[Adonthell-devel] Memory leaks

[Kai Sterker]

Did some debugging with valgrind today. Discovered a few things that I
fixed, but there are still a lot of issues reported that seen to come
from Python. Reading some more about valgrind and Python, it seems
that the two do not play nicely together. More info here:

  http://svn.python.org/projects/python/trunk/Misc/README.valgrind

I might look further into this when I find the time, but for now I
think I'll concentrate on other things. Might be that some of the
issues are actually caused by SWIG. Stuff like this, for example:

==3000== 280 bytes in 1 blocks are possibly lost in loss record 538 of 760
==3000==    at 0x4C2B6CD: malloc (in
/usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==3000==    by 0x74E8578: _PyObject_GC_Malloc (in /usr/lib/libpython2.7.so.1.0)
==3000==    by 0x74E9BEC: _PyObject_GC_New (in /usr/lib/libpython2.7.so.1.0)
==3000==    by 0x74A9043: PyDict_New (in /usr/lib/libpython2.7.so.1.0)
==3000==    by 0x7F32F84: SWIG_Python_NewShadowInstance (py_runtime.cc:2462)
==3000==    by 0x7F331C3: SWIG_Python_NewPointerObj (py_runtime.cc:2599)
==3000==    by 0x7F33CE4: cxx_to_py (py_runtime.cc:3192)
==3000==    by 0x632B4D6: rpg::faction::create_instance(std::string
const&, _object*) (python.h:156)
==3000==    by 0x632C175: rpg::faction::get_state(base::flat&) (faction.cc:202)
==3000==    by 0x632C60C: rpg::faction::load() (faction.cc:128)
==3000==    by 0x5F03536: base::savegame::load(int const&) (callback.h:174)
==3000==    by 0x40348E: world_test::main() (worldtest.cc:148)

Perhaps there's some SWIG cleanup code that we do need to trigger. So
perhaps there's the chance to improve things even more without having
to compile your own Python :-).

Kai