| Summary: | Plasma crashes at logout or shutdown | ||
|---|---|---|---|
| Product: | [Unmaintained] plasma4 | Reporter: | Johannes von Scheidt <linux.vonscheidt> |
| Component: | general | Assignee: | Plasma Bugs List <plasma-bugs> |
| Status: | RESOLVED NOT A BUG | ||
| Severity: | crash | CC: | a.m.p.boelens, ahanssen, alexander.vodomerov, aseigo, auxsvr, eric.alber, fhimpe, gordon.holtslander, kde-bugs, linuxhippy, moabi2000, Ronny.Standtke, sbuehne, theOnce2001 |
| Priority: | NOR | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Platform: | Compiled Sources | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Sentry Crash Report: | |||
| Attachments: | backtrace | ||
|
Description
Johannes von Scheidt
2008-06-09 14:11:04 UTC
Did you update your KDE version and do you still have it? If it's still happening, can you please install the debug packages for your distribution and get the backtrace again so we can use it. Thanks in advance! Johannes, any news on that report? Is the crash still present and can you get a better backtrace please? http://techbase.kde.org/Development/Tutorials/Debugging/How_to_create_useful_crash_reports *** Bug 165675 has been marked as a duplicate of this bug. *** Annma : for the complete backtrace : bug 164861 #5 (Sebsauer also posted a testcase) re comment #4; seems I was only able to reproduce that with Mandriva Cooker. At my other systems - Arch, Kubuntu and openSuse - I failed there with the steps I described at http://bugs.kde.org/show_bug.cgi?id=164861#c6 :-/ Since the other report says that it did happen with debian, it also doesn't seem to be distributors fault but just hard to trigger/reproduce though at Cooker it's 100% reproducable for me with those steps described. Created attachment 25848 [details]
backtrace
and now it did happen also with a checkout from some hours ago on !=Cooker.
The reason seems to be a QEvent::FocusOut that got send on destruction.
and the actual crash happens on;
bool QGraphicsWidget::event(QEvent *event)
{
Q_D(QGraphicsWidget);
// Forward the event to the layout first.
if (d->layout)
d->layout->widgetEvent(event); //<==CRASH
...
}
this is a Qt problem. on deleting a layout it doesn't reset the layout pointer in the parent item. this really must be taken care of within QGraphicsWidget and/or QGraphicsLayout because there are situations, such as on destruction, that there is no rational opportunity to reset the layout but, due to being QObjects, may be automatically deleted. it's either fix it in Qt, or every QGraphicsWidget subclass will have tto do setLayout(0) in its dtor. Andreas: am i reading your code right here? *** Bug 164861 has been marked as a duplicate of this bug. *** *** Bug 165830 has been marked as a duplicate of this bug. *** *** Bug 166101 has been marked as a duplicate of this bug. *** *** Bug 165021 has been marked as a duplicate of this bug. *** *** Bug 166159 has been marked as a duplicate of this bug. *** *** Bug 166333 has been marked as a duplicate of this bug. *** *** Bug 166528 has been marked as a duplicate of this bug. *** I left for vacation on the 5th ;-). Sorry about that. Yes, this is a serious bug in QGraphicsLayout that must be fixed in Qt. The only workaround I can see is to completely recreate the layout without reusing the parent at all. Was this also posted to our bug tracker btw? Andreas On Sat July 5 2008 07:04:31 Aaron J.Seigo wrote: [bugs.kde.org quoted mail] *** Bug 167253 has been marked as a duplicate of this bug. *** I just had the same crash on Kubuntu-8.04.1 with final KDE-4.1 release when logging out from the session. If plasma crashes, all settings of my panel are reset to an empty state next time I login. This empty state is also problematic because plasmoids added to it are either tiny (2 pixels in size) or huge (a clock 400 pixels high :s) and make plasma crash a lot while adding plasmoids or icons to start an application. That plasma crashes is one thing, but that you need to create all your panel settings again is a much bigger problem! Settings shouldn't be reset by the crash. Can it be fixed as a separate fix that your panel settings (including the position of all icons to launch applications and all enabled plasmoids) will not be reset if plasma crashes? That will solve a huge side effect of the crash! upstream bug; upstream is aware of it. @Aardwolf: settings in plasma are flushed to disk 30s after the last change; we may end up shortening that down to something like 10s, but the purpose is to prevent random synchronizations that end up eating more battery life as well as sync'ing in between sets of changes leaving the config in a potentially uncoordinated state. Can someone put this into upstream bugzilla and post a link here? I want to be sure that upstream is really aware of it and the bug won't be lost forever. |