An attempt to fix mypyc tests on MacOS#16520
Conversation
ilevkivskyi
requested review from
AlexWaygood,
JelleZijlstra,
JukkaL and
hauntsaninja
|
OK, another 20 runs passed locally, and a second re-run also passed in GHA. I think we can merge this unless there are objections. |
| @@ -172,12 +172,10 @@ def run_case_inner(self, testcase: DataDrivenTestCase) -> None: | |||
| # new by distutils, shift the mtime of all of the | |||
| # generated artifacts back by a second. | |||
| fudge_dir_mtimes(WORKDIR, -1) | |||
There was a problem hiding this comment.
If we're just doing time.sleep() on all platforms now, maybe we should just delete this line (and adjust the comment above it)? Or comment it out?
| fudge_dir_mtimes(WORKDIR, -1) |
There was a problem hiding this comment.
There is no harm in doing both. I would just keep it as is, and hope someone will find a better solution soon.
msullivan
left a comment
msullivan
left a comment
There was a problem hiding this comment.
Sure, need to stop bleeding.
But maybe something like having fudge_dir_mtimes try to wait for the change to become visible could work?
Hm, I just tried that (in couple variations), and somehow it still doesn't work reliably. Maybe there is some weird rounding behavior. I give up for now. |
|
Thanks for looking into this! |
4 participants
Fixes #16420
Although this is not 100% clear yet, but after 20 runs on a Mac I have it no longer fails (without this patch it failed 20% of times). Btw, contrary to the comment, my Linux Mint (which is an Ubuntu derivative) works perfectly (i.e. test passed 20 times even after I removed the
sleep()). So it is not really Mac vs Linux issue.