https://bugs.kde.org/show_bug.cgi?id=483265
--- Comment #14 from Kacper Paczos <[email protected]> --- Created attachment 195831 --> https://bugs.kde.org/attachment.cgi?id=195831&action=edit coredumpctl report: findModal() infinite recursion stack trace (Plasma 6.7.4, Wayland, Fedora 44) This crash just hit me on Plasma 6.7.4 (Wayland, Fedora 44) — two and a half years after this report was filed. kwin_wayland died with SIGSEGV, and the stack trace is one single frame repeating over and over: KWin::Window::findModal(), all the way to stack overflow. Full coredumpctl report attached. No Java involved in my case. Half a minute before the crash I was clicking through the settings window of a small GTK tray app (a dictation tool) — and nothing else was opening windows at the time. So it looks like any toolkit can produce a loop in the transient chain, not just AWT. One thing struck me: my log has no "caused WM_TRANSIENT_FOR loop" line at all, the one the original reporter saw. That detection seems to live only in the X11Window code — while findModal() itself walks the transients with no cycle protection whatsoever. So on Wayland any client that creates such a cycle takes the compositor down silently, with no trace in the log. Small consolation: crash recovery worked and the session picked itself up — Qt apps reconnected to the new compositor. -- You are receiving this mail because: You are watching all bug changes.
