sightings of crash in createPythonQtClassWrapper

[erno]

Matti at Chiru ran into this after setting up a new 12.10 machine and there was earlier
a mention on #realxtend-dev about it with 2 sightings reported.

Loading plugin 'PythonScriptModule'.

Program received signal SIGSEGV, Segmentation fault.
0xa7bab569 in PythonQtPrivate::createPythonQtClassWrapper(PythonQtClassInfo_,
char const_, object) ()
from /rex/naali-deps/install/lib/libPythonQt.so.1
(gdb) backtrace
0 0xa7bab569 in
PythonQtPrivate::createPythonQtClassWrapper(PythonQtClassInfo_, char
const_, object) ()
from /rex/naali-deps/install/lib/libPythonQt.so.1
1 0xa7bace81 in PythonQtPrivate::registerCPPClass(char const_, char
const_, char const_, QObject_ ()(), void ()(void_,
PythonQtInstanceWrapperStruct_), object, int) () from
/rex/naali-deps/install/lib/libPythonQt.so.1
2 0xa7c30886 in PythonQt_init_QtCoreBuiltin(object) ()
from /rex/naali-deps/install/lib/libPythonQt.so.1
3 0xa7baef67 in PythonQt::init(int, QByteArray const&) ()
from /rex/naali-deps/install/lib/libPythonQt.so.1
4 0xa80152ea in PythonScript::PythonScriptModule::StartPythonQt (
this=this@entry=0x8396dd8)
at /rex/naali/src/Application/PythonScriptModule/PythonScriptModule
5 0xa80163a2 in PythonScript::PythonScriptModule::PythonScriptModule (
this=0x8396dd8)
at /rex/naali/src/Application/PythonScriptModule/PythonScriptModule
6 0xa8016423 in TundraPluginMain (fw=0x8050278)
at /rex/naali/src/Application/PythonScriptModule/PythonScriptModule
7 0xb7df01fe in PluginAPI::LoadPlugin (this=0x81a4cf0, filename=...)
at /rex/naali/src/Core/TundraCore/Framework/PluginAPI.cpp:119

[erno]

I'm seeing this now after updating a laptop to 12.10 (64-bit).

[erno]

I tried git bisect on this but the build was broken in most places in recent history.

[antont]

due to deps changing or? mostly it has been working in history otherwise IIRC, at least we've been many using it all the time?

when i tried to bisect the recent particle system loading bug, noticed that wasn't easy as builds from different points in history didn't just work, i was thinking it was due to deps changes.

[erno]

a couple of bisect points had compile errors in PythonScriptingModule and one failed in a bullet-related
compile error that sounded like a deps mismatch.

[antont]

I made these minimal disables to track the crash a bit, this way the module seems to run on 12.10 - didn't verify yet how much it actually works, possibly not much at all for the qt accessing, but at least doesn't crash.

antont@a4ca32f

the disabling is not minimal, possibly calling the py c api init would not make it crash.

I wonder if this has to do with how we've been using both the py c api init call and the pythonqt init mechanism.

[Stinkfist0]

Closing and migrating this issue to realXtend/TundraAddons#4