pointer, now find the shared pointer form our type list
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
types.InsertUnique (matching_type->shared_from_this());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (types.GetSize() >= max_matches)
break;
}
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_types accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, name.GetCString());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
const uint32_t num_matches = types.GetSize() - initial_types_size;
if (log && num_matches)
{
if (namespace_decl)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindTypes (sc, name=\"%s\", clang::NamespaceDecl(%p) \"%s\", append=%u, max_matches=%u, type_list) => %u",
name.GetCString(),
namespace_decl->GetNamespaceDecl(),
namespace_decl->GetQualifiedName().c_str(),
append,
max_matches,
num_matches);
}
else
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindTypes (sc, name=\"%s\", clang::NamespaceDecl(NULL), append=%u, max_matches=%u, type_list) => %u",
name.GetCString(),
append,
max_matches,
num_matches);
}
}
return num_matches;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
return 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
ClangNamespaceDecl
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
SymbolFileDWARF::FindNamespace (const SymbolContext& sc,
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
const ConstString &name,
const lldb_private::ClangNamespaceDecl *parent_namespace_decl)
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
{
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_LOOKUPS));
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindNamespace (sc, name=\"%s\")",
name.GetCString());
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
}
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
if (!NamespaceDeclMatchesThisSymbolFile(parent_namespace_decl))
return ClangNamespaceDecl();
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
ClangNamespaceDecl namespace_decl;
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
DWARFDebugInfo* info = DebugInfo();
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
if (info)
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DIEArray die_offsets;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
// Index if we already haven't to make sure the compile units
// get indexed and make their global DIE index list
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_namespaces_ap.get())
{
const char *name_cstr = name.GetCString();
m_apple_namespaces_ap->FindByName (name_cstr, die_offsets);
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
if (!m_indexed)
Index ();
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
m_namespace_index.Find (name, die_offsets);
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFCompileUnit* dwarf_cu = NULL;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
const DWARFDebugInfoEntry* die = NULL;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const size_t num_matches = die_offsets.size();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
if (num_matches)
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
for (size_t i=0; i<num_matches; ++i)
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const dw_offset_t die_offset = die_offsets[i];
die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
Completed the glue that passes a ClangNamespaceDecl * down through Module and SymbolVendor into SymbolFile. Added checks to SymbolFileDWARF that restrict symbol searches when a namespace is passed in. llvm-svn: 141847
2011-10-13 01:49:10 +00:00
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (die)
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (parent_namespace_decl && !DIEIsInNamespace (parent_namespace_decl, dwarf_cu, die))
continue;
clang::NamespaceDecl *clang_namespace_decl = ResolveNamespaceDIE (dwarf_cu, die);
if (clang_namespace_decl)
{
namespace_decl.SetASTContext (GetClangASTContext().getASTContext());
namespace_decl.SetNamespaceDecl (clang_namespace_decl);
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
break;
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_namespaces accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, name.GetCString());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
}
}
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
}
Fixed an issue that could occur when looking up functions inside of a namespace where if the NamespaceDecl hadn't been parsed yet, we would say a function wasn't in a namespace. Also improved the logging that happens with "log enable dwarf lookups" is enabled to show when we find matches. llvm-svn: 154352
2012-04-09 22:43:43 +00:00
if (log && namespace_decl.GetNamespaceDecl())
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindNamespace (sc, name=\"%s\") => clang::NamespaceDecl(%p) \"%s\"",
name.GetCString(),
namespace_decl.GetNamespaceDecl(),
namespace_decl.GetQualifiedName().c_str());
}
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
return namespace_decl;
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
SymbolFileDWARF::FindTypes(std::vector<dw_offset_t> die_offsets, uint32_t max_matches, TypeList& types)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// Remember how many sc_list are in the list before we search in case
// we are appending the results to a variable list.
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
uint32_t original_size = types.GetSize();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const uint32_t num_die_offsets = die_offsets.size();
// Parse all of the types we found from the pubtypes matches
uint32_t i;
uint32_t num_matches = 0;
for (i = 0; i < num_die_offsets; ++i)
{
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
Type *matching_type = ResolveTypeUID (die_offsets[i]);
if (matching_type)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
// We found a type pointer, now find the shared pointer form our type list
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
types.InsertUnique (matching_type->shared_from_this());
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
++num_matches;
if (num_matches >= max_matches)
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
// Return the number of variable that were appended to the list
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
return types.GetSize() - original_size;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
size_t
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
SymbolFileDWARF::ParseChildParameters (const SymbolContext& sc,
clang::DeclContext *containing_decl_ctx,
DWARFCompileUnit* dwarf_cu,
const DWARFDebugInfoEntry *parent_die,
bool skip_artificial,
bool &is_static,
TypeList* type_list,
std::vector<clang_type_t>& function_param_types,
std::vector<clang::ParmVarDecl*>& function_param_decls,
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
unsigned &type_quals,
ClangASTContext::TemplateParameterInfos &template_param_infos)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (parent_die == NULL)
return 0;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const uint8_t *fixed_form_sizes = DWARFFormValue::GetFixedFormSizesForAddressSize (dwarf_cu->GetAddressByteSize());
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
size_t arg_idx = 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *die;
for (die = parent_die->GetFirstChild(); die != NULL; die = die->GetSibling())
{
dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_formal_parameter:
{
DWARFDebugInfoEntry::Attributes attributes;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, fixed_form_sizes, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
const char *name = NULL;
Declaration decl;
dw_offset_t param_type_die_offset = DW_INVALID_OFFSET;
Added motheds to C++ classes as we parse them to keep clang happy. llvm-svn: 114616
2010-09-23 01:09:21 +00:00
bool is_artificial = false;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// one of None, Auto, Register, Extern, Static, PrivateExtern
Updated to latest LLVM. Major LLVM changes: - Sema is now exported (and there was much rejoicing.) - Storage classes are now centrally defined. Also fixed some bugs that the new LLVM picked up. llvm-svn: 114622
2010-09-23 03:01:22 +00:00
clang::StorageClass storage = clang::SC_None;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name: name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_type: param_type_die_offset = form_value.Reference(dwarf_cu); break;
Added motheds to C++ classes as we parse them to keep clang happy. llvm-svn: 114616
2010-09-23 01:09:21 +00:00
case DW_AT_artificial: is_artificial = form_value.Unsigned() != 0; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_location:
// if (form_value.BlockData())
// {
// const DataExtractor& debug_info_data = debug_info();
// uint32_t block_length = form_value.Unsigned();
// DataExtractor location(debug_info_data, form_value.BlockData() - debug_info_data.GetDataStart(), block_length);
// }
// else
// {
// }
// break;
case DW_AT_const_value:
case DW_AT_default_value:
case DW_AT_description:
case DW_AT_endianity:
case DW_AT_is_optional:
case DW_AT_segment:
case DW_AT_variable_parameter:
default:
case DW_AT_abstract_origin:
case DW_AT_sibling:
break;
}
}
}
Added motheds to C++ classes as we parse them to keep clang happy. llvm-svn: 114616
2010-09-23 01:09:21 +00:00
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
bool skip = false;
if (skip_artificial)
{
if (is_artificial)
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
{
// In order to determine if a C++ member function is
// "const" we have to look at the const-ness of "this"...
// Ugly, but that
if (arg_idx == 0)
{
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
if (DeclKindIsCXXClass(containing_decl_ctx->getDeclKind()))
Fixed a bug in which the DWARF reader did not distinguish appropriately between C++ static methods and non-static methods. This bug made it impossible to call most static methods, either because Clang did not recognize that a method could be called without providing a "this" parameter, or because Clang did not properly mangle the name of the method when searching for it in the target. Also added a testcase. llvm-svn: 136733
2011-08-02 22:21:50 +00:00
{
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
// Often times compilers omit the "this" name for the
// specification DIEs, so we can't rely upon the name
// being in the formal parameter DIE...
if (name == NULL || ::strcmp(name, "this")==0)
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
{
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
Type *this_type = ResolveTypeUID (param_type_die_offset);
if (this_type)
{
uint32_t encoding_mask = this_type->GetEncodingMask();
if (encoding_mask & Type::eEncodingIsPointerUID)
{
is_static = false;
if (encoding_mask & (1u << Type::eEncodingIsConstUID))
type_quals |= clang::Qualifiers::Const;
if (encoding_mask & (1u << Type::eEncodingIsVolatileUID))
type_quals |= clang::Qualifiers::Volatile;
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
}
}
}
}
}
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
skip = true;
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
}
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
else
{
// HACK: Objective C formal parameters "self" and "_cmd"
// are not marked as artificial in the DWARF...
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompileUnit *comp_unit = GetCompUnitForDWARFCompUnit(dwarf_cu, UINT32_MAX);
if (comp_unit)
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
switch (comp_unit->GetLanguage())
{
case eLanguageTypeObjC:
case eLanguageTypeObjC_plus_plus:
if (name && name[0] && (strcmp (name, "self") == 0 || strcmp (name, "_cmd") == 0))
skip = true;
break;
default:
break;
}
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
}
}
}
Added motheds to C++ classes as we parse them to keep clang happy. llvm-svn: 114616
2010-09-23 01:09:21 +00:00
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
if (!skip)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
Type *type = ResolveTypeUID(param_type_die_offset);
if (type)
{
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
function_param_types.push_back (type->GetClangForwardType());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Use forward types where possible to avoid having to parse extra DWARF when it is not required. llvm-svn: 146418
2011-12-12 21:50:19 +00:00
clang::ParmVarDecl *param_var_decl = GetClangASTContext().CreateParameterDeclaration (name,
type->GetClangForwardType(),
storage);
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
assert(param_var_decl);
function_param_decls.push_back(param_var_decl);
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)param_var_decl, MakeUserID(die->GetOffset()));
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
arg_idx++;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
break;
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
case DW_TAG_template_type_parameter:
case DW_TAG_template_value_parameter:
ParseTemplateDIE (dwarf_cu, die,template_param_infos);
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
default:
break;
}
}
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
return arg_idx;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
size_t
SymbolFileDWARF::ParseChildEnumerators
(
const SymbolContext& sc,
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
clang_type_t enumerator_clang_type,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t enumerator_byte_size,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *parent_die
)
{
if (parent_die == NULL)
return 0;
size_t enumerators_added = 0;
const DWARFDebugInfoEntry *die;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const uint8_t *fixed_form_sizes = DWARFFormValue::GetFixedFormSizesForAddressSize (dwarf_cu->GetAddressByteSize());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
for (die = parent_die->GetFirstChild(); die != NULL; die = die->GetSibling())
{
const dw_tag_t tag = die->Tag();
if (tag == DW_TAG_enumerator)
{
DWARFDebugInfoEntry::Attributes attributes;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_child_attributes = die->GetAttributes(this, dwarf_cu, fixed_form_sizes, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_child_attributes > 0)
{
const char *name = NULL;
bool got_value = false;
int64_t enum_value = 0;
Declaration decl;
uint32_t i;
for (i=0; i<num_child_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_const_value:
got_value = true;
enum_value = form_value.Unsigned();
break;
case DW_AT_name:
name = form_value.AsCString(&get_debug_str_data());
break;
case DW_AT_description:
default:
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_sibling:
break;
}
}
}
if (name && name[0] && got_value)
{
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
GetClangASTContext().AddEnumerationValueToEnumerationType (enumerator_clang_type,
enumerator_clang_type,
decl,
name,
enum_value,
enumerator_byte_size * 8);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
++enumerators_added;
}
}
}
}
return enumerators_added;
}
void
SymbolFileDWARF::ParseChildArrayInfo
(
const SymbolContext& sc,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *parent_die,
int64_t& first_index,
std::vector<uint64_t>& element_orders,
uint32_t& byte_stride,
uint32_t& bit_stride
)
{
if (parent_die == NULL)
return;
const DWARFDebugInfoEntry *die;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const uint8_t *fixed_form_sizes = DWARFFormValue::GetFixedFormSizesForAddressSize (dwarf_cu->GetAddressByteSize());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
for (die = parent_die->GetFirstChild(); die != NULL; die = die->GetSibling())
{
const dw_tag_t tag = die->Tag();
switch (tag)
{
case DW_TAG_enumerator:
{
DWARFDebugInfoEntry::Attributes attributes;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_child_attributes = die->GetAttributes(this, dwarf_cu, fixed_form_sizes, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_child_attributes > 0)
{
const char *name = NULL;
bool got_value = false;
int64_t enum_value = 0;
uint32_t i;
for (i=0; i<num_child_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_const_value:
got_value = true;
enum_value = form_value.Unsigned();
break;
case DW_AT_name:
name = form_value.AsCString(&get_debug_str_data());
break;
case DW_AT_description:
default:
case DW_AT_decl_file:
case DW_AT_decl_line:
case DW_AT_decl_column:
case DW_AT_sibling:
break;
}
}
}
}
}
break;
case DW_TAG_subrange_type:
{
DWARFDebugInfoEntry::Attributes attributes;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_child_attributes = die->GetAttributes(this, dwarf_cu, fixed_form_sizes, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_child_attributes > 0)
{
const char *name = NULL;
bool got_value = false;
uint64_t byte_size = 0;
int64_t enum_value = 0;
uint64_t num_elements = 0;
uint64_t lower_bound = 0;
uint64_t upper_bound = 0;
uint32_t i;
for (i=0; i<num_child_attributes; ++i)
{
const dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_const_value:
got_value = true;
enum_value = form_value.Unsigned();
break;
case DW_AT_name:
name = form_value.AsCString(&get_debug_str_data());
break;
case DW_AT_count:
num_elements = form_value.Unsigned();
break;
case DW_AT_bit_stride:
bit_stride = form_value.Unsigned();
break;
case DW_AT_byte_stride:
byte_stride = form_value.Unsigned();
break;
case DW_AT_byte_size:
byte_size = form_value.Unsigned();
break;
case DW_AT_lower_bound:
lower_bound = form_value.Unsigned();
break;
case DW_AT_upper_bound:
upper_bound = form_value.Unsigned();
break;
default:
case DW_AT_abstract_origin:
case DW_AT_accessibility:
case DW_AT_allocated:
case DW_AT_associated:
case DW_AT_data_location:
case DW_AT_declaration:
case DW_AT_description:
case DW_AT_sibling:
case DW_AT_threads_scaled:
case DW_AT_type:
case DW_AT_visibility:
break;
}
}
}
if (upper_bound > lower_bound)
num_elements = upper_bound - lower_bound + 1;
if (num_elements > 0)
element_orders.push_back (num_elements);
}
}
break;
}
}
}
TypeSP
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolFileDWARF::GetTypeForDIE (DWARFCompileUnit *dwarf_cu, const DWARFDebugInfoEntry* die)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
TypeSP type_sp;
if (die != NULL)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
assert(dwarf_cu != NULL);
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
Type *type_ptr = m_die_to_type.lookup (die);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (type_ptr == NULL)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
CompileUnit* lldb_cu = GetCompUnitForDWARFCompUnit(dwarf_cu);
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
assert (lldb_cu);
SymbolContext sc(lldb_cu);
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
type_sp = ParseType(sc, dwarf_cu, die, NULL);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
else if (type_ptr != DIE_IS_BEING_PARSED)
{
// Grab the existing type from the master types lists
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
type_sp = type_ptr->shared_from_this();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
return type_sp;
}
clang::DeclContext *
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
SymbolFileDWARF::GetClangDeclContextContainingDIEOffset (dw_offset_t die_offset)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (die_offset != DW_INVALID_OFFSET)
{
DWARFCompileUnitSP cu_sp;
const DWARFDebugInfoEntry* die = DebugInfo()->GetDIEPtr(die_offset, &cu_sp);
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
return GetClangDeclContextContainingDIE (cu_sp.get(), die, NULL);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
return NULL;
}
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
clang::DeclContext *
SymbolFileDWARF::GetClangDeclContextForDIEOffset (const SymbolContext &sc, dw_offset_t die_offset)
{
if (die_offset != DW_INVALID_OFFSET)
{
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
if (debug_info)
{
DWARFCompileUnitSP cu_sp;
const DWARFDebugInfoEntry* die = debug_info->GetDIEPtr(die_offset, &cu_sp);
if (die)
return GetClangDeclContextForDIE (sc, cu_sp.get(), die);
}
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
}
return NULL;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
clang::NamespaceDecl *
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
SymbolFileDWARF::ResolveNamespaceDIE (DWARFCompileUnit *dwarf_cu, const DWARFDebugInfoEntry *die)
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
{
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
if (die && die->Tag() == DW_TAG_namespace)
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
{
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
// See if we already parsed this namespace DIE and associated it with a
// uniqued namespace declaration
clang::NamespaceDecl *namespace_decl = static_cast<clang::NamespaceDecl *>(m_die_to_decl_ctx[die]);
if (namespace_decl)
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
return namespace_decl;
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
else
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
const char *namespace_name = die->GetAttributeValueAsString(this, dwarf_cu, DW_AT_name, NULL);
clang::DeclContext *containing_decl_ctx = GetClangDeclContextContainingDIE (dwarf_cu, die, NULL);
Fixed some warnings after enabling some stricter warnings in the Xcode project settings. Also fixed an issue where we weren't creating anonymous namepaces correctly: <rdar://problem/10371295> llvm-svn: 143403
2011-10-31 23:51:19 +00:00
namespace_decl = GetClangASTContext().GetUniqueNamespaceDeclaration (namespace_name, containing_decl_ctx);
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
if (log)
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
{
Fixed some warnings after enabling some stricter warnings in the Xcode project settings. Also fixed an issue where we weren't creating anonymous namepaces correctly: <rdar://problem/10371295> llvm-svn: 143403
2011-10-31 23:51:19 +00:00
if (namespace_name)
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"ASTContext => %p: 0x%8.8llx: DW_TAG_namespace with DW_AT_name(\"%s\") => clang::NamespaceDecl *%p (original = %p)",
GetClangASTContext().getASTContext(),
MakeUserID(die->GetOffset()),
namespace_name,
namespace_decl,
namespace_decl->getOriginalNamespace());
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
}
Fixed some warnings after enabling some stricter warnings in the Xcode project settings. Also fixed an issue where we weren't creating anonymous namepaces correctly: <rdar://problem/10371295> llvm-svn: 143403
2011-10-31 23:51:19 +00:00
else
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"ASTContext => %p: 0x%8.8llx: DW_TAG_namespace (anonymous) => clang::NamespaceDecl *%p (original = %p)",
GetClangASTContext().getASTContext(),
MakeUserID(die->GetOffset()),
namespace_decl,
namespace_decl->getOriginalNamespace());
Fixed some warnings after enabling some stricter warnings in the Xcode project settings. Also fixed an issue where we weren't creating anonymous namepaces correctly: <rdar://problem/10371295> llvm-svn: 143403
2011-10-31 23:51:19 +00:00
}
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
}
Fixed some warnings after enabling some stricter warnings in the Xcode project settings. Also fixed an issue where we weren't creating anonymous namepaces correctly: <rdar://problem/10371295> llvm-svn: 143403
2011-10-31 23:51:19 +00:00
if (namespace_decl)
LinkDeclContextToDIE((clang::DeclContext*)namespace_decl, die);
return namespace_decl;
Added initial support to the lldb_private::SymbolFile for finding namespaces by name given an optional symbol context. I might end up dressing up the "clang::NamespaceDecl" into a lldb_private::Namespace class if we need to do more than is currenlty required of namespaces. Currently we only need to be able to lookup a namespace by name when parsing expressions, so I kept it simple for now. The idea here is even though we are passing around a "clang::NamespaceDecl *", that we always have it be an opaque pointer (it is forward declared inside of "lldb/Core/ClangForward.h") and we only use clang::NamespaceDecl implementations inside of ClangASTContext, or ClangASTType when we need to extract information from the namespace decl object. llvm-svn: 118737
2010-11-10 23:42:09 +00:00
}
}
return NULL;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
clang::DeclContext *
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
SymbolFileDWARF::GetClangDeclContextForDIE (const SymbolContext &sc, DWARFCompileUnit *cu, const DWARFDebugInfoEntry *die)
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
{
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
clang::DeclContext *clang_decl_ctx = GetCachedClangDeclContextForDIE (die);
if (clang_decl_ctx)
return clang_decl_ctx;
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
// If this DIE has a specification, or an abstract origin, then trace to those.
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
dw_offset_t die_offset = die->GetAttributeValueAsReference(this, cu, DW_AT_specification, DW_INVALID_OFFSET);
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
if (die_offset != DW_INVALID_OFFSET)
return GetClangDeclContextForDIEOffset (sc, die_offset);
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
die_offset = die->GetAttributeValueAsReference(this, cu, DW_AT_abstract_origin, DW_INVALID_OFFSET);
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
if (die_offset != DW_INVALID_OFFSET)
return GetClangDeclContextForDIEOffset (sc, die_offset);
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
if (log)
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage(log.get(), "SymbolFileDWARF::GetClangDeclContextForDIE (die = 0x%8.8x) %s '%s'", die->GetOffset(), DW_TAG_value_to_name(die->Tag()), die->GetName(this, cu));
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
// This is the DIE we want. Parse it, then query our map.
