`configure_file` is both an extremely complicated implementation, and
a strange place for copying. It's a bit of a historical artifact, since
the fs module didn't yet exist. It makes more sense to move this to the
fs module and deprecate this `configure_file` version.
This new version works at build time rather than configure time, which
has the disadvantage it can't be passed to `run_command`, but with the
advantage that changes to the input don't require a full reconfigure.
This is ambiguous, if the build directory has the same name as a
subcommand then we end up running the subcommand. It also means we have
a hard time adding *new* subcommands, because if it is a popular name of
a build directory then suddenly scripts that try to set up a build
directory end up running a subcommand instead.
The fact that we support this at all is a legacy design. Back in the
day, the "meson" program was for setting up a build directory and all
other tools were their own entry points, e.g. `mesontest` or
`mesonconf`. Then in commit fa278f351f we
migrated to the subcommand mechanism. So, for backwards compatibility,
we made those tools print a warning and then invoke `meson <tool>`. We
also made the `meson` tool default to setup.
However, we only warned for the other tools whose entry points were
eventually deleted. We never warned for setup itself, we just continued
to silently default to setup if no tool was provided.
`meson setup` has worked since 0.42, which is 5 years old this week.
It's available essentially everywhere. No one needs to use the old
backwards-compatible invocation method, but it continues to drag down
our ability to innovate. Let's finally do what we should have done a
long time ago, and sunset it.
custom_target allows selective installation if it outputs more than one
file. Mention this explicitly in install.
Additionally, fix the types for install_dir.
see: https://github.com/mesonbuild/meson/issues/505
It saves bandwidth and disk usage on downloaded subprojects, and people
very rarely need more than the single commit they’re using.
Add `depth=1` to the `[wrap-git]` examples in the rest of the
documentation, to make it more likely that people will copy-and-paste it
into their `.wrap` files.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
We want to talk about the kwargs to the custom_target() function, but
tried to link to custom_tgt instead, which is not a function. It is an
object, but this was the wrong reference method for a type.
We want the function anyway, so use that.
The end of the paragraph indicates that the options which support --foo
will be listed in the help text. The beginning of the paragraph seems to
suggest the same thing, except it doesn't distinguish between -Dfoo and
--foo style options.
The first mention is redundant and feels like the wrong part of the
paragraph to mention it anyway.
With the previous commit, we made this smartly detect when parentheses
are not needed. But the example was broken, because it doesn't follow
its own documented rules to use `[[#` syntax. Add the missing hash
character.
If we link to
```meson
[[#function]]('posarg')
```
then the ideal linkification would operate on "function" in the
formatted text:
```
function('posarg')
```
Instead, it operated on "function()" in the formatted text:
```
function()('posarg')
```
Fix this by detecting the next character after the reference, and
skipping the usual "automatically append the () for you" stage if it
already has that opening parenthesis.
The unit test was racy but surprisingly never failed on CI. The reason
is we need to ensure ninja build somelib.so before running `make` into
the external project.
Dlang uses both integer version "levels" and arbitrary string
identifiers, and we support both, but don't mention it in the docs.
Also update a test case to pass one via declare_dependency. We already
test this kwarg for build_target.
The type information is clearly wrong as it disagrees with the
description w.r.t. generated headers.
We also rely on it accepting custom targets for the obvious reason that
we accept it in a build target too! In fact, we rely on this in the
testsuite too.
Merging snippets happens in arbitrary order -- whatever filesystem
globbing results in. This didn't matter too much when we ran it once at
release time and checked the resulting release notes into git. However,
now that we generate a temporary version of the release notes for
development versions, the order of the results will periodically change.
Sort the files before processing them in order to guarantee that
whatever order they are in, they stay that way.
As a side effect, it's now technically possible to guarantee an ordering
by judicious use of snippets naming.
Leak sanitizer can be enabled without the whole AddressSanitizer, this
can be done by passing -fsanitize=leak as documented at [1].
Meson doesn't support this, so add support for it.
[1] https://clang.llvm.org/docs/LeakSanitizer.html
Logically, i18n.merge_file cannot ever take a MULTI_OUTPUT_KW, but it
does take a CT_OUTPUT_KW-like interface.
Actually trying to pass multiple merge_file outputs causes the
msgfmthelper script to be entirely malformed in the arguments it
accepts, and treat the broken one like a --flag, then exit with argparse
errors.
Even if we somehow assumed that somehow it was designed to actually
allow this, msgfmt doesn't support conceptually passing multiple outputs
so that would be a msgfmt error instead of an error inside the guts of
`meson --internal msgfmthelper`.
Same logic applies again for the itstool command and the itstool
internal helper.
Catch this error at configuration time by using the single-output kwarg
form.
Likewise, it's totally nonsense to accept multiple install_dir or
install_tags, and ever since commit 11f9638035
the CustomTarget itself won't even check this.
The `install_headers` function now has an optional argument
`preserve_path` that allows installing multi-directory
headerfile structures that live alongside sourcecode with a
single command.
For example, the headerfile structure
headers = [
'one.h',
'two.h',
'alpha/one.h',
'alpha/two.h',
'alpha/three.h'
'beta/one.h'
]
can now be passed to `install_headers(headers, subdir: 'mylib', preserve_path: true)`
and the resulting directory tree will look like
{prefix}
└── include
└── mylib
├── alpha
│ ├── one.h
│ ├── two.h
│ └── three.h
├── beta
│ └── one.h
├── one.h
└── two.h
Fixes#3371
python compiled extensions should never need to expose any symbol other
than PyInit_* which is declared with default visibility via
PyMODINIT_FUNC on supported compilers.
Thus, a reasonably sane default is to mark any other symbols as hidden,
while still respecting any manually specified visibility.
Gate this on the version of python itself, as not all versions decorate
PyMODINIT_FUNC properly.
Otherwise it always returns the value for c++98, starting with MSVC 2017
15.7 or later. Earlier versions are not affected by this mis-feature.
See: https://docs.microsoft.com/en-us/cpp/build/reference/zc-cplusplus?view=msvc-160
This was originally applied as 0b97d58548
but later reverted because it made the CI red. Try it again, now.
Original-patch-by: Dylan Baker <dylan@pnwbakers.com>
Co-authored-by: Dylan Baker <dylan@pnwbakers.com>
Dylan Baker