https://bugs.kde.org/show_bug.cgi?id=348414
--- Comment #70 from A. Wilcox (awilfox) <awil...@adelielinux.org> --- (In reply to RJVB from comment #69) > So I think we may be looking at something akin to GUI widgets being deleted > while events are still pending (or can still be generated for them) because > of `delete widget;` instead using `widget->deleteLater();`. It's weird that > this would be required for alerts with an asynchronous nature (which sound > alerts have by definition if they play through a sound daemon?), but I > presume this could be a side-effect of providing an application's widget as > the parent for a QObject inherited by a (hypothetical) KNotifications > instance. > > I hope I'm not rambling too much here, but this would explain the random > nature of the crash: it depends on system load and to what extent the sound > alert can be processed sufficiently to be able to cope with its > infrastructure being taken down under its feet. With the dolphin crash I > just had something like this was clearly going on: the laptop had just been > woken from sleep, no sounds had been played yet, it was updating the locate > and/or apt databases (etc), so pulseaudio must have taken just long enough > to be swapped in to allow me to accept the warning message before the alert > sound had even started to play. The system that I experienced my most recent crash on is used as a buildbox and was building Firefox when the crash happened. System load was high and my wait time was low; I hit Alt+F4 and immediately Alt+D to 'D'iscard the unsaved changes in KWrite. I'm not sure the dialogue even popped. > Another potentially relevant observation: I use the VLC phonon backend > (latest version, git master/head). I'm not running -HEAD, but I am running Phonon-VLC as well - version 0.9.0 here. -- You are receiving this mail because: You are watching all bug changes.
