GH-109067: fix randomly failing test_async_gen_asyncio_gc_aclose_09 test

[kumaraditya303]

• Issue: test_asyncgen fails randomly on Windows #109067

[gvanrossum]

Great. Let's kill more short sleeps! They always cause problems.

[itamaro]

Thank you!
Can you explain why this was flakily failing on Windows, on why this fixes it? It's not obviously apparent to me from the code change

[kumaraditya303]

Can you explain why this was flakily failing on Windows, on why this fixes it? It's not obviously apparent to me from the code change

On Windows the monotonic clock is low precision so it distorts the test.

[vstinner]

I confirm that the fix works as expected.

On Linux, it's hard for me to reproduce the issue. But on Windows, it's quite easy with the command:

python -m test test_asyncgen -m test_async_gen_asyncio_gc_aclose_09 -j12 --forever 

It fails in a few seconds. I ran the test in a Windows VM with 2 CPUs (running on my laptop which has 12 threads / 6 CPU cores).

With this PR, the test no longer fails. I interrupted the test for a few minutes:

...
0:06:21 load avg: 20.99 [807] test_asyncgen passed
0:06:22 load avg: 20.99 [808] test_asyncgen passed
0:06:23 load avg: 21.02 [809] test_asyncgen passed

To confirm the fix, I ran a second test on the PR on the whole test_asyncgen test module (without filtering on -m test_async_gen_asyncio_gc_aclose_09). I interrupted the test since it worked! test_asyncgen now looks rock solid.

...
0:05:39 load avg: 22.00 [492] test_asyncgen passed
0:05:39 load avg: 22.00 [493] test_asyncgen passed
0:05:40 load avg: 21.95 [494] test_asyncgen passed
0:05:41 load avg: 21.86 [495] test_asyncgen passed
0:05:41 load avg: 21.86 [496] test_asyncgen passed

[vstinner]

LGTM.

[miss-islington]

Thanks @kumaraditya303 for the PR 🌮🎉.. I'm working now to backport this PR to: 3.11, 3.12.
🐍🍒⛏🤖

[bedevere-bot]

GH-109149 is a backport of this pull request to the 3.12 branch.

[bedevere-bot]

GH-109150 is a backport of this pull request to the 3.11 branch.