[FFmpeg-devel] MLP Patch GSoC

[Ian Caulfield]

On Sun, Mar 16, 2008 at 8:37 PM, Justin Ruggles <justinruggles at bellsouth.net>
wrote:

> Michael Niedermayer wrote:
> > On Sun, Mar 16, 2008 at 01:10:02AM -0400, andy g wrote:
> >> I'm reviewing the MLP patch now and I realized that I'm a little
> >> confused on what I should be looking for/fixing. Is the patch
> >> functional at this moment and simply below documentation or coding
> >> standards, or is it non-functional? Thanks.
> >
> > I assume it is functional but i did not test it. Anyway you should
> definitly
> > test your cleaned up version before submiting to ensure it is functional
> and
> > that you didnt break anything.
>
> It is functional as far as I know. It just needs to be tested for
> stability/security issues.  I think it's the same problem as the AC3
> decoder where the coded values cause overreading input and/or
> overwriting output buffers.  It needs to be tested with all the largest
> possible values to make sure all the right checks are in place.  Using a
> damaged stream and testing with valgrind will catch most of these errors.
>

Yep, that's basically it - a couple of other projects have been including
the decoder and I haven't had any bug reports for a while, so I'm confident
about the decoder actually decoding properly (at least for streams actually
in the wild - there's probably some corner cases that I haven't found yet).
The main issue Michael had last review cycle was checking for security holes
etc when the decoder is given broken input. I started to have a look at
this, but haven't had the time to test it rigorously...

Ian