Note: This is a public test instance of Red Hat Bugzilla. The data contained within is a snapshot of the live data so any changes you make will not be reflected in the production Bugzilla. Email is disabled so feel free to test any aspect of the site that you want. File any problems you find or give feedback at bugzilla.redhat.com.
Bug 1096442 - Segmentation fault in QEventDispatcherGlib::processEvents or JSC::BlockAllocator::blockFreeingThreadMain
Summary: Segmentation fault in QEventDispatcherGlib::processEvents or JSC::BlockAlloca...
Keywords:
Status: CLOSED NOTABUG
Alias: None
Product: Fedora
Classification: Fedora
Component: qtwebkit
Version: 20
Hardware: x86_64
OS: Linux
unspecified
unspecified
Target Milestone: ---
Assignee: Than Ngo
QA Contact: Fedora Extras Quality Assurance
URL: http://www.qtcentre.org/archive/index...
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2014-05-10 09:59 UTC by Raphael Groner
Modified: 2014-05-18 09:11 UTC (History)
11 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2014-05-17 21:05:31 UTC
Type: Bug
Embargoed:

Attachments (Terms of Use)
backtrace (63.70 KB, text/plain)
2014-05-10 09:59 UTC, Raphael Groner 		no flags Details
backtrace_2014-05-17 (87.91 KB, text/plain)
2014-05-17 17:45 UTC, Raphael Groner 		no flags Details
View All

Description Raphael Groner 2014-05-10 09:59:22 UTC 
Created attachment 894220 [details]
backtrace

Description of problem:
Program terminated with signal SIGSEGV, Segmentation fault.

Version-Release number of selected component (if applicable):
qt-4.8.6-2.fc20.x86_64
qtwebkit-2.3.3-7.fc20.x86_64

How reproducible:
maybe

Steps to Reproduce:
1. not sure what's going on
2. maybe show notification popup via dbus
3. start to type into a non-focussed input field

Actual results:
crash

Expected results:
no crash

Additional info:
Probably an upstream issue, and we need a small code snippet to reproduce the crash for sure.
Comment 1 Raphael Groner 2014-05-10 10:15:13 UTC 
It seems to happen cause of dirty code that is not thread-safe. 
See also
http://stackoverflow.com/questions/11859227/qt-qcoreapplicationexec-crash-on-ubuntu-12-04
http://www.qtcentre.org/archive/index.php/t-39379.html
Comment 2 Rex Dieter 2014-05-10 13:41:14 UTC 
The references you provide imply this is an application bug:

"SocketNotifier implementation is not thread safe. That means that you are not allowed to control it from different threads."

So, which appplication crashed here?
Comment 3 Raphael Groner 2014-05-10 18:43:55 UTC 
(In reply to Rex Dieter from comment #2)
> So, which appplication crashed here?

It's psi-plus. I use the package from Russian Fedora but I am not sure if it's related to a Qt component or a direct issue in the calling application.
Comment 4 Raphael Groner 2014-05-10 18:45:53 UTC 
I guess this bug happens only when the application is put under stress to let Qt handle too many events synchronously.
Comment 5 Rex Dieter 2014-05-11 01:40:10 UTC 
I'd suggest you file a bug against psi-plus about this.
Comment 6 Raphael Groner 2014-05-11 07:09:05 UTC 
Rex, thanks for the fast response. I got a crash as well with fancy (claws plugin to use webkit) and some javascript in JSC::BlockAllocator yesterday, but don't know how to reproduce. So I guess an issue with the latest webkit updates. Trying to reassing.
Comment 7 Rex Dieter 2014-05-11 18:21:04 UTC 
claws-mail uses webkitgtk
Comment 8 Raphael Groner 2014-05-17 17:45:08 UTC 
Created attachment 896628 [details]
backtrace_2014-05-17

segfault again
Comment 9 Raphael Groner 2014-05-17 17:53:45 UTC 
KDE says it may be a bug in QtWebKit. 
Notice the recent duplication marker of 2014-03-10 ...
https://bugs.kde.org/show_bug.cgi?id=305493

But maybe a bug with localization ...
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=741352
Comment 10 Rex Dieter 2014-05-17 20:44:22 UTC 
Feel free to ignore my suggestions  :-/

Fwiw, the bug you reference, the backtrace looks very different to the one here, afaict.
Comment 11 Rex Dieter 2014-05-17 21:05:31 UTC 
Fyi, here's what appears to be the critical part of the posted backtrace, psidbusnotifier.cpp seems to be passing an invalid object to QObject::disconnect

No qtwebkit or glib involved.

Once psi-plus passes review, we could consider reassigning this there, but until then, closing->notabug

Thread 1 (Thread 0x7f53bd5b48c0 (LWP 1886)):
#0  0x0000000000000060 in ?? ()
No symbol table info available.
#1  0x00007f53bdec2a3b in QObject::disconnect (sender=0x6352ea0, signal=0x6340d19 "destroyed()", signal@entry=0xa05850 "2destroyed()", receiver=receiver@entry=0x7b64040, method=0x7017979 "eventDestroyed()", method@entry=0xa26b2d "1eventDestroyed()") at kernel/qobject.cpp:2911
        method_name = {static shared_null = {ref = {_q_value = 1}, alloc = 0, size = 0, data = 0xe64898 "", array = ""}, static shared_empty = {ref = {_q_value = 22}, alloc = 0, size = 0, data = 0x7f53be215798 <QByteArray::shared_empty+24> "", array = ""}, d = 0x7017960}
        method_arg = 0xa26b2d "1eventDestroyed()"
        membcode = <optimized out>
        method_found = false
        signal_arg = 0xa05850 "2destroyed()"
        signal_name = {static shared_null = {ref = {_q_value = 1}, alloc = 0, size = 0, data = 0xe64898 "", array = ""}, static shared_empty = {ref = {_q_value = 22}, alloc = 0, size = 0, data = 0x7f53be215798 <QByteArray::shared_empty+24> "", array = ""}, d = 0x6340d00}
        signal_found = false
        res = false
        smeta = 0x7fff11c4dd00
#2  0x000000000086feb3 in PsiDBusNotifier::readyToDie (this=this@entry=0x7b64040) at psidbusnotifier.cpp:406
No locals.
#3  0x000000000086ffad in PsiDBusNotifier::popupClosed (this=0x7b64040, id=<optimized out>, reason=<optimized out>) at psidbusnotifier.cpp:385
Comment 12 Raphael Groner 2014-05-18 09:11:41 UTC 
Rex,

sorry for the confusion. I did not find the direct relation to Psi in the backtrace. Thanks for your analysis.

I have asked upstream in the psi-dev jabber room with this bug report and attached backtrace, but ri0n could not detect what's going on.

Since this bug is happening very seldomly, I think NOTABUG is okay for now.
Note You need to log in before you can comment on or make changes to this bug.