Ensure meson's prefix is a valid absolute path

[shoeffner]

The path / is not considered an absolute path on Windows, failing with:

ERROR: prefix value '/' must be an absolute path

os.path.abspath ensures that "/" becomes a valid absolute path on all major platforms, resolving to "/" on Linux and macOS, and to "C:\" (or whichever filesystem is currently active, could be E:\ as well) on Windows.

Closes #17204.

Changelog: Fix: Use a valid prefix path for meson.configure() on Windows, to avoid failures in Python 3.13.
Docs: Omit

• Refer to the issue that supports this Pull Request.
• If the issue has missing info, explain the purpose/use case/pain/need that covers this Pull Request.
• I've read the Contributing guide.
• I've followed the PEP8 style guides for Python code. -> No, line length is longer than 80 or such
• I've opened another PR in the Conan docs repo to the develop branch, documenting this one. -> No

[memsharded]

Thanks for your contribution!

This PR is failing in Windows with:

myhello  undefined
E           
E             User defined options
E               Native files: C:\J\t_v2\9ed6\PY38\tmp_gcx6x3iconans\path with spaces\build-release\conan\conan_meson_native.ini
E               prefix      : C:\
E           
E           Found ninja-1.8.2 at "C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\Common7\IDE\CommonExtensions\Microsoft\CMake\Ninja\ninja.EXE"
E           
E           conanfile.py (demo/0.1): Meson build cmd: meson compile -C "C:\J\t_v2\9ed6\PY38\tmp_gcx6x3iconans\path with spaces\build-release" -j16
E           conanfile.py (demo/0.1): RUN: meson compile -C "C:\J\t_v2\9ed6\PY38\tmp_gcx6x3iconans\path with spaces\build-release" -j16
E           conanvcvars.bat: Activating environment Visual Studio 15 - amd64 - winsdk_version=None - vcvars_ver=14.1
E           [vcvarsall.bat] Environment initialized for: 'x64'
E           [1/2] Compiling C object myhello.exe.p/src_main.c.obj
E           FAILED: myhello.exe.p/src_main.c.obj 
E           "cl" "-Imyhello.exe.p" "-I." "-I.." "-DNDEBUG" "/MD" "/nologo" "/showIncludes" "/utf-8" "/W2" "/O2" "/Gw" "/Fdmyhello.exe.p\src_main.c.pdb" /Fomyhello.exe.p/src_main.c.obj "/c" ../src/main.c
E           .\config.h(8): error C2001: newline in constant
E           ninja: build stopped: subcommand failed.
E           INFO: autodetecting backend as ninja

It is possible that the os.path.abspath result needs to be back to forward slash for Meson to process it?

[shoeffner]

Thanks for the logs! I don't think it's the \ vs / since the native files also use backslashes – and I was able to build inih (see our discussion in #17204) with this change.

I will try to get the full test suite up and running on our build machine and see if I can reproduce and fix the issue.

[memsharded]

I will try to get the full test suite up and running on our build machine and see if I can reproduce and fix the issue.

Probably not worth the effort. The full suite, specially the functional tests, have tons of tools to configure in conftest and conftest_user.py. You are probably good with running just a small subset. The test failing is test_meson_using_prefix_path_in_application, probably running tests in functional/toolchain/meson is more than enough.

[shoeffner]

tl;dr: C:\ is interpreted by the failing test as a newline in constant:

E           .\config.h(8): error C2001: newline in constant

Switching it to C:/ works, and all other test results remain unchanged.

long read:

Okay, that took longer than I expected. I had to change test\utils\tools.py as our mvsc has version 194 (in contrast to the default test version 191) and it took me a while to figure that out.

Here's the result of the unmodified (except for the version as described above) develop2 branch with pip-installed meson:

C:\src\conan>pytest -n 16 test\functional\toolchains\meson
===================================== test session starts ======================================
platform win32 -- Python 3.10.11, pytest-7.4.4, pluggy-1.5.0
rootdir: C:\src\conan
configfile: pytest.ini
plugins: xdist-3.6.1
16 workers [32 items]
ssssssssssssssss..s..s.....F....                                                          [100%]
=========================================== FAILURES ===========================================
___________________________________ test_meson_lib_template ____________________________________
===================== 1 failed, 13 passed, 18 skipped in 74.57s (0:01:14) ======================

I haven't checked yet why this particular test fails, maybe it's because I changed the compiler version locally and this assumes a specific hash?

This is the (shortened) output with the choco-installed meson:

=================================== short test summary info ====================================
FAILED test/functional/toolchains/meson/test_install.py::MesonInstall::test_install - Exception:
FAILED test/functional/toolchains/meson/test_subproject.py::test_subproject - Exception:
FAILED test/functional/toolchains/meson/test_backend.py::test_cross_x86 - Exception:
FAILED test/functional/toolchains/meson/test_meson_and_gnu_deps_flags.py::TestMesonToolchainAndGnuFlags::test_mesondeps_flags_are_being_appended_and_not_replacing_toolchain_ones - Exception:
FAILED test/functional/toolchains/meson/test_v2_meson_template.py::test_meson_lib_template - Exception:
FAILED test/functional/toolchains/meson/test_cross_compilation.py::test_windows_cross_compiling_x86 - Exception:
FAILED test/functional/toolchains/meson/test_meson.py::test_meson_using_prefix_path_in_application - Exception:
FAILED test/functional/toolchains/meson/test_meson.py::MesonToolchainTest::test_definition_of_global_options - Exception:
FAILED test/functional/toolchains/meson/test_v2_meson_template.py::test_meson_exe_template - Exception:
FAILED test/functional/toolchains/meson/test_meson.py::MesonToolchainTest::test_meson_default_dirs - Exception:
FAILED test/functional/toolchains/meson/test_pkg_config_reuse.py::MesonPkgConfigTest::test_reuse - Exception:
FAILED test/functional/toolchains/meson/test_test.py::MesonTest::test_reuse - Exception:
FAILED test/functional/toolchains/meson/test_preprocessor_definitions.py::MesonPreprocessorDefinitionsTest::test_build - Exception:
========================== 13 failed, 1 passed, 18 skipped in 20.62s ===========================

And the error is exactly what I observed initially:

E           ..\meson.build:2:0: ERROR: prefix value '/' must be an absolute path

With my fix, I can reproduce the error observed in the CI pipeline:

FAILED test/functional/toolchains/meson/test_meson.py::test_meson_using_prefix_path_in_application - Exception:

A related issue seems to be #14213 and the test there checks if the prefix is set correctly. It appears that particular test catches problematic backslashes in the prefix path, so C:\, while valid for meson, might be a broken value for some recipes. Using a unix filesep as described in the issue works, i.e., passing C:/ instead.

I updated the code and added a _prefix property for which I documented the constraints and reasoning why the os.path.abspath and replace are required.

[memsharded]

Many thanks for your contribution @shoeffner

Checked that without this the test suite fails in Python 3.13.
We might need to add it to our CI, it seems that 3.13 is changing a few things that are breaking.

Lets merge this to include it in next Conan 2.9 asap