Go back to checking figures for their manager in destroy. - #18184
Conversation
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.
timhoffm
left a comment
There was a problem hiding this comment.
Travis OSX fails. I'm unclear if the error is related to the PR.
|
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. |
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