[FFmpeg-cvslog] r22947 - trunk/libavformat/mpegts.c

[Reimar Döffinger]

On Thu, Apr 22, 2010 at 02:26:10PM -0700, Baptiste Coudurier wrote:
> On 04/22/2010 02:15 PM, Carl Eugen Hoyos wrote:
> >Hi!
> >
> >bcoudurier<subversion<at>  mplayerhq.hu>  writes:
> >
> >>Modified: trunk/libavformat/mpegts.c
> >>==============================================================================
> >>--- trunk/libavformat/mpegts.c	Thu Apr 22 20:58:39 2010	(r22946)
> >>+++ trunk/libavformat/mpegts.c	Thu Apr 22 22:34:30 2010	(r22947)
> >>@@ -497,7 +497,7 @@ static const StreamType ISO_types[] = {
> >>      { 0x04, AVMEDIA_TYPE_AUDIO,        CODEC_ID_MP3 },
> >>      { 0x0f, AVMEDIA_TYPE_AUDIO,        CODEC_ID_AAC },
> >>      { 0x10, AVMEDIA_TYPE_VIDEO,      CODEC_ID_MPEG4 },
> >>-    { 0x11, AVMEDIA_TYPE_AUDIO,        CODEC_ID_AAC }, /* LATM syntax */
> >>+  //{ 0x11, AVMEDIA_TYPE_AUDIO,        CODEC_ID_AAC }, /* LATM syntax */
> >
> >Could you revert this hunk
> >
> >>+    if (pes->stream_type == 0x11)
> >>+        av_log(pes->stream, AV_LOG_WARNING,
> >>+               "AAC LATM not currently supported, patch welcome\n");
> >
> >and leave this one (so I will realize the duplicate the next time)?
> >Decoding works with mplayer -demuxer lavf -ac faad.
> 
> Well, sorry but I don't think this is a good idea. History has
> proven this produced many errors, and users simply only see the
> garbage written on the screen.
> 
> mplayer can still decode the stream, the packets are outputed.
> and codec_tag contains 0x11.

I don't really think this is a good solution.
Either we consider this AAC, then the codec ID should be set as such,
we should not misuse our control over libavformat to work around issues
with libavcodec. The libavcodec could try to detect LATM and then print
appropriate messages.
Or if we consider it a different codec, a different codec id could be used.