Add function to Build class to get targets of type BuildTarget
Update xcode backend to call get_build_targets when iterating over targets.
This resolves crash in xcode backend when using custom targets:
AttributeError: ‘CustomTarget’ object has no attribute ‘objects’
On Windows this would fail because of missing DLL:
```
mylib = library(...)
exe = executable(..., link_with: mylib)
meson.add_install_script(exe)
```
The reason is on Windows we cannot rely on rpath to find libraries from
build directory, they are searched in $PATH. We already have all that
mechanism in place for custom_target() using ExecutableSerialisation
class, so reuse it for install/dist/postconf scripts too.
This has bonus side effect to also use exe_wrapper for those scripts.
Fixes: #8187
There are still caveats here. Rust/cargo handles generated sources by
writing out all targets of a single repo into a single output directory,
setting a path to that via a build-time environment variable, and then
include those files via a set of functions and macros. Meson's build
layout is naturally different, and ninja makes working with environment
variables at compile time difficult.
Fixes#8157
we have two functions to do the exact same thing, and they're basically
implemented the same way. Instead, let's just use the BuildTarget one,
as it's more generally available.
Currently InstallDir is part of the interpreter, and is an Interpreter
object, which is then put in the Build object. This is a layering
violation, the interperter should have a Holder for build data. This
patch fixes that.
This patches takes the options work to it's logical conclusion: A single
flat dictionary of OptionKey: UserOptions. This allows us to simplify a
large number of cases, as we don't need to check if an option is in this
dict or that one (or any of 5 or 6, actually).
I would have prefered to do these seperatately, but they are combined in
some cases, so it was much easier to convert them together.
this eliminates the builtins_per_machine dict, as it's duplicated with
the OptionKey's machine parameter.
This was all layering violations before. Now we have Headers in the
build module, and a holder in the interpreter. All of the type
validation is done in interpreter method for `install_headers`.
So that editors that can fold code (vim, vscode, etc) can correctly fold
functions, instead of getting confused by code that doesn't follow the
current indention. Also, it makes the code easier to read.
The deprecation message for "gui_app" is appearing for every target
rather than just once, and even if the required version is older
than 0.56.0. Use @FeatureDeprecatedKwargs to fix both issues.
If static_library is used as a convenience library (e.g. for link_whole)
it should in principle not need position independent code.
However, if the executables that the libraries is linked to are PIE,
the non-PIC objects in the static library will cause linker errors.
To avoid this, obey b_pie for static libraries if either b_staticpic=false
or they use "pic: false".
Without this patch, QEMU cannot use b_staticpic, which causes a slowdown
on some QEMU benchmarks up to 20%.
extract_objects is repeatedly looking up files in self.sources, which is a list.
Convert it to a set beforehand so that the lookup is O(1).
On a QEMU build, the time spent in extract_objects goes from 3.292s to 0.431s.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This gives the version that the feature was deprecated in, and doesn't
print the warning if the project supports versions of meson in which the
project wasn't deprecated.
in tree like dep structures with a lot of source: declarations, this can
result in a lot of presure on the source list. this saves ~3s out of 7s
in the interpretor stage in efl build.
Generally, we'd want to use str() rather than repr() in error messages
anyhow, as that explicitly gives something designed to be read by
humans.
Sometimes {!r} is being used as a shortcut to avoid writing the quotes
in '{!s}'.
Unfortunately, these things aren't quite the same, as the repr of a
string containing '\' (the path separator on Windows) will have those
escaped.
We don't have a good string representation to use for the arbitrary
internal object used as an argument for install_data() when it's neither
a string nor file (which doesn't lead to a good error message), so drop
that for the moment.
If the feature hadn't been broken in the first place it would have
worked on them anyway, so we might as well expose it. I'm loathe to do
it because one of the best features of meson in a mixed C/C++ code base
is that meson figures out the right linker every time, but there are
cases people have where they want to force a linker. We'll let them keep
the pieces.
Currently it does nothing, as the field is read too late, and additional
languages have already been considered. As such if the language
requested is closer to C (for example you want C but have a C++ source
with only extern C functions) then link_langauge is ignored.
Fixes#6453
A current rather untyped storage of options is one of the things that
contributes to the options code being so complex. This takes a small
step in synching down by storing the compiler options in dicts per
language.
Future work might be replacing the langauge strings with an enum, and
defaultdict with a custom struct, just like `PerMachine` and
`MachineChoice`.
