wayland: Dedupe wayland GL/VK configure code again

[ColinKinloch]

Configurations of xdg_toplevels are pending until acked on xdg_surface configure.

The vulkan backend doesn't need to take additional actions on configure.

Incorporates changes from #17317

I was intending on polishing it a bit more but if it's better than the current state it may be worth merging.

Related Issues

#16198 (comment)

Related Pull Requests

#17309

Reviewers

@Sunderland93

[Sunderland93]

Thanks! On KDE Plasma 6 the issue in #16198 (comment) is gone! On Plasma 5.25 it still there, so maybe it is a bug in Plasma/Kwin itself. The only issue that I found is resizing window in Vulkan backend: black borders around window content while resizing.

_20250101_145650.mp4

[ColinKinloch]

The black boarder seems to be an issue with retroarchs vulkan code as it also happens on x11.

[ColinKinloch]

I pushed some small changes.
They're unlikely to fix the issues on kwin 5.25.

[LibretroAdmin]

When you guys think this is ready for merging, just let me know. No rush though in case stuff still needs to be sorted out.

[ColinKinloch]

I think its in a much better state than master. It's probably worth merging now.

[Sunderland93]

They're unlikely to fix the issues on kwin 5.25.

Issue with maximizing window is totally fixed, but issue with initial window size (window too small) is still in place. In Plasma 6.2 no issues at all.

[ColinKinloch]

Which distro is 5.25?
I'd be interested to know if it's still happening with Debian stable (5.27.5)

[Sunderland93]

Debian stable (5.27.5)

This

[ColinKinloch]

Is it dual monitor (splash screen)?
I've just now replicated it with weston with a single output.

[Sunderland93]

Is it dual monitor (splash screen)?

No, single monitor

[ColinKinloch]

I've got a branch that fixes Weston as well (and hopefully the preview version of kwin-wayland)
It makes some of the changes in this branch unessesary so I made it draft.

[ColinKinloch]

I think I'll remove the splash screen altogether.
On mutter surfaces map on the output the cursor is currently on, unless the surface is fullscreen in which case it is mapped to the leftmost monitor.
I think the splash screen was there to force the window to become fullscreen on the monitor the cursor was on.
I'm not sure this is intended behaviour or a bug, either way I don't think RetroArch should work around it.

I'll create an issue on the mutter repo.