[Eric] Eric4 issue with splitted views

Mon Mar 24 14:01:34 GMT 2014

Dear Phil,

I suffer from a long term Eric 4 problem, where eric crashes consistently on 
exit, when I split just one view. 

Prerequisites:
Running on KDE4 with PyQt4 available
Using tabbed view in Viewmanager

Reproduce:
Just running eric with a single file argument, right click into editor, choose 
"New view (with new split)" and exit reproduces this for me.

Results in:
Thread 1 (Thread 0x7fc162564700 (LWP 4712)):
[KCrash Handler]
#6  0x00007fc15d707aca in QWidgetPrivate::deleteTLSysExtra (this=this at entry=0x44e45f0) at kernel/qwidget_x11.cpp:2885
#7  0x00007fc15d6bea85 in QWidgetPrivate::deleteExtra (this=this at entry=0x44e45f0) at kernel/qwidget.cpp:1832
#8  0x00007fc15d6becad in QWidgetPrivate::~QWidgetPrivate (this=0x44e45f0, __in_chrg=<optimized out>) at kernel/qwidget.cpp:365
#9  0x00007fc15daa8e75 in ~QMenuPrivate (this=0x44e45f0, __in_chrg=<optimized out>) at widgets/qmenu_p.h:171
#10 QMenuPrivate::~QMenuPrivate (this=0x44e45f0, __in_chrg=<optimized out>) at widgets/qmenu_p.h:183
#11 0x00007fc15fc68c35 in cleanup (pointer=<optimized out>) at ../../src/corelib/tools/qscopedpointer.h:62
#12 ~QScopedPointer (this=0x44e4588, __in_chrg=<optimized out>) at ../../src/corelib/tools/qscopedpointer.h:100
#13 QObject::~QObject (this=0x44e4580, __in_chrg=<optimized out>) at kernel/qobject.cpp:816
#14 0x00007fc15d6c76ac in QWidget::~QWidget (this=0x44e4580, __in_chrg=<optimized out>) at kernel/qwidget.cpp:1554
#15 0x00007fc15e5bd989 in sipQMenu::~sipQMenu (this=0x44e4580, __in_chrg=<optimized out>) at sipQtGuiQMenu.cpp:316
#16 0x00007fc15e5bd9fe in release_QMenu (sipCppV=0x44e4580) at sipQtGuiQMenu.cpp:2785
#17 0x00007fc1604ba476 in forgetObject (sw=sw at entry=0x422c2b0) at siplib.c:10891
#18 0x00007fc1604bb599 in sipWrapper_dealloc (self=0x422c2b0) at siplib.c:10443
#19 0x00007fc16209aefa in subtype_dealloc (self=0x422c2b0) at Objects/typeobject.c:1030
#20 0x00007fc1620937e0 in dict_dealloc (mp=0x44f9d50) at Objects/dictobject.c:1010
#21 0x00007fc1620937b9 in dict_dealloc (mp=0x3d30f20) at Objects/dictobject.c:1010
#22 0x00007fc1604b86a7 in sipSimpleWrapper_clear (self=self at entry=0x3c11b00) at siplib.c:10016
#23 0x00007fc1604bb4c4 in sipWrapper_clear (self=0x3c11b00) at siplib.c:10402
#24 0x00007fc16207b232 in delete_garbage (old=0x7fc162370260 <generations+96>, collectable=0x7fff271e2780) at Modules/gcmodule.c:820
#25 collect (generation=2) at Modules/gcmodule.c:984
#26 0x00007fc16207b710 in PyGC_Collect () at Modules/gcmodule.c:1440
#27 0x00007fc1620727b9 in Py_Finalize () at Python/pythonrun.c:449
#28 0x00007fc162072255 in Py_Exit (sts=0) at Python/pythonrun.c:1777
#29 0x00007fc1620723a4 in handle_system_exit () at Python/pythonrun.c:1151
#30 0x00007fc1620723d1 in PyErr_PrintEx (set_sys_last_vars=set_sys_last_vars at entry=1) at Python/pythonrun.c:1161
#31 0x00007fc162072691 in PyErr_Print () at Python/pythonrun.c:1064
#32 0x00007fc162072bf9 in PyRun_SimpleFileExFlags (fp=fp at entry=0x83e4e0, filename=<optimized out>, filename at entry=0x7fff271e4f42 "/usr/lib64/python2.7/site-packages/eric4/eric4.py", closeit=closeit at entry=1, flags=flags at entry=0x7fff271e29e0) at Python/pythonrun.c:952
#33 0x00007fc16207312c in PyRun_AnyFileExFlags (fp=fp at entry=0x83e4e0, filename=filename at entry=0x7fff271e4f42 "/usr/lib64/python2.7/site-packages/eric4/eric4.py", closeit=closeit at entry=1, flags=flags at entry=0x7fff271e29e0) at Python/pythonrun.c:752
#34 0x00007fc16207a81e in Py_Main (argc=<optimized out>, argv=0x7fff271e2b98) at Modules/main.c:640
#35 0x00007fc161a31be5 in __libc_start_main () from /lib64/libc.so.6
#36 0x0000000000400791 in _start ()


As you can see, this crash is a typical case of "dtor called" after python 
exited. Of course, there isn't much, you can do about this. You even provide
sip.setdestroyonexit() for such issues. But that's exactly my problem here.

With this patch applied, shouldn't such an issue not happen anymore:

--- eric4.py.orig       2014-01-07 07:46:19.000000000 +0100
+++ eric4.py    2014-03-24 10:07:04.223139840 +0100
@@ -17,6 +17,7 @@ import traceback
 import cStringIO
 import time
 import logging
+import sip
 
 from PyQt4.QtCore import QTextCodec, SIGNAL, SLOT, qWarning, \
     QLibraryInfo, QTimer
@@ -162,7 +163,12 @@ def main():
     global args, mainWindow, splash, restartArgs
     
     sys.excepthook = excepthook
-    
+
+    try:
+        sip.setdestroyonexit(True)
+    except AttributeError:
+        pass
+
     options = [\
         ("--config=configDir", 
          "use the given directory as the one containing the config files"), 


I checked it also without exception handling, just in case, but problem
persists. Since I always use splitted views, and don't want to miss eric's 
KDE integration, I would like to see this fixed.

I've checked eric's source for suspect operations during the split, but 
couldn't spot any fishy operations. (ViewManager/ViewManager.py, 
Plugins/ViewManagerPlugins/Tabview/Tabview.py). 

Given the environmental conditions, I suspect some timing related issue, 
which might explain, why Detlev cannot reproduce it. 

Is there something wrong with my understanding of sip.setdestroyonexit, or 
do you have an idea, what could lead to this issue?

Thanks in advance,
Pete

System: 
openSUSE 13.1/x86_64
Qt: 4.8.5
KDE: 4.11.5
Python: 2.7.6
sip: 4.15.1
PyQt4: 4.10.2
eric4: 4.5.20
-------------- next part --------------
A non-text attachment was scrubbed...
Name: eric-setdestroyonexit.diff
Type: text/x-patch
Size: 638 bytes
Desc: not available
URL: <http://www.riverbankcomputing.com/pipermail/eric/attachments/20140324/677bb148/attachment.bin>
