Migrate tk backend tests into subprocesses#18261
Migrate tk backend tests into subprocesses#18261tacaswell merged 7 commits intomatplotlib:masterfrom
Conversation
|
I'll take the liberty of converting to draft until you are done experimenting. Feel free to convert back... |
9156b85 to
1b3d407
Compare
|
It seems to be working; you'll need to rebase to get the docs CI working, and also fix the flake8 issues. |
jkseppan
left a comment
There was a problem hiding this comment.
Looks good to me, flake8 has some complaints.
QuLogic
left a comment
There was a problem hiding this comment.
My only question is why sometimes you check subprocess.TimeoutExpired and sometimes subprocess.CalledProcessError?
|
Also, we prefer rebase over merge. |
Try to isolate suspected interactions between tests
add note to continue the concept going forward
fa60aa9 to
030943d
Compare
|
@QuLogic The timeout checks should be there consistently, since you mention it, but some of the tests are better off to check the content of stdout or stderr before checking the return code. Hopefully the new comments make this clear! Also rebased as requested. |
|
Thanks @richardsheridan , I really appreciate your work on hardening the tk code! |
PR Summary
This PR is a first attempt to fix #18246. Try to isolate suspected interactions between tests by using subprocesses liberally. The argument against this might be that subprocess tests are computationally expensive and we might better track down the weird interactions. I'm happy to go with this or try a different idea!
Eagerly making the pull request to engage CI.
PR Checklist