https://pantsbuild.org/ logo
Join Slack
Powered by
<Regarding my one year old question here>, I'd lik...
# general
r
Regarding my one year old question here, I'd like to ensure pytest fixture cleanup when there are "Filesystem changed during run: retrying 
Test
in 500ms.." messages, as retrying the test at a wrong timing causes "port already used" and other resource overlapping errors... Could we do this even without 
--debug
?
Also, could we know which files were modified when the retry occurs?
in 
.pants.d/pants.log
, it often just says without which files became dirty:
w
hm, so actually the “for all paths” indication is something else entirely. that means (on Linux at least), that the inotify buffer filled up, and so everything in memory needs to be invalidated
it’s triggered when we get a Rescan event from the kernel: https://docs.rs/notify/6.0.0/notify/event/enum.Flag.html#variant.Rescan
so you might need to adjust kernel settings to get a longer queue?
see 
/proc/sys/fs/inotify/max_queued_events
in https://man7.org/linux/man-pages/man7/inotify.7.html
r
hmmm how do i set the same config on macOS?;;
w
That shouldn't be possible on macOS. Can you file an issue with information about the macOS version you are using, and the repro steps?
r
i’m on M1 Max with Ventura 13.4. I’ll file an issue. (I’m outside so maybe a few hours later)
btw, would adding more pants_ignore paths help the situation?
w
No, it wouldn't: that's what is strange about the "for all paths" case: it's either in the kernel or a bug in the library we're using
😭 1
r
https://github.com/pantsbuild/pants/issues/19170 I've got a time to write up the issue, so here it is!
b
I was having this same issue and no idea why (it just cropped up this morning for seemingly no reason)--turning my mac on and off again did the trick. I'll follow-up if I'm ever able to reproduce this
Open in Slack
PreviousNext
Powered by