In Debian testing/unstable, there are wrappers available for various
GObject-Introspection tools during cross-builds, using qemu internally.
Signed-off-by: Simon McVittie <smcv@debian.org>
Cross-tools on Debian generally follow the naming convention set by
Autotools AC_CHECK_TOOL, where the name is prefixed with the GNU
architecture tuple for the host architecture. env2mfile was already
using this for pkg-config, but there are many other tools that can
be detected in the same way.
Signed-off-by: Simon McVittie <smcv@debian.org>
`DEB_HOST_ARCH` encodes both the CPU family and the OS, so using it to
get the CPU type gives the wrong answer for non-Linux ports.
However, `DEB_HOST_GNU_CPU` gives less detailed information about the
CPU: it's `arm` for all 32-bit ARM CPUs, and doesn't distinguish between
the differing baselines of `armel` (ARMv5 softfloat) and `armhf`
(ARMv7 hardfloat).
When cross-compiling for x86_64 Linux, this changes the `cpu()` from
`amd64` to `x86_64`, which is consistent with the answer we get during
native builds on that architecture.
When cross-compiling for `ppc64el`, this changes the `cpu()` from
`ppc64el` to `ppc64`, which is a reasonable change but is still not
consistent with what we see during native builds (which is `ppc64le`):
see #13741 for that.
Resolves: https://github.com/mesonbuild/meson/issues/13742
Signed-off-by: Simon McVittie <smcv@debian.org>
All official Debian release architectures use the Linux kernel, but
unofficial ports like hurd-i386 and kfreebsd-amd64 use the Hurd and
FreeBSD kernel, respectively.
Map Linux to 'linux' and kFreeBSD ports to 'freebsd' as per the
reference tables in Meson's documentation. For now, use the Debian
system name such as 'hurd' for anything else (see #13740 for the
question of whether Hurd should identify its kernel differently).
Signed-off-by: Simon McVittie <smcv@debian.org>
As per <https://mesonbuild.com/Reference-tables.html>, and matching what
happens when running Meson for a native build on Debian GNU/Hurd.
Signed-off-by: Simon McVittie <smcv@debian.org>
This makes the frequent pattern of things like "CPU families are the
same as GNU CPUs, with a few known exceptions" less verbose.
Signed-off-by: Simon McVittie <smcv@debian.org>
Separating the part that runs dpkg-architecture from the part that
interprets its results will make it easier to unit-test the latter.
Signed-off-by: Simon McVittie <smcv@debian.org>
This will make it easier to unit-test functions that work with a
MachineInfo, by constructing the expected object in a single call.
Signed-off-by: Simon McVittie <smcv@debian.org>
The clang-tidy-fix target uses run-clang-tidy to do the fixing, however
this script itself spawns `os.cpu_count()` threads as part of its
internal parallelism. When combined with Meson's parallelism this
results in the creation of potentially thousands of unecessary threads.
This commit rewrites the clang-tidy-fix to perform the same task
run-clang-tidy does but exclusively on Meson's thread pool. "Fix-it"
snippets are saved to `meson-private/clang-tidy-fix/` by a parallel
clang-tidy phase, afterwards (to avoid races) all collected fixes are
applied with a single call to clang-apply-replacements.
It was possible (with some frequency) for the clang-tidy/format target
to continue starting new subprocesses after a CTRL-C, because we were
not canceling the already queued tasks and waiting for all of them.
This makes a best-effort attempt to cancel all further subprocesses. It
is not 100%, because there is a race that is hard to avoid (without
major restructuring, at least): new subprocesses may be started after
KeyboardInterrupt (or any other exception) is raised but before we get
to the cancellation.
When the race happens, the calling ninja may exit before Meson exits,
causing some output (from clang-tidy/format and Meson's traceback) to be
printed after returning to the shell prompt. But this is an improvement
over potentially launching all the remaining tasks after having returned
to the shell, which is what used to happen rather often.
In practice, it appears that we cleanly exit with a pretty high
probability unless CTRL-C is hit very early after starting (presumably
before the thread pool has launched subprocesses on all its threads).
The documentation for subprocess.run at https://docs.python.org/3/library/subprocess.html#popen-constructor has a warning,
pointing to using shutil.which() instead of subprocess.run for detecting if exe files exists on the path.
shutil.which() is used in many places already.
Fixes regression in commit 78e9009ff9.
new_rpath is only set when install_rpath appears in meson.build.
Before this commit, we treated new_rpath as a single string to pass to
-add_rpath. This worked as long as new_rpath had a single rpath in it,
though for ELF we explicitly supported multiple rpaths separated with
":" (as binutils ld is quite happy with that too).
install_name_tool does not support paths with colons in it:
```
21:12 <awilfox> Load command 19 cmd LC_RPATH cmdsize 40 path /bar:/baz:/eli:/ldap (offset 12)
21:12 <awilfox> Load command 20 cmd LC_RPATH cmdsize 24 path /foo (offset 12)
21:14 <awilfox> so the result is: do not use colons
```
After commit 78e9009ff9, we simply ended
up with one load command for LC_RPATH per char in new_rpath, which was
wrong in every possible case.