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
bool assert_not_being_parsed = true;
ResolveTypeUID (cu, die, assert_not_being_parsed);
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
clang_decl_ctx = GetCachedClangDeclContextForDIE (die);
return clang_decl_ctx;
This is an overhaul of the expression parser code that detects what context the current expression is meant to execute in. LLDB now properly consults the method declaration in the debug information rather than trying to hunt down the "this" or "self" pointer by name, which can be misleading. Other fixes include: - LLDB now properly detects that it is inside an inlined C++ member function. - LLDB now allows access to non-const members when in const code. - The functions in SymbolFile that locate the DeclContext containing a DIE have been renamed to reflect what they actually do. I have added new functions that find the DeclContext for the DIE itself. I have also introduced testcases for C++ and Objective-C. llvm-svn: 136999
2011-08-05 23:43:37 +00:00
}
clang::DeclContext *
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
SymbolFileDWARF::GetClangDeclContextContainingDIE (DWARFCompileUnit *cu, const DWARFDebugInfoEntry *die, const DWARFDebugInfoEntry **decl_ctx_die_copy)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
if (m_clang_tu_decl == NULL)
m_clang_tu_decl = GetClangASTContext().getASTContext()->getTranslationUnitDecl();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
const DWARFDebugInfoEntry *decl_ctx_die = GetDeclContextDIEContainingDIE (cu, die);
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
if (decl_ctx_die_copy)
*decl_ctx_die_copy = decl_ctx_die;
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
if (decl_ctx_die)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
DIEToDeclContextMap::iterator pos = m_die_to_decl_ctx.find (decl_ctx_die);
if (pos != m_die_to_decl_ctx.end())
return pos->second;
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
switch (decl_ctx_die->Tag())
{
case DW_TAG_compile_unit:
return m_clang_tu_decl;
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
case DW_TAG_namespace:
Make sure we create only unique one namespace per AST when parsing the DWARF. llvm-svn: 142005
2011-10-14 21:34:45 +00:00
return ResolveNamespaceDIE (cu, decl_ctx_die);
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
case DW_TAG_structure_type:
case DW_TAG_union_type:
case DW_TAG_class_type:
{
Type* type = ResolveType (cu, decl_ctx_die);
if (type)
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
{
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
clang::DeclContext *decl_ctx = ClangASTContext::GetDeclContextForType (type->GetClangForwardType ());
if (decl_ctx)
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
{
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
LinkDeclContextToDIE (decl_ctx, decl_ctx_die);
if (decl_ctx)
return decl_ctx;
Fixed an issue that was stopping LLDB from finding complete definitions for types. What was happening was the DWARF parser was almost ignoring definitions (DIEs with the DW_AT_declaration set to 1). It wasn't ignoring declarations that had _some_ children. When this happened, we would treat the declaration as a complete type. Often we would have a declaration of a type with just some enum definitions inside and nothing else. Now we correctly ignore these definitions, and also I added some changes to allow us to figure out what decl context these special declarations actually point to. llvm-svn: 124179
2011-01-25 06:17:32 +00:00
}
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
}
}
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
break;
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
default:
break;
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
}
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
}
return m_clang_tu_decl;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
const DWARFDebugInfoEntry *
SymbolFileDWARF::GetDeclContextDIEContainingDIE (DWARFCompileUnit *cu, const DWARFDebugInfoEntry *die)
{
if (cu && die)
{
const DWARFDebugInfoEntry * const decl_die = die;
while (die != NULL)
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
{
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
// If this is the original DIE that we are searching for a declaration
// for, then don't look in the cache as we don't want our own decl
// context to be our decl context...
if (decl_die != die)
{
switch (die->Tag())
{
case DW_TAG_compile_unit:
case DW_TAG_namespace:
case DW_TAG_structure_type:
case DW_TAG_union_type:
case DW_TAG_class_type:
return die;
default:
break;
}
}
dw_offset_t die_offset = die->GetAttributeValueAsReference(this, cu, DW_AT_specification, DW_INVALID_OFFSET);
if (die_offset != DW_INVALID_OFFSET)
{
DWARFCompileUnit *spec_cu = cu;
const DWARFDebugInfoEntry *spec_die = DebugInfo()->GetDIEPtrWithCompileUnitHint (die_offset, &spec_cu);
const DWARFDebugInfoEntry *spec_die_decl_ctx_die = GetDeclContextDIEContainingDIE (spec_cu, spec_die);
if (spec_die_decl_ctx_die)
return spec_die_decl_ctx_die;
}
die_offset = die->GetAttributeValueAsReference(this, cu, DW_AT_abstract_origin, DW_INVALID_OFFSET);
if (die_offset != DW_INVALID_OFFSET)
{
DWARFCompileUnit *abs_cu = cu;
const DWARFDebugInfoEntry *abs_die = DebugInfo()->GetDIEPtrWithCompileUnitHint (die_offset, &abs_cu);
const DWARFDebugInfoEntry *abs_die_decl_ctx_die = GetDeclContextDIEContainingDIE (abs_cu, abs_die);
if (abs_die_decl_ctx_die)
return abs_die_decl_ctx_die;
}
die = die->GetParent();
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
return NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/10212450> Don't parse function types all the time, only parse them lazily. llvm-svn: 140842
2011-09-30 03:20:47 +00:00
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
Symbol *
SymbolFileDWARF::GetObjCClassSymbol (const ConstString &objc_class_name)
{
Symbol *objc_class_symbol = NULL;
if (m_obj_file)
{
Symtab *symtab = m_obj_file->GetSymtab();
if (symtab)
{
objc_class_symbol = symtab->FindFirstSymbolWithNameAndType (objc_class_name,
eSymbolTypeObjCClass,
Symtab::eDebugNo,
Symtab::eVisibilityAny);
}
}
return objc_class_symbol;
}
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
// Some compilers don't emit the DW_AT_APPLE_objc_complete_type attribute. If they don't
// then we can end up looking through all class types for a complete type and never find
// the full definition. We need to know if this attribute is supported, so we determine
// this here and cache th result. We also need to worry about the debug map DWARF file
// if we are doing darwin DWARF in .o file debugging.
bool
SymbolFileDWARF::Supports_DW_AT_APPLE_objc_complete_type (DWARFCompileUnit *cu)
{
if (m_supports_DW_AT_APPLE_objc_complete_type == eLazyBoolCalculate)
{
m_supports_DW_AT_APPLE_objc_complete_type = eLazyBoolNo;
if (cu && cu->Supports_DW_AT_APPLE_objc_complete_type())
m_supports_DW_AT_APPLE_objc_complete_type = eLazyBoolYes;
else
{
DWARFDebugInfo* debug_info = DebugInfo();
const uint32_t num_compile_units = GetNumCompileUnits();
for (uint32_t cu_idx = 0; cu_idx < num_compile_units; ++cu_idx)
{
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
DWARFCompileUnit* dwarf_cu = debug_info->GetCompileUnitAtIndex(cu_idx);
if (dwarf_cu != cu && dwarf_cu->Supports_DW_AT_APPLE_objc_complete_type())
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
{
m_supports_DW_AT_APPLE_objc_complete_type = eLazyBoolYes;
break;
}
}
}
if (m_supports_DW_AT_APPLE_objc_complete_type == eLazyBoolNo && m_debug_map_symfile)
return m_debug_map_symfile->Supports_DW_AT_APPLE_objc_complete_type (this);
}
return m_supports_DW_AT_APPLE_objc_complete_type == eLazyBoolYes;
}
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
// This function can be used when a DIE is found that is a forward declaration
// DIE and we want to try and find a type that has the complete definition.
TypeSP
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
SymbolFileDWARF::FindCompleteObjCDefinitionTypeForDIE (const DWARFDebugInfoEntry *die,
const ConstString &type_name,
bool must_be_implementation)
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
{
TypeSP type_sp;
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (!type_name || (must_be_implementation && !GetObjCClassSymbol (type_name)))
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
return type_sp;
DIEArray die_offsets;
if (m_using_apple_tables)
{
if (m_apple_types_ap.get())
{
const char *name_cstr = type_name.GetCString();
Fixed an issue where we might not find the one true ObjC definition. Now we do things much smarter by extracting search results more efficiently and by properly obeying the must_be_implementation bool in the SymbolFileDWARF::FindCompleteObjCDefinitionTypeForDIE() function. llvm-svn: 148413
2012-01-18 20:58:12 +00:00
m_apple_types_ap->FindCompleteObjCClassByName (name_cstr, die_offsets, must_be_implementation);
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
}
}
else
{
if (!m_indexed)
Index ();
m_type_index.Find (type_name, die_offsets);
}
const size_t num_matches = die_offsets.size();
DWARFCompileUnit* type_cu = NULL;
const DWARFDebugInfoEntry* type_die = NULL;
if (num_matches)
{
DWARFDebugInfo* debug_info = DebugInfo();
for (size_t i=0; i<num_matches; ++i)
{
const dw_offset_t die_offset = die_offsets[i];
type_die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &type_cu);
if (type_die)
{
bool try_resolving_type = false;
// Don't try and resolve the DIE we are looking for with the DIE itself!
if (type_die != die)
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
switch (type_die->Tag())
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
case DW_TAG_class_type:
case DW_TAG_structure_type:
try_resolving_type = true;
break;
default:
break;
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
}
}
if (try_resolving_type)
{
Fixed a problem where maintaining the ObjCInterfaceMap for each ObjCInterfaceDecl was imposing performance penalties for Objective-C apps. Instead, we now use the normal function query mechanisms, which use the relevant accelerator tables. This fix also includes some modifications to the SymbolFile which allow us to find Objective-C methods and report their Clang Decls correctly. llvm-svn: 148457
2012-01-19 02:17:40 +00:00
if (must_be_implementation && type_cu->Supports_DW_AT_APPLE_objc_complete_type())
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
try_resolving_type = type_die->GetAttributeValueAsUnsigned (this, type_cu, DW_AT_APPLE_objc_complete_type, 0);
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
if (try_resolving_type)
{
Type *resolved_type = ResolveType (type_cu, type_die, false);
if (resolved_type && resolved_type != DIE_IS_BEING_PARSED)
{
DEBUG_PRINTF ("resolved 0x%8.8llx (cu 0x%8.8llx) from %s to 0x%8.8llx (cu 0x%8.8llx)\n",
MakeUserID(die->GetOffset()),
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
MakeUserID(dwarf_cu->GetOffset()),
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
m_obj_file->GetFileSpec().GetFilename().AsCString(),
MakeUserID(type_die->GetOffset()),
MakeUserID(type_cu->GetOffset()));
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (die)
m_die_to_type[die] = resolved_type;
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
type_sp = resolved_type->shared_from_this();
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
break;
}
}
}
}
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_types accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, type_name.GetCString());
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
}
}
}
}
return type_sp;
}
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
//----------------------------------------------------------------------
// This function helps to ensure that the declaration contexts match for
// two different DIEs. Often times debug information will refer to a
// forward declaration of a type (the equivalent of "struct my_struct;".
// There will often be a declaration of that type elsewhere that has the
// full definition. When we go looking for the full type "my_struct", we
// will find one or more matches in the accelerator tables and we will
// then need to make sure the type was in the same declaration context
// as the original DIE. This function can efficiently compare two DIEs
// and will return true when the declaration context matches, and false
// when they don't.
//----------------------------------------------------------------------
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
bool
SymbolFileDWARF::DIEDeclContextsMatch (DWARFCompileUnit* cu1, const DWARFDebugInfoEntry *die1,
DWARFCompileUnit* cu2, const DWARFDebugInfoEntry *die2)
{
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
if (die1 == die2)
return true;
#if defined (LLDB_CONFIGURATION_DEBUG)
// You can't and shouldn't call this function with a compile unit from
// two different SymbolFileDWARF instances.
assert (DebugInfo()->ContainsCompileUnit (cu1));
assert (DebugInfo()->ContainsCompileUnit (cu2));
#endif
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
DWARFDIECollection decl_ctx_1;
DWARFDIECollection decl_ctx_2;
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
//The declaration DIE stack is a stack of the declaration context
// DIEs all the way back to the compile unit. If a type "T" is
// declared inside a class "B", and class "B" is declared inside
// a class "A" and class "A" is in a namespace "lldb", and the
// namespace is in a compile unit, there will be a stack of DIEs:
//
// [0] DW_TAG_class_type for "B"
// [1] DW_TAG_class_type for "A"
// [2] DW_TAG_namespace for "lldb"
// [3] DW_TAG_compile_unit for the source file.
//
// We grab both contexts and make sure that everything matches
// all the way back to the compiler unit.
// First lets grab the decl contexts for both DIEs
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
die1->GetDeclContextDIEs (this, cu1, decl_ctx_1);
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
die2->GetDeclContextDIEs (this, cu2, decl_ctx_2);
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
// Make sure the context arrays have the same size, otherwise
// we are done
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
const size_t count1 = decl_ctx_1.Size();
const size_t count2 = decl_ctx_2.Size();
if (count1 != count2)
return false;
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
// Make sure the DW_TAG values match all the way back up the the
// compile unit. If they don't, then we are done.
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
const DWARFDebugInfoEntry *decl_ctx_die1;
const DWARFDebugInfoEntry *decl_ctx_die2;
size_t i;
for (i=0; i<count1; i++)
{
decl_ctx_die1 = decl_ctx_1.GetDIEPtrAtIndex (i);
decl_ctx_die2 = decl_ctx_2.GetDIEPtrAtIndex (i);
if (decl_ctx_die1->Tag() != decl_ctx_die2->Tag())
return false;
}
#if defined LLDB_CONFIGURATION_DEBUG
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
// Make sure the top item in the decl context die array is always
// DW_TAG_compile_unit. If it isn't then something went wrong in
// the DWARFDebugInfoEntry::GetDeclContextDIEs() function...
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
assert (decl_ctx_1.GetDIEPtrAtIndex (count1 - 1)->Tag() == DW_TAG_compile_unit);
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
#endif
// Always skip the compile unit when comparing by only iterating up to
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
// "count - 1". Here we compare the names as we go.
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
for (i=0; i<count1 - 1; i++)
{
decl_ctx_die1 = decl_ctx_1.GetDIEPtrAtIndex (i);
decl_ctx_die2 = decl_ctx_2.GetDIEPtrAtIndex (i);
const char *name1 = decl_ctx_die1->GetName(this, cu1);
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
const char *name2 = decl_ctx_die2->GetName(this, cu2);
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
// If the string was from a DW_FORM_strp, then the pointer will often
// be the same!
Removed all of the "#ifndef SWIG" from the SB header files since we are using interface (.i) files for each class. Changed the FindFunction class from: uint32_t SBTarget::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) uint32_t SBModule::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) To: lldb::SBSymbolContextList SBTarget::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); lldb::SBSymbolContextList SBModule::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); This makes the API easier to use from python. Also added the ability to append a SBSymbolContext or a SBSymbolContextList to a SBSymbolContextList. Exposed properties for lldb.SBSymbolContextList in python: lldb.SBSymbolContextList.modules => list() or all lldb.SBModule objects in the list lldb.SBSymbolContextList.compile_units => list() or all lldb.SBCompileUnits objects in the list lldb.SBSymbolContextList.functions => list() or all lldb.SBFunction objects in the list lldb.SBSymbolContextList.blocks => list() or all lldb.SBBlock objects in the list lldb.SBSymbolContextList.line_entries => list() or all lldb.SBLineEntry objects in the list lldb.SBSymbolContextList.symbols => list() or all lldb.SBSymbol objects in the list This allows a call to the SBTarget::FindFunctions(...) and SBModule::FindFunctions(...) and then the result can be used to extract the desired information: sc_list = lldb.target.FindFunctions("erase") for function in sc_list.functions: print function for symbol in sc_list.symbols: print symbol Exposed properties for the lldb.SBSymbolContext objects in python: lldb.SBSymbolContext.module => lldb.SBModule lldb.SBSymbolContext.compile_unit => lldb.SBCompileUnit lldb.SBSymbolContext.function => lldb.SBFunction lldb.SBSymbolContext.block => lldb.SBBlock lldb.SBSymbolContext.line_entry => lldb.SBLineEntry lldb.SBSymbolContext.symbol => lldb.SBSymbol Exposed properties for the lldb.SBBlock objects in python: lldb.SBBlock.parent => lldb.SBBlock for the parent block that contains lldb.SBBlock.sibling => lldb.SBBlock for the sibling block to the current block lldb.SBBlock.first_child => lldb.SBBlock for the first child block to the current block lldb.SBBlock.call_site => for inline functions, return a lldb.declaration object that gives the call site file, line and column lldb.SBBlock.name => for inline functions this is the name of the inline function that this block represents lldb.SBBlock.inlined_block => returns the inlined function block that contains this block (might return itself if the current block is an inlined block) lldb.SBBlock.range[int] => access the address ranges for a block by index, a list() with start and end address is returned lldb.SBBlock.ranges => an array or all address ranges for this block lldb.SBBlock.num_ranges => the number of address ranges for this blcok SBFunction objects can now get the SBType and the SBBlock that represents the top scope of the function. SBBlock objects can now get the variable list from the current block. The value list returned allows varaibles to be viewed prior with no process if code wants to check the variables in a function. There are two ways to get a variable list from a SBBlock: lldb::SBValueList SBBlock::GetVariables (lldb::SBFrame& frame, bool arguments, bool locals, bool statics, lldb::DynamicValueType use_dynamic); lldb::SBValueList SBBlock::GetVariables (lldb::SBTarget& target, bool arguments, bool locals, bool statics); When a SBFrame is used, the values returned will be locked down to the frame and the values will be evaluated in the context of that frame. When a SBTarget is used, global an static variables can be viewed without a running process. llvm-svn: 149853
2012-02-06 01:44:54 +00:00
if (name1 == name2)
continue;
// Name pointers are not equal, so only compare the strings
// if both are not NULL.
if (name1 && name2)
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
{
Removed all of the "#ifndef SWIG" from the SB header files since we are using interface (.i) files for each class. Changed the FindFunction class from: uint32_t SBTarget::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) uint32_t SBModule::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) To: lldb::SBSymbolContextList SBTarget::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); lldb::SBSymbolContextList SBModule::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); This makes the API easier to use from python. Also added the ability to append a SBSymbolContext or a SBSymbolContextList to a SBSymbolContextList. Exposed properties for lldb.SBSymbolContextList in python: lldb.SBSymbolContextList.modules => list() or all lldb.SBModule objects in the list lldb.SBSymbolContextList.compile_units => list() or all lldb.SBCompileUnits objects in the list lldb.SBSymbolContextList.functions => list() or all lldb.SBFunction objects in the list lldb.SBSymbolContextList.blocks => list() or all lldb.SBBlock objects in the list lldb.SBSymbolContextList.line_entries => list() or all lldb.SBLineEntry objects in the list lldb.SBSymbolContextList.symbols => list() or all lldb.SBSymbol objects in the list This allows a call to the SBTarget::FindFunctions(...) and SBModule::FindFunctions(...) and then the result can be used to extract the desired information: sc_list = lldb.target.FindFunctions("erase") for function in sc_list.functions: print function for symbol in sc_list.symbols: print symbol Exposed properties for the lldb.SBSymbolContext objects in python: lldb.SBSymbolContext.module => lldb.SBModule lldb.SBSymbolContext.compile_unit => lldb.SBCompileUnit lldb.SBSymbolContext.function => lldb.SBFunction lldb.SBSymbolContext.block => lldb.SBBlock lldb.SBSymbolContext.line_entry => lldb.SBLineEntry lldb.SBSymbolContext.symbol => lldb.SBSymbol Exposed properties for the lldb.SBBlock objects in python: lldb.SBBlock.parent => lldb.SBBlock for the parent block that contains lldb.SBBlock.sibling => lldb.SBBlock for the sibling block to the current block lldb.SBBlock.first_child => lldb.SBBlock for the first child block to the current block lldb.SBBlock.call_site => for inline functions, return a lldb.declaration object that gives the call site file, line and column lldb.SBBlock.name => for inline functions this is the name of the inline function that this block represents lldb.SBBlock.inlined_block => returns the inlined function block that contains this block (might return itself if the current block is an inlined block) lldb.SBBlock.range[int] => access the address ranges for a block by index, a list() with start and end address is returned lldb.SBBlock.ranges => an array or all address ranges for this block lldb.SBBlock.num_ranges => the number of address ranges for this blcok SBFunction objects can now get the SBType and the SBBlock that represents the top scope of the function. SBBlock objects can now get the variable list from the current block. The value list returned allows varaibles to be viewed prior with no process if code wants to check the variables in a function. There are two ways to get a variable list from a SBBlock: lldb::SBValueList SBBlock::GetVariables (lldb::SBFrame& frame, bool arguments, bool locals, bool statics, lldb::DynamicValueType use_dynamic); lldb::SBValueList SBBlock::GetVariables (lldb::SBTarget& target, bool arguments, bool locals, bool statics); When a SBFrame is used, the values returned will be locked down to the frame and the values will be evaluated in the context of that frame. When a SBTarget is used, global an static variables can be viewed without a running process. llvm-svn: 149853
2012-02-06 01:44:54 +00:00
// If the strings don't compare, we are done...
if (strcmp(name1, name2) != 0)
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
return false;
Removed all of the "#ifndef SWIG" from the SB header files since we are using interface (.i) files for each class. Changed the FindFunction class from: uint32_t SBTarget::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) uint32_t SBModule::FindFunctions (const char *name, uint32_t name_type_mask, bool append, lldb::SBSymbolContextList& sc_list) To: lldb::SBSymbolContextList SBTarget::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); lldb::SBSymbolContextList SBModule::FindFunctions (const char *name, uint32_t name_type_mask = lldb::eFunctionNameTypeAny); This makes the API easier to use from python. Also added the ability to append a SBSymbolContext or a SBSymbolContextList to a SBSymbolContextList. Exposed properties for lldb.SBSymbolContextList in python: lldb.SBSymbolContextList.modules => list() or all lldb.SBModule objects in the list lldb.SBSymbolContextList.compile_units => list() or all lldb.SBCompileUnits objects in the list lldb.SBSymbolContextList.functions => list() or all lldb.SBFunction objects in the list lldb.SBSymbolContextList.blocks => list() or all lldb.SBBlock objects in the list lldb.SBSymbolContextList.line_entries => list() or all lldb.SBLineEntry objects in the list lldb.SBSymbolContextList.symbols => list() or all lldb.SBSymbol objects in the list This allows a call to the SBTarget::FindFunctions(...) and SBModule::FindFunctions(...) and then the result can be used to extract the desired information: sc_list = lldb.target.FindFunctions("erase") for function in sc_list.functions: print function for symbol in sc_list.symbols: print symbol Exposed properties for the lldb.SBSymbolContext objects in python: lldb.SBSymbolContext.module => lldb.SBModule lldb.SBSymbolContext.compile_unit => lldb.SBCompileUnit lldb.SBSymbolContext.function => lldb.SBFunction lldb.SBSymbolContext.block => lldb.SBBlock lldb.SBSymbolContext.line_entry => lldb.SBLineEntry lldb.SBSymbolContext.symbol => lldb.SBSymbol Exposed properties for the lldb.SBBlock objects in python: lldb.SBBlock.parent => lldb.SBBlock for the parent block that contains lldb.SBBlock.sibling => lldb.SBBlock for the sibling block to the current block lldb.SBBlock.first_child => lldb.SBBlock for the first child block to the current block lldb.SBBlock.call_site => for inline functions, return a lldb.declaration object that gives the call site file, line and column lldb.SBBlock.name => for inline functions this is the name of the inline function that this block represents lldb.SBBlock.inlined_block => returns the inlined function block that contains this block (might return itself if the current block is an inlined block) lldb.SBBlock.range[int] => access the address ranges for a block by index, a list() with start and end address is returned lldb.SBBlock.ranges => an array or all address ranges for this block lldb.SBBlock.num_ranges => the number of address ranges for this blcok SBFunction objects can now get the SBType and the SBBlock that represents the top scope of the function. SBBlock objects can now get the variable list from the current block. The value list returned allows varaibles to be viewed prior with no process if code wants to check the variables in a function. There are two ways to get a variable list from a SBBlock: lldb::SBValueList SBBlock::GetVariables (lldb::SBFrame& frame, bool arguments, bool locals, bool statics, lldb::DynamicValueType use_dynamic); lldb::SBValueList SBBlock::GetVariables (lldb::SBTarget& target, bool arguments, bool locals, bool statics); When a SBFrame is used, the values returned will be locked down to the frame and the values will be evaluated in the context of that frame. When a SBTarget is used, global an static variables can be viewed without a running process. llvm-svn: 149853
2012-02-06 01:44:54 +00:00
}
else
{
// One name was NULL while the other wasn't
return false;
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
}
}
Added some extra comments for the declaration context comparison function in the DWARF plug-in. llvm-svn: 149811
2012-02-05 06:12:47 +00:00
// We made it through all of the checks and the declaration contexts
// are equal.
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
return true;
}
Tested a theory on the where when we lookup things in the accelerator tables that if we prefer the current compile unit, followed by any compile units that already had their DIEs parsed, followed by the rest of the matches, that we might save some memory. This turned out not to help much. The code is commented out, but I want to check it in so I don't lose the code in case it could help later. Added the ability to efficiently find the objective C class implementation when using the new .apple_types acclerator tables with the type flags. If the type flags are not available, we default back to what we were doing before. llvm-svn: 146250
2011-12-09 08:48:30 +00:00
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
// This function can be used when a DIE is found that is a forward declaration
// DIE and we want to try and find a type that has the complete definition.
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
// "cu" and "die" must be from this SymbolFileDWARF
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
TypeSP
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
SymbolFileDWARF::FindDefinitionTypeForDIE (DWARFCompileUnit* cu,
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const DWARFDebugInfoEntry *die,
const ConstString &type_name)
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
{
TypeSP type_sp;
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
#if defined (LLDB_CONFIGURATION_DEBUG)
// You can't and shouldn't call this function with a compile unit from
// another SymbolFileDWARF instance.
assert (DebugInfo()->ContainsCompileUnit (cu));
#endif
Enabled extra warnings and fixed a bunch of small issues. llvm-svn: 124250
2011-01-25 23:55:37 +00:00
if (cu == NULL || die == NULL || !type_name)
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
return type_sp;
Added logging so we can see when we are trying to complete a forward type and pull in the world. This is due to a compiler bug we are tracking (<rdar://problem/11291658>) where forward decls to classes and types are not properly scoped in namespaces, which results in the current LLDB looking for a type it will find many times in the accelerator tables, but never match. For example, when debugging with clang we get a forward decl like: class AnalysisResolver; And we will look for it everywhere and find many many matches, but the decl context of those matching DIEs is "clang::AnalysisResolver", so we never match anything, yet we pull in waaayyy too much DWARF in the process. To enable this logging enable the "lookups" category in the "dwarf" log channel: (lldb) log enable dwarf lookups llvm-svn: 155233
2012-04-20 20:35:47 +00:00
LogSP log (LogChannelDWARF::GetLogIfAny(DWARF_LOG_TYPE_COMPLETION|DWARF_LOG_LOOKUPS));
if (log)
{
std::string qualified_name;
die->GetQualifiedName(this, cu, qualified_name);
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDIE(die=0x%8.8x (%s), name='%s')",
die->GetOffset(),
qualified_name.c_str(),
type_name.GetCString());
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DIEArray die_offsets;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_types_ap.get())
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
if (m_apple_types_ap->GetHeader().header_data.atoms.size() > 1)
{
We finalized on the new .apple_types accelerator table format where we don't emit fully qualified names, so now we make sure the DW_TAG values match and still lookup using the basename. llvm-svn: 147634
2012-01-06 00:17:16 +00:00
m_apple_types_ap->FindByNameAndTag (type_name.GetCString(), die->Tag(), die_offsets);
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
}
else
{
m_apple_types_ap->FindByName (type_name.GetCString(), die_offsets);
}
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
if (!m_indexed)
Index ();
m_type_index.Find (type_name, die_offsets);
}
const size_t num_matches = die_offsets.size();
Make sure to index the DWARF if we already haven't so we can resolve forward declarations when they haven't been found yet. llvm-svn: 120840
2010-12-03 21:42:06 +00:00
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
const dw_tag_t die_tag = die->Tag();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFCompileUnit* type_cu = NULL;
const DWARFDebugInfoEntry* type_die = NULL;
if (num_matches)
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
for (size_t i=0; i<num_matches; ++i)
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const dw_offset_t die_offset = die_offsets[i];
type_die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &type_cu);
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (type_die)
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
{
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
bool try_resolving_type = false;
// Don't try and resolve the DIE we are looking for with the DIE itself!
if (type_die != die)
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
{
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
const dw_tag_t type_die_tag = type_die->Tag();
// Make sure the tags match
if (type_die_tag == die_tag)
{
// The tags match, lets try resolving this type
try_resolving_type = true;
}
else
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
{
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
// The tags don't match, but we need to watch our for a
// forward declaration for a struct and ("struct foo")
// ends up being a class ("class foo { ... };") or
// vice versa.
switch (type_die_tag)
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
{
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
case DW_TAG_class_type:
// We had a "class foo", see if we ended up with a "struct foo { ... };"
try_resolving_type = (die_tag == DW_TAG_structure_type);
break;
case DW_TAG_structure_type:
// We had a "struct foo", see if we ended up with a "class foo { ... };"
try_resolving_type = (die_tag == DW_TAG_class_type);
break;
default:
// Tags don't match, don't event try to resolve
// using this type whose name matches....
