https://bugs.kde.org/show_bug.cgi?id=492506
Nicolai Weitkemper <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] | |m --- Comment #33 from Nicolai Weitkemper <[email protected]> --- I suspect I hit the same issue on 2026-09-28, and thus support reopening this report. ███ The following summary is mostly generated by GPT-6 Sol; I want to be maximally transparent about this and hope it contains relevant information. ███ My Wayland desktop froze and stopped accepting input, including Ctrl+Alt+Fn, while SSH continued to work. KWin's main thread consumed approximately one CPU core, and its D-Bus supportInformation call timed out. The selected task switcher was Sidebar (`[TabBox] LayoutName=sidebar`). I captured a live all-thread trace with eu-stack. The main thread was in this path (caller → callee): `TabBoxHandlerPrivate::show()` → QML binding update → `KWin::QPA::Window::map()` → expose/render → `QQuickWindowPrivate::polishItems()` → `QQuickLayout::updatePolish()` → implicit-size changes / layout rearrangement → `QGridLayoutEngine::setGeometries()` → label resizing / `QQuickTextPrivate::setupTextLayout()`. This looks consistent with the non-converging layout cycle in comment 32. The DRM commit thread and Mesa worker threads were waiting in the captured trace. I do not have a reliable reproducer or proof of which sizing binding triggers the loop. Environment: - NixOS, Wayland session, Intel graphics using i915 (Lenovo 21QDS60T00). - Running kernel at the time: 6.18.52. - The installed KWin systemd unit pointed to KWin 6.7.5 (version at the time of the crash is not 100% clear due to occasional use of `nixos-rebuild test`) - The installed Sidebar QML contained the earlier MR !663 changes. The exact QML loaded by the already-running process was not verified. I have saved the full thread trace, relevant journals, and the installed Sidebar QML, and will happily attach them or provide more details if useful. -- You are receiving this mail because: You are watching all bug changes.