Adjust the handling of a name_prefix: [] kwarg to be the same as
name_suffix: [] kwarg, i.e. identically to the case where it's omitted,
so BuildTarget.prefix doesn't get set (so the default is used).
Also clarify the error reported when a non-empty list is used.
Similar to meson.override_find_program() but overrides the result of the
dependency() function.
Also ensure that dependency() always returns the same result when
looking for the same dependency, this fixes cases where parts of the
project could be using a system library and other parts use the library
provided by a subproject.
This makes the typing annotations basically impossible to get right, but
if we only have one key then it's easy. Fortunately python provides
comprehensions, so we don't even need the ability to pass multiple keys,
we can just [extract_as_list(kwargs, c) for c in ('a', 'b', 'c')] and
get the same result.
listify shouldn't be unholdering, it's a function to turn scalar values
into lists, or flatten lists. Having a separate function is clearer,
easier to understand, and can be run recursively if necessary.
This gives consistent reporting of this error for all platforms.
Also, reporting this error when constructing the BuildTarget, rather
than discovering the problem during backend generation means that the
error is reported against with a location.
declare_dependencies
This allows dependencies declared in subprojects to set variables, and
for those variables to be accessed via the get_variable method, just
like those from pkg-config and cmake. This makes it easier to use
projects from subprojects in a polymorphic manner, lowering the
distinction between a subproject and an external dependency every
further.
There are three problems:
1) Dunders like `__lt__` and `__gt__` don't return bool, they return
either a bool or the NotImplemented singleton to signal that they don't
know how to be compared.
2) The don't take type object, the take `typing.Any`
3) They need to return NotImplemented if the comparison is not
implemented, this allows python to try the inverse dunder from the
other object. If that object returns NotImplemented as well a
TypeError is raised.
t.pic won't be defined. We can only hope it has been built with -fPIC.
Linker will complain otherwise any way.
t.extract_all_objects_recurse() won't be defined. We could support this
case by extracting the archive somewhere and pick object files.
Some things, like `method[...](...)` or `x: ... = ...` python 3.5
doesn't support, so I made a comment instead with the intention that it
can someday be made into a real annotation.
In most cases instead pass `for_machine`, the name of the relevant
machines (what compilers target, what targets run on, etc). This allows
us to use the cross code path in the native case, deduplicating the
code.
As one can see, environment got bigger as more information is kept
structured there, while ninjabackend got a smaller. Overall a few amount
of lines were added, but the hope is what's added is a lot simpler than
what's removed.
This function is currently setup with keyword arguments defaulting to
None. However, it is never called without passing all of it's arguments
explicitly, and only one of it's arguments would actually be valid as
None. So just drop that, and make them all positional. And annotate
them.
Some things, like `method[...](...)` or `x: ... = ...` python 3.5
doesn't support, so I made a comment instead with the intention that it
can someday be made into a real annotation.
This isn't safe given the way python implements default arguments.
Basically python store a reference to the instance it was passed, and
then if that argument is not provided it uses the default. That means
that two calls to the same function get the same instance, if one of
them mutates that instance every subsequent call that gets the default
will receive the mutated instance. The idiom to this in python is to use
None and replace the None,
def in(value: str, container: Optional[List[str]]) -> boolean:
return src in (container or [])
if there is no chance of mutation it's less code to use or and take
advantage of None being falsy. If you may want to mutate the value
passed in you need a ternary (this example is stupid):
def add(value: str, container: Optional[List[str]]) -> None:
container = container if container is not None else []
container.append(value)
I've used or everywhere I'm sure that the value will not be mutated by
the function and erred toward caution by using ternaries for the rest.
If find_program() returns a file from the source directory, anything
that uses it should add the file to the dependencies, so that they are
rebuilt whenever the script changes. Generator is not doing that.
While at it, I am doing two related fixes:
- Generator is not checking whther the generator actually was found,
resulting in a Python error involving NoneType if it isn't. To minimize
backwards compatibility issues, I am only raising the error when
g.process() is acutally called.
- the error message for custom_target with a nonexisting program
erroneously mention a not-found external program "nonexistingprogram".
The new error is similar to the one I am adding for generators.
Warn when someone tries to use append() or prepend() on an env var
which already has an operation set on it. People seem to think that
multiple append/prepend operations stack, but they don't.
