Re: [Swig-user] python/swig complaining about a memory leak when there is none

[Danny C. <dan...@gm...>]

Matthias,

I still get the same  false warning for the simple testcase.

However, some other code that essentially does the same thing (returns
a heap allocated array) produces weird results.

I have %typemap(newfree) u_char * "delete [] ($1);";
defined.

If I use %newobject obj::get_uchar;
I dont get the error message, however, the memory foot print keeps
increasing, indicating that the delete is not called.

And If I use %newobject get_uchar; w/o the class qualifier
I get the error message saying that a mem leak has been detected. But
the memory foot print does not increase, indicating NO leak.

I dont know what gives? Any insight would be helpful.

thanks,

On Jan 29, 2008 7:13 PM, Nitro <ni...@dr...> wrote:
> Am 29.01.2008, 23:34 Uhr, schrieb Danny Crasto <dan...@gm...>:
>
> > I have a C++ member function that returns a heap allocated array
> > pointer u_char*
> >
> > u_char* obj::get_uchar(int size){ return new u_char[size]}
> >
> > I specify in the interface file
> > %typemap(newfree) u_char * "delete [] ($1);";
> > %newobject get_uchar;
> >
> > I call this function in an infinite loop. The memory usage (in top) is
> > constant. So there is no leak. However, python says
> > "swig/python detected a memory leak of type 'u_char *', no destructor
> > found"
> >
> > how do I get rid of this message? Or am I doing something wrong?
>
> Not sure, doesn't this have to be
>
> %newobject obj::get_uchar;
>
> ?
>
> -Matthias
>

Matthias,

I get the same thing either way.


-- 
- danny
"for(i=0;i<1000000;i++) {;} //speed up loop, drop a zero"