[Ffmpeg-devel] Third Try: Set bit_rate for asf format

[Zuxy Meng]

Hi,

2007/4/16, Michael Niedermayer <michaelni at gmx.at>:
> Hi
>
> On Mon, Apr 02, 2007 at 01:26:40PM +0800, Zuxy Meng wrote:
> > Hi,
> >
> > 2007/3/31, Zuxy Meng <zuxy.meng at gmail.com>:
> > >Hi,
> > >
> > >2007/3/30, Zuxy Meng <zuxy.meng at gmail.com>:
> > >> Hi,
> > >>
> > >> 2007/3/30, Michael Niedermayer <michaelni at gmx.at>:
> > >> > Hi
> > >> >
> > >> > still i have a bad feeling with setting it just for video knowing it
> > >> > breakes audio if it would be set there too ...
> > >>
> > >> Me too. I'll dig into wma.c this weekend.
> > >>
> > >> > somehow this isnt a clean solution at all
> > >
> > >The reason is that for audio streams, there will be a wave header
> > >defined at the asf container level, and get_wav_header will set
> > >bit_rate as 8*nAvgBytesPerSec. So unlike video streams, we don't need
> > >to look for the optional extended stream property object for bitrate.
> > >
> > >And something more about the fulltest: asf files used/generated in
> > >fulltest don't contain extended stream property objects so bitrate[i]
> > >== 0, and the subsequent assignment will effectively overwrite what
> > >has been found in get_wav_header. This is absolutely my fault; I
> > >forgot that such objects are optional, not mandatory.
> >
> > Please take a look at this patch; hopefully this is cleaner. And it
> > passes fulltest, too.
> [...]
> > @@ -500,11 +505,14 @@
> >
> >      for(i=0; i<128; i++){
> >          int stream_num= asf->asfid2avid[i];
> > -        if(stream_num>=0 && dar[i].num>0 && dar[i].den>0){
> > +        if(stream_num>=0){
> >              AVCodecContext *codec= s->streams[stream_num]->codec;
> > -            av_reduce(&codec->sample_aspect_ratio.num,
> > -                    &codec->sample_aspect_ratio.den,
> > -                    dar[i].num, dar[i].den, INT_MAX);
> > +            if (bitrate[i])
> > +                codec->bit_rate = bitrate[i];
>
> this still overrides the bitrate and IMHO will break WMA like the previous
> patches, its just that bitrate[i] is 0 during the regression tests but this
> is not guranteed to be true for files in the wild

bitrate[i] is 0 because asf files used in the regression tests don't
contain a bitrate object. IMHO the encoder should guanrantee that the
bitrate in the wave header and the one in the object (if present) be
the same.

Is codec->bit_rate always initialized to zero? If so we can use
if (codec->bit_rate)
so bitrate in the bitrate object won't overwrite the wave header info.
-- 
Zuxy
Beauty is truth,
While truth is beauty.
PGP KeyID: E8555ED6