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
break;
}
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
}
}
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
if (try_resolving_type)
{
Added logging so we can see when we are trying to complete a forward type and pull in the world. This is due to a compiler bug we are tracking (<rdar://problem/11291658>) where forward decls to classes and types are not properly scoped in namespaces, which results in the current LLDB looking for a type it will find many times in the accelerator tables, but never match. For example, when debugging with clang we get a forward decl like: class AnalysisResolver; And we will look for it everywhere and find many many matches, but the decl context of those matching DIEs is "clang::AnalysisResolver", so we never match anything, yet we pull in waaayyy too much DWARF in the process. To enable this logging enable the "lookups" category in the "dwarf" log channel: (lldb) log enable dwarf lookups llvm-svn: 155233
2012-04-20 20:35:47 +00:00
if (log)
{
std::string qualified_name;
type_die->GetQualifiedName(this, cu, qualified_name);
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDIE(die=0x%8.8x, name='%s') trying die=0x%8.8x (%s)",
die->GetOffset(),
type_name.GetCString(),
type_die->GetOffset(),
qualified_name.c_str());
}
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
// Make sure the decl contexts match all the way up
if (DIEDeclContextsMatch(cu, die, type_cu, type_die))
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
{
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
Type *resolved_type = ResolveType (type_cu, type_die, false);
if (resolved_type && resolved_type != DIE_IS_BEING_PARSED)
{
DEBUG_PRINTF ("resolved 0x%8.8llx (cu 0x%8.8llx) from %s to 0x%8.8llx (cu 0x%8.8llx)\n",
MakeUserID(die->GetOffset()),
<rdar://problem/11160171> Fixed an issue where there were more than one way to get a CompileUnitSP created when using SymbolFileDWARF with SymbolFileDWARFDebugMap. This led to an assertion that would fire under certain conditions. Now there is only one way to create the compile unit and it will "do the right thing". llvm-svn: 153908
2012-04-02 22:59:12 +00:00
MakeUserID(dwarf_cu->GetOffset()),
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
m_obj_file->GetFileSpec().GetFilename().AsCString(),
MakeUserID(type_die->GetOffset()),
MakeUserID(type_cu->GetOffset()));
m_die_to_type[die] = resolved_type;
Fixed a build breakage when trying to assign a shared pointer using a raw pointer. llvm-svn: 149609
2012-02-02 18:16:59 +00:00
type_sp = resolved_type->shared_from_this();
Fixed an issue where we might accept the wrong type when completing a type when we have a forward declaration. We always have found a type by basename, but now we also compare the decl context of the die we are trying to complete with the matches we find from the accelerator tables to ensure we get the right one. llvm-svn: 149593
2012-02-02 05:48:16 +00:00
break;
}
<rdar://problem/10522194> Fixed an issue where if we have the DWARF equivalent of: struct foo; class foo { ... }; Or vice versa, we wouldn't be able to find the complete type. Since many compilers allow forward declarations to have struct and definitions to have class, we need to be able to deal with both cases. This commit fixes this in the DWARF parser. llvm-svn: 145733
2011-12-03 00:27:05 +00:00
}
}
Added logging so we can see when we are trying to complete a forward type and pull in the world. This is due to a compiler bug we are tracking (<rdar://problem/11291658>) where forward decls to classes and types are not properly scoped in namespaces, which results in the current LLDB looking for a type it will find many times in the accelerator tables, but never match. For example, when debugging with clang we get a forward decl like: class AnalysisResolver; And we will look for it everywhere and find many many matches, but the decl context of those matching DIEs is "clang::AnalysisResolver", so we never match anything, yet we pull in waaayyy too much DWARF in the process. To enable this logging enable the "lookups" category in the "dwarf" log channel: (lldb) log enable dwarf lookups llvm-svn: 155233
2012-04-20 20:35:47 +00:00
else
{
if (log)
{
std::string qualified_name;
type_die->GetQualifiedName(this, cu, qualified_name);
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDIE(die=0x%8.8x, name='%s') ignoring die=0x%8.8x (%s)",
die->GetOffset(),
type_name.GetCString(),
type_die->GetOffset(),
qualified_name.c_str());
}
}
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_types accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, type_name.GetCString());
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Modified the DWARF parser for both the single DWARF file and for the case where the DWARF is in the .o files so they can track down the actual type for a forward declaration. This was working before for just DWARF files, but not for DWARF in .o files where the actual definition was in another .o file. Modified the main thread name in the driver to be more consistent with the other LLDB thread names. llvm-svn: 118383
2010-11-07 21:02:03 +00:00
}
}
return type_sp;
}
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
TypeSP
SymbolFileDWARF::FindDefinitionTypeForDWARFDeclContext (const DWARFDeclContext &dwarf_decl_ctx)
{
TypeSP type_sp;
const uint32_t dwarf_decl_ctx_count = dwarf_decl_ctx.GetSize();
if (dwarf_decl_ctx_count > 0)
{
const ConstString type_name(dwarf_decl_ctx[0].name);
const dw_tag_t tag = dwarf_decl_ctx[0].tag;
if (type_name)
{
LogSP log (LogChannelDWARF::GetLogIfAny(DWARF_LOG_TYPE_COMPLETION|DWARF_LOG_LOOKUPS));
if (log)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDWARFDeclContext(tag=%s, qualified-name='%s')",
DW_TAG_value_to_name(dwarf_decl_ctx[0].tag),
dwarf_decl_ctx.GetQualifiedName());
}
DIEArray die_offsets;
if (m_using_apple_tables)
{
if (m_apple_types_ap.get())
{
if (m_apple_types_ap->GetHeader().header_data.atoms.size() > 1)
{
m_apple_types_ap->FindByNameAndTag (type_name.GetCString(), tag, die_offsets);
}
else
{
m_apple_types_ap->FindByName (type_name.GetCString(), die_offsets);
}
}
}
else
{
if (!m_indexed)
Index ();
m_type_index.Find (type_name, die_offsets);
}
const size_t num_matches = die_offsets.size();
DWARFCompileUnit* type_cu = NULL;
const DWARFDebugInfoEntry* type_die = NULL;
if (num_matches)
{
DWARFDebugInfo* debug_info = DebugInfo();
for (size_t i=0; i<num_matches; ++i)
{
const dw_offset_t die_offset = die_offsets[i];
type_die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &type_cu);
if (type_die)
{
bool try_resolving_type = false;
// Don't try and resolve the DIE we are looking for with the DIE itself!
const dw_tag_t type_tag = type_die->Tag();
// Make sure the tags match
if (type_tag == tag)
{
// The tags match, lets try resolving this type
try_resolving_type = true;
}
else
{
// The tags don't match, but we need to watch our for a
// forward declaration for a struct and ("struct foo")
// ends up being a class ("class foo { ... };") or
// vice versa.
switch (type_tag)
{
case DW_TAG_class_type:
// We had a "class foo", see if we ended up with a "struct foo { ... };"
try_resolving_type = (tag == DW_TAG_structure_type);
break;
case DW_TAG_structure_type:
// We had a "struct foo", see if we ended up with a "class foo { ... };"
try_resolving_type = (tag == DW_TAG_class_type);
break;
default:
// Tags don't match, don't event try to resolve
// using this type whose name matches....
break;
}
}
if (try_resolving_type)
{
DWARFDeclContext type_dwarf_decl_ctx;
type_die->GetDWARFDeclContext (this, type_cu, type_dwarf_decl_ctx);
if (log)
{
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDWARFDeclContext(tag=%s, qualified-name='%s') trying die=0x%8.8x (%s)",
DW_TAG_value_to_name(dwarf_decl_ctx[0].tag),
dwarf_decl_ctx.GetQualifiedName(),
type_die->GetOffset(),
type_dwarf_decl_ctx.GetQualifiedName());
}
// Make sure the decl contexts match all the way up
if (dwarf_decl_ctx == type_dwarf_decl_ctx)
{
Type *resolved_type = ResolveType (type_cu, type_die, false);
if (resolved_type && resolved_type != DIE_IS_BEING_PARSED)
{
type_sp = resolved_type->shared_from_this();
break;
}
}
}
else
{
if (log)
{
std::string qualified_name;
type_die->GetQualifiedName(this, type_cu, qualified_name);
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::FindDefinitionTypeForDWARFDeclContext(tag=%s, qualified-name='%s') ignoring die=0x%8.8x (%s)",
DW_TAG_value_to_name(dwarf_decl_ctx[0].tag),
dwarf_decl_ctx.GetQualifiedName(),
type_die->GetOffset(),
qualified_name.c_str());
}
}
}
else
{
if (m_using_apple_tables)
{
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_types accelerator table had bad die 0x%8.8x for '%s')\n",
die_offset, type_name.GetCString());
}
}
}
}
}
}
return type_sp;
}
<rdar://problem/11240464> Correctly unique a class' methods when we detect that a class has been uniqued to another. llvm-svn: 156795
2012-05-15 02:33:01 +00:00
bool
SymbolFileDWARF::CopyUniqueClassMethodTypes (Type *class_type,
DWARFCompileUnit* src_cu,
const DWARFDebugInfoEntry *src_class_die,
DWARFCompileUnit* dst_cu,
const DWARFDebugInfoEntry *dst_class_die)
{
if (!class_type || !src_cu || !src_class_die || !dst_cu || !dst_class_die)
return false;
if (src_class_die->Tag() != dst_class_die->Tag())
return false;
// We need to complete the class type so we can get all of the method types
// parsed so we can then unique those types to their equivalent counterparts
// in "dst_cu" and "dst_class_die"
class_type->GetClangFullType();
const DWARFDebugInfoEntry *src_die;
const DWARFDebugInfoEntry *dst_die;
UniqueCStringMap<const DWARFDebugInfoEntry *> src_name_to_die;
UniqueCStringMap<const DWARFDebugInfoEntry *> dst_name_to_die;
for (src_die = src_class_die->GetFirstChild(); src_die != NULL; src_die = src_die->GetSibling())
{
if (src_die->Tag() == DW_TAG_subprogram)
{
const char *src_name = src_die->GetMangledName (this, src_cu);
if (src_name)
src_name_to_die.Append(ConstString(src_name).GetCString(), src_die);
}
}
for (dst_die = dst_class_die->GetFirstChild(); dst_die != NULL; dst_die = dst_die->GetSibling())
{
if (dst_die->Tag() == DW_TAG_subprogram)
{
const char *dst_name = dst_die->GetMangledName (this, dst_cu);
if (dst_name)
dst_name_to_die.Append(ConstString(dst_name).GetCString(), dst_die);
}
}
const uint32_t src_size = src_name_to_die.GetSize ();
const uint32_t dst_size = dst_name_to_die.GetSize ();
LogSP log (LogChannelDWARF::GetLogIfAny(DWARF_LOG_DEBUG_INFO | DWARF_LOG_TYPE_COMPLETION));
if (src_size && dst_size)
{
uint32_t idx;
for (idx = 0; idx < src_size; ++idx)
{
src_die = src_name_to_die.GetValueAtIndexUnchecked (idx);
dst_die = dst_name_to_die.GetValueAtIndexUnchecked (idx);
if (src_die->Tag() != dst_die->Tag())
{
if (log)
log->Printf("warning: tried to unique class DIE 0x%8.8x to 0x%8.8x, but 0x%8.8x (%s) tags didn't match 0x%8.8x (%s)",
src_class_die->GetOffset(),
dst_class_die->GetOffset(),
src_die->GetOffset(),
DW_TAG_value_to_name(src_die->Tag()),
dst_die->GetOffset(),
DW_TAG_value_to_name(src_die->Tag()));
return false;
}
const char *src_name = src_die->GetMangledName (this, src_cu);
const char *dst_name = dst_die->GetMangledName (this, dst_cu);
// Make sure the names match
if (src_name == dst_name || (strcmp (src_name, dst_name) == 0))
continue;
if (log)
log->Printf("warning: tried to unique class DIE 0x%8.8x to 0x%8.8x, but 0x%8.8x (%s) names didn't match 0x%8.8x (%s)",
src_class_die->GetOffset(),
dst_class_die->GetOffset(),
src_die->GetOffset(),
src_name,
dst_die->GetOffset(),
dst_name);
return false;
}
for (idx = 0; idx < src_size; ++idx)
{
src_die = src_name_to_die.GetValueAtIndexUnchecked (idx);
dst_die = dst_name_to_die.GetValueAtIndexUnchecked (idx);
clang::DeclContext *src_decl_ctx = m_die_to_decl_ctx[src_die];
if (src_decl_ctx)
{
if (log)
log->Printf ("uniquing decl context %p from 0x%8.8x for 0x%8.8x\n", src_decl_ctx, src_die->GetOffset(), dst_die->GetOffset());
LinkDeclContextToDIE (src_decl_ctx, dst_die);
}
else
{
if (log)
log->Printf ("warning: tried to unique decl context from 0x%8.8x for 0x%8.8x, but none was found\n", src_die->GetOffset(), dst_die->GetOffset());
}
Type *src_child_type = m_die_to_type[src_die];
if (src_child_type)
{
if (log)
log->Printf ("uniquing type %p (uid=0x%llx) from 0x%8.8x for 0x%8.8x\n", src_child_type, src_child_type->GetID(), src_die->GetOffset(), dst_die->GetOffset());
m_die_to_type[dst_die] = src_child_type;
}
else
{
if (log)
log->Printf ("warning: tried to unique lldb_private::Type from 0x%8.8x for 0x%8.8x, but none was found\n", src_die->GetOffset(), dst_die->GetOffset());
}
}
return true;
}
else
{
if (log)
log->Printf("warning: tried to unique class DIE 0x%8.8x to 0x%8.8x, but 0x%8.8x has %u methods and 0x%8.8x has %u",
src_class_die->GetOffset(),
dst_class_die->GetOffset(),
src_die->GetOffset(),
src_size,
dst_die->GetOffset(),
dst_size);
}
return false;
}
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
TypeSP
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
SymbolFileDWARF::ParseType (const SymbolContext& sc, DWARFCompileUnit* dwarf_cu, const DWARFDebugInfoEntry *die, bool *type_is_new_ptr)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
TypeSP type_sp;
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
if (type_is_new_ptr)
*type_is_new_ptr = false;
Added a logging helper class for SymbolFileDWARF::ParseType() that will enable us to track the depth of parsing and what is being parsed. This helps when trying to track down difficult type parsing issues and is only enabled in non-production builds. llvm-svn: 150203
2012-02-09 20:03:49 +00:00
#if defined(LLDB_CONFIGURATION_DEBUG) or defined(LLDB_CONFIGURATION_RELEASE)
static DIEStack g_die_stack;
DIEStack::ScopedPopper scoped_die_logger(g_die_stack);
#endif
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
Fixed namespace visibility problems that were breaking the build for me on a non-internal SnowLeopard system. llvm-svn: 110412
2010-08-06 00:32:49 +00:00
AccessType accessibility = eAccessNone;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (die != NULL)
{
Added a new logging channel to the DWARF called "lookups": (lldb) log enable dwarf lookups This allows us to see when lookups are being done on functions, addresses, and types by both name and regular expresssion. llvm-svn: 141259
2011-10-06 00:09:08 +00:00
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
if (log)
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
{
const DWARFDebugInfoEntry *context_die;
clang::DeclContext *context = GetClangDeclContextContainingDIE (dwarf_cu, die, &context_die);
Added a logging helper class for SymbolFileDWARF::ParseType() that will enable us to track the depth of parsing and what is being parsed. This helps when trying to track down difficult type parsing issues and is only enabled in non-production builds. llvm-svn: 150203
2012-02-09 20:03:49 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(), "SymbolFileDWARF::ParseType (die = 0x%8.8x, decl_ctx = %p (die 0x%8.8x)) %s name = '%s')",
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
die->GetOffset(),
context,
context_die->GetOffset(),
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
DW_TAG_value_to_name(die->Tag()),
Added a logging helper class for SymbolFileDWARF::ParseType() that will enable us to track the depth of parsing and what is being parsed. This helps when trying to track down difficult type parsing issues and is only enabled in non-production builds. llvm-svn: 150203
2012-02-09 20:03:49 +00:00
die->GetName(this, dwarf_cu));
#if defined(LLDB_CONFIGURATION_DEBUG) or defined(LLDB_CONFIGURATION_RELEASE)
scoped_die_logger.Push (dwarf_cu, die);
g_die_stack.LogDIEs(log.get(), this);
#endif
I have brought LLDB up-to-date with top of tree LLVM/Clang. This brings in several fixes, including: - Improvements in the Just-In-Time compiler's allocation of memory: the JIT now allocates memory in chunks of sections, improving its ability to generate relocations. I have revamped the RecordingMemoryManager to reflect these changes, as well as to get the memory allocation and data copying out fo the ClangExpressionParser code. Jim Grosbach wrote the updates to the JIT on the LLVM side. - A new ExternalASTSource interface to allow LLDB to report accurate structure layout information to Clang. Previously we could only report the sizes of fields, not their offsets. This meant that if data structures included field alignment directives, we could not communicate the necessary alignment to Clang and accesses to the data would fail. Now we can (and I have update the relevant test case). Thanks to Doug Gregor for implementing the Clang side of this fix. - The way Objective-C interfaces are completed by Clang has been made consistent with RecordDecls; with help from Doug Gregor and Greg Clayton I have ensured that this still works. - I have eliminated all local LLVM and Clang patches, committing the ones that are still relevant to LLVM and Clang as needed. I have tested the changes extensively locally, but please let me know if they cause any trouble for you. llvm-svn: 149775
2012-02-04 08:49:35 +00:00
}
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
//
// LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
// if (log && dwarf_cu)
// {
// StreamString s;
// die->DumpLocation (this, dwarf_cu, s);
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
// GetObjectFile()->GetModule()->LogMessage (log.get(), "SymbolFileDwarf::%s %s", __FUNCTION__, s.GetData());
<rdar://problem/10559329> An assertion was firing when parsing types due to trying to complete parent class decl contenxt types too often. Also, relax where "dsymutil" binary can come from in the Makefile.rules. llvm-svn: 146310
2011-12-10 02:15:28 +00:00
//
// }
Add logging to SymbolFileDWARF::ParseType so that we can tell the type/comp_unit/obj_file were being processed when something goes wrong... llvm-svn: 138600
2011-08-25 23:21:43 +00:00
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
Type *type_ptr = m_die_to_type.lookup (die);
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
TypeList* type_list = GetTypeList();
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
if (type_ptr == NULL)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ClangASTContext &ast = GetClangASTContext();
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
if (type_is_new_ptr)
*type_is_new_ptr = true;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
const dw_tag_t tag = die->Tag();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
bool is_forward_declaration = false;
DWARFDebugInfoEntry::Attributes attributes;
const char *type_name_cstr = NULL;
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
ConstString type_name_const_str;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Type::ResolveState resolve_state = Type::eResolveStateUnresolved;
size_t byte_size = 0;
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
bool byte_size_valid = false;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Declaration decl;
Cleaned up a unused member variable in Debugger. Added the start of Host specific launch services, though it currently isn't hookup up to anything. We want to be able to launch a process and use the native launch services to launch an app like it would be launched by the user double clicking on the app. We also eventually want to be able to run a command line app in a newly spawned terminal to avoid terminal sharing. Fixed an issue with the new DWARF forward type declaration stuff. A crasher was found that was happening when trying to properly expand the forward declarations. llvm-svn: 115213
2010-09-30 21:49:03 +00:00
Type::EncodingDataType encoding_data_type = Type::eEncodingIsUID;
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
clang_type_t clang_type = NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
dw_attr_t attr;
switch (tag)
{
case DW_TAG_base_type:
case DW_TAG_pointer_type:
case DW_TAG_reference_type:
case DW_TAG_typedef:
case DW_TAG_const_type:
case DW_TAG_restrict_type:
case DW_TAG_volatile_type:
Added support for the C++Ox "nullptr_t" type. llvm-svn: 143225
2011-10-28 21:00:00 +00:00
case DW_TAG_unspecified_type:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
// Set a bit that lets us know that we are currently parsing this
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = DIE_IS_BEING_PARSED;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint32_t encoding = 0;
lldb::user_id_t encoding_uid = LLDB_INVALID_UID;
if (num_attributes > 0)
{
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name:
Work around a llvm gcc bug where the name of a reference doesn't include the "&". llvm-svn: 129620
2011-04-15 23:41:23 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
type_name_cstr = form_value.AsCString(&get_debug_str_data());
Work around a llvm gcc bug where the name of a reference doesn't include the "&". llvm-svn: 129620
2011-04-15 23:41:23 +00:00
// Work around a bug in llvm-gcc where they give a name to a reference type which doesn't
// include the "&"...
if (tag == DW_TAG_reference_type)
{
if (strchr (type_name_cstr, '&') == NULL)
type_name_cstr = NULL;
}
if (type_name_cstr)
type_name_const_str.SetCString(type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
case DW_AT_byte_size: byte_size = form_value.Unsigned(); byte_size_valid = true; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_encoding: encoding = form_value.Unsigned(); break;
case DW_AT_type: encoding_uid = form_value.Reference(dwarf_cu); break;
default:
case DW_AT_sibling:
break;
}
}
}
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DEBUG_PRINTF ("0x%8.8llx: %s (\"%s\") type => 0x%8.8x\n", MakeUserID(die->GetOffset()), DW_TAG_value_to_name(tag), type_name_cstr, encoding_uid);
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
switch (tag)
{
default:
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
break;
Added support for the C++Ox "nullptr_t" type. llvm-svn: 143225
2011-10-28 21:00:00 +00:00
case DW_TAG_unspecified_type:
if (strcmp(type_name_cstr, "nullptr_t") == 0)
{
resolve_state = Type::eResolveStateFull;
clang_type = ast.getASTContext()->NullPtrTy.getAsOpaquePtr();
break;
}
// Fall through to base type below in case we can handle the type there...
