[FFmpeg-devel] [PATCH] oggparsevorbis: use av_realloc consistently

[Paweł Hajdan]

On Fri, Jan 11, 2013 at 9:12 PM, Michael Niedermayer <michaelni at gmx.at>wrote:

> yes, we need to do something about this, is there some way by which
> debugallocation can be detected ?
>

I'm not aware of such mechanism. By the way, it seems wrong to do a
different thing depending on whether debugallocation is present or not,
since then you'd be debugging a different program.


> It would also be interresting to know how many cases of
> posix_memalign + realloc() exist in the code.
>

At least counting what's covered by "make fate", I only needed to make one
other change:
http://ffmpeg.org/pipermail/ffmpeg-devel/2013-January/138150.html


> ATM i see as solutions
> A. detect debugallocation and do a malloc+memcpy+free for av_realloc
> B. change the code to never mix av_malloc* and av_realloc
> C. ignore it and declare debugallocation unsupported
> D. mark memory so the used allocator can be detected
>
> B is only practical if there are very few such cases
>

It seems B would be quite practical - just this patch and the one linked
above. There might be some more cases, but so far these two patches fix
things covered by Chromium test suite and "make fate".