Simon Ser authored 2 years ago and Matt Turner committed 2 years ago

This explicitly indicates that GNU extensions (like asm) are used.
This fixes build errors when Pixman is used as a Meson subproject.

Signed-off-by: Simon Ser <contact@emersion.fr>

Signed-off-by: Simon Ser <contact@emersion.fr>

Signed-off-by: Simon Ser <contact@emersion.fr>

Benjamin Gilbert authored 2 years ago and Matt Turner committed 2 years ago

When compiling with MinGW, use of the __thread attribute causes pixman
to gain a dependency on the winpthread DLL.  With Autotools, this could
be avoided by configuring with ac_cv_tls=none, causing pixman to fall
back to TlsSetValue() instead.

Add a Meson 'tls' option that can be 'disabled' to skip support for TLS
compiler attributes, or 'enabled' to require a working TLS attribute.

The library that the pkgconfig file is for should be the first
positional argument. The `libraries:` kwarg is for libraries that the
user must also link against, and which meson does not know about (and
hence cannot automatically add to the `Libs:` or `Requires:` section
in the .pc file).

Fixes:
```
subprojects/pixman/meson.build:564: DEPRECATION: Library pixman-1 was
passed to the "libraries" keyword argument of a previous call to
generate() method instead of first positional argument. Adding
pixman-1 to "Requires" field, but this is a deprecated behaviour that
will change in a future version of Meson. Please report the issue if
this warning cannot be avoided in your case.
```

Nirbheek Chauhan authored 3 years ago and Simon Ser committed 3 years ago

We use this because of a meson bug that was fixed in 0.52:

https://mesonbuild.com/Release-notes-for-0-52-0.html#improved-support-for-static-libraries

Bump the requirement and remove the extract_all_objects workaround.
This gets rid of a meson warning:

WARNING: extract_all_objects called without setting recursive
keyword argument. Meson currently defaults to
non-recursive to maintain backward compatibility but
the default will be changed in the future.

GTK2 has reached end of life, and GTK3 has been available for a
almost a decade.

Signed-off-by: Manuel Stoeckl <code@mstoeckl.com>
Reviewed-by: Simon Ser <contact@emersion.fr>

Since aarch64 has different neon syntax from aarch32 and has no
support for (older) arm-simd,
there are no SIMD accelerations for pixman on aarch64.

We need new implementations.

This patch also contains Ben Avions's series of patches for aarch32
and now the benchmark results are fine to aarch64.

Please find the result at the below ticket.

Added: https://bugs.freedesktop.org/show_bug.cgi?id=94758


Signed-off-by: Mizuki Asakura <ed6e117f@gmail.com>

AFAICT from the git history, what happened is that the gtk demos rely on
gtk being built with pixman support. pkg-config isn't really expressive
enough to have that information, so the solution that was come up with
was to search for pixman as well as gtk+ and hope that pixman being
installed was.

This isn't actually used anywhere in the meson build anyway, and it's
causing problems for projects that want to use pixman as a supproject
(there's a port of cairo underway that's hitting this), because it
confuses meson.

Tim-Philipp Müller authored 4 years ago and Matt Turner committed 4 years ago

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

This should resolve pixman/pixman#22
and make the tests pass with clang.

-ftrapping-math is already the default[1] for gcc, so this should not change
behavior when compiling with gcc.  However, clang defaults[2] to -fno-trapping-math,
so -ftrapping-math is needed to avoid floating-point expceptions when running the
combiner and stress tests.

The root causes of this issue is that that pixman-combine-float.c guards floating-point
division operations with a FLOAT_IS_ZERO check e.g.

if (FLOAT_IS_ZERO (sa))
	f = 1.0f;
else
	f = CLAMP (da / sa);

With -fno-trapping-math, the compiler assumes that division will never trap, so it may
re-order the division and the guard and execute the division first.  In most cases,
this would not be an issue, because floating-point exceptions are ignored.  However,
these tests call enable_divbyzero_exceptions() which causes the SIGFPE signal to
be sent to the program when a divide by zero exception is raised.

