Created attachment 150802 [details] Kwin Debug Logs SUMMARY On master my main monitor stopped working, I tried enabling it with kscreen-doctor, but kwin logs say it just get's disabled again instantly. With git-bisect i have narrowed it down to this commit https://invent.kde.org/plasma/kwin/-/commit/f50547de1ee1e892dbaa6e020db3818db4333e14 STEPS TO REPRODUCE 1. Have a displayport monitor connected to your (NVIDIA?) GPU 2. Launch the Plasma Wayland session OBSERVED RESULT The monitor does not display anything EXPECTED RESULT The monitor displays something and receives a signal SOFTWARE/OS VERSIONS Linux/KDE Plasma: Arch Linux (5.18.12) KDE Plasma Version: 5.25.80 KDE Frameworks Version: 5.97.0 Qt Version: 5.15.5
Comment on attachment 150802 [details] Kwin Debug Logs a
Comment on attachment 150802 [details] Kwin Debug Logs 3e4yh34hthe4rhteherhreher
Created attachment 150803 [details] KWIN Debug Log
The content of attachment 150802 [details] has been deleted for the following reason: Requested by submitter
I don't see anything that would cause the output to be disabled like that but as KWin crashes afterwards, that's most likely related. Can you provide a backtrace with debug symbols?
Created attachment 150836 [details] KWIN Crash Backtrace Here you go (took a bit since I had written RelWithDebug instead of RelWithDebInfo)
It's crashing in qdebug, that's too late for getting anything useful out of it. Can you try running kwin with valgrind? It'll likely be super slow, but it finds out exactly where things go wrong.
Created attachment 150848 [details] KWIN Valgrind Output It seems to not crash anymore when i run with valgrind, nor kwin standalone, here is valgrind with kwin though, still has a few errors Sidenote: I couldn't run kwin or plasma from a tty, but i just ran a script that uses valgrind with my display manager instead
Found a workaround at least, if I disable one of my other monitors first (doesn't matter which one) and then enable the main monitor, and then enable the other monitor, all with kscreen-doctor, it works again.
Created attachment 150859 [details] KWIN Crash Backtrace for !2711
Created attachment 150865 [details] KWIN Crash Backtrace for !2711
Created attachment 150867 [details] KWIN Valgrind Output for !2711
Git commit 5aef41663ab7da5b02e3c5871adb8b39213778f9 by Vlad Zahorodnii. Committed on 24/07/2022 at 14:20. Pushed by vladz into branch 'master'. backends/drm: Fix potential use-after-free When the output is removed, there are legit cases when the connector can be still accessed. M +1 -1 src/backends/drm/drm_gpu.cpp https://invent.kde.org/plasma/kwin/commit/5aef41663ab7da5b02e3c5871adb8b39213778f9
Git commit 96f3a5eec287ac81e84ae0442d2f8a1945471420 by Vlad Zahorodnii. Committed on 24/07/2022 at 14:20. Pushed by vladz into branch 'master'. backends/drm: Fix dangling connectors in m_allObjects We add connectors to m_allObjects but never remove them. M +2 -0 src/backends/drm/drm_gpu.cpp https://invent.kde.org/plasma/kwin/commit/96f3a5eec287ac81e84ae0442d2f8a1945471420
*** Bug 457519 has been marked as a duplicate of this bug. ***
In progress with https://invent.kde.org/plasma/kwin/-/merge_requests/2824!.
Git commit 506e2d6f9f677b9975068b5d425e49273b2dd14c by Xaver Hugl, on behalf of Janet Blackquill. Committed on 16/08/2022 at 13:04. Pushed by zamundaaa into branch 'master'. drm_gpu: don't bother trying to assign a crtc to a connector that isn't connected M +1 -1 src/backends/drm/drm_gpu.cpp https://invent.kde.org/plasma/kwin/commit/506e2d6f9f677b9975068b5d425e49273b2dd14c
Git commit ac232be9681d53bd09cded87313c1c0b0a4d7214 by Xaver Hugl, on behalf of Janet Blackquill. Committed on 16/08/2022 at 13:56. Pushed by zamundaaa into branch 'Plasma/5.25'. drm_gpu: don't bother trying to assign a crtc to a connector that isn't connected (cherry picked from commit 506e2d6f9f677b9975068b5d425e49273b2dd14c) M +1 -1 src/backends/drm/drm_gpu.cpp https://invent.kde.org/plasma/kwin/commit/ac232be9681d53bd09cded87313c1c0b0a4d7214