https://bugs.kde.org/show_bug.cgi?id=527991
[email protected] changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #5 from [email protected] --- Created attachment 198114 --> https://bugs.kde.org/attachment.cgi?id=198114&action=edit Full symbolicated GDB backtrace (all threads), Arch Linux, Plasma 6.8.2, PID 615689 Confirming this on a different distribution and different hardware, with current Plasma 6.8.2. Environment: - Arch Linux, Wayland - Plasma 6.8.2 (PowerDevil 6.8.2, libkscreen 6.8.2, KWin 6.8.2), KF 6.31.0, Qt 6.12.0 - Kernel 7.2.14-arch1-1, Mesa 26.2.6 - AMD Phoenix1 iGPU (amdgpu) - Internal eDP-1 2880x1920@120 + two external 3840x2160@60 monitors (DP-9, DP-10) via a USB-C dock using DisplayPort MST with DSC (Synaptics MST hub) Reproduction pattern: The crash happens on resume from suspend while docked. When the MST outputs come back after "PM: suspend exit" (kernel logs "[drm] MST_DSC Send DSC enable to synaptics"), PowerDevil crashes in the same place within about a second. systemd restarts it, so the only visible effect is the DrKonqi notification. 4 identical crashes in 3 days (2026-11-03 to 2026-11-05). Crash location (6.8.2, so the line is 344 instead of 339): #6 QtWayland::kde_output_device_mode_v2::object (this=0x10) at qwayland-kde-output-device-v2.h:391 #7 KScreen::WaylandOutputDevice::setWlConfig (...) at backends/kwayland/waylandoutputdevice.cpp:344 #8 KScreen::WaylandConfig::applyConfig (...) at backends/kwayland/waylandconfig.cpp:223 #9 KScreen::WaylandBackend::setConfig (...) at backends/kwayland/waylandbackend.cpp:83 #10 KScreen::SetConfigOperation::start (...) at src/setconfigoperation.cpp:81 The faulting line is: wlConfig->mode(object(), deviceModeFromId(mode->id())->object()); deviceModeFromId() returns nullptr because the config being applied still references a mode id that no longer exists on the output device. After the MST outputs reconnect, KWin announces new kde_output_device_mode_v2 objects, so the old id is stale. The result is not checked before ->object() is called. A null check there (skip the mode change, or fail the operation cleanly) would avoid the crash. Full symbolicated backtrace of all threads (Arch debuginfod) is attached. -- You are receiving this mail because: You are watching all bug changes.