[1] https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
[2] https://clang.llvm.org/docs/UsersManual.html#controlling-floating-point-behavior

Signed-off-by: Matt Turner <mattst88@gmail.com>

Signed-off-by: Matt Turner <mattst88@gmail.com>

Shiyou authored 5 years ago and Matt Turner committed 4 years ago

It's recommended to use '-mloongson-mmi' for MMI.

Reviewed-by: Matt Turner <mattst88@gmail.com>

.. and add gettimeofday to the list of funcs to check instead
of having a separate check for it.

Those flags are set by the configure.ac script

The build system for libpng for MSVC does not generate a pkg-config file
for us, and CMake support in Meson does not work very well.  So, look
for libpng manually on MSVC builds if depedency discovery did not work
out via pkg-config or the CMake config files.

Look also for pthread.h if threading support is found by Meson, as the
underlying threading support may not be PThreads, depending on platform.

For now, disable the thread-test test program if pthread.h and if
necessary, the PThreads library, cannot be found, as the current
implementation assumes the use of PThreads.

Also bump the required Meson version to 0.50.0 since we need it for
-cc.get_argument_syntax()
-For a later commit, the has_headers sub-method for cc.find_library()

The implementation of OpenMP is not compliant for our uses, so disable
it for now by just not checking for it on MSVC builds, as we implicitly
add an /openmp switch to the build, which will cause linking the tests
programs to fail, as the OpenMP implementation is not enough.

-For MSVC builds, do not use the GCC-specific CFlags when checking for
 these features.

-For the MMX check, assume that we have good enough MMX intrinsics and
 inline assembly support (on ix86), since MSVC provides sufficient
 support for those since before the times of MSVC 2008, and 2008 is the
 oldest version that we can support, as with the pre-C99 GTK+ stack.

Unfortunately due to x64 compiler issues, pre-Visual Studio 2010 will
crash when building SSSE3 code, so we do not enable building SSSE3 code
on pre-2010 Visual Studio.

Also, for all x64 Visual Studio builds, we do not enable USE_X86_MMX
as inline assembly is not allowed for x64 Visual Studio builds, and
instead use the compatibility instrinsics that we already have in the
code.

Christoph Reiter authored 5 years ago and Matt Turner committed 5 years ago

GCC on Windows complains that "__declspec(thread)" doesn't work, but still
compiles it, so the meson check doesn't work. The warning printed by gcc:
"warning: 'thread' attribute directive ignored [-Wattributes]"

Pass -Werror=attributes to make the check fail instead.

This fixes the test suite (minus gtk tests) on Windows with mingw.

Christoph Reiter authored 5 years ago and Matt Turner committed 5 years ago

meson builds defaulted to SIZEOF_LONG=0 in various places

Signed-off-by: Matt Turner <mattst88@gmail.com>

Signed-off-by: Matt Turner <mattst88@gmail.com>

Signed-off-by: Matt Turner <mattst88@gmail.com>

Signed-off-by: Matt Turner <mattst88@gmail.com>

Signed-off-by: Matt Turner <mattst88@gmail.com>

This is unfortunately required to make the tests work correctly, as
otherwise meson assumes that the files are C code not assembly. I've
opened https://github.com/mesonbuild/meson/issues/5151, to discuss
fixing the issue in meson upstream.

Fixes #29

mach -> march

This issue causes openmp arguments to be injected into compilers that
can support openmp, even if they don't. This issue will be fixed in
0.51 (code already landed in mesonbuild#5116), for older versions lets
work around the issue.

And update RELEASING for the new meson build system.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

Dylan Baker authored 6 years ago and Adam Jackson committed 6 years ago

This commit adds a meson build system for pixman. It carries the usual
improvements of meson, better clean build time, much better incremental
build times, while being simpler and easier to understand.

This takes advantage of some features from the most recent versions of
meson: the builtin openmp dependency and the feature option type.

There are a couple of things that I've done a bit differently than the
autotools build system, I've built a libdemos which is the utilities
from the demos folder, and I've linked the demos with libtestutils from
tetsts, otherwise I expect that most things will be the same.

I've tested so far cross compiling from x86_64 -> x86, x86_64 ->
Aarch64, and Linux to Windows via mingw, as well as native x86_64 Linux
builds which all work. I've also built with mingw nativly, there are
some test failures there. An MSVC build can be generated, but fails.

v2: - set WORDS_BIGENDIAN in the config for big endian systems.