Summary: | systemsettings crash after viewing effects | ||
---|---|---|---|
Product: | qtatspi | Reporter: | Stéphane Gourichon <stephane_kde> |
Component: | general | Assignee: | Frederik Gladhorn <gladhorn> |
Status: | RESOLVED UNMAINTAINED | ||
Severity: | crash | CC: | cfeck, h.k.ghost, ifantastici4, kwin-bugs-null, nicoelyorugua, xplayer05 |
Priority: | NOR | ||
Version: | 0.3.1 | ||
Target Milestone: | --- | ||
Platform: | Ubuntu | ||
OS: | Linux | ||
Latest Commit: | Version Fixed In: | ||
Attachments: |
Valgrind log, annotated on test on fresh account no crash though invalid operations
Crash reproduced under valgrind control. Valgrind log: after purged libqspiaccessiblebridge.so . Bug no longer observed. |
Description
Stéphane Gourichon
2012-08-14 17:23:27 UTC
*** Bug 305843 has been marked as a duplicate of this bug. *** If the crash is reproducible, could you run system settings in valgrind, and add the valgrind log? For more information, see http://techbase.kde.org/Development/Tutorials/Debugging/How_to_create_useful_crash_reports#Retrieving_a_backtrace_with_Valgrind > #17 QAccessible::queryAccessibleInterface (object=0x28bde20) at accessible/qaccessible.cpp:593
> #18 0x00007f4b6172951d in QAccessible::updateAccessibility (o=<optimized out>, who=0, reason=QAccessible::ObjectDestroyed) at accessible/qaccessible_unix.cpp:102
The backtrace also indicates, that you may be use Qt with accessibility enabled. If this is the case, please ask in the forums of your distribution how to disable Qt accessibility, and report back, if this fixes the issue.
Created attachment 73537 [details]
Valgrind log, annotated on test on fresh account no crash though invalid operations
1) On test account : not reproduced, valgrind log provided anyway
I have a test account for bug reporting, which did not exhibit the bug. Typing systemsettings from command show this on stderr:
QDBusConnection: session D-Bus connection created before QCoreApplication. Application may misbehave.
QDBusConnection: session D-Bus connection created before QCoreApplication. Application may misbehave.
Still a valgrind log may be useful so I ran systemsettings again with this :
valgrind --leak-check=full --track-origins=yes --trace-children=yes systemsettings
2) Information on regular account
My regular account (used when seeing the crash) was never regularly used for full kde session.
Here's its timeline :
* account created, used with unity, gnome, then xfce for many sessions.
* no specific hack or anything that may interfere
* only KDE app used was probably k3b, never a full kde sesson (full kde not even installed)
* installed full kde (from Ubuntu standard repo) for a try, tried to login with kde, saw this bug, reported
I'll try to reproduce the bug on the regular account soon.
Thank you.
Please tell if valgrind log provided now is useful to anything.
No we'd require a valgrind log of what actually (probably) leads to the memory corruption. If it's reproducible for sure with the regular account but not at all with the test one, it's very likely that the critical code path is not entered on the latter, therefore doesn't show up in valgrind. It would be great, if you could provide a valgrind log of the actual crash. Also you might want to MOVE ~/.kde/share/config/* away and put back the settings one by one, seeing what file triggers the crash, but using valgrind is likely simpler ;-) Created attachment 73540 [details]
Crash reproduced under valgrind control.
Actions performed :
* login to kde session using regular account. Not changed accessibility (not found in preferece, but there is a "KDE accessible" icon on tray.
* Run : valgrind --leak-check=full --track-origins=yes --trace-children=yes systemsettings 2>&1 | tee valgrind_log_crash.txt
* Added annotations (lines starting with #).
* The part of valgrind log specific to the crash seems to be between lines 7225 and 7294.
* Notice that valgrind says, juste before the click that triggers crash :
==4677== More than 100 errors detected. Subsequent errors
==4677== will still be recorded, but in less detail than before.
Regards.
Thanks for the valgrind log. Please ask in an Ubuntu forum how you can disable Qt accessibility, and report back if this fixes the bug. *** Bug 300027 has been marked as a duplicate of this bug. *** *** Bug 297413 has been marked as a duplicate of this bug. *** *** Bug 299974 has been marked as a duplicate of this bug. *** > Please ask in an Ubuntu forum how you can disable Qt accessibility, and report back if this fixes the bug. Done http://askubuntu.com/questions/181815/how-to-disable-qt-accessibility-in-12-04 . I tested again with valgrind with two differences : * from my regular XFCE session instead of KDE session * after purging package "kaccessible". Bug still occurs. I can provide valgrind log if useful. If everything fails, check which package installs /usr/lib/x86_64-linux-gnu/qt4/plugins/accessiblebridge/libqspiaccessiblebridge.so and try to remove it. If you cannot, because of dependencies, you can manually move the file away as root, and restart the desktop. This should remove support for Qt accessibility. Created attachment 73546 [details] Valgrind log: after purged libqspiaccessiblebridge.so . Bug no longer observed. > check which package installs /usr/lib/x86_64-linux-gnu/qt4/plugins/accessiblebridge/libqspiaccessiblebridge.so $ dlocate /usr/lib/x86_64-linux-gnu/qt4/plugins/accessiblebridge/libqspiaccessiblebridge.so qt-at-spi:amd64: /usr/lib/x86_64-linux-gnu/qt4/plugins/accessiblebridge/libqspiaccessiblebridge.so $ sudo dpkg --purge qt-at-spi (Lecture de la base de données... 284271 fichiers et répertoires déjà installés.) Suppression de qt-at-spi ... (from my regular XFCE session) $ valgrind --leak-check=full --track-origins=yes --trace-children=yes systemsettings 2>&1 | tee valgrind_log_without_libqspiaccessiblebridge_so.txt $ grep q.*access valgrind_log_without_libqspiaccessiblebridge_so.txt || echo no match no match So, removing this package and library fixed the bug. Thank you, Stéphane. Reassigning to qtatspi developers. Hello! Sorry to be the bearer of bad news, but this project has been unmaintained for many years so I am closing this bug. |