Go back to checking figures for their manager in destroy. #18184
Conversation
| num = next((manager.num for manager in cls.figs.values() | ||
| if manager.canvas.figure == fig), None) | ||
| if num is not None: | ||
| cls.destroy(num) |
There was a problem hiding this comment.
Can we add a comment, why we have to destroy by number?
There was a problem hiding this comment.
It actually works either way (manager or manager.num); it's just that the code in destroy is simpler for num, i.e., we know that cls.figs.get(manager.num) is manager because we just pulled manager from cls.figs.
|
Yea, that's weird, a restart was not effective; I'll try it out on the mac. |
|
It worked on the Mac Mini, so I restarted Travis again. |
4d8e0d1 to
f65eb17
Compare
|
I tried increasing the timeout, which didn't work, but it's appearing on other tests now, so I've reverted that as the problem doesn't seem strictly related to this PR. |
|
As discussed on the call to get 3.3.1 out the door we are going to disable this test on travis (it should be noted that we do run this test on azure) so the current working theory is that this is a travis-specific issues with the GUI and is not worth the time right now to track down the root problem (but we should open a new issue and label it "good first issue / medium" localized, but involves defeating CI ;) ). |
This is a partial revert of a small part of matplotlib#13581. Fixes matplotlib#18163.
We will follow up on this later, matplotlib#18213.
f65eb17 to
db6fb4d
Compare
…anager in destroy.
…184-on-v3.3.x Backport PR #18184 on branch v3.3.x (Go back to checking figures for their manager in destroy.)
PR Summary
This is a partial revert of a small part of #13581.
Fixes #18163.
PR Checklist