GeezerVison

Some tests are very sensitive to timing and will fail on a loaded system

Which tests? Do you have examples?

The Python test suite is reliable on busy / heavy loaded systems.

The latest example which triggered this:

FAIL: test_timerfd_initval (test.test_os.TimerfdTests.test_timerfd_initval)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python3.13/test/test_os.py", line 4166, in test_timerfd_initval
    self.assertAlmostEqual(next_expiration, initial_expiration, places=self.CLOCK_RES_PLACES)
    ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: 0.24403756000000001 != 0.25 within 3 places (0.0059624399999999855 difference)

The test is checking how accurately the timer is.

Would be be possible to tune TimerfdTests to accept a large time difference, rather than skipping the test?

Isn't the point of the test to check the timings though? Would it be considered a pass if the request was a 0.1 second timer and a bug meant it waited for 0.1 hours instead?

Found an existing bug for that specific test on another Linux system: #126112.

Found an existing bug for that specific test on another Linux system: #126112

These are two different timerfd tests. The one quoted above isn't really testing the timer at all – it never expires. It's really testing that a few lines of Python code can run within 1 millisecond. This has already been increased to 10 ms on Android emulators (see the comment at the top of the test class), and I think it wouldn't be unreasonable to extend that to all platforms.

By comparison, the test mentioned in #126112 has a much wider timing margin – 125 ms if I understand correctly – so either it's caused by something else, or my understanding is wrong.

This PR is stale because it has been open for 30 days with no activity.