Ethan Lee

Odd, do we need to resize the buffer as part of setting fullscreen? I figured setting fullscreen would do that for us, via configure or a similar mechanism...

The second bit is probably the resize in SetWindowFullscreen, which unfortunately was necessary to get compositors to not blow apart the window size when leaving fullscreen (test with Super Hexagon)....

I'd be up for reintroducing the first one - the second one basically only works because we put off resizing _so late_ that it skips over a lot of activity...

That makes perfect sense to me - in messing around with this I think the block should be replaced with a flush, as that was what I was desperately trying...

Vulkan handles presentation on its own via `vkQueuePresentKHR` and `vkAcquireNextImageKHR`, so it doesn't need to touch SDL like OpenGL does. I wonder if the glitch has to do with the...

I did Even More Super Hexagon Tests(TM) and it seems like a root cause may have been us ignoring configure events...? https://github.com/libsdl-org/SDL/compare/2506729...1fb1aa1 With this I get the right window size...

Interesting, if this related to eglSwapBuffers, it might be worth checking with the swap interval set to both 0 and 1. The weird custom vsync waiting has some weird side...

Oof, yeah that is a messy stack... I think that might have been a carryover from before we had the shared CommitMinMax function, I wonder if moving the commit out...

Has anyone had a chance to dig into this since the last post? I'll try to start looking around this week but didn't want to retread anything by accident.

This one is at risk of being pushed back to 2.0.20, unless #4821 ends up being used or something finishes the WIP in https://github.com/libsdl-org/SDL/commit/785dfdfbe2633c203c1fe74445c98019cb8052dc.