It's only useful to use those when you have to override include dirs
or library paths by appending them from various sources according to
the priority order, or if the compiler args need to be converted from
Unix/GCC-style to native (MSVC, for instance) style.
Sanity checks match neither of these.
Closes https://github.com/mesonbuild/meson/issues/1351
To reproduce, one could write:
foo = files('foo.c')
foo[0].path()
and get:
Traceback (most recent call last):
[snip]
File "/home/trhd/Projects/meson/mesonbuild/interpreterbase.py", line 398, in method_call
raise InvalidArguments('Variable "%s" is not callable.' % object_name)
UnboundLocalError: local variable 'object_name' referenced before assignment
Fix this by handling file objects separately.
os.path.commonpath was added in Python 3.5, so just write our own for
now. pathlib was added in Python 3.4, so this should be ok. We need to
use that instead of doing str.split() etc because Windows path handling
has a lot of exceptions and pathlib handles all that for us.
Also adds a unit test for this.
Also, now the linker options are added from various sources in the same
order as compiler arguments for compile commands. As a result, all
libraries and library paths from external and internal sources are added
after all the linker options have been added.
As a result option_link_args() are added when libraries are added to the
list since currently the only thing they add are the libraries specific
in cpp_winlibs/c_winlibs. This fixes an issue where compilation with the
MinGW toolchain (which uses static libraries for winlibs) would result
in undefined symbol errors because the static libraries would be added
in the very beginning and hence would not be scanned for symbols.
Detailed comments have been added that explain where each option is
coming from and why it's been added at that specific point.
More improvements are necessary here because we currently still
unnecessarily repeat libraries from dependencies over and over, which
is a major problem in gst-build because inter-subproject dependencies
cause linker command-lines to almost exceed the argument list length
limit imposed by the kernel. It is also causing us to unnecessarily
add static libraries which have already been linked into a shared
library. See: self.build_target_link_arguments()
At the same time, also fix the order in which compile arguments are
added. Detailed comments have been added concerning the priority and
order of the arguments.
Also adds a unit test and an integration test for the same.
While reading shebangs, when we detect an attempt to run 'python3', use
sys.executable instead. For example:
#!/usr/bin/python3
#!python3
#!/usr/bin/env python3
Back in November when this broke, we didn't notice because our tests
are run in-process, so we don't check that `msbuild RUN_TESTS.vcxproj`
and `ninja test` actually work.
Now we do.
Always generate the vcxproj file, but only add it to the build
configuration if it's either supposed to be built by default, or is
a dependency of another target that is built by default.
Also add a test() that can be run on all platforms.
Currently unit tests are only run on Linux, so this was only testing the
Ninja backend. This change reveals that build-by-default was broken with
the Visual Studio backend.
Previously, build_by_default=false targets would not be built at all
even if they were used in a test (as the exe or as a command-line
argument) which would lead to a test failure.
Now we look into all the defined tests and add all CustomTargets and
BuildTargets used in them to the list of build_by_default targets.
The purpose of this class is to make it possible to sanely generate
compiler command-lines by ensuring that new arguments appended or added
to a list of arguments properly override previous arguments.
For instance:
>>> a = CompilerArgs(['-Lfoo', '-DBAR'])
>>> a += ['-Lgah', '-DTAZ']
>>> print(a)
['-Lgah', '-Lfoo', '-DBAR', '-DTAZ']
Arguments will be de-duped if it is safe to do so. Currently, this is
only done for -I and -L arguments (previous occurances are removed when
a new one is added) and arguments that once added cannot be overriden
such as -pipe are removed completely.
Use a single check for both cases when we have includes and when we
don't. This way we ensure three things:
1. Built-in checks are 100% reliable with clang and on macOS since clang
implements __has_builtin
2. When the #include is present, this ensures that __builtin_func is not
checked for (because of MSYS, and because it is faster)
3. We fallback to checking __builtin_func when all else fails
prefix might define _GNU_SOURCE, which *must* be defined before your
first include of limits.h, so we must define it first.
There's not really any downsides to including limits.h after the
prefix.
Our "43 has function" test should also work with clang and icc on Linux,
so enable them. Also detect builtins with __has_builtin if available,
which is much faster on clang.
There is a feature request for the same with GCC too:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66970
If it's available, it's much much faster than doing #include
+ preprocess. Speeds up has_header checks by about 40% for me.
Currently, only Clang supports this
Jussi Pakkanen
Nirbheek Chauhan
Hemmo Nieminen
Tim-Philipp Müller
Thibault Saunier