TST: xfail test_kind::test_quad_precision on AIX/PPC #29693
Conversation
numpy/f2py/tests/test_kind.py
Outdated
| @@ -37,7 +38,7 @@ def test_real(self): | |||
| i | |||
| ), f"selectedrealkind({i}): expected {selected_real_kind(i)!r} but got {selectedrealkind(i)!r}" | |||
|
|
|||
| @pytest.mark.xfail(platform.machine().lower().startswith("ppc"), | |||
| @pytest.mark.xfail(IS_PPC, | |||
There was a problem hiding this comment.
cmake's platform.machine()
What does CMake have to do with NumPy in this context? It is being used to compile CPython somewhere or? I thought CPython was still using an older build system.
Also, if the root of the problem is AIX OS itself, would it make sense to just check for it directly with sys.platform, which works just fine to pick up aix? https://docs.python.org/3/library/sys.html#sys.platform
Can you provide some context--you're working on an AIX machine and this test failed?
There was a problem hiding this comment.
Thanks for the review. My apologies. I meant python's platform.machine(). On my AIX machine, platform.machine() returns 00FA13034C00 which is not useful. Anyway, checking sys.platform == 'aix' or platform.system() == 'AIX' is fine for this purpose.
Can you provide some context--you're working on an AIX machine and this test failed?
Yes, I am working on AIX and trying to use flang in numpy.f2py. On AIX, Fortran's selected_real_kind(32) returns -1. This test is already xfailed for ppc platform.
On AIX, cmake's
platform.machine()does not return a meaningful string that can be used to determine the architecture. This PR is to add the check ofplatform.processor()to xfail the test.