Skip to content

Conversation

@minrk
Copy link
Member

@minrk minrk commented Oct 20, 2025

also fixes %matplotlib osx which gets converted to macosx on its way through matplotlib and back

This should hopefully be the regression test for #1458

@minrk
Copy link
Member Author

minrk commented Oct 20, 2025

all the matplotlib tests are passing on mac and windows, need to figure out why linux isn't enabling the event loops

and increase timeout for tk
@ccordoba12
Copy link
Member

need to figure out why linux isn't enabling the event loops

Perhaps you need to run that test with xvfb?

@minrk
Copy link
Member Author

minrk commented Oct 20, 2025

Perhaps you need to run that test with xvfb?

Probably right, looking at that.

@ianthomas23
Copy link
Collaborator

Changes look good to me, and tests pass locally on both macos and linux. I don't understand the CI failures but I assume they are related to xvfb? Matplotlib itself pip installs pytest-xvfb which might be worth a try as with it installed pytest automatically uses xvfb without having to set the PRE_TEST stuff.

@minrk
Copy link
Member Author

minrk commented Oct 22, 2025

with xvfb, the tests are hanging after embed_kernel, which makes me think it might be during setup of test_eventloops

@ianthomas23
Copy link
Collaborator

Is it worth just disabling the reliably-failing test_matplotlib_eventloops.py on qt5/python3.14 so we can merge and come back to this later?

@minrk
Copy link
Member Author

minrk commented Oct 27, 2025

Yes, I think so. Sorry I couldn't figure out how to get these more reliable. They pass 100% every time I test locally, so I think it's something about how slow ci environments are

minrk added 2 commits October 26, 2025 19:53
they pass locally, so keep the tests, but don't keep holding up release with slow, flaky tests
@minrk
Copy link
Member Author

minrk commented Oct 27, 2025

I give up, I just skipped the tests altogether on CI. They are too flaky. I'm sure there's some race somewhere, perhaps in where the kernel runs the cell activating the event loop and then enters the eventloop mode, maybe a message can get dropped or stuck waiting somewhere? It could be a real bug, but I don't want to hold up the 7.1 release anymore, and this does have the %matplotlib osx fix that should get in there.

But these tests pass 100% for me locally, so I at least know that they verify %matplotlib osx work, so I don't want to delete them. I'll come back and try to track down the problem another time with less pressure.

Copy link
Collaborator

@ianthomas23 ianthomas23 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @minrk, I completely agree. Approving and merging this with CI disabled so that I can proceed with the 7.1.0 release and we can circle back to this later.

@ianthomas23 ianthomas23 merged commit 6f61a68 into ipython:main Oct 27, 2025
80 of 85 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants