Hello, can someone give a quick review to this one? We need it to be able to use official winit instead of our fork!

cc @madsmtm because you merged the previous one 🙏

r @daxpedda

I think removing the assert here is actually wrong, I can only come up with two reasons this fails:

1. It's a race condition, like @madsmtm suspected, in which case we actually have to recreate the handler instead of just moving on like nothing happened.
2. Our .0001 offset to deal with FP precision isn't enough.

So I would suggest the following:

• Instead of removing the assert!, we should include the past and current scale factor, so we can make sure that the issue is not the offset (problem 2.).
• If reproduction is an issue, we can merge it and open an issue tracking this awaiting further feedback.
• @kchibisov I'm leaving a proper fix for this for another time because of the merge freeze, let me know if you want me to address this nevertheless.

If you know how to fix you can post a fix, it's just will be merged with delay. Though, if it's short and doesn't conflict with KB v2, we could merge.

@maximetinu could you do what I suggested above and figure out if this isn't an issue with the offset but just a racing condition? Unfortunately I don't seem to be able to reproduce this on my end.

@kchibisov the fix shouldn't conflict with KB v2, I will make one as soon as we know more.

Replaced by #2850.

Sorry @daxpedda I never answered 🙏 I was never able to reproduce neither, just got crash reports from users hitting this randomly when resizing. I hope #2850 fixes it, thx!

Hi @daxpedda , sorry to revive such an old thread, but, months later, I got a way to consistently reproduce this crash: #3579

Or, at least, it's a crash in what seems to be the same assert.

When this happened to me months ago, it was caused by resizing the screen (but not always).

As I say in this new #3579, the same assert can also be hit by opening the print dialog