[FFmpeg-devel] [PATCH] fix H.264 crash (issue 706)
Fri Apr 10 12:18:03 CEST 2009
On Fri, Apr 10, 2009 at 09:17:43AM +0200, Reimar D?ffinger wrote:
> On Fri, Apr 10, 2009 at 12:11:52AM +0200, Michael Niedermayer wrote:
> > On Thu, Apr 09, 2009 at 11:31:46PM +0200, Reimar D?ffinger wrote:
> > > On Thu, Apr 09, 2009 at 10:42:37PM +0200, Michael Niedermayer wrote:
> > > > On Thu, Apr 09, 2009 at 10:33:49PM +0200, Reimar D?ffinger wrote:
> > > > > Well, but that code will only run if there is either either a DC or a MV
> > > > > error (see above), whereas the problematic code will only run if there
> > > > > is no MV error. So if there is only an AC error I think the code I added
> > > > > above would be necessary I think...
> > > >
> > > > short awnser
> > > > our h264 decoder doesnt set AC_ERROR without MV_ERROR
> > >
> > > But that code is not used only by H.264...
> > > Also, I sure don't think at all that more undocumented, obscure, not
> > > generally-correct and inconsistenly applied assumptions is really what
> > > any of the MPEG-related code needs.
> > >
> > > > the code is called "/* set unknown mb-type to most likely */"
> > >
> > > Well, _that_ one is easily fixed.
> > > Note that attached patch is completely untested right now.
> > the patch does look accpetable assuming it works and is tested and it
> > does fix something.
> It seems to work (I just don't have many test cases for error
> resilience, those I tried seem unchanged),
if you want more test cases, /tools/trasher or any other fuzzer should be
able to help
also enabling slices / data partitioning when encoding allows more stuff
tested. For example the different XY_ERROR differ only when data partitioning
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
The worst form of inequality is to try to make unequal things equal.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
More information about the ffmpeg-devel