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_TAG_base_type:
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
resolve_state = Type::eResolveStateFull;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
clang_type = ast.GetBuiltinTypeForDWARFEncodingAndBitSize (type_name_cstr,
encoding,
byte_size * 8);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
case DW_TAG_pointer_type: encoding_data_type = Type::eEncodingIsPointerUID; break;
case DW_TAG_reference_type: encoding_data_type = Type::eEncodingIsLValueReferenceUID; break;
case DW_TAG_typedef: encoding_data_type = Type::eEncodingIsTypedefUID; break;
case DW_TAG_const_type: encoding_data_type = Type::eEncodingIsConstUID; break;
case DW_TAG_restrict_type: encoding_data_type = Type::eEncodingIsRestrictUID; break;
case DW_TAG_volatile_type: encoding_data_type = Type::eEncodingIsVolatileUID; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (clang_type == NULL && (encoding_data_type == Type::eEncodingIsPointerUID || encoding_data_type == Type::eEncodingIsTypedefUID))
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (type_name_cstr != NULL && sc.comp_unit != NULL &&
(sc.comp_unit->GetLanguage() == eLanguageTypeObjC || sc.comp_unit->GetLanguage() == eLanguageTypeObjC_plus_plus))
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
static ConstString g_objc_type_name_id("id");
static ConstString g_objc_type_name_Class("Class");
static ConstString g_objc_type_name_selector("SEL");
if (type_name_const_str == g_objc_type_name_id)
{
if (log)
GetObjectFile()->GetModule()->LogMessage (log.get(), "SymbolFileDWARF::ParseType (die = 0x%8.8x) %s '%s' is Objective C 'id' built-in type.",
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
die->GetName(this, dwarf_cu));
clang_type = ast.GetBuiltInType_objc_id();
encoding_data_type = Type::eEncodingIsUID;
encoding_uid = LLDB_INVALID_UID;
resolve_state = Type::eResolveStateFull;
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
}
else if (type_name_const_str == g_objc_type_name_Class)
{
if (log)
GetObjectFile()->GetModule()->LogMessage (log.get(), "SymbolFileDWARF::ParseType (die = 0x%8.8x) %s '%s' is Objective C 'Class' built-in type.",
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
die->GetName(this, dwarf_cu));
clang_type = ast.GetBuiltInType_objc_Class();
encoding_data_type = Type::eEncodingIsUID;
encoding_uid = LLDB_INVALID_UID;
resolve_state = Type::eResolveStateFull;
}
else if (type_name_const_str == g_objc_type_name_selector)
{
if (log)
GetObjectFile()->GetModule()->LogMessage (log.get(), "SymbolFileDWARF::ParseType (die = 0x%8.8x) %s '%s' is Objective C 'selector' built-in type.",
die->GetOffset(),
DW_TAG_value_to_name(die->Tag()),
die->GetName(this, dwarf_cu));
clang_type = ast.GetBuiltInType_objc_selector();
encoding_data_type = Type::eEncodingIsUID;
encoding_uid = LLDB_INVALID_UID;
resolve_state = Type::eResolveStateFull;
}
Added support for objective C built-in types: id, Class, and SEL. This involved watching for the objective C built-in types in DWARF and making sure when we convert the DWARF types into clang types that we use the appropriate ASTContext types. Added a way to find and dump types in lldb (something equivalent to gdb's "ptype" command): image lookup --type <TYPENAME> This only works for looking up types by name and won't work with variables. It also currently dumps out verbose internal information. I will modify it to dump more appropriate user level info in my next submission. Hookup up the "FindTypes()" functions in the SymbolFile and SymbolVendor so we can lookup types by name in one or more images. Fixed "image lookup --address <ADDRESS>" to be able to correctly show all symbol context information, but it will only show this extra information when the new "--verbose" flag is used. Updated to latest LLVM to get a few needed fixes. llvm-svn: 110089
2010-08-03 00:35:52 +00:00
}
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset( new Type (MakeUserID(die->GetOffset()),
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
this,
type_name_const_str,
byte_size,
NULL,
encoding_uid,
encoding_data_type,
&decl,
clang_type,
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
resolve_state));
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = type_sp.get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
// Type* encoding_type = GetUniquedTypeForDIEOffset(encoding_uid, type_sp, NULL, 0, 0, false);
// if (encoding_type != NULL)
// {
// if (encoding_type != DIE_IS_BEING_PARSED)
// type_sp->SetEncodingType(encoding_type);
// else
// m_indirect_fixups.push_back(type_sp.get());
// }
}
break;
case DW_TAG_structure_type:
case DW_TAG_union_type:
case DW_TAG_class_type:
{
// Set a bit that lets us know that we are currently parsing this
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = DIE_IS_BEING_PARSED;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
LanguageType class_language = eLanguageTypeUnknown;
<rdar://problem/10500242> Find the one true objective C type definition if there is one within a module. llvm-svn: 145457
2011-11-29 23:40:34 +00:00
bool is_complete_objc_class = false;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
//bool struct_is_class = false;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
case DW_AT_decl_file:
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (dwarf_cu->DW_AT_decl_file_attributes_are_invalid())
{
// llvm-gcc outputs invalid DW_AT_decl_file attributes that always
// point to the compile unit file, so we clear this invalid value
// so that we can still unique types efficiently.
decl.SetFile(FileSpec ("<invalid>", false));
}
else
decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned()));
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
break;
case DW_AT_decl_line:
decl.SetLine(form_value.Unsigned());
break;
case DW_AT_decl_column:
decl.SetColumn(form_value.Unsigned());
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_name:
type_name_cstr = form_value.AsCString(&get_debug_str_data());
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
type_name_const_str.SetCString(type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
case DW_AT_byte_size:
byte_size = form_value.Unsigned();
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
byte_size_valid = true;
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
break;
case DW_AT_accessibility:
accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned());
break;
case DW_AT_declaration:
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
is_forward_declaration = form_value.Unsigned() != 0;
Created lldb::LanguageType by moving an enumeration from the lldb_private::Language class into the enumerations header so it can be freely used by other interfaces. Added correct objective C class support to the DWARF symbol parser. Prior to this fix we were parsing objective C classes as C++ classes and now that the expression parser is ready to call functions we need to make sure the objective C classes have correct AST types. llvm-svn: 109574
2010-07-28 02:04:09 +00:00
break;
case DW_AT_APPLE_runtime_class:
class_language = (LanguageType)form_value.Signed();
break;
<rdar://problem/10500242> Find the one true objective C type definition if there is one within a module. llvm-svn: 145457
2011-11-29 23:40:34 +00:00
case DW_AT_APPLE_objc_complete_type:
is_complete_objc_class = form_value.Signed();
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_allocated:
case DW_AT_associated:
case DW_AT_data_location:
case DW_AT_description:
case DW_AT_start_scope:
case DW_AT_visibility:
default:
case DW_AT_sibling:
break;
}
}
}
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
UniqueDWARFASTType unique_ast_entry;
Made SymbolFileDWARF be less strict when looking for types that can be uniqued to the given type. This is especially helpful when types are missing file and line information. llvm-svn: 150004
2012-02-07 21:13:38 +00:00
<rdar://problem/10967107> Don't try and unique anonymous struct/union/class types. llvm-svn: 151863
2012-03-02 00:07:15 +00:00
// Only try and unique the type if it has a name.
if (type_name_const_str &&
GetUniqueDWARFASTTypeMap().Find (type_name_const_str,
Made SymbolFileDWARF be less strict when looking for types that can be uniqued to the given type. This is especially helpful when types are missing file and line information. llvm-svn: 150004
2012-02-07 21:13:38 +00:00
this,
dwarf_cu,
die,
decl,
byte_size_valid ? byte_size : -1,
unique_ast_entry))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Made SymbolFileDWARF be less strict when looking for types that can be uniqued to the given type. This is especially helpful when types are missing file and line information. llvm-svn: 150004
2012-02-07 21:13:38 +00:00
// We have already parsed this type or from another
// compile unit. GCC loves to use the "one definition
// rule" which can result in multiple definitions
// of the same class over and over in each compile
// unit.
type_sp = unique_ast_entry.m_type_sp;
if (type_sp)
Fixed an issue in the DWARF parser that was causing forward declarations to not get resolved. Fixed the "void **isa_ptr" variable inside the objective C verifier to start with a '$' character so we don't go looking for it in our program. Moved the lookup for "$__lldb_class" into the part that knows we are looking for internal types that start with a '$'. llvm-svn: 118488
2010-11-09 04:42:43 +00:00
{
Made SymbolFileDWARF be less strict when looking for types that can be uniqued to the given type. This is especially helpful when types are missing file and line information. llvm-svn: 150004
2012-02-07 21:13:38 +00:00
m_die_to_type[die] = type_sp.get();
return type_sp;
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DEBUG_PRINTF ("0x%8.8llx: %s (\"%s\")\n", MakeUserID(die->GetOffset()), DW_TAG_value_to_name(tag), type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
int tag_decl_kind = -1;
AccessType default_accessibility = eAccessNone;
if (tag == DW_TAG_structure_type)
{
tag_decl_kind = clang::TTK_Struct;
default_accessibility = eAccessPublic;
}
else if (tag == DW_TAG_union_type)
{
tag_decl_kind = clang::TTK_Union;
default_accessibility = eAccessPublic;
}
else if (tag == DW_TAG_class_type)
{
tag_decl_kind = clang::TTK_Class;
default_accessibility = eAccessPrivate;
}
Add a work around to deal with incorrect forward class definitions in objective C DWARF emitted by clang. llvm-svn: 145471
2011-11-30 02:48:28 +00:00
if (byte_size_valid && byte_size == 0 && type_name_cstr &&
die->HasChildren() == false &&
sc.comp_unit->GetLanguage() == eLanguageTypeObjC)
{
// Work around an issue with clang at the moment where
// forward declarations for objective C classes are emitted
// as:
// DW_TAG_structure_type [2]
// DW_AT_name( "ForwardObjcClass" )
// DW_AT_byte_size( 0x00 )
// DW_AT_decl_file( "..." )
// DW_AT_decl_line( 1 )
//
// Note that there is no DW_AT_declaration and there are
// no children, and the byte size is zero.
is_forward_declaration = true;
}
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
Recognize Objective-C classes with runtime class ObjCPlusPlus as Objective-C classes. Really the compiler should say they have Objective-C runtime class, but we should be a little more resilient (we were refusing to find ivars in those classes before). Also added a test case. llvm-svn: 155515
2012-04-25 01:03:57 +00:00
if (class_language == eLanguageTypeObjC ||
class_language == eLanguageTypeObjC_plus_plus)
<rdar://problem/10500242> Find the one true objective C type definition if there is one within a module. llvm-svn: 145457
2011-11-29 23:40:34 +00:00
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (!is_complete_objc_class && Supports_DW_AT_APPLE_objc_complete_type(dwarf_cu))
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
{
// We have a valid eSymbolTypeObjCClass class symbol whose
// name matches the current objective C class that we
// are trying to find and this DIE isn't the complete
// definition (we checked is_complete_objc_class above and
// know it is false), so the real definition is in here somewhere
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
type_sp = FindCompleteObjCDefinitionTypeForDIE (die, type_name_const_str, true);
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
if (!type_sp && m_debug_map_symfile)
{
// We weren't able to find a full declaration in
// this DWARF, see if we have a declaration anywhere
// else...
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
type_sp = m_debug_map_symfile->FindCompleteObjCDefinitionTypeForDIE (die, type_name_const_str, true);
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
}
if (type_sp)
{
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF(%p) - 0x%8.8x: %s type \"%s\" is an incomplete objc type, complete type is 0x%8.8llx",
this,
die->GetOffset(),
DW_TAG_value_to_name(tag),
type_name_cstr,
type_sp->GetID());
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
}
// We found a real definition for this type elsewhere
// so lets use it and cache the fact that we found
// a complete type for this die
m_die_to_type[die] = type_sp.get();
return type_sp;
}
}
<rdar://problem/10500242> Find the one true objective C type definition if there is one within a module. llvm-svn: 145457
2011-11-29 23:40:34 +00:00
}
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
Added code to make sure we don't recursively try to find an objective C class. The thing with Objective C classes is the debug info might have a definition that isn't just a forward decl, but it is incomplete. So we need to look and see if we can find the complete definition and avoid recursing a lot due to the fact that our accelerator tables will have many versions of the type, but only one complete one. We might not also have the complete type and we need to deal with this correctly. llvm-svn: 145759
2011-12-03 04:40:03 +00:00
if (is_forward_declaration)
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
{
// We have a forward declaration to a type and we need
// to try and find a full declaration. We look in the
// current type index just in case we have a forward
// declaration followed by an actual declarations in the
// DWARF. If this fails, we need to look elsewhere...
CommandObjectProcess was recently changed to automatically use the platform to launch a process for debugging. Since this isn't supported on all platforms, we need to do what we used to do if this isn't supported. I added: bool Platform::CanDebugProcess (); This will get checked before trying to launch a process for debugging and then fall back to launching the process through the current host debugger. This should solve the issue for linux and keep the platform code clean. Centralized logging code for logging errors, warnings and logs when reporting things for modules or symbol files. Both lldb_private::Module and lldb_private::SymbolFile now have the following member functions: void LogMessage (Log *log, const char *format, ...); void ReportWarning (const char *format, ...); void ReportError (const char *format, ...); These will all output the module name and object (if any) such as: "error: lldb.so ...." "warning: my_archive.a(foo.o) ...." This will keep the output consistent and stop a lot of logging calls from having to try and output all of the information that uniquely identifies a module or symbol file. Many places in the code were grabbing the path to the object file manually and if the module represented a .o file in an archive, we would see log messages like: error: foo.a - some error happened llvm-svn: 145219
2011-11-28 01:45:00 +00:00
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF(%p) - 0x%8.8x: %s type \"%s\" is a forward declaration, trying to find complete type",
this,
die->GetOffset(),
DW_TAG_value_to_name(tag),
type_name_cstr);
CommandObjectProcess was recently changed to automatically use the platform to launch a process for debugging. Since this isn't supported on all platforms, we need to do what we used to do if this isn't supported. I added: bool Platform::CanDebugProcess (); This will get checked before trying to launch a process for debugging and then fall back to launching the process through the current host debugger. This should solve the issue for linux and keep the platform code clean. Centralized logging code for logging errors, warnings and logs when reporting things for modules or symbol files. Both lldb_private::Module and lldb_private::SymbolFile now have the following member functions: void LogMessage (Log *log, const char *format, ...); void ReportWarning (const char *format, ...); void ReportError (const char *format, ...); These will all output the module name and object (if any) such as: "error: lldb.so ...." "warning: my_archive.a(foo.o) ...." This will keep the output consistent and stop a lot of logging calls from having to try and output all of the information that uniquely identifies a module or symbol file. Many places in the code were grabbing the path to the object file manually and if the module represented a .o file in an archive, we would see log messages like: error: foo.a - some error happened llvm-svn: 145219
2011-11-28 01:45:00 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
DWARFDeclContext die_decl_ctx;
die->GetDWARFDeclContext(this, dwarf_cu, die_decl_ctx);
//type_sp = FindDefinitionTypeForDIE (dwarf_cu, die, type_name_const_str);
type_sp = FindDefinitionTypeForDWARFDeclContext (die_decl_ctx);
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
if (!type_sp && m_debug_map_symfile)
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
{
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
// We weren't able to find a full declaration in
// this DWARF, see if we have a declaration anywhere
// else...
<rdar://problem/11291668> Fixed an issue that would happen when using debug map with DWARF in the .o files where we wouldn't ever track down the actual definition for a type when things were in namespaces. We now serialize the decl context information into an intermediate format which allows us to track down the correct definition for a type regardless of which DWARF symbol file it comes from. We do this by creating a "DWARFDeclContext" object that contains the DW_TAG + name for each item in a decl context which we can then use to veto potential accelerator table matches. For example, the accelerator tables store the basename of the type, so if you have "std::vector<int>", we would end up with an accelerator table entry for the type that contained "vector<int>", which we would then search for using a DWARFDeclContext object that contained: [0] DW_TAG_class_type "vector<int>" [1] DW_TAG_namespace "std" This is currently used to track down forward declarations for things like "class a::b::Foo;". llvm-svn: 155488
2012-04-24 21:22:41 +00:00
type_sp = m_debug_map_symfile->FindDefinitionTypeForDWARFDeclContext (die_decl_ctx);
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
if (type_sp)
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
{
CommandObjectProcess was recently changed to automatically use the platform to launch a process for debugging. Since this isn't supported on all platforms, we need to do what we used to do if this isn't supported. I added: bool Platform::CanDebugProcess (); This will get checked before trying to launch a process for debugging and then fall back to launching the process through the current host debugger. This should solve the issue for linux and keep the platform code clean. Centralized logging code for logging errors, warnings and logs when reporting things for modules or symbol files. Both lldb_private::Module and lldb_private::SymbolFile now have the following member functions: void LogMessage (Log *log, const char *format, ...); void ReportWarning (const char *format, ...); void ReportError (const char *format, ...); These will all output the module name and object (if any) such as: "error: lldb.so ...." "warning: my_archive.a(foo.o) ...." This will keep the output consistent and stop a lot of logging calls from having to try and output all of the information that uniquely identifies a module or symbol file. Many places in the code were grabbing the path to the object file manually and if the module represented a .o file in an archive, we would see log messages like: error: foo.a - some error happened llvm-svn: 145219
2011-11-28 01:45:00 +00:00
if (log)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF(%p) - 0x%8.8x: %s type \"%s\" is a forward declaration, complete type is 0x%8.8llx",
this,
die->GetOffset(),
DW_TAG_value_to_name(tag),
type_name_cstr,
type_sp->GetID());
CommandObjectProcess was recently changed to automatically use the platform to launch a process for debugging. Since this isn't supported on all platforms, we need to do what we used to do if this isn't supported. I added: bool Platform::CanDebugProcess (); This will get checked before trying to launch a process for debugging and then fall back to launching the process through the current host debugger. This should solve the issue for linux and keep the platform code clean. Centralized logging code for logging errors, warnings and logs when reporting things for modules or symbol files. Both lldb_private::Module and lldb_private::SymbolFile now have the following member functions: void LogMessage (Log *log, const char *format, ...); void ReportWarning (const char *format, ...); void ReportError (const char *format, ...); These will all output the module name and object (if any) such as: "error: lldb.so ...." "warning: my_archive.a(foo.o) ...." This will keep the output consistent and stop a lot of logging calls from having to try and output all of the information that uniquely identifies a module or symbol file. Many places in the code were grabbing the path to the object file manually and if the module represented a .o file in an archive, we would see log messages like: error: foo.a - some error happened llvm-svn: 145219
2011-11-28 01:45:00 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
// We found a real definition for this type elsewhere
// so lets use it and cache the fact that we found
// a complete type for this die
m_die_to_type[die] = type_sp.get();
return type_sp;
Modified the PluginManager to be ready for loading plug-ins from a system LLDB plugin directory and a user LLDB plugin directory. We currently still need to work out at what layer the plug-ins will be, but at least we are prepared for plug-ins. Plug-ins will attempt to be loaded from the "/Developer/Library/PrivateFrameworks/LLDB.framework/Resources/Plugins" folder, and from the "~/Library/Application Support/LLDB/Plugins" folder on MacOSX. Each plugin will be scanned for: extern "C" bool LLDBPluginInitialize(void); extern "C" void LLDBPluginTerminate(void); If at least LLDBPluginInitialize is found, the plug-in will be loaded. The LLDBPluginInitialize function returns a bool that indicates if the plug-in should stay loaded or not (plug-ins might check the current OS, current hardware, or anything else and determine they don't want to run on the current host). The plug-in is uniqued by path and added to a static loaded plug-in map. The plug-in scanning happens during "lldb_private::Initialize()" which calls to the PluginManager::Initialize() function. Likewise with termination lldb_private::Terminate() calls PluginManager::Terminate(). The paths for the plug-in directories is fetched through new Host calls: bool Host::GetLLDBPath (ePathTypeLLDBSystemPlugins, dir_spec); bool Host::GetLLDBPath (ePathTypeLLDBUserPlugins, dir_spec); This way linux and other systems can define their own appropriate locations for plug-ins to be loaded. To allow dynamic shared library loading, the Host layer has also been modified to include shared library open, close and get symbol: static void * Host::DynamicLibraryOpen (const FileSpec &file_spec, Error &error); static Error Host::DynamicLibraryClose (void *dynamic_library_handle); static void * Host::DynamicLibraryGetSymbol (void *dynamic_library_handle, const char *symbol_name, Error &error); lldb_private::FileSpec also has been modified to support directory enumeration in an attempt to abstract the directory enumeration into one spot in the code. The directory enumertion function is static and takes a callback: typedef enum EnumerateDirectoryResult { eEnumerateDirectoryResultNext, // Enumerate next entry in the current directory eEnumerateDirectoryResultEnter, // Recurse into the current entry if it is a directory or symlink, or next if not eEnumerateDirectoryResultExit, // Exit from the current directory at the current level. eEnumerateDirectoryResultQuit // Stop directory enumerations at any level }; typedef FileSpec::EnumerateDirectoryResult (*EnumerateDirectoryCallbackType) (void *baton, FileSpec::FileType file_type, const FileSpec &spec); static FileSpec::EnumerateDirectoryResult FileSpec::EnumerateDirectory (const char *dir_path, bool find_directories, bool find_files, bool find_other, EnumerateDirectoryCallbackType callback, void *callback_baton); This allow clients to specify the directory to search, and specifies if only files, directories or other (pipe, symlink, fifo, etc) files will cause the callback to be called. The callback also gets to return with the action that should be performed after this directory entry. eEnumerateDirectoryResultNext specifies to continue enumerating through a directory with the next entry. eEnumerateDirectoryResultEnter specifies to recurse down into a directory entry, or if the file is not a directory or symlink/alias to a directory, then just iterate to the next entry. eEnumerateDirectoryResultExit specifies to exit the current directory and skip any entries that might be remaining, yet continue enumerating to the next entry in the parent directory. And finally eEnumerateDirectoryResultQuit means to abort all directory enumerations at all levels. Modified the Declaration class to not include column information currently since we don't have any compilers that currently support column based declaration information. Columns support can be re-enabled with the additions of a #define. Added the ability to find an EmulateInstruction plug-in given a target triple and optional plug-in name in the plug-in manager. Fixed a few cases where opendir/readdir was being used, but yet not closedir was being used. Soon these will be deprecated in favor of the new directory enumeration call that was added to the FileSpec class. llvm-svn: 124716
2011-02-02 02:24:04 +00:00
}
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
assert (tag_decl_kind != -1);
bool clang_type_was_created = false;
clang_type = m_forward_decl_die_to_clang_type.lookup (die);
if (clang_type == NULL)
{
Caching the DIE for the DeclContext as reported by GetClangDeclContextContainingDIE, for better debuggability. llvm-svn: 150211
2012-02-09 22:54:11 +00:00
const DWARFDebugInfoEntry *decl_ctx_die;
clang::DeclContext *decl_ctx = GetClangDeclContextContainingDIE (dwarf_cu, die, &decl_ctx_die);
Fixed an issue where a class that resides inside another class wasn't getting an access specifier set on it, causing an assertion to fire when building with a Debug+Asserts build of clang. llvm-svn: 143010
2011-10-26 03:31:36 +00:00
if (accessibility == eAccessNone && decl_ctx)
{
// Check the decl context that contains this class/struct/union.
// If it is a class we must give it an accessability.
const clang::Decl::Kind containing_decl_kind = decl_ctx->getDeclKind();
if (DeclKindIsCXXClass (containing_decl_kind))
accessibility = default_accessibility;
}
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
if (type_name_cstr && strchr (type_name_cstr, '<'))
{
ClangASTContext::TemplateParameterInfos template_param_infos;
if (ParseTemplateParameterInfos (dwarf_cu, die, template_param_infos))
{
clang::ClassTemplateDecl *class_template_decl = ParseClassTemplateDecl (decl_ctx,
Fixed an issue where a class that resides inside another class wasn't getting an access specifier set on it, causing an assertion to fire when building with a Debug+Asserts build of clang. llvm-svn: 143010
2011-10-26 03:31:36 +00:00
accessibility,
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
type_name_cstr,
tag_decl_kind,
template_param_infos);
clang::ClassTemplateSpecializationDecl *class_specialization_decl = ast.CreateClassTemplateSpecializationDecl (decl_ctx,
class_template_decl,
tag_decl_kind,
template_param_infos);
clang_type = ast.CreateClassTemplateSpecializationType (class_specialization_decl);
clang_type_was_created = true;
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)class_template_decl, MakeUserID(die->GetOffset()));
GetClangASTContext().SetMetadata((uintptr_t)class_specialization_decl, MakeUserID(die->GetOffset()));
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
}
}
if (!clang_type_was_created)
{
clang_type_was_created = true;
Fixed an issue where a class that resides inside another class wasn't getting an access specifier set on it, causing an assertion to fire when building with a Debug+Asserts build of clang. llvm-svn: 143010
2011-10-26 03:31:36 +00:00
clang_type = ast.CreateRecordType (decl_ctx,
accessibility,
type_name_cstr,
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
tag_decl_kind,
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
class_language,
We now record metadata for Objective-C interfaces, Objective-C methods, and Objective-C properties. llvm-svn: 154972
2012-04-18 01:06:17 +00:00
MakeUserID(die->GetOffset()));
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
}
// Store a forward declaration to this class type in case any
// parameters in any class methods need it for the clang
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
// types for function prototypes.
