Robert Bragg

> I think the thing what winit is doing is not entirely wrong, and I guess most platforms don't require sync events. Though, I have my own issues with the...

> Be aware that frame callbacks on Wayland require a request(I'd assume the same on the Web), and to do that request you need to actually submit a buffer right...

>I mean that current winit is sane with single window, but with multi windows, you have a global RedrawEventsCleared, and some clients draw on it. So the API should change...

> App can draw whenever, we don't need an extension for that, what we want is a recommended place to perform drawing operations. This isn't always exactly true. Mobile OSs...

> > The idea of redraw events (plural) seems pretty odd in the first place (I'd only expect there to be one event for redrawing each frame. > > [`RedrawEventsCleared`](https://docs.rs/winit/latest/winit/event/enum.Event.html#variant.RedrawEventsCleared)...

Adding to this... I realized yesterday that my corresponding Windows changes for `run`, `run_ondemand` and `pump_events` weren't correct and I had to take a second pass over those and found...

Hmm, curious. I wonder if you have tried reproducing this with the jni master branch - all of the thread attachment details have been re-implemented since jni 0.21 - but...

Btw, thanks for creating the repro repo - that makes this much easier to investigate. As an initial experiment, I updated the test repo for jni git master here: https://github.com/rib/jni-rs-issue-701-repo/tree/rib/wip/windows-fls-alloc...

I'm noticing CI hang on Windows quite often now, e.g. this job has been stuck for > 40 minutes: https://github.com/jni-rs/jni-rs/actions/runs/20099832737/job/57667475308 and I'm thinking now that these hangs quite likely come...

No worries, yeah `FlsAttach` seems to be the way to go, though I'm not quite sure atm how to document that the jni crate will probably have to assume that...