[FFmpeg-devel] [PATCH] fix H.264 crash (issue 706)

[Michael Niedermayer]

On Thu, Apr 09, 2009 at 09:51:56PM +0200, Reimar D?ffinger wrote:
> On Thu, Apr 09, 2009 at 09:03:31PM +0200, Michael Niedermayer wrote:
> > On Thu, Apr 09, 2009 at 07:46:19PM +0200, Reimar D?ffinger wrote:
> > > Index: libavcodec/error_resilience.c
> > > ===================================================================
> > > --- libavcodec/error_resilience.c	(revision 18389)
> > > +++ libavcodec/error_resilience.c	(working copy)
> > > @@ -343,6 +343,7 @@
> > >  
> > >                  if(IS_INTRA(s->current_picture.mb_type[mb_xy]))  continue;
> > >                  if(!(s->error_status_table[mb_xy]&MV_ERROR)) continue;
> > > +                if(!s->last_picture.data[0]) continue;
> > >  
> > >                  s->mv_dir = MV_DIR_FORWARD;
> > >                  s->mb_intra=0;
> > 
> > this should use MV_DIR_BACKWARD i suspect instead of skiping it completely
> 
> Do I need to flip the sign of the mvs or is that correct as-is (well, in
> one of those cases those mvs are 0 anyway).

you should use the other MV, MBs have 2 (its likely a [0] that should be [1])


> 
> > > @@ -908,8 +910,11 @@
> > >                  if(IS_INTRA(mb_type)) continue;
> > >                  if(!(error&MV_ERROR)) continue;           //inter with undamaged MV
> > >                  if(!(error&AC_ERROR)) continue;           //undamaged inter
> > > +                if(!s->last_picture.data[0] && !s->next_picture.data[0]) continue;
> > 
> > that should not happen, i mean mb_typeshould be intra if there are no reference
> > frames
> 
> Not sure, did you meant even currently mb_type will be intra if no
> reference frames are available? I can't really see that, so I assumed
> you want me to change this...

> Index: libavcodec/error_resilience.c
> ===================================================================
> --- libavcodec/error_resilience.c	(revision 18389)
> +++ libavcodec/error_resilience.c	(working copy)
> @@ -344,7 +344,7 @@
>                  if(IS_INTRA(s->current_picture.mb_type[mb_xy]))  continue;
>                  if(!(s->error_status_table[mb_xy]&MV_ERROR)) continue;
>  
> -                s->mv_dir = MV_DIR_FORWARD;
> +                s->mv_dir = s->last_picture.data[0] ? MV_DIR_FORWARD : MV_DIR_BACKWARD;
>                  s->mb_intra=0;
>                  s->mv_type = MV_TYPE_16X16;
>                  s->mb_skipped=0;

ok


> @@ -850,6 +850,12 @@
>      /* set unknown mb-type to most likely */
>      for(i=0; i<s->mb_num; i++){
>          const int mb_xy= s->mb_index2xy[i];
> +        // change inter with no references to intra
> +        if (!IS_INTRA(s->current_picture.mb_type[mb_xy]) &&
> +            !s->last_picture.data[0] && !s->next_picture.data[0]) {
> +            s->current_picture.mb_type[mb_xy]= MB_TYPE_INTRA4x4;
> +            continue;
> +        }
>          error= s->error_status_table[mb_xy];
>          if(!((error&DC_ERROR) && (error&MV_ERROR)))
>              continue;

is_intra_more_likely() should return 1
and that should cause all to be set to intra

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Many that live deserve death. And some that die deserve life. Can you give
it to them? Then do not be too eager to deal out death in judgement. For
even the very wise cannot see all ends. -- Gandalf
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090409/b39848d7/attachment.pgp>