LinkDeclContextToDIE(ClangASTContext::GetDeclContextForType(clang_type), die);
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset (new Type (MakeUserID(die->GetOffset()),
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
this,
type_name_const_str,
byte_size,
NULL,
LLDB_INVALID_UID,
Type::eEncodingIsUID,
&decl,
clang_type,
Type::eResolveStateForward));
Added support for looking up the complete type for Objective-C classes. This allows LLDB to find ivars declared in class extensions in modules other than where the debugger is currently stopped (we already supported this when the debugger was stopped in the same module as the definition). This involved the following main changes: - The ObjCLanguageRuntime now knows how to hunt for the authoritative version of an Objective-C type. It looks for the symbol indicating a definition, and then gets the type from the module containing that symbol. - ValueObjects now report their type with a potential override, and the override is set if the type of the ValueObject is an Objective-C class or pointer type that is defined somewhere other than the original reported type. This means that "frame variable" will always use the complete type if one is available. - The ClangASTSource now looks for the complete type when looking for ivars. This means that "expr" will always use the complete type if one is available. - I added a testcase that verifies that both "frame variable" and "expr" work. llvm-svn: 151214
2012-02-22 23:57:45 +00:00
type_sp->SetIsCompleteObjCClass(is_complete_objc_class);
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
// Add our type to the unique type map so we don't
// end up creating many copies of the same type over
// and over in the ASTContext for our module
unique_ast_entry.m_type_sp = type_sp;
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
unique_ast_entry.m_symfile = this;
unique_ast_entry.m_cu = dwarf_cu;
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
unique_ast_entry.m_die = die;
unique_ast_entry.m_declaration = decl;
Tightened up type uniq'ing so we don't uniq two anonymous types to each other unless they have the same byte_size. llvm-svn: 150422
2012-02-13 22:30:16 +00:00
unique_ast_entry.m_byte_size = byte_size;
All UnwindPlan objects are now passed around as shared pointers. ArchDefaultUnwindPlan plug-in interfaces are now cached per architecture instead of being leaked for every frame. Split the ArchDefaultUnwindPlan_x86 into ArchDefaultUnwindPlan_x86_64 and ArchDefaultUnwindPlan_i386 interfaces. There were sporadic crashes that were due to something leaking or being destroyed when doing stack crawls. This patch should clear up these issues. llvm-svn: 125541
2011-02-15 00:19:15 +00:00
GetUniqueDWARFASTTypeMap().Insert (type_name_const_str,
unique_ast_entry);
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
If the expression parser is unable to complete a TagDecl in the context in which it was originally found, the expression parser now goes hunting for it in all modules (in the appropriate namespace, if applicable). This means that forward-declared types that exist in another shared library will now be resolved correctly. Added a test case to cover this. The test case also tests "frame variable," which does not have this functionality yet. llvm-svn: 146204
2011-12-08 23:45:45 +00:00
if (!is_forward_declaration)
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
{
// Always start the definition for a class type so that
// if the class has child classes or types that require
// the class to be created for use as their decl contexts
// the class will be ready to accept these child definitions.
If the expression parser is unable to complete a TagDecl in the context in which it was originally found, the expression parser now goes hunting for it in all modules (in the appropriate namespace, if applicable). This means that forward-declared types that exist in another shared library will now be resolved correctly. Added a test case to cover this. The test case also tests "frame variable," which does not have this functionality yet. llvm-svn: 146204
2011-12-08 23:45:45 +00:00
if (die->HasChildren() == false)
{
// No children for this struct/union/class, lets finish it
ast.StartTagDeclarationDefinition (clang_type);
ast.CompleteTagDeclarationDefinition (clang_type);
}
else if (clang_type_was_created)
{
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
// Start the definition if the class is not objective C since
// the underlying decls respond to isCompleteDefinition(). Objective
// C decls dont' respond to isCompleteDefinition() so we can't
// start the declaration definition right away. For C++ classs/union/structs
// we want to start the definition in case the class is needed as the
// declaration context for a contained class or type without the need
// to complete that type..
Recognize Objective-C classes with runtime class ObjCPlusPlus as Objective-C classes. Really the compiler should say they have Objective-C runtime class, but we should be a little more resilient (we were refusing to find ivars in those classes before). Also added a test case. llvm-svn: 155515
2012-04-25 01:03:57 +00:00
if (class_language != eLanguageTypeObjC &&
class_language != eLanguageTypeObjC_plus_plus)
<rdar://problem/11082392> Fixed an issue that could cause circular type parsing that will assert and kill LLDB. Prior to this fix the DWARF parser would always create class types and not start their definitions (for both C++ and ObjC classes) until we were asked to complete the class later. When we had cases like: class A { class B { }; }; We would alway try to complete A before specifying "A" as the decl context for B. Turns out we can just start the definition and still not complete the class since we can check the TagDecl::isCompleteDefinition() function. This only works for C++ types. This means we will not be pulling in the full definition of parent classes all the time and should help with our memory consumption and also reduce the amount of debug info we have to parse. I also reduced redundant code that was checking in a lldb::clang_type_t was a possible C++ dynamic type since it was still completing the type, just to see if it was dynamic. This was fixed in another function that was checking for a type being dynamic as an ObjC or a C++ type, but there was dedicated fucntion for C++ that we missed. llvm-svn: 153713
2012-03-30 00:51:13 +00:00
ast.StartTagDeclarationDefinition (clang_type);
If the expression parser is unable to complete a TagDecl in the context in which it was originally found, the expression parser now goes hunting for it in all modules (in the appropriate namespace, if applicable). This means that forward-declared types that exist in another shared library will now be resolved correctly. Added a test case to cover this. The test case also tests "frame variable," which does not have this functionality yet. llvm-svn: 146204
2011-12-08 23:45:45 +00:00
// Leave this as a forward declaration until we need
// to know the details of the type. lldb_private::Type
// will automatically call the SymbolFile virtual function
// "SymbolFileDWARF::ResolveClangOpaqueTypeDefinition(Type *)"
// When the definition needs to be defined.
m_forward_decl_die_to_clang_type[die] = clang_type;
m_forward_decl_clang_type_to_die[ClangASTType::RemoveFastQualifiers (clang_type)] = die;
ClangASTContext::SetHasExternalStorage (clang_type, true);
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
}
Fixed an issue where we are asking to get the decl context for a function that is in a class from the expression parser, and it was causing an assertion. There is now a function that will correctly resolve a type even if it is in a class. llvm-svn: 146141
2011-12-08 05:16:30 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
break;
case DW_TAG_enumeration_type:
{
// Set a bit that lets us know that we are currently parsing this
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = DIE_IS_BEING_PARSED;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
lldb::user_id_t encoding_uid = DW_INVALID_OFFSET;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_name:
type_name_cstr = form_value.AsCString(&get_debug_str_data());
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
type_name_const_str.SetCString(type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
case DW_AT_type: encoding_uid = form_value.Reference(dwarf_cu); break;
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
case DW_AT_byte_size: byte_size = form_value.Unsigned(); byte_size_valid = true; break;
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
case DW_AT_accessibility: accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned()); break;
case DW_AT_declaration: is_forward_declaration = form_value.Unsigned() != 0; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_allocated:
case DW_AT_associated:
case DW_AT_bit_stride:
case DW_AT_byte_stride:
case DW_AT_data_location:
case DW_AT_description:
case DW_AT_start_scope:
case DW_AT_visibility:
case DW_AT_specification:
case DW_AT_abstract_origin:
case DW_AT_sibling:
break;
}
}
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DEBUG_PRINTF ("0x%8.8llx: %s (\"%s\")\n", MakeUserID(die->GetOffset()), DW_TAG_value_to_name(tag), type_name_cstr);
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
clang_type_t enumerator_clang_type = NULL;
clang_type = m_forward_decl_die_to_clang_type.lookup (die);
if (clang_type == NULL)
{
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
enumerator_clang_type = ast.GetBuiltinTypeForDWARFEncodingAndBitSize (NULL,
DW_ATE_signed,
byte_size * 8);
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
clang_type = ast.CreateEnumerationType (type_name_cstr,
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
GetClangDeclContextContainingDIE (dwarf_cu, die, NULL),
Fixed an error in the type map for "char **" that was a bad memory smasher. Anytime we had a valid python list that was trying to go from Python down into our C++ API, it was allocating too little memory and it ended up smashing whatever was next to the allocated memory. Added typemap conversions for "void *, size_t" so we can get SBProcess::ReadMemory() working. Also added a typemap for "const void *, size_t" so we can get SBProcess::WriteMemory() to work. Fixed an issue in the DWARF parser where we weren't correctly calculating the DeclContext for all types and classes. We now should be a lot more accurate. Fixes include: enums should now be setting their parent decl context correctly. We saw a lot of examples where enums in classes were not being properly namespace scoped. Also, classes within classes now get properly scoped. Fixed the objective C runtime pointer checkers to let "nil" pointers through since these are accepted by compiled code. We also now don't call "abort()" when a pointer doesn't validate correctly since this was wreaking havoc on the process due to the way abort() works. We now just dereference memory which should give us an exception from which we can easily and reliably recover. llvm-svn: 123428
2011-01-14 04:54:56 +00:00
decl,
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
enumerator_clang_type);
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
}
else
{
enumerator_clang_type = ClangASTContext::GetEnumerationIntegerType (clang_type);
assert (enumerator_clang_type != NULL);
}
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
LinkDeclContextToDIE(ClangASTContext::GetDeclContextForType(clang_type), die);
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset( new Type (MakeUserID(die->GetOffset()),
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
this,
type_name_const_str,
byte_size,
NULL,
encoding_uid,
Type::eEncodingIsUID,
&decl,
clang_type,
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Type::eResolveStateForward));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
ast.StartTagDeclarationDefinition (clang_type);
if (die->HasChildren())
{
Enabled extra warnings and fixed a bunch of small issues. llvm-svn: 124250
2011-01-25 23:55:37 +00:00
SymbolContext cu_sc(GetCompUnitForDWARFCompUnit(dwarf_cu));
ParseChildEnumerators(cu_sc, clang_type, type_sp->GetByteSize(), dwarf_cu, die);
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
}
ast.CompleteTagDeclarationDefinition (clang_type);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
break;
Few little fixes to reading in inlined functions. Also added a test case with some inlining. llvm-svn: 110892
2010-08-12 01:20:14 +00:00
case DW_TAG_inlined_subroutine:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_TAG_subprogram:
case DW_TAG_subroutine_type:
{
// Set a bit that lets us know that we are currently parsing this
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = DIE_IS_BEING_PARSED;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const char *mangled = NULL;
dw_offset_t type_die_offset = DW_INVALID_OFFSET;
Added motheds to C++ classes as we parse them to keep clang happy. llvm-svn: 114616
2010-09-23 01:09:21 +00:00
bool is_variadic = false;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
bool is_inline = false;
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
bool is_static = false;
bool is_virtual = false;
Make C++ constructors and destructors correctly within the clang types we generate from DWARF. llvm-svn: 115268
2010-10-01 02:31:07 +00:00
bool is_explicit = false;
Sometimes the debug information includes artifically- generated special member functions (constructors, destructors, etc.) for classes that don't really have them. We needed to mark these as artificial to reflect the debug information; this bug does that for constructors and destructors. The "etc." case (certain assignment operators, mostly) remains to be fixed. llvm-svn: 143526
2011-11-02 01:38:59 +00:00
bool is_artificial = false;
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
dw_offset_t specification_die_offset = DW_INVALID_OFFSET;
dw_offset_t abstract_origin_die_offset = DW_INVALID_OFFSET;
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
unsigned type_quals = 0;
Updated to latest LLVM. Major LLVM changes: - Sema is now exported (and there was much rejoicing.) - Storage classes are now centrally defined. Also fixed some bugs that the new LLVM picked up. llvm-svn: 114622
2010-09-23 03:01:22 +00:00
clang::StorageClass storage = clang::SC_None;//, Extern, Static, PrivateExtern
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
Enabled extra warnings and fixed a bunch of small issues. llvm-svn: 124250
2011-01-25 23:55:37 +00:00
attr = attributes.AttributeAtIndex(i);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name:
type_name_cstr = form_value.AsCString(&get_debug_str_data());
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
type_name_const_str.SetCString(type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
case DW_AT_MIPS_linkage_name: mangled = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_type: type_die_offset = form_value.Reference(dwarf_cu); break;
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
case DW_AT_accessibility: accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned()); break;
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
case DW_AT_declaration: is_forward_declaration = form_value.Unsigned() != 0; break;
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
case DW_AT_inline: is_inline = form_value.Unsigned() != 0; break;
case DW_AT_virtuality: is_virtual = form_value.Unsigned() != 0; break;
Make C++ constructors and destructors correctly within the clang types we generate from DWARF. llvm-svn: 115268
2010-10-01 02:31:07 +00:00
case DW_AT_explicit: is_explicit = form_value.Unsigned() != 0; break;
Sometimes the debug information includes artifically- generated special member functions (constructors, destructors, etc.) for classes that don't really have them. We needed to mark these as artificial to reflect the debug information; this bug does that for constructors and destructors. The "etc." case (certain assignment operators, mostly) remains to be fixed. llvm-svn: 143526
2011-11-02 01:38:59 +00:00
case DW_AT_artificial: is_artificial = form_value.Unsigned() != 0; break;
Make C++ constructors and destructors correctly within the clang types we generate from DWARF. llvm-svn: 115268
2010-10-01 02:31:07 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_external:
if (form_value.Unsigned())
{
Updated to latest LLVM. Major LLVM changes: - Sema is now exported (and there was much rejoicing.) - Storage classes are now centrally defined. Also fixed some bugs that the new LLVM picked up. llvm-svn: 114622
2010-09-23 03:01:22 +00:00
if (storage == clang::SC_None)
storage = clang::SC_Extern;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
else
Updated to latest LLVM. Major LLVM changes: - Sema is now exported (and there was much rejoicing.) - Storage classes are now centrally defined. Also fixed some bugs that the new LLVM picked up. llvm-svn: 114622
2010-09-23 03:01:22 +00:00
storage = clang::SC_PrivateExtern;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
break;
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
case DW_AT_specification:
specification_die_offset = form_value.Reference(dwarf_cu);
break;
case DW_AT_abstract_origin:
abstract_origin_die_offset = form_value.Reference(dwarf_cu);
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_allocated:
case DW_AT_associated:
case DW_AT_address_class:
case DW_AT_calling_convention:
case DW_AT_data_location:
case DW_AT_elemental:
case DW_AT_entry_pc:
case DW_AT_frame_base:
case DW_AT_high_pc:
case DW_AT_low_pc:
case DW_AT_object_pointer:
case DW_AT_prototyped:
case DW_AT_pure:
case DW_AT_ranges:
case DW_AT_recursive:
case DW_AT_return_addr:
case DW_AT_segment:
case DW_AT_start_scope:
case DW_AT_static_link:
case DW_AT_trampoline:
case DW_AT_visibility:
case DW_AT_vtable_elem_location:
case DW_AT_description:
case DW_AT_sibling:
break;
}
}
}
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DEBUG_PRINTF ("0x%8.8llx: %s (\"%s\")\n", MakeUserID(die->GetOffset()), DW_TAG_value_to_name(tag), type_name_cstr);
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
clang_type_t return_clang_type = NULL;
Type *func_type = NULL;
if (type_die_offset != DW_INVALID_OFFSET)
func_type = ResolveTypeUID(type_die_offset);
Make C++ constructors and destructors correctly within the clang types we generate from DWARF. llvm-svn: 115268
2010-10-01 02:31:07 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (func_type)
Use forward types where possible to avoid having to parse extra DWARF when it is not required. llvm-svn: 146418
2011-12-12 21:50:19 +00:00
return_clang_type = func_type->GetClangForwardType();
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
else
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
return_clang_type = ast.GetBuiltInType_void();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Make C++ constructors and destructors correctly within the clang types we generate from DWARF. llvm-svn: 115268
2010-10-01 02:31:07 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
std::vector<clang_type_t> function_param_types;
std::vector<clang::ParmVarDecl*> function_param_decls;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
// Parse the function children for the parameters
Fixed a bug in which the DWARF reader did not distinguish appropriately between C++ static methods and non-static methods. This bug made it impossible to call most static methods, either because Clang did not recognize that a method could be called without providing a "this" parameter, or because Clang did not properly mangle the name of the method when searching for it in the target. Also added a testcase. llvm-svn: 136733
2011-08-02 22:21:50 +00:00
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
const DWARFDebugInfoEntry *decl_ctx_die = NULL;
clang::DeclContext *containing_decl_ctx = GetClangDeclContextContainingDIE (dwarf_cu, die, &decl_ctx_die);
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
const clang::Decl::Kind containing_decl_kind = containing_decl_ctx->getDeclKind();
Added template support when parsing DWARF into types. We can now use STL classes in the expression parser. llvm-svn: 142717
2011-10-22 03:33:13 +00:00
const bool is_cxx_method = DeclKindIsCXXClass (containing_decl_kind);
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
// Start off static. This will be set to false in ParseChildParameters(...)
// if we find a "this" paramters as the first parameter
if (is_cxx_method)
Fixed a bug in which the DWARF reader did not distinguish appropriately between C++ static methods and non-static methods. This bug made it impossible to call most static methods, either because Clang did not recognize that a method could be called without providing a "this" parameter, or because Clang did not properly mangle the name of the method when searching for it in the target. Also added a testcase. llvm-svn: 136733
2011-08-02 22:21:50 +00:00
is_static = true;
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
ClangASTContext::TemplateParameterInfos template_param_infos;
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (die->HasChildren())
{
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
bool skip_artificial = true;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ParseChildParameters (sc,
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
containing_decl_ctx,
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
dwarf_cu,
die,
Fixed a bug in which the DWARF reader did not distinguish appropriately between C++ static methods and non-static methods. This bug made it impossible to call most static methods, either because Clang did not recognize that a method could be called without providing a "this" parameter, or because Clang did not properly mangle the name of the method when searching for it in the target. Also added a testcase. llvm-svn: 136733
2011-08-02 22:21:50 +00:00
skip_artificial,
is_static,
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
type_list,
function_param_types,
First attempt and getting "const" C++ method function signatures correct. It currently isn't working, but it should be close. I will work on this more when I figure out what I am not doing correctly. llvm-svn: 119324
2010-11-16 02:10:54 +00:00
function_param_decls,
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
type_quals,
template_param_infos);
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
// clang_type will get the function prototype clang type after this call
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
clang_type = ast.CreateFunctionType (return_clang_type,
&function_param_types[0],
function_param_types.size(),
is_variadic,
type_quals);
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (type_name_cstr)
{
bool type_handled = false;
if (tag == DW_TAG_subprogram)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Be sure to use the one and only trusted source for chopping up objective C class names when extracting the class name, selector and name without category for objective C full class and instance method names. llvm-svn: 148435
2012-01-18 23:40:49 +00:00
ConstString class_name;
Fixed an issue with the Instruction subclasses where the strings might be fetched too many times and the DisassemblerLLVM was appending to strings when the opcode, mnemonic and comment accessors were called multiple times and if any of the strings were empty. Also fixed the test suite failures from recent Objective C modifications. llvm-svn: 148460
2012-01-19 03:24:53 +00:00
ConstString class_name_no_category;
if (ObjCLanguageRuntime::ParseMethodName (type_name_cstr, &class_name, NULL, NULL, &class_name_no_category))
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
{
Fixed an issue with the Instruction subclasses where the strings might be fetched too many times and the DisassemblerLLVM was appending to strings when the opcode, mnemonic and comment accessors were called multiple times and if any of the strings were empty. Also fixed the test suite failures from recent Objective C modifications. llvm-svn: 148460
2012-01-19 03:24:53 +00:00
// Use the class name with no category if there is one
if (class_name_no_category)
class_name = class_name_no_category;
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
SymbolContext empty_sc;
clang_type_t class_opaque_type = NULL;
Be sure to use the one and only trusted source for chopping up objective C class names when extracting the class name, selector and name without category for objective C full class and instance method names. llvm-svn: 148435
2012-01-18 23:40:49 +00:00
if (class_name)
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
{
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
TypeList types;
Added an extra way to chop up an objective C prototype and use it where necessary. llvm-svn: 148445
2012-01-19 00:52:59 +00:00
TypeSP complete_objc_class_type_sp (FindCompleteObjCDefinitionTypeForDIE (NULL, class_name, false));
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
if (complete_objc_class_type_sp)
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
{
<rdar://problem/10681814> Fix DWARF parsing issue we can run into when using llvm-gcc based dSYM files. Also fix the parsing of objective C built-in types (Class, id and SEL) so they don't parse more information that is not needed due to the way they are represented in DWARF. llvm-svn: 148016
2012-01-12 04:33:28 +00:00
clang_type_t type_clang_forward_type = complete_objc_class_type_sp->GetClangForwardType();
if (ClangASTContext::IsObjCClassType (type_clang_forward_type))
class_opaque_type = type_clang_forward_type;
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
}
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (class_opaque_type)
{
// If accessibility isn't set to anything valid, assume public for
// now...
if (accessibility == eAccessNone)
accessibility = eAccessPublic;
clang::ObjCMethodDecl *objc_method_decl;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
objc_method_decl = ast.AddMethodToObjCObjectType (class_opaque_type,
type_name_cstr,
clang_type,
accessibility);
Make sure we track CXX and objc method decls. llvm-svn: 136920
2011-08-04 21:02:57 +00:00
LinkDeclContextToDIE(ClangASTContext::GetAsDeclContext(objc_method_decl), die);
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
type_handled = objc_method_decl != NULL;
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)objc_method_decl, MakeUserID(die->GetOffset()));
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
}
Fixed some issues with parsing C++ methods where our detection was failing if the DWARF was laid out in a certain way. The way we detect C++ classes is now more robust so that a class method can be defined outside of the class and refer to a definition inside the class with a DW_AT_specification or DW_AT_abstract_origin attribute. Fixed a case in Thread.cpp where we were looking up info in the frame when we didn't need to. This was from some changes to support external editors. Now the info is only looked up if needed. llvm-svn: 137436
2011-08-12 06:47:54 +00:00
else if (is_cxx_method)
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
{
// Look at the parent of this DIE and see if is is
// a class or struct and see if this is actually a
// C++ method
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
Type *class_type = ResolveType (dwarf_cu, decl_ctx_die);
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (class_type)
{
<rdar://problem/11240464> Correctly unique a class' methods when we detect that a class has been uniqued to another. llvm-svn: 156795
2012-05-15 02:33:01 +00:00
if (class_type->GetID() != MakeUserID(decl_ctx_die->GetOffset()))
{
// We uniqued the parent class of this function to another class
// so we now need to associate all dies under "decl_ctx_die" to
// DIEs in the DIE for "class_type"...
DWARFCompileUnitSP class_type_cu_sp;
const DWARFDebugInfoEntry *class_type_die = DebugInfo()->GetDIEPtr(class_type->GetID(), &class_type_cu_sp);
if (class_type_die)
{
if (CopyUniqueClassMethodTypes (class_type,
class_type_cu_sp.get(),
class_type_die,
dwarf_cu,
decl_ctx_die))
{
type_ptr = m_die_to_type[die];
if (type_ptr)
{
type_sp = type_ptr->shared_from_this();
break;
}
}
}
}
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
if (specification_die_offset != DW_INVALID_OFFSET)
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
{
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
// We have a specification which we are going to base our function
// prototype off of, so we need this type to be completed so that the
// m_die_to_decl_ctx for the method in the specification has a valid
// clang decl context.
Use forward declarations more of the time to save on things that we need to parse. llvm-svn: 146473
2011-12-13 04:34:06 +00:00
class_type->GetClangForwardType();
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
// If we have a specification, then the function type should have been
// made with the specification and not with this die.
DWARFCompileUnitSP spec_cu_sp;
const DWARFDebugInfoEntry* spec_die = DebugInfo()->GetDIEPtr(specification_die_offset, &spec_cu_sp);
Use GetClangDeclContextForDIE, it'll find the cached one if it's there and we may not have a cached die yet. This fixes a bunch of false positives on "die has no decl". llvm-svn: 153417
2012-03-25 19:37:33 +00:00
clang::DeclContext *spec_clang_decl_ctx = GetClangDeclContextForDIE (sc, dwarf_cu, spec_die);
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
if (spec_clang_decl_ctx)
{
LinkDeclContextToDIE(spec_clang_decl_ctx, die);
}
else
Centralize the warning reporting and use ReportWarning & ReportError everywhere we were using fprintf(stderr directly. llvm-svn: 140813
2011-09-29 22:12:35 +00:00
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportWarning ("0x%8.8llx: DW_AT_specification(0x%8.8x) has no decl\n",
MakeUserID(die->GetOffset()),
specification_die_offset);
Centralize the warning reporting and use ReportWarning & ReportError everywhere we were using fprintf(stderr directly. llvm-svn: 140813
2011-09-29 22:12:35 +00:00
}
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
type_handled = true;
}
else if (abstract_origin_die_offset != DW_INVALID_OFFSET)
{
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
// We have a specification which we are going to base our function
// prototype off of, so we need this type to be completed so that the
// m_die_to_decl_ctx for the method in the abstract origin has a valid
// clang decl context.
