Fix flaky CI tests - #26680
Conversation
There was a problem hiding this comment.
Can you explain this song and dance?
There was a problem hiding this comment.
Yes, it's in the commit message:
On some systems/pytest versions, the skip in an exception handler does not skip, but is treated as an exception. Namely, the ARM test machine in Cirrus and on my WSL Ubuntu.
I assume that is a bug somewhere in pytest but the subprocess avoids that. Also, changing the locale is a global modification, and it seems safer to do so in a subprocess, and not have to worry about fixing it up in the main process.
There was a problem hiding this comment.
Perhaps comment the code as well? This seems an odd thing to do.
On some systems/pytest versions, the skip in an exception handler does not skip, but is treated as an exception. Namely, the ARM test machine in Cirrus and on my WSL Ubuntu.
The main parent process is using Agg, so the subprocess should as well, to be able to unpickle correctly.
d0988c0 to
02263d2
Compare
jklymak
left a comment
There was a problem hiding this comment.
Feel free to self merge in case you want to squash
…680-on-v3.8.x Backport PR #26680 on branch v3.8.x (Fix flaky CI tests)

PR summary
I've pulled these out of #24597 as that is taking some time to be merged, and it seems like these are failing more often now. Some of these were also triggered on #26621 (while it had bugs) as well.
PR checklist