JIT CI is broken due to test_pathlib and test_posixpath

[savannahostrowski]

Looks like something in emulation/QEMU changed and now test_pathlib and test_posixpath are causing the GHA workflow to fail for Linux Arm.

Linked PRs

• gh-118246: Move test_pathlib and test_posixpath to excluded tests #118247

[nineteendo]

Can it not be fixed? Or can we skip certain tests?

[barneygale]

Could you share a link to a failing GHA workflow please?

[brandtbucher]

Can it not be fixed? Or can we skip certain tests?

Yeah, we could probably use the --ignorefile option and maintain a text file with a list of individual failures under emulation... any volunteers?

The current PR is a good stopgap though, just to get things passing on PRs (we already skip a bunch of tests, so adding these isn't a huge change).

[savannahostrowski]

Yes, definitely - https://github.com/python/cpython/actions/runs/8820363755/job/24213816403

I'd like to spend time digging into the emulation failures but this is intended to provide fix in the meantime.

[nineteendo]

======================================================================
FAIL: test_ismount (test.test_posixpath.PosixPathTest.test_ismount)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython/Lib/test/test_posixpath.py", line 215, in test_ismount
    self.assertIs(posixpath.ismount("/"), True)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: False is not True

======================================================================
FAIL: test_is_mount_root (test.test_pathlib.test_pathlib.PathSubclassTest.test_is_mount_root)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython/Lib/test/test_pathlib/test_pathlib.py", line 1121, in test_is_mount_root
    self.assertTrue(R.is_mount())
    ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^
AssertionError: False is not true

======================================================================
FAIL: test_is_mount_root (test.test_pathlib.test_pathlib.PathTest.test_is_mount_root)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython/Lib/test/test_pathlib/test_pathlib.py", line 1121, in test_is_mount_root
    self.assertTrue(R.is_mount())
    ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^
AssertionError: False is not true

======================================================================
FAIL: test_is_mount_root (test.test_pathlib.test_pathlib.PosixPathTest.test_is_mount_root)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython/Lib/test/test_pathlib/test_pathlib.py", line 1121, in test_is_mount_root
    self.assertTrue(R.is_mount())
    ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^
AssertionError: False is not true

[barneygale]

Thanks for the link! It might have been aggravated by #117447 - perhaps under emulation, lstat('/') and lstat('/..') are returning different st_inos for some reason.

[brandtbucher]

Ah, I missed that change. Thanks for digging into it.

[brandtbucher]

Yeah, we could probably use the --ignorefile option and maintain a text file with a list of individual failures under emulation... any volunteers?

GH-118251

[brandtbucher]

@barneygale, is this something that you think is worth fixing in os.path.ismount? I don't think Python technically supports emulated environments, so it's not worth reverting any speed improvement or whatever.

[barneygale]

POSIX says:

The special filename dot-dot shall refer to the parent directory of its predecessor directory. As a special case, in the root directory, dot-dot may refer to the root directory itself.

I think / and /.. always refer to the same directory in practice, and os.path.ismount() uses that fact to detect the root directory. In all other cases, joining .. onto a path and calling lstat() will produce a different stat result. This all seems quite reasonable to me -- I'm not sure if there's anything to fix in ismount().

[brandtbucher]

This is done, then!