Use forward declarations more of the time to save on things that we need to parse. llvm-svn: 146473
2011-12-13 04:34:06 +00:00
class_type->GetClangForwardType();
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
DWARFCompileUnitSP abs_cu_sp;
const DWARFDebugInfoEntry* abs_die = DebugInfo()->GetDIEPtr(abstract_origin_die_offset, &abs_cu_sp);
Use GetClangDeclContextForDIE, it'll find the cached one if it's there and we may not have a cached die yet. This fixes a bunch of false positives on "die has no decl". llvm-svn: 153417
2012-03-25 19:37:33 +00:00
clang::DeclContext *abs_clang_decl_ctx = GetClangDeclContextForDIE (sc, dwarf_cu, abs_die);
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
if (abs_clang_decl_ctx)
{
LinkDeclContextToDIE (abs_clang_decl_ctx, die);
}
else
Centralize the warning reporting and use ReportWarning & ReportError everywhere we were using fprintf(stderr directly. llvm-svn: 140813
2011-09-29 22:12:35 +00:00
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportWarning ("0x%8.8llx: DW_AT_abstract_origin(0x%8.8x) has no decl\n",
MakeUserID(die->GetOffset()),
abstract_origin_die_offset);
Centralize the warning reporting and use ReportWarning & ReportError everywhere we were using fprintf(stderr directly. llvm-svn: 140813
2011-09-29 22:12:35 +00:00
}
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
type_handled = true;
}
else
{
clang_type_t class_opaque_type = class_type->GetClangForwardType();
if (ClangASTContext::IsCXXClassType (class_opaque_type))
Changed the SymbolFile::FindFunction() function calls to only return lldb_private::Function objects. Previously the SymbolFileSymtab subclass would return lldb_private::Symbol objects when it was asked to find functions. The Module::FindFunctions (...) now take a boolean "bool include_symbols" so that the module can track down functions and symbols, yet functions are found by the SymbolFile plug-ins (through the SymbolVendor class), and symbols are gotten through the ObjectFile plug-ins. Fixed and issue where the DWARF parser might run into incomplete class member function defintions which would make clang mad when we tried to make certain member functions with invalid number of parameters (such as an operator= operator that had no parameters). Now we just avoid and don't complete these incomplete functions. llvm-svn: 124359
2011-01-27 06:44:37 +00:00
{
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
if (ClangASTContext::IsBeingDefined (class_opaque_type))
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
{
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
// Neither GCC 4.2 nor clang++ currently set a valid accessibility
// in the DWARF for C++ methods... Default to public for now...
if (accessibility == eAccessNone)
accessibility = eAccessPublic;
if (!is_static && !die->HasChildren())
{
// We have a C++ member function with no children (this pointer!)
// and clang will get mad if we try and make a function that isn't
// well formed in the DWARF, so we will just skip it...
type_handled = true;
}
else
{
clang::CXXMethodDecl *cxx_method_decl;
// REMOVE THE CRASH DESCRIPTION BELOW
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
Host::SetCrashDescriptionWithFormat ("SymbolFileDWARF::ParseType() is adding a method %s to class %s in DIE 0x%8.8llx from %s/%s",
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
type_name_cstr,
class_type->GetName().GetCString(),
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
MakeUserID(die->GetOffset()),
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
m_obj_file->GetFileSpec().GetDirectory().GetCString(),
m_obj_file->GetFileSpec().GetFilename().GetCString());
Added the capability (turned off for now) to mark a method as __attribute__ ((used)) when adding it to a class. This functionality is useful when stopped in anonymous namespaces: expressions attached to classes in anonymous namespaces are typically elided by Clang's CodeGen because they have no namespaces are intended not to be externally visible. __attribute__ ((used)) forces CodeGen to emit the function. Right now, __attribute__ ((used)) causes the JIT not to emit the function, so we're not enabling it until we fix that. llvm-svn: 143469
2011-11-01 18:07:13 +00:00
const bool is_attr_used = false;
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
cxx_method_decl = ast.AddMethodToCXXRecordType (class_opaque_type,
type_name_cstr,
clang_type,
accessibility,
is_virtual,
is_static,
is_inline,
Added the capability (turned off for now) to mark a method as __attribute__ ((used)) when adding it to a class. This functionality is useful when stopped in anonymous namespaces: expressions attached to classes in anonymous namespaces are typically elided by Clang's CodeGen because they have no namespaces are intended not to be externally visible. __attribute__ ((used)) forces CodeGen to emit the function. Right now, __attribute__ ((used)) causes the JIT not to emit the function, so we're not enabling it until we fix that. llvm-svn: 143469
2011-11-01 18:07:13 +00:00
is_explicit,
Sometimes the debug information includes artifically- generated special member functions (constructors, destructors, etc.) for classes that don't really have them. We needed to mark these as artificial to reflect the debug information; this bug does that for constructors and destructors. The "etc." case (certain assignment operators, mostly) remains to be fixed. llvm-svn: 143526
2011-11-02 01:38:59 +00:00
is_attr_used,
is_artificial);
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
LinkDeclContextToDIE(ClangASTContext::GetAsDeclContext(cxx_method_decl), die);
Made the Host::SetCrashDescription(const char *) function copy the incoming string to avoid possible later crashes. Modified the locations that do set the crash description to NULL out the string when they are done doing their tasks. llvm-svn: 144297
2011-11-10 18:31:53 +00:00
Host::SetCrashDescription (NULL);
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
type_handled = cxx_method_decl != NULL;
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)cxx_method_decl, MakeUserID(die->GetOffset()));
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
}
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
}
else
{
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
// We were asked to parse the type for a method in a class, yet the
// class hasn't been asked to complete itself through the
// clang::ExternalASTSource protocol, so we need to just have the
// class complete itself and do things the right way, then our
// DIE should then have an entry in the m_die_to_type map. First
// we need to modify the m_die_to_type so it doesn't think we are
// trying to parse this DIE anymore...
m_die_to_type[die] = NULL;
// Now we get the full type to force our class type to complete itself
// using the clang::ExternalASTSource protocol which will parse all
// base classes and all methods (including the method for this DIE).
class_type->GetClangFullType();
// The type for this DIE should have been filled in the function call above
type_ptr = m_die_to_type[die];
if (type_ptr)
{
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
type_sp = type_ptr->shared_from_this();
Fixed a case where we might end up trying to parse a type in the DWARF parser for a method whose class isn't currently in the process of completing itself. Currently, methods of a class, must be parsed when the class type that contains the method is asked to complete itself through the clang::ExternalASTSource virtual functions. Now we "do the right thing" by checking if the class is being defined, and if so we parse it, else we tell the class to complete itself so everything happens correctly. llvm-svn: 141908
2011-10-13 23:13:20 +00:00
break;
}
Fixed a problem where LLDB inserted regular C FunctionDecls into classes if it looked up a method in a different DWARF context than the one where it found the parent class's definition. The symptom of this was, for a method A::B(), 1) LLDB finds A in context 1, creating a CXXRecordDecl for A and marking it as needing completion 2) LLDB looks up B in context 2, finds that its parent A already has a CXXRecordDecl, but can't find a CXXMethodDecl for B 3) Not finding a CXXMethodDecl for B, LLDB doesn't set the flag indicating that B was resolved 4) Because the flag wasn't set, LLDB's fallthrough code creates a FunctionDecl for B and sticks it in the DeclContext -- in this case, A. 5) Clang crashes on finding a FunctionDecl inside a CXXRecordDecl. llvm-svn: 154627
2012-04-12 23:10:00 +00:00
// FIXME This is fixing some even uglier behavior but we really need to
// uniq the methods of each class as well as the class itself.
// <rdar://problem/11240464>
type_handled = true;
Fixed an issue where we could end up creating multiple C++ methods for a function depending on how the DWARF was created. Now we parse the class type from the definition, and all methods that use DW_AT_specification or DW_AT_abstract_origin attributes to point to the definition, now won't create duplicate entries. This is in response to how clang++ creates much different DWARF than gcc. llvm-svn: 137737
2011-08-16 18:40:23 +00:00
}
Changed the SymbolFile::FindFunction() function calls to only return lldb_private::Function objects. Previously the SymbolFileSymtab subclass would return lldb_private::Symbol objects when it was asked to find functions. The Module::FindFunctions (...) now take a boolean "bool include_symbols" so that the module can track down functions and symbols, yet functions are found by the SymbolFile plug-ins (through the SymbolVendor class), and symbols are gotten through the ObjectFile plug-ins. Fixed and issue where the DWARF parser might run into incomplete class member function defintions which would make clang mad when we tried to make certain member functions with invalid number of parameters (such as an operator= operator that had no parameters). Now we just avoid and don't complete these incomplete functions. llvm-svn: 124359
2011-01-27 06:44:37 +00:00
}
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
}
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (!type_handled)
{
// We just have a function that isn't part of a class
Add function decls to their parent decl context. llvm-svn: 142011
2011-10-14 22:47:18 +00:00
clang::FunctionDecl *function_decl = ast.CreateFunctionDeclaration (containing_decl_ctx,
type_name_cstr,
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
clang_type,
storage,
is_inline);
Almost have templatized functions working (templatized classes are already working, but not functions). I need to check on a few things to make sure I am registering everything correctly in the right order and in the right contexts. llvm-svn: 149858
2012-02-06 06:42:51 +00:00
// if (template_param_infos.GetSize() > 0)
// {
// clang::FunctionTemplateDecl *func_template_decl = ast.CreateFunctionTemplateDecl (containing_decl_ctx,
// function_decl,
// type_name_cstr,
// template_param_infos);
//
// ast.CreateFunctionTemplateSpecializationInfo (function_decl,
// func_template_decl,
// template_param_infos);
// }
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
// Add the decl to our DIE to decl context map
assert (function_decl);
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
LinkDeclContextToDIE(function_decl, die);
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
if (!function_param_decls.empty())
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
ast.SetFunctionParameters (function_decl,
&function_param_decls.front(),
function_param_decls.size());
Added a mechanism for keeping track of where in the debug information individual Decls came from. We've had a metadata infrastructure for a while, which was intended to solve a problem we've since dealt with in a different way. (It was meant to keep track of which definition of an Objective-C class was the "true" definition, but we now find it by searching the symbols for the class symbol.) The metadata is attached to the ExternalASTSource, which means it has a one-to-one correspondence with AST contexts. I've repurposed the metadata infrastructure to hold the object file and DIE offset for the DWARF information corresponding to a Decl. There are methods in ClangASTContext that get and set this metadata, and the ClangASTImporter is capable of tracking down the metadata for Decls that have been copied out of the debug information into the parser's AST context without using any additional memory. To see the metadata, you just have to enable the expression log: - (lldb) log enable lldb expr - and watch the import messages. The high 32 bits of the metadata indicate the index of the object file in its containing DWARFDebugMap; I have also added a log which you can use to track that mapping: - (lldb) log enable dwarf map - This adds 64 bits per Decl, which in my testing hasn't turned out to be very much (debugging Clang produces around 6500 Decls in my tests). To track how much data is being consumed, I've also added a global variable g_TotalSizeOfMetadata which tracks the total number of Decls that have metadata in all active AST contexts. Right now this metadata is enormously useful for tracking down bugs in the debug info parser. In the future I also want to use this information to provide more intelligent error messages instead of printing empty source lines wherever Clang refers to the location where something is defined. llvm-svn: 154634
2012-04-13 00:10:03 +00:00
GetClangASTContext().SetMetadata((uintptr_t)function_decl, MakeUserID(die->GetOffset()));
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset( new Type (MakeUserID(die->GetOffset()),
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
this,
type_name_const_str,
0,
NULL,
LLDB_INVALID_UID,
Type::eEncodingIsUID,
&decl,
clang_type,
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
Type::eResolveStateFull));
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
assert(type_sp.get());
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
break;
case DW_TAG_array_type:
{
// Set a bit that lets us know that we are currently parsing this
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
m_die_to_type[die] = DIE_IS_BEING_PARSED;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
lldb::user_id_t type_die_offset = DW_INVALID_OFFSET;
int64_t first_index = 0;
uint32_t byte_stride = 0;
uint32_t bit_stride = 0;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (num_attributes > 0)
{
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name:
type_name_cstr = form_value.AsCString(&get_debug_str_data());
Fixed C++ class clang type creation and display by making sure we omit artifical members (like the vtable pointer member that shows up in the DWARF). We were adding this to each class which was making all member variables be off by a pointer size. Added a test case so we can track this with "test/forward". Fixed the type name index in DWARF to include all the types after finding some types were being omitted due to the DW_AT_specification having the DW_AT_declaration attribute which was being read into the real type instances when there were forward declarations in the DWARF, causing the type to be omitted. We now check to make sure any DW_AT_declaration values are only respected when parsing types if the attribute is from the current DIE. After fixing the missing types, we ran into some issues with the expression parser finding duplicate entries for __va_list_tag since they are built in types and would result in a "duplicate __va_list_tag definition" error. We are now just ignoring this name during lookup, but we will need to see if we can get the name lookup function to not get called in these cases. Fixed an issue that would cause an assertion where DW_TAG_subroutine_types that had no children, would not properly make a clang function type of: "void (*) (void)". llvm-svn: 116392
2010-10-13 03:15:28 +00:00
type_name_const_str.SetCString(type_name_cstr);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
case DW_AT_type: type_die_offset = form_value.Reference(dwarf_cu); break;
Added a fix that should help incorrect type uniquing. There was an issue for templatized types that could cause parts of a std::vector (and I am sure other STL types) to be incorrectly uniqued to each other wreaking havoc on variable display for types within the same executable module. llvm-svn: 127662
2011-03-15 04:38:20 +00:00
case DW_AT_byte_size: byte_size = form_value.Unsigned(); byte_size_valid = true; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_byte_stride: byte_stride = form_value.Unsigned(); break;
case DW_AT_bit_stride: bit_stride = form_value.Unsigned(); break;
Added a new enumeration named "ClangASTContext::AccessType" that abstracts the type creation from the various access enumerations in Clang. Currently there are clang::AccessSpecifier and the objective C ivars have their own enumeration. So I added a new enumeration that will allow a consistent interface when creating types through ClangASTContext. I also added new functions to create an Objective C class, ivar and set an objective C superclass. They aren't hooked up in the DWARF parser yet. That is the next step, though I am unsure if I will do this in the DWARF parser or try and do it generically in the existing Record manipulation functions. llvm-svn: 109130
2010-07-22 18:30:50 +00:00
case DW_AT_accessibility: accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned()); break;
Did a lot of code cleanup. Fixed the DWARF plug-in such that when it gets all attributes for a DIE, that it omits the DW_AT_sibling and DW_AT_declaration when getting attributes from a DW_AT_abstract_origin or DW_AT_specification DIE. llvm-svn: 118654
2010-11-09 23:46:37 +00:00
case DW_AT_declaration: is_forward_declaration = form_value.Unsigned() != 0; break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
case DW_AT_allocated:
case DW_AT_associated:
case DW_AT_data_location:
case DW_AT_description:
case DW_AT_ordering:
case DW_AT_start_scope:
case DW_AT_visibility:
case DW_AT_specification:
case DW_AT_abstract_origin:
case DW_AT_sibling:
break;
}
}
}
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
DEBUG_PRINTF ("0x%8.8llx: %s (\"%s\")\n", MakeUserID(die->GetOffset()), DW_TAG_value_to_name(tag), type_name_cstr);
Fixed an issue where if a method funciton was asked to be parsed before its containing class was parsed, we would crash. llvm-svn: 115343
2010-10-01 20:48:32 +00:00
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Type *element_type = ResolveTypeUID(type_die_offset);
if (element_type)
{
std::vector<uint64_t> element_orders;
ParseChildArrayInfo(sc, dwarf_cu, die, first_index, element_orders, byte_stride, bit_stride);
Added a work in the DWARF parser when we parse an array that ends up having no elements so that they at least have 1 element. Added the ability to show the declaration location of variables to the "frame variables" with the "--show-declaration" option ("-c" for short). Changed the "frame variables" command over to use the value object code so that we use the same code path as the public API does when accessing and displaying variable values. llvm-svn: 113733
2010-09-13 02:37:44 +00:00
// We have an array that claims to have no members, lets give it at least one member...
if (element_orders.empty())
element_orders.push_back (1);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (byte_stride == 0 && bit_stride == 0)
byte_stride = element_type->GetByteSize();
Use forward types where possible to avoid having to parse extra DWARF when it is not required. llvm-svn: 146418
2011-12-12 21:50:19 +00:00
clang_type_t array_element_type = element_type->GetClangForwardType();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
uint64_t array_element_bit_stride = byte_stride * 8 + bit_stride;
uint64_t num_elements = 0;
std::vector<uint64_t>::const_reverse_iterator pos;
std::vector<uint64_t>::const_reverse_iterator end = element_orders.rend();
for (pos = element_orders.rbegin(); pos != end; ++pos)
{
num_elements = *pos;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
clang_type = ast.CreateArrayType (array_element_type,
num_elements,
num_elements * array_element_bit_stride);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
array_element_type = clang_type;
array_element_bit_stride = array_element_bit_stride * num_elements;
}
ConstString empty_name;
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset( new Type (MakeUserID(die->GetOffset()),
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
this,
empty_name,
array_element_bit_stride / 8,
NULL,
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
type_die_offset,
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
Type::eEncodingIsUID,
&decl,
clang_type,
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Type::eResolveStateFull));
type_sp->SetEncodingType (element_type);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
break;
Anders Carlsson patch for member pointers. Thanks Anders. llvm-svn: 105868
2010-06-12 01:20:30 +00:00
case DW_TAG_ptr_to_member_type:
{
dw_offset_t type_die_offset = DW_INVALID_OFFSET;
dw_offset_t containing_type_die_offset = DW_INVALID_OFFSET;
15-20% speed improvement when parsing DWARF. I used instruments to find the hotspots in our code when indexing the DWARF. A combination of using SmallVector to avoid collection allocations, using fixed form sizes when possible, and optimizing the hot loops contributed to the speedup. llvm-svn: 113961
2010-09-15 08:33:30 +00:00
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
Anders Carlsson patch for member pointers. Thanks Anders. llvm-svn: 105868
2010-06-12 01:20:30 +00:00
if (num_attributes > 0) {
uint32_t i;
for (i=0; i<num_attributes; ++i)
{
attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
{
switch (attr)
{
case DW_AT_type:
type_die_offset = form_value.Reference(dwarf_cu); break;
case DW_AT_containing_type:
containing_type_die_offset = form_value.Reference(dwarf_cu); break;
}
}
}
Type *pointee_type = ResolveTypeUID(type_die_offset);
Type *class_type = ResolveTypeUID(containing_type_die_offset);
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
clang_type_t pointee_clang_type = pointee_type->GetClangForwardType();
clang_type_t class_clang_type = class_type->GetClangLayoutType();
Anders Carlsson patch for member pointers. Thanks Anders. llvm-svn: 105868
2010-06-12 01:20:30 +00:00
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
clang_type = ast.CreateMemberPointerType(pointee_clang_type,
class_clang_type);
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
byte_size = ClangASTType::GetClangTypeBitWidth (ast.getASTContext(),
clang_type) / 8;
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
type_sp.reset( new Type (MakeUserID(die->GetOffset()),
Modified lldb_private::SymboleFile to be able to override where its TypeList comes from by using a virtual function to provide it from the Module's SymbolVendor by default. This allows the DWARF parser, when being used to parse DWARF in .o files with a parent DWARF + debug map parser, to get its type list from the DWARF + debug map parser so when we go and find full definitions for types (that might come from other .o files), we can use the type list from the debug map parser. Otherwise we ended up mixing clang types from one .o file (say a const pointer to a forward declaration "class A") with the a full type from another .o file. This causes expression parsing, when copying the clang types from those parsed by the DWARF parser into the expression AST, to fail -- for good reason. Now all types are created in the same list. Also added host support for crash description strings that can be set before doing a piece of work. On MacOSX, this ties in with CrashReporter support that allows a string to be dispalyed when the app crashes and allows LLDB.framework to print a description string in the crash log. Right now this is hookup up the the CommandInterpreter::HandleCommand() where each command notes that it is about to be executed, so if we crash while trying to do this command, we should be able to see the command that caused LLDB to exit. For all other platforms, this is a nop. llvm-svn: 118672
2010-11-10 04:57:04 +00:00
this,
type_name_const_str,
byte_size,
NULL,
LLDB_INVALID_UID,
Type::eEncodingIsUID,
NULL,
clang_type,
Modified the lldb_private::Type clang type resolving code to handle three cases when getting the clang type: - need only a forward declaration - need a clang type that can be used for layout (members and args/return types) - need a full clang type This allows us to partially parse the clang types and be as lazy as possible. The first case is when we just need to declare a type and we will complete it later. The forward declaration happens only for class/union/structs and enums. The layout type allows us to resolve the full clang type _except_ if we have any modifiers on a pointer or reference (both R and L value). In this case when we are adding members or function args or return types, we only need to know how the type will be laid out and we can defer completing the pointee type until we later need it. The last type means we need a full definition for the clang type. Did some renaming of some enumerations to get rid of the old "DC" prefix (which stands for DebugCore which is no longer around). Modified the clang namespace support to be almost ready to be fed to the expression parser. I made a new ClangNamespaceDecl class that can carry around the AST and the namespace decl so we can copy it into the expression AST. I modified the symbol vendor and symbol file plug-ins to use this new class. llvm-svn: 118976
2010-11-13 03:52:47 +00:00
Type::eResolveStateForward));
Anders Carlsson patch for member pointers. Thanks Anders. llvm-svn: 105868
2010-06-12 01:20:30 +00:00
}
break;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
default:
Anders Carlsson patch for member pointers. Thanks Anders. llvm-svn: 105868
2010-06-12 01:20:30 +00:00
assert(false && "Unhandled type tag!");
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
break;
}
if (type_sp.get())
{
const DWARFDebugInfoEntry *sc_parent_die = GetParentSymbolContextDIE(die);
dw_tag_t sc_parent_tag = sc_parent_die ? sc_parent_die->Tag() : 0;
SymbolContextScope * symbol_context_scope = NULL;
if (sc_parent_tag == DW_TAG_compile_unit)
{
symbol_context_scope = sc.comp_unit;
}
else if (sc.function != NULL)
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
symbol_context_scope = sc.function->GetBlock(true).FindBlockByID(MakeUserID(sc_parent_die->GetOffset()));
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (symbol_context_scope == NULL)
symbol_context_scope = sc.function;
}
if (symbol_context_scope != NULL)
{
type_sp->SetSymbolContextScope(symbol_context_scope);
}
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
// We are ready to put this type into the uniqued list up at the module level
type_list->Insert (type_sp);
Fixed the Objective C method prototypes to be correct (the selectors weren't being chopped up correctly). The DWARF plug-in also keeps a map of the ObjC class names to selectors for easy parsing of all class selectors when we parse the class type. llvm-svn: 116290
2010-10-12 02:24:53 +00:00
Added the DWARF unique type map such that we only create a type once in the module's AST context. Prior to this fix, with gcc binaries, we end up with a full class definition for any used classes in each compile unit due to the one definition rule. This would result in us making N copies of class T, where N is the number of compile units that use class T, in the module AST. When an expression would then try and use any types that were duplicated, it would quickly confuse clang and make expression evaluation fail due to all of the duplicate types that got copied over. This is now fixed by making a map of types in the DWARF that maps type names to a collection of types + declaration (file + line number) + DIE. Then later when we find a type we look in this module map and find any already cached types that we can just use. 8935777 llvm-svn: 125207
2011-02-09 19:06:17 +00:00
m_die_to_type[die] = type_sp.get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Hooked up detach for ProcessGDBRemote. Remove the GetUserData()/SetUserData() from the DWARFDebugInfoEntry class. We now track everything with dense maps. llvm-svn: 114876
2010-09-27 21:07:38 +00:00
