[datingesp-hash]

Fixes the second problem in #136: after screens: from a text moves a surface to another monitor, its first layer can stay on the monitor it left. It stays mapped, frozen on its last frame, and does not take input. It looks as if the scene is on both monitors at once.

Cause

configured() stores a clone of the first configured layer as the popups' parent ("while the mouse hasn't been seen on any, popups hang from the first one"). That clone is only replaced when the pointer enters another surface. rescreen drops the layer everywhere else, but this clone keeps it alive. In a launcher driven by a controller, with the pointer hidden, it stayed there until the scene was closed.

Change

forget_parent(wl): when a layer goes (rescreen, remove_from, surface_gone), it stops being the popups' parent. The next layer configured takes that place, as at start.

Tested

Hyprland 0.57, three 1920×1080 monitors. Layers of the scene's pid from hyprctl layers -j, sampled every 20 ms:

	slow moves	12 moves at 150 ms
0.7.22	the first layer stays (DP-1/20cf0 beside each new one)	2 layers in 135 of 153 samples
0.7.22 + this	one layer each time	never 2 (0 of 151)

• The same with GlassLauncher, a full-screen launcher moved with screens: on_screen: 22 moves between two monitors (16 at 150 ms), 0 of 315 samples with two layers. Then tried by hand switching fast with the controller, and it no longer happens.
• cargo test --release: 193 passed, 2 ignored.
• I have not tested popups themselves after a move. They should hang from the next layer configured, as they do before the pointer is first seen.

🤖 Generated with Claude Code

[k4ditano]

Thanks, merged. A layer that leaves should never stay the popups' parent, and the fix is right where it belongs. I checked that a missing parent only skips a popup until the next layer configured takes the place, and ran the two-monitor move bench on it: same moves, the scene kept its state. It's out in 0.7.23.