[PD] fux_kinect

[Mathieu Bouchard]

Le 2011-11-11 à 19:52:00, tim vets a écrit :

> the 'where' output remains the same afaict...

Well, you're lucky, because it crashes really quickly after the bug :

> #7  0xb42514da in freenect_init () from /usr/local/lib/libfreenect.so.0.0
> #8  0xb43824cf in fux_kinect::fux_kinect (this=0x8612638, argc=0, 
>     argv=0xbfffec4c) at fux_kinect.cpp:84

This (and the Valgrind report that you didn't post) all point to the first 
half of the fux_kinect constructor. I just tried reading it for the first 
time. It contains the line :

   gl_frame_cond = (pthread_cond_t*) malloc(sizeof(pthread_mutex_t));

but the size of a pthread mutex is either 24 bytes or 40 bytes (32-bit vs 
64-bit), and the size of a pthread cond is 48 bytes (though some bytes 
might be unused in 32-bit mode). This causes memory corruption in the next 
line already :

     pthread_cond_init(gl_frame_cond, NULL);

So the fix is to use pthread_cond_t in the malloc call above.

But a better fix might be to not use malloc at all for those things, 
unless there's a good reason to use it. Less opportunities for bugs.

  ______________________________________________________________________
| Mathieu BOUCHARD ----- téléphone : +1.514.383.3801 ----- Montréal, QC