else if (type_ptr != DIE_IS_BEING_PARSED)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Switching back to using std::tr1::shared_ptr. We originally switched away due to RTTI worries since llvm and clang don't use RTTI, but I was able to switch back with no issues as far as I can tell. Once the RTTI issue wasn't an issue, we were looking for a way to properly track weak pointers to objects to solve some of the threading issues we have been running into which naturally led us back to std::tr1::weak_ptr. We also wanted the ability to make a shared pointer from just a pointer, which is also easily solved using the std::tr1::enable_shared_from_this class. The main reason for this move back is so we can start properly having weak references to objects. Currently a lldb_private::Thread class has a refrence to its parent lldb_private::Process. This doesn't work well when we now hand out a SBThread object that contains a shared pointer to a lldb_private::Thread as this SBThread can be held onto by external clients and if they end up using one of these objects we can easily crash. So the next task is to start adopting std::tr1::weak_ptr where ever it makes sense which we can do with lldb_private::Debugger, lldb_private::Target, lldb_private::Process, lldb_private::Thread, lldb_private::StackFrame, and many more objects now that they are no longer using intrusive ref counted pointer objects (you can't do std::tr1::weak_ptr functionality with intrusive pointers). llvm-svn: 149207
2012-01-29 20:56:30 +00:00
type_sp = type_ptr->shared_from_this();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
return type_sp;
}
size_t
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
SymbolFileDWARF::ParseTypes
(
const SymbolContext& sc,
DWARFCompileUnit* dwarf_cu,
const DWARFDebugInfoEntry *die,
bool parse_siblings,
bool parse_children
)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
size_t types_added = 0;
while (die != NULL)
{
bool type_is_new = false;
Fixed the forward declaration issue that was present in the DWARF parser after adding methods to C++ and objective C classes. In order to make methods, we need the function prototype which means we need the arguments. Parsing these could cause a circular reference that caused an assertion. Added a new typedef for the clang opaque types which are just void pointers: lldb::clang_type_t. This appears in lldb-types.h. This was fixed by enabling struct, union, class, and enum types to only get a forward declaration when we make the clang opaque qual type for these types. When they need to actually be resolved, lldb_private::Type will call a new function in the SymbolFile protocol to resolve a clang type when it is not fully defined (clang::TagDecl::getDefinition() returns NULL). This allows us to be a lot more lazy when parsing clang types and keeps down the amount of data that gets parsed into the ASTContext for each module. Getting the clang type from a "lldb_private::Type" object now takes a boolean that indicates if a forward declaration is ok: clang_type_t lldb_private::Type::GetClangType (bool forward_decl_is_ok); So function prototypes that define parameters that are "const T&" can now just parse the forward declaration for type 'T' and we avoid circular references in the type system. llvm-svn: 115012
2010-09-29 01:12:09 +00:00
if (ParseType(sc, dwarf_cu, die, &type_is_new).get())
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
if (type_is_new)
++types_added;
}
if (parse_children && die->HasChildren())
{
if (die->Tag() == DW_TAG_subprogram)
{
SymbolContext child_sc(sc);
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
child_sc.function = sc.comp_unit->FindFunctionByUID(MakeUserID(die->GetOffset())).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
types_added += ParseTypes(child_sc, dwarf_cu, die->GetFirstChild(), true, true);
}
else
types_added += ParseTypes(sc, dwarf_cu, die->GetFirstChild(), true, true);
}
if (parse_siblings)
die = die->GetSibling();
else
die = NULL;
}
return types_added;
}
size_t
SymbolFileDWARF::ParseFunctionBlocks (const SymbolContext &sc)
{
assert(sc.comp_unit && sc.function);
size_t functions_added = 0;
DWARFCompileUnit* dwarf_cu = GetDWARFCompileUnitForUID(sc.comp_unit->GetID());
if (dwarf_cu)
{
dw_offset_t function_die_offset = sc.function->GetID();
const DWARFDebugInfoEntry *function_die = dwarf_cu->GetDIEPtr(function_die_offset);
if (function_die)
{
Fixed the issue of a DW_TAG_subprogram in a DW_TAG_subprogram correctly this time after recently backing out another fix. llvm-svn: 137475
2011-08-12 17:54:33 +00:00
ParseFunctionBlocks(sc, &sc.function->GetBlock (false), dwarf_cu, function_die, LLDB_INVALID_ADDRESS, 0);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
return functions_added;
}
size_t
SymbolFileDWARF::ParseTypes (const SymbolContext &sc)
{
// At least a compile unit must be valid
assert(sc.comp_unit);
size_t types_added = 0;
DWARFCompileUnit* dwarf_cu = GetDWARFCompileUnitForUID(sc.comp_unit->GetID());
if (dwarf_cu)
{
if (sc.function)
{
dw_offset_t function_die_offset = sc.function->GetID();
const DWARFDebugInfoEntry *func_die = dwarf_cu->GetDIEPtr(function_die_offset);
if (func_die && func_die->HasChildren())
{
types_added = ParseTypes(sc, dwarf_cu, func_die->GetFirstChild(), true, true);
}
}
else
{
const DWARFDebugInfoEntry *dwarf_cu_die = dwarf_cu->DIE();
if (dwarf_cu_die && dwarf_cu_die->HasChildren())
{
types_added = ParseTypes(sc, dwarf_cu, dwarf_cu_die->GetFirstChild(), true, true);
}
}
}
return types_added;
}
size_t
SymbolFileDWARF::ParseVariablesForContext (const SymbolContext& sc)
{
if (sc.comp_unit != NULL)
{
Fixed globals not showing up for any but the first compile unit due to a mismatch: I was using the compile unit user ID (the compile unit offset in the DWARF) as the compile unit index. llvm-svn: 117948
2010-11-01 20:32:12 +00:00
DWARFDebugInfo* info = DebugInfo();
if (info == NULL)
return 0;
uint32_t cu_idx = UINT32_MAX;
DWARFCompileUnit* dwarf_cu = info->GetCompileUnit(sc.comp_unit->GetID(), &cu_idx).get();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
if (dwarf_cu == NULL)
return 0;
if (sc.function)
{
const DWARFDebugInfoEntry *function_die = dwarf_cu->GetDIEPtr(sc.function->GetID());
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 02:20:48 +00:00
dw_addr_t func_lo_pc = function_die->GetAttributeValueAsUnsigned (this, dwarf_cu, DW_AT_low_pc, DW_INVALID_ADDRESS);
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
if (func_lo_pc != DW_INVALID_ADDRESS)
{
const size_t num_variables = ParseVariables(sc, dwarf_cu, func_lo_pc, function_die->GetFirstChild(), true, true);
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
// Let all blocks know they have parse all their variables
sc.function->GetBlock (false).SetDidParseVariables (true, true);
return num_variables;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
else if (sc.comp_unit)
{
uint32_t vars_added = 0;
VariableListSP variables (sc.comp_unit->GetVariableList(false));
if (variables.get() == NULL)
{
variables.reset(new VariableList());
sc.comp_unit->SetVariableList(variables);
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFCompileUnit* match_dwarf_cu = NULL;
const DWARFDebugInfoEntry* die = NULL;
DIEArray die_offsets;
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_using_apple_tables)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Fixed some issues where we might not have one of the new apple accelerator tables (like the .apple_namespaces) and it would cause us to index DWARF that didn't need to be indexed. Updated the MappedHash.h (generic Apple accelerator table) and the DWARF specific one (HashedNameToDIE.h) to be up to date with the latest and greatest hash table format. llvm-svn: 142627
2011-10-20 22:30:33 +00:00
if (m_apple_names_ap.get())
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
{
DWARFMappedHash::DIEInfoArray hash_data_array;
if (m_apple_names_ap->AppendAllDIEsInRange (dwarf_cu->GetOffset(),
dwarf_cu->GetNextCompileUnitOffset(),
hash_data_array))
{
DWARFMappedHash::ExtractDIEArray (hash_data_array, die_offsets);
}
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
else
{
// Index if we already haven't to make sure the compile units
// get indexed and make their global DIE index list
if (!m_indexed)
Index ();
m_global_index.FindAllEntriesForCompileUnit (dwarf_cu->GetOffset(),
dwarf_cu->GetNextCompileUnitOffset(),
die_offsets);
}
const size_t num_matches = die_offsets.size();
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
if (num_matches)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
DWARFDebugInfo* debug_info = DebugInfo();
for (size_t i=0; i<num_matches; ++i)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
const dw_offset_t die_offset = die_offsets[i];
die = debug_info->GetDIEPtrWithCompileUnitHint (die_offset, &match_dwarf_cu);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
if (die)
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
{
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
VariableSP var_sp (ParseVariableDIE(sc, dwarf_cu, die, LLDB_INVALID_ADDRESS));
if (var_sp)
{
variables->AddVariableIfUnique (var_sp);
++vars_added;
}
Huge memory and performance improvements in the DWARF parser. Address ranges are now split up into two different tables: - one in DWARFDebugInfo that is compile unit specific - one in each DWARFCompileUnit that has exact function DIE offsets This helps keep the size of the aranges down since the main table will get uniqued and sorted and have consecutive ranges merged. We then only parse the compile unit one on demand once we have determined that a compile unit contains the address in question. We also now use the .debug_aranges section if there is one instead of always indexing the DWARF manually. NameToDIE now uses a UniqueCStringMap<dw_offset> map instead of a std::map. std::map is very bulky as each node has 3 pointers and the key and value types. This gets our NameToDIE entry down to 12 bytes each instead of 48 which saves us a lot of memory when we have very large DWARF. DWARFDebugAranges now has a smaller footprint for each range it contains to save on memory. llvm-svn: 139557
2011-09-12 23:21:58 +00:00
}
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
else
{
if (m_using_apple_tables)
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportErrorIfModifyDetected ("the DWARF debug information has been modified (.apple_names accelerator table had bad die 0x%8.8x)\n", die_offset);
Added report errors for when the .apple_XXXX hashed name table lookups contain invalid information. This will help us perfect the tables that are emitted by clang. llvm-svn: 144359
2011-11-11 03:16:25 +00:00
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
}
return vars_added;
}
}
return 0;
}
VariableSP
SymbolFileDWARF::ParseVariableDIE
(
const SymbolContext& sc,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 02:20:48 +00:00
const DWARFDebugInfoEntry *die,
const lldb::addr_t func_low_pc
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
)
{
Just like functions can have a basename and a mangled/demangled name, variable can too. So now the lldb_private::Variable class has support for this. Variables now have support for having a basename ("i"), and a mangled name ("_ZN12_GLOBAL__N_11iE"), and a demangled name ("(anonymous namespace)::i"). Nowwhen searching for a variable by name, users might enter the fully qualified name, or just the basename. So new test functions were added to the Variable and Mangled classes as: bool NameMatches (const ConstString &name); bool NameMatches (const RegularExpression &regex); I also modified "ClangExpressionDeclMap::FindVariableInScope" to also search for global variables that are not in the current file scope by first starting with the current module, then moving on to all modules. Fixed an issue in the DWARF parser that could cause a varaible to get parsed more than once. Now, once we have parsed a VariableSP for a DIE, we cache the result even if a variable wasn't made so we don't do any re-parsing. Some DW_TAG_variable DIEs don't have locations, or are missing vital info that stops a debugger from being able to display anything for it, we parse a NULL variable shared pointer for these DIEs so we don't keep trying to reparse it. llvm-svn: 119085
2010-11-14 22:13:40 +00:00
VariableSP var_sp (m_die_to_variable_sp[die]);
if (var_sp)
return var_sp; // Already been parsed!
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const dw_tag_t tag = die->Tag();
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if ((tag == DW_TAG_variable) ||
(tag == DW_TAG_constant) ||
(tag == DW_TAG_formal_parameter && sc.function))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DWARFDebugInfoEntry::Attributes attributes;
const size_t num_attributes = die->GetAttributes(this, dwarf_cu, NULL, attributes);
if (num_attributes > 0)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const char *name = NULL;
const char *mangled = NULL;
Declaration decl;
uint32_t i;
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
lldb::user_id_t type_uid = LLDB_INVALID_UID;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
DWARFExpression location;
bool is_external = false;
bool is_artificial = false;
bool location_is_const_value_data = false;
AccessType accessibility = eAccessNone;
for (i=0; i<num_attributes; ++i)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
dw_attr_t attr = attributes.AttributeAtIndex(i);
DWARFFormValue form_value;
if (attributes.ExtractFormValueAtIndex(this, i, form_value))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
switch (attr)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
case DW_AT_decl_file: decl.SetFile(sc.comp_unit->GetSupportFiles().GetFileSpecAtIndex(form_value.Unsigned())); break;
case DW_AT_decl_line: decl.SetLine(form_value.Unsigned()); break;
case DW_AT_decl_column: decl.SetColumn(form_value.Unsigned()); break;
case DW_AT_name: name = form_value.AsCString(&get_debug_str_data()); break;
case DW_AT_MIPS_linkage_name: mangled = form_value.AsCString(&get_debug_str_data()); break;
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
case DW_AT_type: type_uid = form_value.Reference(dwarf_cu); break;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
case DW_AT_external: is_external = form_value.Unsigned() != 0; break;
case DW_AT_const_value:
location_is_const_value_data = true;
// Fall through...
case DW_AT_location:
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (form_value.BlockData())
{
const DataExtractor& debug_info_data = get_debug_info_data();
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
uint32_t block_offset = form_value.BlockData() - debug_info_data.GetDataStart();
uint32_t block_length = form_value.Unsigned();
location.SetOpcodeData(get_debug_info_data(), block_offset, block_length);
}
else
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const DataExtractor& debug_loc_data = get_debug_loc_data();
const dw_offset_t debug_loc_offset = form_value.Unsigned();
size_t loc_list_length = DWARFLocationList::Size(debug_loc_data, debug_loc_offset);
if (loc_list_length > 0)
{
location.SetOpcodeData(debug_loc_data, debug_loc_offset, loc_list_length);
assert (func_low_pc != LLDB_INVALID_ADDRESS);
location.SetLocationListSlide (func_low_pc - dwarf_cu->GetBaseAddress());
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
break;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
case DW_AT_artificial: is_artificial = form_value.Unsigned() != 0; break;
case DW_AT_accessibility: accessibility = DW_ACCESS_to_AccessType(form_value.Unsigned()); break;
case DW_AT_declaration:
case DW_AT_description:
case DW_AT_endianity:
case DW_AT_segment:
case DW_AT_start_scope:
case DW_AT_visibility:
default:
case DW_AT_abstract_origin:
case DW_AT_sibling:
case DW_AT_specification:
break;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (location.IsValid())
{
ValueType scope = eValueTypeInvalid;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
const DWARFDebugInfoEntry *sc_parent_die = GetParentSymbolContextDIE(die);
dw_tag_t parent_tag = sc_parent_die ? sc_parent_die->Tag() : 0;
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
SymbolContextScope * symbol_context_scope = NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
// DWARF doesn't specify if a DW_TAG_variable is a local, global
// or static variable, so we have to do a little digging by
// looking at the location of a varaible to see if it contains
// a DW_OP_addr opcode _somewhere_ in the definition. I say
// somewhere because clang likes to combine small global variables
// into the same symbol and have locations like:
// DW_OP_addr(0x1000), DW_OP_constu(2), DW_OP_plus
// So if we don't have a DW_TAG_formal_parameter, we can look at
// the location to see if it contains a DW_OP_addr opcode, and
// then we can correctly classify our variables.
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
if (tag == DW_TAG_formal_parameter)
scope = eValueTypeVariableArgument;
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
else
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
<rdar://problem/10507811> Be better at detecting when DWARF changes and handle this more gracefully than asserting and exiting. Also fixed up a bunch of system calls that weren't properly checking for EINTR. llvm-svn: 147559
2012-01-04 22:56:43 +00:00
bool op_error = false;
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
// Check if the location has a DW_OP_addr with any address value...
addr_t location_has_op_addr = false;
if (!location_is_const_value_data)
<rdar://problem/10507811> Be better at detecting when DWARF changes and handle this more gracefully than asserting and exiting. Also fixed up a bunch of system calls that weren't properly checking for EINTR. llvm-svn: 147559
2012-01-04 22:56:43 +00:00
{
location_has_op_addr = location.LocationContains_DW_OP_addr (LLDB_INVALID_ADDRESS, op_error);
if (op_error)
{
StreamString strm;
location.DumpLocationForAddress (&strm, eDescriptionLevelFull, 0, 0, NULL);
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportError ("0x%8.8x: %s has an invalid location: %s", die->GetOffset(), DW_TAG_value_to_name(die->Tag()), strm.GetString().c_str());
<rdar://problem/10507811> Be better at detecting when DWARF changes and handle this more gracefully than asserting and exiting. Also fixed up a bunch of system calls that weren't properly checking for EINTR. llvm-svn: 147559
2012-01-04 22:56:43 +00:00
}
}
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
if (location_has_op_addr)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
if (is_external)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
scope = eValueTypeVariableGlobal;
if (m_debug_map_symfile)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
// When leaving the DWARF in the .o files on darwin,
// when we have a global variable that wasn't initialized,
// the .o file might not have allocated a virtual
// address for the global variable. In this case it will
// have created a symbol for the global variable
// that is undefined and external and the value will
// be the byte size of the variable. When we do the
// address map in SymbolFileDWARFDebugMap we rely on
// having an address, we need to do some magic here
// so we can get the correct address for our global
// variable. The address for all of these entries
// will be zero, and there will be an undefined symbol
// in this object file, and the executable will have
// a matching symbol with a good address. So here we
// dig up the correct address and replace it in the
// location for the variable, and set the variable's
// symbol context scope to be that of the main executable
// so the file address will resolve correctly.
<rdar://problem/10507811> Be better at detecting when DWARF changes and handle this more gracefully than asserting and exiting. Also fixed up a bunch of system calls that weren't properly checking for EINTR. llvm-svn: 147559
2012-01-04 22:56:43 +00:00
if (location.LocationContains_DW_OP_addr (0, op_error))
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
// we have a possible uninitialized extern global
Symtab *symtab = m_obj_file->GetSymtab();
if (symtab)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
ConstString const_name(name);
Symbol *undefined_symbol = symtab->FindFirstSymbolWithNameAndType (const_name,
eSymbolTypeUndefined,
Symtab::eDebugNo,
Symtab::eVisibilityExtern);
if (undefined_symbol)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
ObjectFile *debug_map_objfile = m_debug_map_symfile->GetObjectFile();
if (debug_map_objfile)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
Symtab *debug_map_symtab = debug_map_objfile->GetSymtab();
Symbol *defined_symbol = debug_map_symtab->FindFirstSymbolWithNameAndType (const_name,
eSymbolTypeData,
Symtab::eDebugYes,
Symtab::eVisibilityExtern);
if (defined_symbol)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
<rdar://problem/10997402> This fix really needed to happen as a previous fix I had submitted for calculating symbol sizes made many symbols appear to have zero size since the function that was calculating the symbol size was calling another function that would cause the calculation to happen again. This resulted in some symbols having zero size when they shouldn't. This could then cause infinite stack traces and many other side affects. llvm-svn: 152244
2012-03-07 21:03:09 +00:00
if (defined_symbol->ValueIsAddress())
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
<rdar://problem/10997402> This fix really needed to happen as a previous fix I had submitted for calculating symbol sizes made many symbols appear to have zero size since the function that was calculating the symbol size was calling another function that would cause the calculation to happen again. This resulted in some symbols having zero size when they shouldn't. This could then cause infinite stack traces and many other side affects. llvm-svn: 152244
2012-03-07 21:03:09 +00:00
const addr_t defined_addr = defined_symbol->GetAddress().GetFileAddress();
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
if (defined_addr != LLDB_INVALID_ADDRESS)
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
{
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
if (location.Update_DW_OP_addr (defined_addr))
{
symbol_context_scope = defined_symbol;
}
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
}
}
}
}
}
}
}
}
}
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
else
{
scope = eValueTypeVariableStatic;
}
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
}
else
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
{
scope = eValueTypeVariableLocal;
}
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
if (symbol_context_scope == NULL)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
switch (parent_tag)
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
{
<rdar://problem/10338439> This is the actual fix for the above radar where global variables that weren't initialized were not being shown correctly when leaving the DWARF in the .o files. Global variables that aren't intialized have symbols in the .o files that specify they are undefined and external to the .o file, yet document the size of the variable. This allows the compiler to emit a single copy, but makes it harder for our DWARF in .o files with the executable having a debug map because the symbol for the global in the .o file doesn't exist in a section that we can assign a fixed up linked address to, and also the DWARF contains an invalid address in the "DW_OP_addr" location (always zero). This means that the DWARF is incorrect and actually maps all such global varaibles to the first file address in the .o file which is usually the first function. So we can fix this in either of two ways: make a new fake section in the .o file so that we have a file address in the .o file that we can relink, or fix the the variable as it is created in the .o file DWARF parser and actually give it the file address from the executable. Each variable contains a SymbolContextScope, or a single pointer that helps us to recreate where the variables came from (which module, file, function, etc). This context helps us to resolve any file addresses that might be in the location description of the variable by pointing us to which file the file address comes from, so we can just replace the SymbolContextScope and also fix up the location, which we would have had to do for the other case as well, and update the file address. Now globals display correctly. The above changes made it possible to determine if a variable is a global or static variable when parsing DWARF. The DWARF emits a DW_TAG_variable tag for each variable (local, global, or static), yet DWARF provides no way for us to classify these variables into these categories. We can now detect when a variable has a simple address expressions as its location and this will help us classify these correctly. While making the above changes I also noticed that we had two symbol types: eSymbolTypeExtern and eSymbolTypeUndefined which mean essentially the same thing: the symbol is not defined in the current object file. Symbol objects also have a bit that specifies if a symbol is externally visible, so I got rid of the eSymbolTypeExtern symbol type and moved all code locations that used it to use the eSymbolTypeUndefined type. llvm-svn: 144489
2011-11-13 04:15:56 +00:00
case DW_TAG_subprogram:
case DW_TAG_inlined_subroutine:
case DW_TAG_lexical_block:
if (sc.function)
{
symbol_context_scope = sc.function->GetBlock(true).FindBlockByID(MakeUserID(sc_parent_die->GetOffset()));
if (symbol_context_scope == NULL)
symbol_context_scope = sc.function;
}
break;
default:
symbol_context_scope = sc.comp_unit;
break;
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
if (symbol_context_scope)
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
{
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
var_sp.reset (new Variable (MakeUserID(die->GetOffset()),
name,
mangled,
Added a new class called lldb_private::SymbolFileType which is designed to take a SymbolFile reference and a lldb::user_id_t and be used in objects which represent things in debug symbols that have types where we don't need to know the true type yet, such as in lldb_private::Variable objects. This allows us to defer resolving the type until something is used. More specifically this allows us to get 1000 local variables from the current function, and if the user types "frame variable argc", we end up _only_ resolving the type for "argc" and not for the 999 other local variables. We can expand the use of this as needed in the future. Modified the DWARFMappedHash class to be able to read the HashData that has more than just the DIE offset. It currently will read the atoms in the header definition and read the data correctly. Currently only the DIE offset and type flags are supported. This is needed for adding type flags to the .apple_types hash accelerator tables. Fixed a assertion crash that would happen if we have a variable that had a DW_AT_const_value instead of a location where "location.LocationContains_DW_OP_addr()" would end up asserting when it tried to parse the variable location as a DWARF opcode list. Decreased the amount of memory that LLDB would use when evaluating an expression by 3x - 4x for clang. There was a place in the namespace lookup code that was parsing all namespaces with a certain name in a DWARF file instead of stopping when it found the first match. This was causing all of the compile units with a matching namespace to get parsed into memory and causing unnecessary memory bloat. Improved "Target::EvaluateExpression(...)" to not try and find a variable when the expression contains characters that would certainly cause an expression to need to be evaluated by the debugger. llvm-svn: 146130
2011-12-08 02:13:16 +00:00
SymbolFileTypeSP (new SymbolFileType(*this, type_uid)),
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
scope,
symbol_context_scope,
&decl,
location,
is_external,
is_artificial));
Stop spurious "DW_AT_specification(0x%8.8x) has no decl" warnings that were due to not parsing the function types when the function blocks are made and parsed. Cached the Function object's m_type value after it has been calculated. llvm-svn: 141225
2011-10-05 22:22:08 +00:00
var_sp->SetLocationIsConstantValueData (location_is_const_value_data);
}
else
{
// Not ready to parse this variable yet. It might be a global
// or static variable that is in a function scope and the function
// in the symbol context wasn't filled in yet
return var_sp;
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
}
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Enable all the new accelerator tables if they are present and don't manually index the DWARF. Also fixed an issue with memory accelerator tables with a size of 1 where we would loop infinitely. Added support for parsing the new .apple_namespaces section which gives us a memory hash table for looking up namespaces. llvm-svn: 141128
2011-10-04 22:41:51 +00:00
// Cache var_sp even if NULL (the variable was just a specification or
// was missing vital information to be able to be displayed in the debugger
// (missing location due to optimization, etc)) so we don't re-parse
// this DIE over and over later...
