Preview available at https://egui-pr-preview.github.io/pr/7565-badclose
Note that it might take a couple seconds for the update to show up after the preview_build workflow has completed.

Thanks for workin on this! This looks very promising.

Please run cargo run -p test_viewports and make sure that still works as expected 🙏

It works as best I can tell

The other example that I was able to run on my computer behaved the same as when they were run on the main branch (excluding "serial_windows", which works as expected on Wayland with this PR).

I researched into the why the viewport ID check was added and it to resolve this issue with this commit.

if viewport_id == Some(ViewportId::ROOT) {
    return EventResult::Exit;
} else {
    return EventResult::Wait;
}

However, I cannot verify that this is not a regression as the close button cannot be hidden on Wayland, and the example did not run at all on X11. Although if this the window is recreated (which it appears to be as my closing animation is played each time the button in the example is clicked) on Wayland a crash did not occur.

I believe that this is not a regression, as the viewport does close and reopen on Wayland without any issues. However, the example did not run at all on X11, either on the main branch or my branch.

I wonder if this also solves

• eframe: create multiple windows in sequence #1918

I believe that this issue will be solved, on Linux at least, as the behavior described in that issue matches exactly what I have fixed.
Though I am not able to verify if it is solved on MacOS, since I do not have access to any Apple devices. I may be able to test on a Windows device in about 10 or so hours, as I do not have any to hand.

I tested it on macOS, and this works great:

• Closes eframe: create multiple windows in sequence #1918
• cargo run -p test_viewports still works

Thank you so much, @tye-exe!
Because showing the window after it's been hidden is broken in egui on Windows and Wayland, I spent 25mins implementing opening a new window as a solution for those platforms in my project using a loop and a wait for a tokio::sync::Notify.
However when it just went black and unresponsive my heart sank and I was fully ready to abandon the whole effort, even relegating it to a separate branch and scrapping it on main, but I thankfully decided to just look up if there had been any reports of this or fixes.
Turns out I just needed to update from 0.32 to 0.33 :)