Repository navigation
MAINT: testing: specify python executable to use in extbuild - #28183
Conversation
This would cause failures during testing distribution packages on Cygwin. The code is copied from meson-python, as advised in mesonbuild/meson-python#51 and amended in mesonbuild/meson-python#137. Note the changed organization in many of the links.
|
close/reopen |
|
I have reenabled the cygwin test, let's see how it does. |
|
My PR #28149 fails |
|
DWesl#68 tracks my progress in getting Cygwin's Python 3.12 test package working, just re-enabling the current Python 3.9 CI on Cygwin is likely to fail. I don't expect either Cygwin Python version to pass CI at this time. This particular error would primarily affect distribution package managers testing packages for multiple Python versions at once. The test I found it in is currently skipped on Cygwin, and will likely remain so until I convince CI to run to completion, rather than hanging for six hours at around 26% completion. I found that test because a test run hung after saying it started that file. Running just that test succeeded, but I decided to run the test anyway and find out why it failed. After a few times staring at the test output saying it successfully created a Py3.9 extension but couldn't find the 3.12 extension, I worked out what kind of problem it was and found the meson-python issue with the solution. |
|
Going to disable the test again. @DWesl keep us posted on your progress. |
rgommers
left a comment
rgommers
left a comment
There was a problem hiding this comment.
Thanks @DWesl. These changes look fine to me. This same pattern is present in several other test files that build extension modules:
Details
$ rg -C2 native-file
_core/tests/test_limited_api.py
43- # Ensure we use the correct Python interpreter even when `meson` is
44- # installed in a different Python environment (see gh-24956)
45: native_file = str(build_dir / 'interpreter-native-file.ini')
46- with open(native_file, 'w') as f:
47- f.write("[binaries]\n")
--
57- "--werror",
58- "--buildtype=release",
59: "--vsenv", "--native-file", native_file,
60- str(srcdir)],
61- cwd=build_dir,
--
63- else:
64- subprocess.check_call(["meson", "setup", "--werror",
65: "--native-file", native_file, str(srcdir)],
66- cwd=build_dir
67- )
_core/tests/test_cython.py
44- # Ensure we use the correct Python interpreter even when `meson` is
45- # installed in a different Python environment (see gh-24956)
46: native_file = str(build_dir / 'interpreter-native-file.ini')
47- with open(native_file, 'w') as f:
48- f.write("[binaries]\n")
--
57- subprocess.check_call(["meson", "setup",
58- "--buildtype=release",
59: "--vsenv", "--native-file", native_file,
60- str(srcdir)],
61- cwd=build_dir,
--
63- else:
64- subprocess.check_call(["meson", "setup",
65: "--native-file", native_file, str(srcdir)],
66- cwd=build_dir
67- )
--
71- print("----------------")
72- print("meson build failed when doing")
73: print(f"'meson setup --native-file {native_file} {srcdir}'")
74- print("'meson compile -vv'")
75- print(f"in {build_dir}")
random/tests/test_extending.py
66- # Ensure we use the correct Python interpreter even when `meson` is
67- # installed in a different Python environment (see gh-24956)
68: native_file = str(build_dir / 'interpreter-native-file.ini')
69- with open(native_file, 'w') as f:
70- f.write("[binaries]\n")
--
74- subprocess.check_call(["meson", "setup",
75- "--buildtype=release",
76: "--vsenv", "--native-file", native_file,
77- str(build_dir)],
78- cwd=target_dir,
--
80- else:
81- subprocess.check_call(["meson", "setup",
82: "--native-file", native_file, str(build_dir)],
83- cwd=target_dir
84- )
For some reason it was overlooked here and didn't cause a problem before. Making this change can only improve correctness.
Are you happy with this, or still iterating?
| include_directories: {include_dirs}, | ||
| ) | ||
| """)) | ||
| # https://github.com/FFY00/meson-python/blob/2510cde44342f3167c44ea05d650c8a524a43879/mesonpy/__init__.py#L615-L618 |
There was a problem hiding this comment.
Fine to leave out this comment in my opinion
| ) | ||
| else: | ||
| subprocess.check_call(["meson", "setup", "--vsenv", ".."], | ||
| # https://github.com/FFY00/meson-python/blob/2510cde44342f3167c44ea05d650c8a524a43879/mesonpy/__init__.py#L665-L667 |
There was a problem hiding this comment.
same here, no need for the comment
I'll likely have a few more changes as I try to get NumPy compiling on Cygwin's Python 3.12 test release, but it'll be a while before I get python to stop hanging for long enough to get test results. This is complete as it stands, so I'm fine with it going in. |
rgommers
added
03 - Maintenance
component: numpy.testing
and removed
01 - Enhancement
labels
rgommers
changed the title
4 participants
This would cause failures during testing distribution packages on Cygwin. The code is copied from meson-python, as advised in mesonbuild/meson-python#51 and amended in mesonbuild/meson-python#137.
Note the changed organization in many of the links.
I found the failure specifically in
numpy._core.tests.test_array_interface.pyon Cygwin, where the system meson has a python3.9 shebang, so the test failed when it couldn't find a 3.12 extension module.