kdevelop crashes often -- valgrind reports oxygen invalid reads

[Sundberg Pauli]

Hello.

System: kubuntu - natty,
sundberg@hp-k64:~/eigenor/system-utilities$ uname -a
Linux hp-k64 2.6.37-12-generic #26-Ubuntu SMP Wed Jan 5 18:35:17 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

I am using kdevelop as developing ide. It crashes now to then. Example of crash can be found from here:
http://susundberg.pastebin.com/bFi9AgQ2

Some crashes report from possibly corrupted stack.

Kdevelop developers say that this is probably caused by non kdevelop issues. Running with valgrind produces invalid reads from what seems to be oxygen:
http://susundberg.pastebin.com/Vh2KeLTp

Running kdevelop with:
export QT_NO_GLIB=1
makes 'invalid reads' disapear. (QT_NO_GLIB) makes first invalid read of 4 to disapear:
http://susundberg.pastebin.com/tG4hjnNA

Running kdevelop with:
kdevelop -graphicssystem raster -style plastique
makes all invalid reads dissapear.

So to me it seems like there might be something wrong with oxygen style.

[Sundberg Pauli (susundberg)]

Just for the record: the crash happens even with:
kdevelop -graphicssystem raster -style plastique

http://susundberg.pastebin.com/w4EJAWju

and there you can find 'corrupted stack?' hints.

[Sundberg Pauli (susundberg)]

Another example crash: http://susundberg.pastebin.com/raugAAzh

[Bart Janssens (bart-janssens)]

Attaching a stack trace, using todays version of kdevelop git (kdevplatform cae5d37f9309564a3704db882518b84739760252 and kdevelop ebe44ee16d31929b9d76e0f0097c1c2f1f0ddca5).

Seems this always happens inside KDevelop::DUChainPrivate::CleanupThread::run as far as kdevelop is concerned, but the code there seems fine. Given how often this happens for me, and how few people are watching this, I'm wondering if this is some very platform specific bug (running on a pentium D here)? The crash ultimatly points to nptl/sysdeps/unix/sysv/linux/x86_64/pthread_cond_timedwait.S:216 which is in libc and is assembly for x86_64 CPUs.

[Sundberg Pauli (susundberg)]

I am using machine with "AMD Athlon(tm) 64 X2 Dual Core Processor 4000+"

In kdevelop irc channel there was short discussion if this would be QT-related problem and even bug was highlighted, but unfortunately i have lost the link. The bug was upstream bug related with memory fragmentation.

I must say i have encountered few crashes in other applications, and even X-systems hangs (system seems still be running properly, as pressing power button initiates clean shutdown), that would give hint of deeper problem.

Currently i am testing qtcreator -- it should probably crash also if problem is with QT.

[Sundberg Pauli (susundberg)]

At some library updates made this problem to go away. Now Kdevelop runs stable.