Closes https://github.com/mesonbuild/meson/issues/5087
I'll be using this later, but it seems useful to allow dependencies to
that have special handlers to declare that they depend on other
dependencies. This should allow us to stop treating threads special
internally and just make it a normal dependency.
Since the "-l<lib>" flags in the build.ninja file are passed in
"--start-group"/"--end-group" flags, there should be no need to have any
library listed twice, even if there are circular dependencies. Therefore we
can eliminate duplicates. For speed, rather than deduplicating at the end
of the process, it's faster to not add the duplicate flags in the first
place.
This should help fix#2150
A custom_target, if install is set to true, will always be built by
default even if build_by_default is explicitly set to false.
Ensure that this does not happen if it's set explicitly. To keep
backward compatibility, if build_by_default is not set explicitly and
install is true, set build_by_default to true.
Fixes#4107
Previously, the configuration worked fine, but the compiler raised an
error. Now, we explicitly check for the existence of files and print a
useful error message if they do not exist.
Fixed-size hash makes paths shorter and prevents doubling of path length
because of subdir usage in target id: "subdir/id" would generate
"subdir/{subdir-without-slashes}@@id" target otherwise.
Export construct_id_from_path() to aid tests.
Add a separate unit test for this function to make sure it is not broken unexpectedly.
Closes#4226.
It's a (presumably unintentional) quirk of the current implementation of
SharedLibrary.determine_filenames() that if both name_suffix and
name_prefix are set, an import library isn't generated.
Adjust test 'common/25 library versions': Make the library have exports,
so an implib is generated with MSVC. Add implib to set of files expected
to be installed
Adjust test 'common/122 shared module': Add libnosyms implib to set of
files expected to be installed, except for MSVC, where none is generated
as it has no exports
Use the specified name_prefix for implib, rather than hardcoding it.
(This is needed to allow an installed library given name_prefix:'' and a
name starting with 'lib' to be linked with using -l. (This case is
handled specially in the pkgconfig module))
Handle clang's cl or clang-cl being in PATH, or set in CC/CXX
Future work: checking the name of the executable here seems like a bad idea.
These compilers will fail to be detected if they are renamed.
v2:
Update compiler.get_argument_type() test
Fix comparisons of id inside CCompiler, backends and elsewhere
v3:
ClangClCPPCompiler should be a subclass of ClangClCCompier, as well
Future work: mocking in test_find_library_patterns() is effected, as we
now test for a subclass, rather than self.id in CCompiler.get_library_naming()
If a subproject is not required and fails during its configuration, the
parent project continues, but should not include any target or state set
by the failed subproject. This fix ninja still trying to build targets
generated by subprojects before they fail in their configuration.
The 'build' object is now per-interpreter instead of being global. Once
a subproject interpreter succeed, values from its 'build' object are
merged back into its parent 'build' object.
Before, the mappings has been created over all the links, while it
actaully only used the Shared or Static Targets. This structure now is
tree like structured and cached, thus the results can be computed a lot
faster.
The generator step generate_install is now for EFL from 6 sec. down to
0.3s. Which improves the overall build time from ~20 sec. to ~14 sec.
there is a huge amount of isinstance calls, this reduces the amount of
these calls while splitting up a rather big function. It also assosiates
every target type with theire default install directory.
The problem with the earlier position of the generation code was, that
the results could not be cached, because the list of all link_deps was
overall different. However, it shared a special kind of subsets with
other build build targets.
Generating the set of subdirs that are required for linking, alongside
with the link dependencies brings the possibility of caching this.
This reduces the buildting from 1 min. in efl down to 20 sec. And
reduces the amount of 30872534 calls down.
this saves ~40 sec.
With this it is now possible to do
foobar = executable('foobar', ...)
meson.override_find_program('foobar', foobar)
Which is convenient for a project like protobuf which produces both a
dependency and a tool. If protobuf is updated to use
override_find_program, it can be used as
protobuf_dep = dependency('protobuf', version : '>=3.3.1',
fallback : ['protobuf', 'protobuf_dep'])
protoc_prog = find_program('protoc')
We now use the soversion to set compatibility_version and
current_version by default. This is the only sane thing we can do by
default because of the restrictions on the values that can be used for
compatibility and current version.
Users can override this value with the `darwin_versions:` kwarg, which
can be a single value or a two-element list of values. The first one
is the compatibility version and the second is the current version.
