https://pantsbuild.org/ logo
Join Slack
Powered by
I'm seeing a test failure in CI for <https://githu...
# development
h
I'm seeing a test failure in CI for https://github.com/pantsbuild/pants/pull/10758 that I'm not seeing locally
namely, "Exception message: I/O operation on closed file." appearing in pantsd logs, which we explicitly test for. I think this is a std lib python error message from trying to read a closed file
@witty-crayon-22786 or @aloof-angle-91616 do you have any ideas about where theproblematic file open that these tests were added for might live?
w
that is an OSX shard, i’ll see if i can repro it on my machine in a bit.
unless you have an OSX box available
h
ah, no I don't
w
yea, i repro on OSX. don’t have time to debug right now, but can look Monday
h
ok. so seems to be an OSX-specific bug. I think I'll abandon trying to get this into the release today and just cut it in the next several hours
w
sounds good. i would recommend leaving this as alpha.
👍 1
with my fingers crossed very hard, i can say that we might consider beta early next week.
✨ 2
a
i missed your tag but i will try to check out this PR and debug it to make it more likely stu can follow up on monday
agreed with stu's recommendation
h
mostly I was curious if you or stu remembered what the issue was that caused the test for the absence of the string " I/O operation on closed file." to be written to begin with
a
no i didn't but i've seen it and it really annoys me. i was under the impression it was relating to log file handles that we touch in the signal handler but that's no further than you've got
h
@witty-crayon-22786 @aloof-angle-91616 https://github.com/pantsbuild/pants/pull/10758 jsut passed in CI now (with the brnach rebased on whatever changes were released over the weekend)
but I'm reluctant to merge in case the tests are flaky rather than always failing
(although I still haven't been able to trigger a failure on my own system locally)
okay I finally was able to reproduce this test failure locally, so it is flaky
sometimes, not consistently
this generated traceback has that "Exception message: I/O operation on closed file." message in it
and then that gets propagated to log files that the exception sink integration test is looking for
I want to dig more into making the ExceptionSink code cleaner and make it more clear exactly what it's doing, but I also want to get this merged, so I'm pushing a commit that just removes that particular check in the test for now: https://github.com/pantsbuild/pants/pull/10758 @witty-crayon-22786 @aloof-angle-91616
w
k, lemme know when it’s reviewable
h
it's reviewable right now I just pushed
👍 1
w
🚢
Open in Slack
PreviousNext
Powered by