m_die_to_variable_sp[die] = var_sp;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
return var_sp;
}
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
const DWARFDebugInfoEntry *
SymbolFileDWARF::FindBlockContainingSpecification (dw_offset_t func_die_offset,
dw_offset_t spec_block_die_offset,
DWARFCompileUnit **result_die_cu_handle)
{
// Give the concrete function die specified by "func_die_offset", find the
// concrete block whose DW_AT_specification or DW_AT_abstract_origin points
// to "spec_block_die_offset"
DWARFDebugInfo* info = DebugInfo();
const DWARFDebugInfoEntry *die = info->GetDIEPtrWithCompileUnitHint(func_die_offset, result_die_cu_handle);
if (die)
{
assert (*result_die_cu_handle);
return FindBlockContainingSpecification (*result_die_cu_handle, die, spec_block_die_offset, result_die_cu_handle);
}
return NULL;
}
const DWARFDebugInfoEntry *
SymbolFileDWARF::FindBlockContainingSpecification(DWARFCompileUnit* dwarf_cu,
const DWARFDebugInfoEntry *die,
dw_offset_t spec_block_die_offset,
DWARFCompileUnit **result_die_cu_handle)
{
if (die)
{
switch (die->Tag())
{
case DW_TAG_subprogram:
case DW_TAG_inlined_subroutine:
case DW_TAG_lexical_block:
{
if (die->GetAttributeValueAsReference (this, dwarf_cu, DW_AT_specification, DW_INVALID_OFFSET) == spec_block_die_offset)
{
*result_die_cu_handle = dwarf_cu;
return die;
}
if (die->GetAttributeValueAsReference (this, dwarf_cu, DW_AT_abstract_origin, DW_INVALID_OFFSET) == spec_block_die_offset)
{
*result_die_cu_handle = dwarf_cu;
return die;
}
}
break;
}
// Give the concrete function die specified by "func_die_offset", find the
// concrete block whose DW_AT_specification or DW_AT_abstract_origin points
// to "spec_block_die_offset"
for (const DWARFDebugInfoEntry *child_die = die->GetFirstChild(); child_die != NULL; child_die = child_die->GetSibling())
{
const DWARFDebugInfoEntry *result_die = FindBlockContainingSpecification (dwarf_cu,
child_die,
spec_block_die_offset,
result_die_cu_handle);
if (result_die)
return result_die;
}
}
*result_die_cu_handle = NULL;
return NULL;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
size_t
SymbolFileDWARF::ParseVariables
(
const SymbolContext& sc,
Added the ability to create an objective C method for an objective C interface in ClangASTContext. Also added two bool returning functions that indicated if an opaque clang qual type is a CXX class type, and if it is an ObjC class type. Objective C classes now will get their methods added lazily as they are encountered. The reason for this is currently, unlike C++, the DW_TAG_structure_type and owns the ivars, doesn't not also contain the member functions. This means when we parse the objective C class interface we either need to find all functions whose names start with "+[CLASS_NAME" or "-[CLASS_NAME" and add them all to the class, or when we parse each objective C function, we slowly add it to the class interface definition. Since objective C's class doesn't change internal bits according to whether it has certain types of member functions (like C++ does if it has virtual functions, or if it has user ctors/dtors), I currently chose to lazily populate the class when each functions is parsed. Another issue we run into with ObjC method declarations is the "self" and "_cmd" implicit args are not marked as artificial in the DWARF (DW_AT_artifical), so we currently have to look for the parameters by name if we are trying to omit artificial function args if the language of the compile unit is ObjC or ObjC++. llvm-svn: 114722
2010-09-24 05:15:53 +00:00
DWARFCompileUnit* dwarf_cu,
Looking at some of the test suite failures in DWARF in .o files with the debug map showed that the location lists in the .o files needed some refactoring in order to work. The case that was failing was where a function that was in the "__TEXT.__textcoal_nt" in the .o file, and in the "__TEXT.__text" section in the main executable. This made symbol lookup fail due to the way we were finding a real address in the debug map which was by finding the section that the function was in in the .o file and trying to find this in the main executable. Now the section list supports finding a linked address in a section or any child sections. After fixing this, we ran into issue that were due to DWARF and how it represents locations lists. DWARF makes a list of address ranges and expressions that go along with those address ranges. The location addresses are expressed in terms of a compile unit address + offset. This works fine as long as nothing moves around. When stuff moves around and offsets change between the remapped compile unit base address and the new function address, then we can run into trouble. To deal with this, we now store supply a location list slide amount to any location list expressions that will allow us to make the location list addresses into zero based offsets from the object that owns the location list (always a function in our case). With these fixes we can now re-link random address ranges inside the debugger for use with our DWARF + debug map, incremental linking, and more. Another issue that arose when doing the DWARF in the .o files was that GCC 4.2 emits a ".debug_aranges" that only mentions functions that are externally visible. This makes .debug_aranges useless to us and we now generate a real address range lookup table in the DWARF parser at the same time as we index the name tables (that are needed because .debug_pubnames is just as useless). llvm-gcc doesn't generate a .debug_aranges section, though this could be fixed, we aren't going to rely upon it. Renamed a bunch of "UINT_MAX" to "UINT32_MAX". llvm-svn: 113829
2010-09-14 02:20:48 +00:00
const lldb::addr_t func_low_pc,
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *orig_die,
bool parse_siblings,
bool parse_children,
VariableList* cc_variable_list
)
{
if (orig_die == NULL)
return 0;
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
VariableListSP variable_list_sp;
size_t vars_added = 0;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
const DWARFDebugInfoEntry *die = orig_die;
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
while (die != NULL)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
dw_tag_t tag = die->Tag();
// Check to see if we have already parsed this variable or constant?
if (m_die_to_variable_sp[die])
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (cc_variable_list)
cc_variable_list->AddVariableIfUnique (m_die_to_variable_sp[die]);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
else
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
// We haven't already parsed it, lets do that now.
if ((tag == DW_TAG_variable) ||
(tag == DW_TAG_constant) ||
(tag == DW_TAG_formal_parameter && sc.function))
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (variable_list_sp.get() == NULL)
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
const DWARFDebugInfoEntry *sc_parent_die = GetParentSymbolContextDIE(orig_die);
dw_tag_t parent_tag = sc_parent_die ? sc_parent_die->Tag() : 0;
switch (parent_tag)
{
case DW_TAG_compile_unit:
if (sc.comp_unit != NULL)
{
variable_list_sp = sc.comp_unit->GetVariableList(false);
if (variable_list_sp.get() == NULL)
{
variable_list_sp.reset(new VariableList());
sc.comp_unit->SetVariableList(variable_list_sp);
}
}
else
{
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportError ("parent 0x%8.8llx %s with no valid compile unit in symbol context for 0x%8.8llx %s.\n",
MakeUserID(sc_parent_die->GetOffset()),
DW_TAG_value_to_name (parent_tag),
MakeUserID(orig_die->GetOffset()),
DW_TAG_value_to_name (orig_die->Tag()));
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
}
break;
case DW_TAG_subprogram:
case DW_TAG_inlined_subroutine:
case DW_TAG_lexical_block:
if (sc.function != NULL)
{
// Check to see if we already have parsed the variables for the given scope
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
Block *block = sc.function->GetBlock(true).FindBlockByID(MakeUserID(sc_parent_die->GetOffset()));
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (block == NULL)
{
// This must be a specification or abstract origin with
// a concrete block couterpart in the current function. We need
// to find the concrete block so we can correctly add the
// variable to it
DWARFCompileUnit *concrete_block_die_cu = dwarf_cu;
const DWARFDebugInfoEntry *concrete_block_die = FindBlockContainingSpecification (sc.function->GetID(),
sc_parent_die->GetOffset(),
&concrete_block_die_cu);
if (concrete_block_die)
Moved lldb::user_id_t values to be 64 bit. This was going to be needed for process IDs, and thread IDs, but was mainly needed for for the UserID's for Types so that DWARF with debug map can work flawlessly. With DWARF in .o files the type ID was the DIE offset in the DWARF for the .o file which is not unique across all .o files, so now the SymbolFileDWARFDebugMap class will make the .o file index part (the high 32 bits) of the unique type identifier so it can uniquely identify the types. llvm-svn: 142534
2011-10-19 18:09:39 +00:00
block = sc.function->GetBlock(true).FindBlockByID(MakeUserID(concrete_block_die->GetOffset()));
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
}
if (block != NULL)
{
const bool can_create = false;
variable_list_sp = block->GetBlockVariableList (can_create);
if (variable_list_sp.get() == NULL)
{
variable_list_sp.reset(new VariableList());
block->SetVariableList(variable_list_sp);
}
}
}
break;
default:
Added code in the Host layer that can report system log messages so that we don't have "fprintf (stderr, ...)" calls sprinkled everywhere. Changed all needed locations over to using this. For non-darwin, we log to stderr only. On darwin, we log to stderr _and_ to ASL (Apple System Log facility). This will allow GUI apps to have a place for these error and warning messages to go, and also allows the command line apps to log directly to the terminal. llvm-svn: 147596
2012-01-05 03:57:59 +00:00
GetObjectFile()->GetModule()->ReportError ("didn't find appropriate parent DIE for variable list for 0x%8.8llx %s.\n",
MakeUserID(orig_die->GetOffset()),
DW_TAG_value_to_name (orig_die->Tag()));
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
break;
}
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
}
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (variable_list_sp)
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
{
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
VariableSP var_sp (ParseVariableDIE(sc, dwarf_cu, die, func_low_pc));
if (var_sp)
{
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
variable_list_sp->AddVariableIfUnique (var_sp);
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
if (cc_variable_list)
cc_variable_list->AddVariableIfUnique (var_sp);
++vars_added;
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
}
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
}
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
bool skip_children = (sc.function == NULL && tag == DW_TAG_subprogram);
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (!skip_children && parse_children && die->HasChildren())
{
vars_added += ParseVariables(sc, dwarf_cu, func_low_pc, die->GetFirstChild(), true, true, cc_variable_list);
Improved the packet throughput when debugging with GDB remote by over 3x on darwin (not sure about other platforms). Modified the communication and connection classes to not require the BytesAvailable function. Now the "Read(...)" function has a timeout in microseconds. Fixed a lot of assertions that were firing off in certain cases and replaced them with error output and code that can deal with the assertion case. llvm-svn: 133224
2011-06-17 01:22:15 +00:00
}
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
if (parse_siblings)
die = die->GetSibling();
else
die = NULL;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
Fixed variable parsing to not parse block variables over and over due to an issue in the way block variables are marked as parsed. In the DWARF parser we always parse all blocks for a function at once, so we can mark all blocks as having all variables parsed and avoid recursive function calls to try and reparse things that have already been handled. Fixed an issue with how variables get scoped into blocks. The DWARF parser can now handle abtract class definitions that contain concrete static variables. When the concrete instance of the class functions get instantiated, they will track down the concrete block for the abtract block and add the variable to each block. llvm-svn: 133302
2011-06-17 22:10:16 +00:00
return vars_added;
Initial checkin of lldb code from internal Apple repo. llvm-svn: 105619
2010-06-08 16:52:24 +00:00
}
//------------------------------------------------------------------
// PluginInterface protocol
//------------------------------------------------------------------
const char *
SymbolFileDWARF::GetPluginName()
{
return "SymbolFileDWARF";
}
const char *
SymbolFileDWARF::GetShortPluginName()
{
return GetPluginNameStatic();
}
uint32_t
SymbolFileDWARF::GetPluginVersion()
{
return 1;
}
A few of the issue I have been trying to track down and fix have been due to the way LLDB lazily gets complete definitions for types within the debug info. When we run across a class/struct/union definition in the DWARF, we will only parse the full definition if we need to. This works fine for top level types that are assigned directly to variables and arguments, but when we have a variable with a class, lets say "A" for this example, that has a member: "B *m_b". Initially we don't need to hunt down a definition for this class unless we are ever asked to do something with it ("expr m_b->getDecl()" for example). With my previous approach to lazy type completion, we would be able to take a "A *a" and get a complete type for it, but we wouldn't be able to then do an "a->m_b->getDecl()" unless we always expanded all types within a class prior to handing out the type. Expanding everything is very costly and it would be great if there were a better way. A few months ago I worked with the llvm/clang folks to have the ExternalASTSource class be able to complete classes if there weren't completed yet: class ExternalASTSource { .... virtual void CompleteType (clang::TagDecl *Tag); virtual void CompleteType (clang::ObjCInterfaceDecl *Class); }; This was great, because we can now have the class that is producing the AST (SymbolFileDWARF and SymbolFileDWARFDebugMap) sign up as external AST sources and the object that creates the forward declaration types can now also complete them anywhere within the clang type system. This patch makes a few major changes: - lldb_private::Module classes now own the AST context. Previously the TypeList objects did. - The DWARF parsers now sign up as an external AST sources so they can complete types. - All of the pure clang type system wrapper code we have in LLDB (ClangASTContext, ClangASTType, and more) can now be iterating through children of any type, and if a class/union/struct type (clang::RecordType or ObjC interface) is found that is incomplete, we can ask the AST to get the definition. - The SymbolFileDWARFDebugMap class now will create and use a single AST that all child SymbolFileDWARF classes will share (much like what happens when we have a complete linked DWARF for an executable). We will need to modify some of the ClangUserExpression code to take more advantage of this completion ability in the near future. Meanwhile we should be better off now that we can be accessing any children of variables through pointers and always be able to resolve the clang type if needed. llvm-svn: 123613
2011-01-17 03:46:26 +00:00
void
SymbolFileDWARF::CompleteTagDecl (void *baton, clang::TagDecl *decl)
{
SymbolFileDWARF *symbol_file_dwarf = (SymbolFileDWARF *)baton;
clang_type_t clang_type = symbol_file_dwarf->GetClangASTContext().GetTypeForDecl (decl);
if (clang_type)
symbol_file_dwarf->ResolveClangOpaqueTypeDefinition (clang_type);
}
void
SymbolFileDWARF::CompleteObjCInterfaceDecl (void *baton, clang::ObjCInterfaceDecl *decl)
{
SymbolFileDWARF *symbol_file_dwarf = (SymbolFileDWARF *)baton;
clang_type_t clang_type = symbol_file_dwarf->GetClangASTContext().GetTypeForDecl (decl);
if (clang_type)
symbol_file_dwarf->ResolveClangOpaqueTypeDefinition (clang_type);
}
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
void
This change brings in the latest LLVM/Clang, and completes the support in the LLDB expression parser for incomplete types. Clang now imports types lazily, and we complete those types as necessary. Changes include: - ClangASTSource now supports three APIs which it passes to ClangExpressionDeclMap. CompleteType completes a TagDecl or an ObjCInterfaceDecl when needed; FindExternalVisibleDecls finds named entities that are visible in the expression's scope; and FindExternalLexicalDecls performs a (potentially restricted) search for entities inside a lexical scope like a namespace. These changes mean that entities in namespaces should work normally. - The SymbolFileDWARF code for searching a context for a specific name is now more general, and can search arbitrary contexts. - We are continuing to adapt our calls into LLVM from interfaces that take start and end iterators when accepting multiple items to interfaces that use ArrayRef. - I have cleaned up some code, especially our use of namespaces. This change is neutral for our testsuite and greatly improves correctness for large programs (like Clang) with complicated type systems. It should also lay the groundwork for improving the expression parser's performance as we are lazier and lazier about providing type information. llvm-svn: 136555
2011-07-30 02:42:06 +00:00
SymbolFileDWARF::DumpIndexes ()
{
StreamFile s(stdout, false);
s.Printf ("DWARF index for (%s) '%s/%s':",
GetObjectFile()->GetModule()->GetArchitecture().GetArchitectureName(),
GetObjectFile()->GetFileSpec().GetDirectory().AsCString(),
GetObjectFile()->GetFileSpec().GetFilename().AsCString());
s.Printf("\nFunction basenames:\n"); m_function_basename_index.Dump (&s);
s.Printf("\nFunction fullnames:\n"); m_function_fullname_index.Dump (&s);
s.Printf("\nFunction methods:\n"); m_function_method_index.Dump (&s);
s.Printf("\nFunction selectors:\n"); m_function_selector_index.Dump (&s);
s.Printf("\nObjective C class selectors:\n"); m_objc_class_selectors_index.Dump (&s);
s.Printf("\nGlobals and statics:\n"); m_global_index.Dump (&s);
s.Printf("\nTypes:\n"); m_type_index.Dump (&s);
s.Printf("\nNamepaces:\n"); m_namespace_index.Dump (&s);
}
void
SymbolFileDWARF::SearchDeclContext (const clang::DeclContext *decl_context,
const char *name,
llvm::SmallVectorImpl <clang::NamedDecl *> *results)
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
{
This change brings in the latest LLVM/Clang, and completes the support in the LLDB expression parser for incomplete types. Clang now imports types lazily, and we complete those types as necessary. Changes include: - ClangASTSource now supports three APIs which it passes to ClangExpressionDeclMap. CompleteType completes a TagDecl or an ObjCInterfaceDecl when needed; FindExternalVisibleDecls finds named entities that are visible in the expression's scope; and FindExternalLexicalDecls performs a (potentially restricted) search for entities inside a lexical scope like a namespace. These changes mean that entities in namespaces should work normally. - The SymbolFileDWARF code for searching a context for a specific name is now more general, and can search arbitrary contexts. - We are continuing to adapt our calls into LLVM from interfaces that take start and end iterators when accepting multiple items to interfaces that use ArrayRef. - I have cleaned up some code, especially our use of namespaces. This change is neutral for our testsuite and greatly improves correctness for large programs (like Clang) with complicated type systems. It should also lay the groundwork for improving the expression parser's performance as we are lazier and lazier about providing type information. llvm-svn: 136555
2011-07-30 02:42:06 +00:00
DeclContextToDIEMap::iterator iter = m_decl_ctx_to_die.find(decl_context);
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
if (iter == m_decl_ctx_to_die.end())
return;
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
for (DIEPointerSet::iterator pos = iter->second.begin(), end = iter->second.end(); pos != end; ++pos)
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
{
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
const DWARFDebugInfoEntry *context_die = *pos;
if (!results)
return;
DWARFDebugInfo* info = DebugInfo();
DIEArray die_offsets;
DWARFCompileUnit* dwarf_cu = NULL;
const DWARFDebugInfoEntry* die = NULL;
Tested a theory on the where when we lookup things in the accelerator tables that if we prefer the current compile unit, followed by any compile units that already had their DIEs parsed, followed by the rest of the matches, that we might save some memory. This turned out not to help much. The code is commented out, but I want to check it in so I don't lose the code in case it could help later. Added the ability to efficiently find the objective C class implementation when using the new .apple_types acclerator tables with the type flags. If the type flags are not available, we default back to what we were doing before. llvm-svn: 146250
2011-12-09 08:48:30 +00:00
if (m_using_apple_tables)
{
if (m_apple_types_ap.get())
m_apple_types_ap->FindByName (name, die_offsets);
}
else
{
if (!m_indexed)
Index ();
m_type_index.Find (ConstString(name), die_offsets);
}
const size_t num_matches = die_offsets.size();
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
if (num_matches)
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
{
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
for (size_t i = 0; i < num_matches; ++i)
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
{
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
const dw_offset_t die_offset = die_offsets[i];
die = info->GetDIEPtrWithCompileUnitHint (die_offset, &dwarf_cu);
if (die->GetParent() != context_die)
continue;
Type *matching_type = ResolveType (dwarf_cu, die);
Use forward types where possible to avoid having to parse extra DWARF when it is not required. llvm-svn: 146418
2011-12-12 21:50:19 +00:00
lldb::clang_type_t type = matching_type->GetClangForwardType();
Modified the DWARF parser to take care of a 1 clang::DeclContext to many DIEs. Prior to this fix, there was a 1 to 1 mapping. llvm-svn: 141917
2011-10-13 23:49:28 +00:00
clang::QualType qual_type = clang::QualType::getFromOpaquePtr(type);
if (const clang::TagType *tag_type = llvm::dyn_cast<clang::TagType>(qual_type.getTypePtr()))
{
clang::TagDecl *tag_decl = tag_type->getDecl();
results->push_back(tag_decl);
}
else if (const clang::TypedefType *typedef_type = llvm::dyn_cast<clang::TypedefType>(qual_type.getTypePtr()))
{
clang::TypedefNameDecl *typedef_decl = typedef_type->getDecl();
results->push_back(typedef_decl);
}
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
}
}
}
}
void
SymbolFileDWARF::FindExternalVisibleDeclsByName (void *baton,
Changed lldb_private::Type over to use the intrusive ref counted pointers so we don't have to lookup types in a type list by ID. Changed the DWARF parser to remove the "can externally complete myself" bits from the type when we are in the process of completing the type itself to avoid an onslaught of external visible decl requests from the clang::ExternalASTSource. llvm-svn: 142461
2011-10-18 23:36:41 +00:00
const clang::DeclContext *decl_context,
clang::DeclarationName decl_name,
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
llvm::SmallVectorImpl <clang::NamedDecl *> *results)
{
Changed lldb_private::Type over to use the intrusive ref counted pointers so we don't have to lookup types in a type list by ID. Changed the DWARF parser to remove the "can externally complete myself" bits from the type when we are in the process of completing the type itself to avoid an onslaught of external visible decl requests from the clang::ExternalASTSource. llvm-svn: 142461
2011-10-18 23:36:41 +00:00
switch (decl_context->getDeclKind())
{
case clang::Decl::Namespace:
case clang::Decl::TranslationUnit:
{
SymbolFileDWARF *symbol_file_dwarf = (SymbolFileDWARF *)baton;
symbol_file_dwarf->SearchDeclContext (decl_context, decl_name.getAsString().c_str(), results);
}
break;
default:
break;
}
This commit adds broad architectural support for hierarchical inspection of namespaces in the expression parser. ClangExpressionDeclMap hitherto reported that namespaces had been completely imported, even though the namespaces are returned empty. To deal with this situation, ClangASTSource was recently extended with an API to complete incomplete type definitions, and, for greater efficiency, to complete these definitions partially, returning only those objects that have a given name. This commit supports these APIs on LLDB's side, and uses it to provide information on types resident in namespaces. Namespaces are now imported as they were -- that is to say, empty -- but with minimal import mode on. This means that Clang will come back and request their contents by name as needed. We now respond with information on the contained types; this will be followed soon by information on functions and variables. llvm-svn: 133852
2011-06-25 00:44:06 +00:00
}
Adding the DWARF parser side for assited layout where the AST context will ask ExternalASTSource objects to help laying out a type. This is needed because the DWARF typically doesn't contain alignement or packing attribute values, and we need to be able to match up types that the compiler uses in expressions. llvm-svn: 149160
2012-01-28 00:48:57 +00:00
bool
SymbolFileDWARF::LayoutRecordType (void *baton,
const clang::RecordDecl *record_decl,
uint64_t &size,
uint64_t &alignment,
llvm::DenseMap <const clang::FieldDecl *, uint64_t> &field_offsets,
llvm::DenseMap <const clang::CXXRecordDecl *, clang::CharUnits> &base_offsets,
llvm::DenseMap <const clang::CXXRecordDecl *, clang::CharUnits> &vbase_offsets)
{
SymbolFileDWARF *symbol_file_dwarf = (SymbolFileDWARF *)baton;
return symbol_file_dwarf->LayoutRecordType (record_decl, size, alignment, field_offsets, base_offsets, vbase_offsets);
}
bool
SymbolFileDWARF::LayoutRecordType (const clang::RecordDecl *record_decl,
uint64_t &bit_size,
uint64_t &alignment,
llvm::DenseMap <const clang::FieldDecl *, uint64_t> &field_offsets,
llvm::DenseMap <const clang::CXXRecordDecl *, clang::CharUnits> &base_offsets,
llvm::DenseMap <const clang::CXXRecordDecl *, clang::CharUnits> &vbase_offsets)
{
LogSP log (LogChannelDWARF::GetLogIfAll(DWARF_LOG_DEBUG_INFO));
RecordDeclToLayoutMap::iterator pos = m_record_decl_to_layout_map.find (record_decl);
bool success = false;
base_offsets.clear();
vbase_offsets.clear();
if (pos != m_record_decl_to_layout_map.end())
{
bit_size = pos->second.bit_size;
alignment = pos->second.alignment;
field_offsets.swap(pos->second.field_offsets);
m_record_decl_to_layout_map.erase(pos);
success = true;
}
else
{
bit_size = 0;
alignment = 0;
field_offsets.clear();
}
if (log)
GetObjectFile()->GetModule()->LogMessage (log.get(),
"SymbolFileDWARF::LayoutRecordType (record_decl = %p, bit_size = %llu, alignment = %llu, field_offsets[%u],base_offsets[%u], vbase_offsets[%u]) success = %i",
record_decl,
bit_size,
alignment,
(uint32_t)field_offsets.size(),
(uint32_t)base_offsets.size(),
(uint32_t)vbase_offsets.size(),
success);
return success;
}
Reference in New Issue Copy Permalink