Don't associate Wx timers with the parent frame. #11590
Conversation
|
This looks like it also fixes #11578 for wxagg and changes the failure on wx to a timeout. Can you rebase this on master and re-enable the wxagg tests? |
|
done |
03ed370 to
9d4cc11
Compare
|
Doesn't segfault travis anymore, but still fails (as you noted, on a timeout). Let's keep this wx test disabled? |
|
Yes, leave |
|
I'm puzzled why wx fails when wxagg works (they should share the same code re: interactivity), but sure... |
9d4cc11 to
27e9c2b
Compare
|
Given that it is timing out probably something subtle in the draw / render / put the window on the (fake) screen process? |
|
Well, that still fails. Let's just get rid of the wx test. |
This is consistent with the behavior on Qt and GTK, and avoids a segfault due to lack of disconnection of the timer after the parent widget is destroyed (otherwise, we'd need to keep track of timers associated with each widget and tear them down when the widget is destroyed).
27e9c2b to
4270fd0
Compare
|
@DietmarSchwertberger Can you review this? |
|
Looks good to me. Pending events are the most common source of segfaults... |
|
Thanks @DietmarSchwertberger ! |
This is consistent with the behavior on Qt and GTK, and avoids a
segfault due to lack of disconnection of the timer after the parent
widget is destroyed (otherwise, we'd need to keep track of timers
associated with each widget and tear them down when the widget is
destroyed).
Closes #11582.
Marking as release-critical per "fixes a segfault", though not a regression so not insisting on it.
PR Summary
PR Checklist