https://bugs.kde.org/show_bug.cgi?id=524048
--- Comment #2 from Wehrwolfmann <[email protected]> --- DRAFT for bugs.kde.org/524048 — NOT SENT. Prepared 2026-08-18 by Wehrwolfmann. ---8<--- comment text starts here ---8<--- Thanks for the pointer to bug 524392 — same root state, and it adds what my report was missing. That reporter got there by switching sessions (kwin_x11 -> Cinnamon -> kwin_wayland), not by running a nested compositor. So the stale display environment is not exotic; an ordinary session switch reaches it. Their variables are only half wrong (DISPLAY=:1 while WAYLAND_DISPLAY=wayland-0), which is already enough to abort a Qt GUI process, and graphical-session.target is active throughout. The difference between the two reports is worth spelling out: baloo_file_extractor just crashes over and over because something keeps restarting it. drkonqi is the only component that feeds its own core back into itself, so it is the one that turns a handful of crashes into hundreds. Stale environment is the trigger; self-recursion is what makes it unbounded. I re-checked master today (6.7.80): all three points from comment 0 stand unchanged. The processor forwards every matching dump without looking at the executable, tryDrkonqi() has no self-check, main.cpp:248 still builds a QGuiApplication, and the socket still has TriggerLimitBurst=0 with no StartLimitIntervalSec/StartLimitBurst on the service. Only the Requisite= line from 522477 landed. Also worth flagging: bug 517353 was closed FIXED on 2026-08-07, but the QGuiApplication promotion it introduced is still in place and the FATAL_ERROR reminder in the top-level CMakeLists still points at that now-closed bug. My machine has been clean since I fixed the environment by hand (drkonqi 6.7.4, Qt 6.11.1, systemd 261.2, Wayland). With a sane environment the pipeline is fine — a kioworker SIGSEGV on 18 Aug produced exactly one launcher instance, which ran drkonqi and exited normally. Only the stale case has no brake. Could this move off UNCONFIRMED? Two machines, two distributions, two different routes into the same state. ---8<--- comment text ends here ---8<--- -- You are receiving this mail because: You are watching all bug changes.
