-
-
Notifications
You must be signed in to change notification settings - Fork 18.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CI: Pin PYTEST_WORKERS=1 for Windows builds due to memory errors #47318
Conversation
im not seeing a way to download the log file (which is too big to display in the browser) |
https://github.com/pandas-dev/pandas/suites/6902339888/logs?attempt=2 is the direct download link for the 3.9 build |
Looks like you're doing trial/error to figure out what test is failing? LMK if i can be helpful |
Yeah, I think there's a memory leak somewhere but for now I'm just skipping tests to hopefully skip the one that leaks. If you have time/ideas how to identify the leaking test(s) that would be helpful |
@jbrockmendel the first PR merged into main that exhibited these failures was #47299. If you have a chance could you comb over that PR again to see if any change there could cause a memory leak? |
The changes in tzconversion.pyx are the most plausible candidate, though
not a slam dunk
…On Sun, Jun 19, 2022 at 1:27 PM Matthew Roeschke ***@***.***> wrote:
@jbrockmendel <https://github.com/jbrockmendel> the first PR merged into
main that exhibited these failures was #47299
<#47299>. If you have a change
could you comb over that PR again to see if any change there could cause a
memory leak?
—
Reply to this email directly, view it on GitHub
<#47318 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AB5UM6AXF6OJIHNI4KNSIPDVP57BVANCNFSM5YSCUS4A>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
lgtm to get ci green(ish) for now. |
@simonjayhawkins should we backport to 1.4.x or just merge to main? |
probably should backport as could miss things when ci is not green? |
Okay merging to get the CI back to green for Windows builds |
…ilds due to memory errors
sure. have not seen windows failures on 1.4.x but if we don't know the root cause then no harm. |
…ows builds due to memory errors) (#47445) Backport PR #47318: CI: Pin PYTEST_WORKERS=1 for Windows builds due to memory errors Co-authored-by: Matthew Roeschke <[email protected]>
…das-dev#47318) * CI/DEBUG: Windows 3.9 memory error * Try skipping some suspicious tests * Try skipping this hanging test * Check if can run okay on single process * Trigger CI * run in parallel again * Log even more * a little less debugging * dont try new pyarrow * Try one more time with pyarrow<8, simple * Comment out pyarrow hack * Restart, show setup and full traceback * Also try the older windows * generate debug log again * change to latest * dump log if failure * Create separate log per os/version * See if it was any dependencies bumped recently * See if same errors happen without dependencies pinnning * Skip plotting tests on windows * Skip in frame as well * Trigger another run * Is it the non-nano? * skip resample tests * Just use 1 thread for windows * Full run
windows-2019
aswindows-latest