Fix fullscreen configure events sending size 0,0 during state transitions

[Copilot]

Related: pop-os/cosmic-comp#1642, pop-os/cosmic-comp#2099

What's new?

Chrome video fullscreen fails because XdgToplevelStable::send_toplevel_configure() sends size 0,0 when requested_window_size() isn't set. This occurs when state change observers fire before resize observers update the cached size.

Fix: Fall back to scene_surface()->window_size() when requested_window_size() is unavailable. Scene surface is updated synchronously by the window manager, so it reflects correct size even when async observer callbacks haven't run.

// Before: defaults to 0,0 if requested_window_size() not set
geom::Size size = requested_window_size().value_or(geom::Size{0, 0});

// After: uses scene surface's actual window size as fallback
if (auto const requested_size = requested_window_size())
    size = requested_size.value();
else if (auto const surface = scene_surface())
    size = surface.value()->window_size();
else
    size = geom::Size{0, 0};

Applied to both xdg_shell_stable.cpp and xdg_shell_v6.cpp.

How to test

1. Open Chrome on a non-fullscreen window
2. Navigate to any video site (YouTube, etc.)
3. Click fullscreen on the video player
4. Expected: Video goes fullscreen, not the browser window

Checklist

• Tests added and pass
• Adequate documentation added
• (optional) Added Screenshots or videos

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Making videos fullscreen in Google Chrome fails if the window isn't already fullscreen</issue_title>
<issue_description>### Version

v2.25.2

Environment

OS: Ubuntu 24.04

What happened? What did you expect to happen?

When you fullscreen a video in Google Chrome it will instead just make the window fullscreen (like using F11). If the window is already fullscreen then the video does become fullscreen. This doesn't occur with Firefox, but it also doesn't occur when using Google Chrome in GNOME.

I think this issue for the Pop OS Cosmic compositor is the same: [BUG] Maximized chromium browsers: clicking video fullscreen makes browser fullscreen instead and their fix is here: fix: window state preservation

Fixes a bunch of issues where the maximized/snapped state of a window was lost in scenarios involving:

• transitions to full-screen mode,
• from full-screen mode to snapped state,
• moving full-screen/snapped window between workspaces.

Exact fixes:

• For Chromium and similar clients, implements strict protocol compliance via pre-set fullscreen before unmap. Unmap sends intermediate configures considered by Chromium as a refusal to transition to full-screen mode, as they did not contain full-screen. This fixes the issue: [BUG] Maximized chromium browsers: clicking video fullscreen makes browser fullscreen instead #1642 (for both maximized and snapped windows).
• Preserving snapped state when transitioning to fullscreen/minimizing. FloatingRestoreData now stores was_snapped: Option<TiledCorners>. The information is captured before floating_layer.unmap() clears it and is used later for restoration.
• Preserving maximized/snapped state when moving between workspaces. After moving the window, the maximized or snapped state from FloatingRestoreData is re-applied in the target workspace (Maximized windows don't stay maximized when they are moved to a different display or workspace #528, Maximized window loses its state when moved to another workspace cosmic-epoch#2608)

What I tested:

• Maximize Chrome > Click fullscreen on any video > Single-click fullscreen
• Maximize Firefox > Click fullscreen on any video > Single-click fullscreen (regression check)
• Snap Chrome to a corner/side > Click fullscreen on any video > Single-click fullscreen
• Snap window to a corner/side > Go fullscreen > Exit fullscreen > Window returns to the snap position
• Maximize window > Move window to the other workspace > Window stays maximized
• Snap window to a corner/side > Move window to the other workspace > Window stays snapped to the same corner/side

I tried to make the changes as small as possible.

I also wanted to fix the issue with returning the window to the stack from full-screen state in this PR, but it appears to be much more complicated than I initially imagined and has a lot of edge cases.

Reproduction steps

No response

Relevant log output

```</issue_description>

## Comments on the Issue (you are @copilot in this section)

<comments>
<comment_new><author>@Saviq</author><body>
Thanks for the detailed report!</body></comment_new>
<comment_new><author>@AlanGriffiths</author><body>
I'm having trouble reproducing. Testing with Miriway 26.01-mir2.25.2 and Chromium:

With window (on https://www.youtube.com/watch?v=7fGB-hjc2Gc) in a restored state, press "F" or click "Fullscreen" icon: video fullscreens

With window (on https://www.youtube.com/watch?v=7fGB-hjc2Gc) in a fullscreen state, press "F" or click "Fullscreen" icon: video fullscreens

Maybe I have misunderstood the reproduction steps? It seems unlikely Chromium behaves differently to Google Chrome in this respect?</body></comment_new>
<comment_new><author>@AlanGriffiths</author><body>
I'm also starting Miriway from GDM but don't have Nvidia. (That shouldn't make a difference.)

On restarting chromium I could see the problem until I fullscreened the window - after which things worked regardless of the fullscreen/restored window state.

If chromium starts fullscreen, I don't see a problem, nor after it has been fullscreen at least once. (Fullscreen is typically how I use it, which explains why I've not noticed a problem and couldn't reproduce above.)

Could "randomly stop occurring" be after you fullscreen the browser window?</body></comment_new>
<comment_new><author>@AlanGriffiths</author><body>
The Chromium snap just updated to the latest build (3365) and I can no longer reproduce.</body></comment_new>
<comment_new><author>@AlanGriffiths</author><body>
Argh! On my main laptop I can no longer reproduce. On my "testing s...

</details>



<!-- START COPILOT CODING AGENT SUFFIX -->

- Fixes canonical/mir#4688

<!-- START COPILOT CODING AGENT TIPS -->
---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more [Copilot coding agent tips](https://gh.io/copilot-coding-agent-tips) in the docs.

[Saviq]

Fixed in #4698 instead