What we actually need to do is pass every distinct rpath as a separate
`-add_rpath` argument, so we split it on the colons instead. We do the
same splitting to ensure proper diff'ability for ELF anyways...
Fixes#13355
This does a couple of things:
1. Scrape the `Py_GIL_DISABLED` sysconfig var, which is the best way as of today
to determine whether the target interpreter was built with `--disable-gil`
2. link against the correct libpython
3. On Windows, work around a known issue in the python.org installer with a
missing define in `pyconfig.h`, which the CPython devs have said is unlikely
to be fixed since headers are shared between the regular and free-threaded
builds in a single NSIS installer.
This adds the `-quiet` option when invoking clang-tidy for the generated
`clang-tidy` target. (Note that the `clang-tidy-fix` target already does
so.)
This prevents messages like
```
Suppressed 1084 warnings (1084 in non-user code).
Use -header-filter=.* to display errors from all non-system headers. Use -system-headers to display errors from system headers as well.
```
from being repeated for every file, which drowns out the actual
warnings/errors from clang-tidy when more than a few files are
processed.
Also the tip about `-header-fileter` and `-system-headers` is not very
useful here because Meson doesn't currently provide a way to supply
these options to clang-tidy.
Even with `-quiet`, clang-tidy still prints a line like `1084 warnings
generated.` for each file.
--html-nested option is used to create a separate web page for each file and directory. Each of these web pages includes the contents of file with annotations that summarize code coverage.
Signed-off-by: Ewelina Walkusz <ewelinax.walkusz@intel.com>
This fixes the unit test `TestAllPlatformTests.test_noop_changes_cause_no_rebuilds`,
when run with an `nm` binary from `cctools-port` (as shipped by conda-forge, see
https://github.com/conda-forge/cctools-and-ld64-feedstock).
It also addresses the issue discussed in https://github.com/mesonbuild/meson/discussions/11131,
and this build warning:
```
[48/1383] Generating symbol file scipy/special/libsf_error_state.dylib.p/libsf_error_state.dylib.symbols
WARNING: ['arm64-apple-darwin20.0.0-nm'] does not work. Relinking will always happen on source changes.
error: arm64-apple-darwin20.0.0-nm: invalid argument --
```
as reported in scipy#20740.
The unit test traceback was:
```
> self.assertBuildRelinkedOnlyTarget('mylib')
E AssertionError: Lists differ: ['mylib', 'prog'] != ['mylib']
E
E First list contains 1 additional elements.
E First extra element 1:
E 'prog'
E
E - ['mylib', 'prog']
E + ['mylib']
unittests/allplatformstests.py:1292: AssertionError
```
The `nm` shipped by Apple yields the exact same results either way; the man page for `nm`
only lists the single-character form so this seems preferred either way.
This somewhat aligns "darwin" (Mach-O) with how ELF RPATHs are treated.
Instead of blindly removing all RPATHs, only remove the ones that are in
the rpath_dirs_to_remove set. This way, RPATHs that were added by the
toolchain or user are left untouched.
It is important not to remove RPATHs as they may be vital for the
executable at runtime. Issues #12045 and #12288 are examples of this.
Issue: https://github.com/mesonbuild/meson/issues/12045
Signed-off-by: Fabian Groffen <grobian@gentoo.org>
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
This basically existed for an assert which we don't need, as mypy would
catch that issue for us anyway. Removing the function entirely has some
small performance advantages
We already have to decide whether to scan a file at configure time, so
we don't want to have to do it again at compile time, every time the
depscan rule is run. We can do this by saving and passing the language
to use in the pickle, so depscan doesn't have to re-calculate it. As an
added bonus, this removes an import from depscan
We don't need to write and pass two separate files to the depscanner,
I've used the pickle because the pickle serializer/deserializer should
be faster than JSON, thought I haven't tested.
The .lcovrc file can contain several directives for genhtml, mostly
pertaining the visualisation of the generated HTML report. Passing the
config file to genhtml allows the user to customize their report.
Fixes#12863
In recursive scanning, a script variable was overwritten that caused the
logic *intended* to make the embedded path be trimmed based on $DESTDIR,
to no longer apply. This resulted in embedding the staging path, but
only when install_subdir() was used instead of specifying each file as
the argument to install_sources.
This replaces all of the Apache blurbs at the start of each file with an
`# SPDX-License-Identifier: Apache-2.0` string. It also fixes existing
uses to be consistent in capitalization, and to be placed above any
copyright notices.
This removes nearly 3000 lines of boilerplate from the project (only
python files), which no developer cares to look at.
SPDX is in common use, particularly in the Linux kernel, and is the
recommended format for Meson's own `project(license: )` field
Version 2.0 of lcov triggers an error when an exclude pattern is unused. This
can happen when the project has subprojects, but no code in them ends up
covered:
lcov: ERROR: 'exclude' pattern '/[...]/subprojects/*' is unused.
(use "lcov --ignore-errors unused ..." to bypass this error)
So, simply do as it says. Unused patterns doesn't seem problematic (or even
interesting) here, so I don't think there's any risk in simply turning this
off, which matches what happened with earlier versions of lcov anyway.
