test.test_time.TimeTestCase.test_thread_time fails with Python 3.15 on s390x

[mcepl]

Bug report

Bug description:

When building Python 3.15 packages for openSUSE/Factory I got test.test_time.TimeTestCase.test_thread_time failing on s390x. It passes even on minority architectures like aarch64, ppc64le, or armv7l, but on s390x it fails this one test:

[ 4378s] ======================================================================
[ 4378s] FAIL: test_thread_time (test.test_time.TimeTestCase.test_thread_time)
[ 4378s] ----------------------------------------------------------------------
[ 4378s] Traceback (most recent call last):
[ 4378s]   File "/home/abuild/rpmbuild/BUILD/python315-core-3.15.0_a3-build/Python-3.15.0a3/Lib/test/test_time.py", line 587, in test_thread_time
[ 4378s]     self.assertLess(stop - start, 0.020)
[ 4378s]     ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
[ 4378s] AssertionError: 0.07135823999999502 not less than 0.02

Is it just that our Hercules emulators are too slow and I should just increase that 0.020 to 0.1 or something, or there could be something more there there? The only other test_thread_time issue I found is this on NetBSD, but I don’t think it applies here, does it?

Complete build log

CPython versions tested on:

3.15

Operating systems tested on:

Linux

Linked PRs

• gh-143528: Fix test_time.test_thread_time() #143558
• [3.14] gh-143528: Fix test_time.test_thread_time() (GH-143558) #143559
• [3.13] gh-143528: Fix test_time.test_thread_time() (GH-143558) #143560

[vstinner]

I wrote #143558 to tolerate 100 ms instead of 20 ms.

[vstinner]

I merged my change. If you still hit the issue with new 100 ms threshold, you can try to reduce the concurrency when running tests in parallel: -jN option of python -m test.

Another option would be to remove the test.