I'm not able to reproduce the segfault, but I do see the flickering. Oddly the flickering isn't present at all on my normal testing Mac (an M2 mini).

I borrowed two other Macs and both flicker during resize, making it impossible to see the rendered window contents while resizing. At first I thought maybe a queue_wait_idle(..) might be a more appropriate fix, but no, it has no effect, just like the MoltenVK issue suggests.

The fix in this PR does work (on all platforms, too) but it is probably not what you expect because that function is called every frame. Also I still see a small amount of flicker during resize. Moving the idle into the contained null-check made the flickering worse than this PR, but better than nothing.

I found that doing a device_wait_idle() right when the window detects a resize event (so inside the Swapchain::set_info(..) function) works well, but there still is a very small amount of flicker. Combining this with an idle inside the null-check of Swapchain::destroy_swapchain(..) eliminates the flicker for me.

Other things I tried:

• Multiple repeated device_wait_idle() calls: Works and solves the flicker completely, but obviously that's just delaying the MoltenVK code from running which allows the bug to be fixed by timing
• std::thread::yield_now() in various places: no effect

Windows and Linux don't seem to be affected by the presence of these fixes in any way, but it would be hygienic to wrap any changes in #[cfg(target_os = "macos")] or something.

I tested these variations on:

• MacBook Pro 2016 Intel HD 530 & AMD Radeon Pro 450 (Monterey 12.7.6)
• MacBook Pro 2023 M3 Max (Sequoia 15.4)
• Mac mini 2023 M2 (Sonoma 14.4.1)
• Windows 11 Intel Iris Xe
• Windows 10 AMD Radeon Pro 6500 XT (also Ubuntu 25.04 X11 & Wayland)
• Windows 10 NVIDIA GeForce RTX 3090 (also Ubuntu 24.04 X11 & Wayland)

How do these changes work for you? I can push up what I'm talking about if that helps. And thank you for the detailed report!!

I have added changes based on my understanding of your reply, let me know if you meant something else.

For reference, I have Macbook M1 Max wit OSX 15.3.2 - perhaps it behaves a little different from your machines.

the additional device.device_wait_idle() in Swapchaing::set_info is unfortunately not making much differences on my side. However, the crash is avoided as long as I have one of those two device.device_wait_idle().