QGIS » QGIS Application

I did not add any layer. simply opening and closing QGIS would produce this. Tested on Ubuntu 12.04

Do you have any plugins enabled? Especially an open plugin dock or window would explain something.

Confirmed here also with the qgis --noplugins command

here the stacktrace with gdb:

Program received signal SIGSEGV, Segmentation fault.
new_threadstate (init=1, interp=0x0) at ../Python/pystate.c:199
199    ../Python/pystate.c: File o directory non esistente.
(gdb) bt full
#0  new_threadstate (init=1, interp=0x0) at ../Python/pystate.c:199
        tstate = 0x2bb3fe0
#1  PyThreadState_New (interp=0x0) at ../Python/pystate.c:210
No locals.
#2  0x00007fffd1d8955c in PyGILState_Ensure () at ../Python/pystate.c:597
        current = <optimized out>
        tcur = 0x0
#3  0x00007fffd173ed56 in ?? () from /usr/lib/python2.7/dist-packages/PyQt4/QtCore.so
No symbol table info available.
#4  0x00007fffd0fa457e in ?? () from /usr/lib/python2.7/dist-packages/PyQt4/QtGui.so
No symbol table info available.
#5  0x00007ffff288ec27 in QApplication::palette(QWidget const*) ()
   from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
No symbol table info available.
#6  0x00007ffff28cfe0d in QWidgetPrivate::naturalWidgetPalette(unsigned int) const ()
   from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
No symbol table info available.
#7  0x00007ffff28dbb95 in QWidget::setPalette(QPalette const&) ()
   from /usr/lib/x86_64-linux-gnu/libQtGui.so.4
No symbol table info available.
#8  0x00007ffff2bb237d in ?? () from /usr/lib/x86_64-linux-gnu/libQtGui.so.4

Commenting out the lines 799 and 800 in qgisapp.cpp does not produce the crash.

We can comment them again, that's not a big deal, the OS will clean up (at least the memory) on app exit anyway.

But I wonder, what is really causing that problem, because here python / pyqt cleanup and exit nice and clean.

What's left of python when running with --noplugins?
Only the python console and part of the plugin manager?

Matthias Kuhn wrote:

We can comment them again, that's not a big deal, the OS will clean up (at least the memory) on app exit anyway.

But I wonder, what is really causing that problem, because here python / pyqt cleanup and exit nice and clean.

What's left of python when running with --noplugins?
Only the python console and part of the plugin manager?

Today I'm getting a lot of

gio@sibirica ~ $ qgis
Warning: loading of qt translation failed [/usr/share/qt4/translations/qt_en_US]
Fatal: QGIS died on signal 11
Stacktrace (piped through c++filt):
Segmentation fault

when clicking on "new project"

This bug only concerns a crash when exiting qgis.

Matthias Kuhn wrote:

This bug only concerns a crash when exiting qgis.

Hi Matthias, right. I got it also on exit. Cheers!

Matthias Kuhn wrote:

What's left of python when running with --noplugins?
Only the python console and part of the plugin manager?

The Plugin Manager is fully loaded, so you can disable/uninstall/update crashing plugins.

In my case the crash only happens if Value Tool is loaded (or was loaded in the session, even if I unload it prior to closing). Can anybody confirm? See #8184

Borys Jurgiel wrote:

In my case the crash only happens if Value Tool is loaded (or was loaded in the session, even if I unload it prior to closing). Can anybody confirm? See #8184

Hi Borys,
here the crash is not dependent from plugins (I guess) as I run with --noplugins command.
I did try by installing Value Tool plugin (it was not installed yet on my machine) and I got the same stacktrace.

ouf of curiosity, which Qt version?

Salvatore Larosa wrote:

here the crash is not dependent from plugins (I guess) as I run with --noplugins command.

Ok, so it's something another.

ouf of curiosity, which Qt version?

Qt 4.8.4 + PyQt 4.10.1 @ Debian Jessie.

Hi,

Here the problem is also not dependent of plugins.

QGIS compiled against Qt 4.8.1 and Python-Qt 4.9.1.

Matthias Kuhn wrote:

Would you be able to debug the crash in gdb and then enter "py-bt" in the gdb console?
I hope this gives valuable additional information about this problem.

I already had tried that but unfortunately it doesn't give any additional info.

I tracked down the issue in the ZoomToCoordinates plugin:

The issue is related to creating a QMainWindow in init (Line 57) and then reparenting this window to iface.mainWindow() (Line 59), if instead ownership remains with the plugin ( pass None instead of iface.mainWindow() to setParent() ) there is no crash.

That it is in init is important, because this gets also executed with --noparent and if the plugin is disabled.

Therefore a question to the plugin department (borys?):
Why does this code need to be executed (I see it's executed once and destroyed almost immediately again when loading the plugins)?
If this code would not need to be executed, it would be much easier to find buggy plugins instead of suspecting QGIS of being responsible for the crash.
I would vote for closing this bug as soon as the QGIS application does not crash any more because of a buggy but disabled plugin.

Concerning this particular issue. There seems to be an ownership issue with PyQt, such that certain widgets get deleted but references to these remain present in other widgets. I guess, that this issue has been resolved in a recent version of PyQt or depend on a certain configuration because it does not happen on every system. So I don't see, what we could do from the QGIS side. (This paragraph is based on assumptions, I would be happy to be corrected by somebody with a better understanding of this problem)

Also interesting:

When the Python interpreter exits it garbage collects those objects that it can.
This means that any corresponding C++ instances and C structures owned by Python are destroyed.
Unfortunately this happens in an unpredictable order and so can cause memory faults within the wrapped library.
Calling this function with a value of False disables the automatic destruction of C++ instances and C structures.

sip.setdestroyonexit is unfortunately only available starting from SIP 4.14.2
Current OSGeo4W version is 4.12.1

One of the following package updates solves the problem, if I downgrade I get the crash also:

PyQt4-4.9.5 => PyQt4-4.10.1
sip-4.14.1 => sip-4.14.6

Hi Matthias,

I no longer have the crash. I think I can confirm the fix!

Thank you very much!