[FFmpeg-devel] [PATCH] fix issue434

[Michael Niedermayer]

On Mon, Apr 21, 2008 at 02:04:08AM +0200, Michael Niedermayer wrote:
> On Mon, Apr 21, 2008 at 12:21:07AM +0100, M?ns Rullg?rd wrote:
> > Michael Niedermayer <michaelni at gmx.at> writes:
> > 
> > > Hi
> > >
> > > The patch below fixes issue434, i will commit it in 24h unless mans objects
> > >
> > > Index: libavformat/mpeg.c
> > > ===================================================================
> > > --- libavformat/mpeg.c	(revision 12867)
> > > +++ libavformat/mpeg.c	(working copy)
> > > @@ -341,12 +341,13 @@
> > >          if (flags & 0x01) { /* PES extension */
> > >              pes_ext = get_byte(s->pb);
> > >              header_len--;
> > > -            if (pes_ext & 0x40) { /* pack header - should be zero in PS */
> > > -                goto error_redo;
> > > -            }
> > >              /* Skip PES private data, program packet sequence counter and P-STD buffer */
> > >              skip = (pes_ext >> 4) & 0xb;
> > >              skip += skip & 0x9;
> > > +            if (pes_ext & 0x40 || skip > header_len){
> > > +                av_log(s, AV_LOG_WARNING, "pes_ext %X is invalid\n", pes_ext);
> > > +                pes_ext=skip=0;
> > > +            }
> > >              url_fskip(s->pb, skip);
> > >              header_len -= skip;
> > 
> > Workarounds for out-of-spec streams are only acceptable if explicitly
> > requested.  Does lavf have something equivalent to lavc's
> > AVCodecContext.strict_std_compliance?
> 
> No, also strict_std_compliance is a pure encoder side thing. Used for
> preventing users to shoot themselfs in their foot (encoding expermental
> codecs like snow, storing non jpeg YUV in jpeg, ...).
> 
> The variable with a matching purpose is error_resilience (bad name, yes ...)
> It does select between how much spec non compliance should be ignored and
> how much should be considered an error. Thus its a tradeoff between early
> detection of errors in damaged streams vs. supporting streams which dont
> conform to the specs.
> 
> We certainly could add such a variable to AVFormatContext as well.

ping
What shall be done about this issue?
* This does not affect valid files
* It does print a big and nasty "blah is invalid"
* Its ffmpegs "policy" to support broken files by default

I can add a error_resilience variable to AVFormatContext if you want, but
its default would be what has the highest chance to result in correct
playaback.

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

I am the wisest man alive, for I know one thing, and that is that I know
nothing. -- Socrates
-------------- 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/20080428/df09589f/attachment.pgp>