[Metakit] [Fwd: memory leaks in python interface [was Hash view
ralf at brainbot.com
Sat Mar 4 19:49:59 CET 2006
Ralf Schmitt schrieb:
> I'm pretty much convinced that the python wrapper code contains even
> more leaks...
> Maybe I'll do some more tests...
Turns out that the patch I sent may trigger segfaults .
The problem is that the PWOMappingMmbr class does not have a copy
constructor and consequently just copies memory instead of counting
references. C++ is such a pig. I didn't notice this the first time,
since I think the compiler may or may not use a copy constructor for
statements such as
PWOMappingMmbr res = PWOMappingMmbr(rslt, *this, _key);
depending on the compiler and optimizations being used.
In case it uses the "memcpy copy constructor", the _key member is being
decrefed one too many times.
I'll send another patch.
More information about the Metakit