Weston 8.0.0 does not appear to clean up overlay planes on switch-out
Switching between Mutter and Weston on different ttys, sessions from different users, under certain circumstances subsurfaces from Weston stick around in Mutter (which does not clear overlay planes itself so far).
Relevant IRC log:
robert.mader: yeah, me too. Fun fact: just opened obs in weston on tty3 and the right subsurface now stays on screen
even when I switch back to tt2 - like cross user, cross wayland session content leaking :/
I suppose a mesa gdm bug or so?
...
robert.mader: oh wait, weston supports hardware planes, right? That must be it. So most likely a weston bug....or a kernel bug?
robert.mader: I suppose it must be a kernel / intel driver bug
...
MrCooper: sounds like a bug in what's on tty2
or weston
but at the end of the day each KMS user has to make ensure the KMS state they want, can't rely on the previous one cleaning up after themselves
pq: MrCooper, indeed. OTOH, Weston definitely has code that calls drmModeSetPlane() to turn all overlays off on switch-away.
and cursor planes
would need to sprinkle a little more drm_debug() calls in weston and look at the drm debug log to see if weston actually leaves planes around
...
robert.mader: MrCooper Concerning KMS: on tty2 was my GS session - then mutter does not ensure it cleares all hardware planes IIUC?
MrCooper: sounds like it; or, since weston (which version are you using?) is supposed to disable overlay planes, there could be a kernel bug
robert.mader: Currently on weston-8.0.0-4.fc32
pq: I suppose with legacy KMS it is uncommon to go and reset the overlay planes you don't use on switch-in...Mutter issue: https://gitlab.gnome.org/GNOME/mutter/-/issues/1383
Edited by Robert Mader