When a user invokes the scan-build target that Meson generates
all subprojects are included in the resulting report. This commit
modifies the invocation of scan-build to exclude all bugs that
scan-build finds in the subprojects from the final report.
A release note has also been added describing the changed behaviour.
Certain envs may not support invoking itstool by itself directly as a script
as shebang lines are not supported, such as under cmd.exe shells on Windows,
that are normally used for Visual Studio (and the like, such as clang-cl)
builds.
This will call the corresponding interpreter to invoke itstool when needed, so
that itstool can be properly run, even if shebang lines are not supported by
the env.
This will fix building appstream on Windows using clang-cl, for instance.
lcov 2.0 deprecates `--rc lcov_branch_coverage=1` for `--rc branch_coverage=1` and
gives an error when an exclude is used on a non existing directory.
I added a version check for lcov and removed the subprojects directory from the
exclusion list if it does not exist.
Fixes#11995
On python >=3.8, this information is expected to be encoded in the
sysconfig vars.
In distutils, it is always necessary to link to libpython on Windows;
for posix platforms, it depends on the value of LIBPYTHON (which is the
library to link to, possibly the empty string) as generated by
configure.ac and embedded into python.pc and python-config.sh, and then
coded a second time in the distutils python sources.
There are a couple of caveats which have ramifications for Cygwin and
Android:
- python.pc and python-config.sh disagree with distutils when python is
not built shared. In that case, the former act the same as a shared
build, while the latter *never* links to libpython
- python.pc disagrees with python-config.sh and distutils when python is
built shared. The former never links to libpython, while the latter do
The disagreement is resolved in favor of distutils' behavior in all
cases, and python.pc is correct for our purposes on python 3.12; see:
https://github.com/python/cpython/pull/100356https://github.com/python/cpython/pull/100967
Although it was not backported to older releases, Cygwin at least has
always patched in a fix for python.pc, which behavior is now declared
canonical. We can reliably assume it is always correct.
This is the other half of the fix for #7702
Since 3.10.3, Debian finally started patching sysconfig with custom
paths, instead of just distutils. This means we can now go use that
instead. It reduces our reliance on the deprecated distutils module.
Partial fix for #7702
The former is unix-only, which complains when running on the Windows
platform -- even though this code is in a unix-specific branch.
But per the docs:
> As of Python 3.3, this is equivalent to os.chmod(fd, mode).
So we can just use the identical function that is available in more
places.
Add the `clang-tidy-fix` target to apply clang-tidy fixes to the source
code.
This is done by calling `run-clang-tidy` with `-fix` argument.
Add a test case to run `clang-tidy-fix` and verify the file is changed.
Signed-off-by: Lei YU <yulei.sh@bytedance.com>
This commit adds a new keyword arg to extension_module() that enables
a user to target the Python Limited API, declaring the version of the
limited API that they wish to target.
Two new unittests have been added to test this functionality.
Performed using https://github.com/ilevkivskyi/com2ann
This has no actual effect on the codebase as type checkers (still)
support both and negligible effect on runtime performance since
__future__ annotations ameliorates that. Technically, the bytecode would
be bigger for non function-local annotations, of which we have many
either way.
So if it doesn't really matter, why do a large-scale refactor? Simple:
because people keep wanting to, but it's getting nickle-and-dimed. If
we're going to do this we might as well do it consistently in one shot,
using tooling that guarantees repeatability and correctness.
Repeat with:
```
com2ann mesonbuild/
```
We support this in a machine file:
```
[binaries]
pkgconfig = 'pkg-config'
pkg-config = 'pkg-config'
```
and you can use either one, because internally we look up both. If you
only set *one* of them, this plays awkwardly with setting $PKG_CONFIG,
since we don't know which one you set in the machine file and the
*other* one will be initialized from the environment instead.
In commit 22df45a319 we changed program
lookup of config-tool style dependencies to use the regular tool names
and only fall back on the strange internal names. This affected the
pkg-config class too.
The result is that instead of preferring `pkgconfig =` followed by
$PKG_CONFIG followed by `pkg-config =`, we inverted the lookup order.
This is a good idea anyway, because now it behaves consistently with
`find_program('pkg-config')`.
Unfortunately, we documented the wrong name in a bunch of places, and
also used the wrong name in various testsuite bits, which meant that if
you set $PKG_CONFIG and then ran the testsuite, it would fail.
Correct these references, because they are buggy.
One test case expected to find_program() a native copy for convenience
of testing against natively installed glib. Force it to resolve a native
copy.
Simon McVittie
Samuel Thibault
Jussi Pakkanen
Jonathon Anderson
Mark A. Tsuchida
Dylan Baker
Monson Shao
Mads Andreasen
Eli Schwartz
Ralf Gommers
Walkusz
Ole André Vadla Ravnås
Fabian Groffen
dragonmux
Esther Dalhuisen
David Robillard
Andrew McNulty
Luke Elliott
Chun-wei Fan
Eli Schwartz
Nigel Kukard
Charles Brunet
Lei YU
L. E. Segovia