Fixes https://github.com/mesonbuild/meson/issues/3555
Fixes https://github.com/mesonbuild/meson/issues/1451
Ninja buffers all commands and prints them only after they are
complete. Because of this, long-running commands such as `cargo
build` show no output at all and it's impossible to know if the
command is merely taking too long or is stuck somewhere.
To cater to such use-cases, Ninja has a 'pool' with depth 1 called
'console', and all processes in this pool have the following
properties:
1. stdout is connected to the program, so output can be seen in
real-time
2. The output of all other commands is buffered and displayed after
a command in this pool finishes running
3. Commands in this pool are executed serially (normal commands
continue to run in the background)
This feature is available since Ninja v1.5
https://ninja-build.org/manual.html#_the_literal_console_literal_pool
We now pass the current subproject to every FeatureNew and
FeatureDeprecated call. This requires a bunch of rework to:
1. Ensure that we have access to the subproject in the list of
arguments when used as a decorator (see _get_callee_args).
2. Pass the subproject to .use() when it's called manually.
3. We also can't do feature checks for new features in
meson_options.txt because that's parsed before we know the
meson_version from project()
* Use _get_callee_args to unwrap function call arguments, needed for
module functions.
* Move some FeatureNewKwargs from build.py to interpreter.py
* Print a summary for featurenew only if conflicts were found. The
summary now only prints conflicting features.
* Report and store featurenew/featuredeprecated only once
* Fix version comparison: use le/ge and resize arrays to not fail on
'0.47.0>=0.47'
Closes https://github.com/mesonbuild/meson/issues/3660
If the external program is a string that is meant to be searched in
PATH, we can't add a dependency on it at configure time because we don't
know where it will be at compile time.
D is not a 'c-like' language, but it can link to C libraries. The same
might be true of Rust in the future and Go when we add support for it.
This contains no functionality changes.
Since `build_always` also adds a target to the set of default targets,
this option is marked deprecated in favour of the new option
`build_always_stale`.
`build_always_stale` *only* marks the target to be always considered out
of date, but does *not* add it to the set of default targets.
The old behaviour can still be achieved by combining
`build_always_stale` with `build_by_default`.
fixes#1942
On Windows, if we are going to link with a shared module, we need the
implib.
Use case: The Xorg server builds some X protocol extensions as modules. The
implibs for these modules need to be shipped as part of the SDK, to enable
building of 3rd party extensions which reference symbols in (and hence on
Windows, need to be linked with) these modules.
Refine #3277
According to what I read on the internet, on OSX, both MH_BUNDLE (module)
and MH_DYLIB (shared library) can be dynamically loaded using dlopen(), but
it is not possible to link against MH_BUNDLE as if they were shared
libraries.
Metion this as an issue in the documentation.
Emitting a warning, and then going on to fail during the build with
mysterious errors in symbolextractor isn't very helpful, so make attempting
this an error on OSX.
Add a test for that.
See also:
https://docstore.mik.ua/orelly/unix3/mac/ch05_03.htmhttps://stackoverflow.com/questions/2339679/what-are-the-differences-between-so-and-dylib-on-osx
This makes it possible to customize permissions of all installable
targets, such as executable(), libraries, man pages, header files and
custom or generated targets.
This is useful, for instance, to install setuid/setgid binaries, which
was hard to accomplish without access to this attribute.
Libraries that have been linked with link_whole: are internal
implementation details and should never be exposed to the outside
world in either Libs: or Libs.private:
Closes https://github.com/mesonbuild/meson/issues/3509
To maintain backward compatibility we cannot add recursive objects by
default. Print a warning when there are recursive objects to be pulled
and the argument is not set. After a while we'll do pull recursive
objects by default.
- determine_ext_objs: What matters is if extobj.target is a unity build,
not if the target using those objects is a unity build.
- determine_ext_objs: Return one object file per compiler, taking into
account generated sources.
- object_filename_from_source: No need to special-case unity build, it
does the same thing in both code paths.
- check_unity_compatible: For each compiler we must extract either none
or all its sources, taking into account generated sources.
when flattening the chained dependencies of an object, we don't need to
create any new internal dependencies if all the fields to be added to it
are empty.
For projects with a lot of libraries and dependency objects this can lead
to noticeable performance improvements.
fixup
When getting dependencies, we don't need to get the same dependencies and
dependency chains multiple times. If library a depends on x, y and z, and
library b depends on a, then we should not have to iterate through x, y and
z multiple times. Pruning at the stage of scanning the dependencies leads
to significant time savings when running meson
Jeff Moguillansky