[FFmpeg-devel] [PATCH] ADTS AAC with ID3v2

[Michael Niedermayer]

On Thu, Jan 15, 2009 at 02:37:11AM -0500, Alex Converse wrote:
> On Wed, Jan 14, 2009 at 1:53 PM, Michael Niedermayer <michaelni at gmx.at> wrote:
> >
> > On Wed, Jan 14, 2009 at 01:39:05PM -0500, Alex Converse wrote:
> > > On Wed, Jan 14, 2009 at 12:57 PM, Michael Niedermayer <michaelni at gmx.at>wrote:
> > >
> > > > On Wed, Jan 14, 2009 at 10:17:33AM -0500, Alex Converse wrote:
> > > > > On Wed, Jan 14, 2009 at 10:10 AM, Michael Niedermayer <michaelni at gmx.at
> > > > >wrote:
> > > > >
> > > > > > On Wed, Jan 14, 2009 at 09:47:29AM -0500, Alex Converse wrote:
> > > > > > > On Wed, Jan 14, 2009 at 9:17 AM, Michael Niedermayer <
> > > > michaelni at gmx.at
> > > > > > >wrote:
> > > > > > >
> > > > > > > > On Wed, Jan 14, 2009 at 12:23:00AM -0500, Alex Converse wrote:
> > > > > > > > > Hi,
> > > > > > > > >
> > > > > > > > > The attached patch adds support to the ADTS AAC probe to detect
> > > > AAC
> > > > > > files
> > > > > > > > > with ID3v2 tags at a higher level than the MP3 probe and the MPC
> > > > > > probe.
> > > > > > > > It
> > > > > > > > > refactors some id3 detection from mp3.c into id3v2.c.
> > > > > > > > >
> > > > > > > > [...]
> > > > > > > >
> > > > > > > > >  #define ID3v1_TAG_SIZE 128
> > > > > > > > > -
> > > > > > > > >  #define ID3v1_GENRE_MAX 125
> > > > > > > > >
> > > > > > > > >  static const char * const id3v1_genre_str[ID3v1_GENRE_MAX + 1] =
> > > > {
> > > > > > > >
> > > > > > > > cosmetic
> > > > > > > >
> > > > > > > > [...]
> > > > > > > > > @@ -487,7 +472,7 @@ static int mp3_read_header(AVFormatContext
> > > > *s,
> > > > > > > > >      ret = get_buffer(s->pb, buf, ID3v2_HEADER_SIZE);
> > > > > > > > >      if (ret != ID3v2_HEADER_SIZE)
> > > > > > > > >          return -1;
> > > > > > > > > -    if (id3v2_match(buf)) {
> > > > > > > > > +    if (ff_id3v2_match(buf)) {
> > > > > > > > >          /* parse ID3v2 header */
> > > > > > > > >          len = ((buf[6] & 0x7f) << 21) |
> > > > > > > > >              ((buf[7] & 0x7f) << 14) |
> > > > > > > >
> > > > > > > > the split out of (ff_)id3v2_match is ok but should be a seperate
> > > > patch
> > > > > > > > and commit
> > > > > > > >
> > > > > > > > [...]
> > > > > > > > >      }
> > > > > > > > > -    if   (first_frames>=3) return AVPROBE_SCORE_MAX/2+1;
> > > > > > > > > +    if   (first_frames>=3) return AVPROBE_SCORE_MAX/2+2;
> > > > > > > > >      else if(max_frames>500)return AVPROBE_SCORE_MAX/2;
> > > > > > > > >      else if(max_frames>=3) return AVPROBE_SCORE_MAX/4;
> > > > > > > > >      else if(max_frames>=1) return 1;
> > > > > > > >
> > > > > > > > why?
> > > > > > >
> > > > > > >
> > > > > > > Just having an ID3v2 tag gives a score of AVPROBE_SCORE_MAX/2+1 for
> > > > mp3.
> > > > > >
> > > > > > That sounds like a bug ...
> > > > > >
> > > > >
> > > > > The comment about mpeg-ps in mp3.c made me think it was a hackish design
> > > > > decision, not a bug.
> > > >
> > > > well one could call it that way, still the existence of hackish code is no
> > > > argument to add more hackish code like the +1 -> +2
> > > >
> > >
> > > What would be the correct solution for MP3 then? skipping the tag and trying
> > > again?
> >
> > i think so
> 
> I fixed the MP3 hackishness in the attached patch. I also fixed a typo
> in the old patch. Once I fixed MP3, musepack became the next greedy
> culprit so to prevent MP3 regressions I had to fix him too. I don't
> have any MPC files with ID3v2 so that portion of that patch is
> untested.
> 
> In a related note, the tag skipper in mpc_read_header seems to not
> check for a footer, which I think could be trouble.
[...]
> diff --git a/libavformat/id3v2.c b/libavformat/id3v2.c
> index c27443b..649fafa 100644
> --- a/libavformat/id3v2.c
> +++ b/libavformat/id3v2.c
> @@ -33,3 +33,15 @@ int ff_id3v2_match(const uint8_t *buf)
>              (buf[8] & 0x80) == 0 &&
>              (buf[9] & 0x80) == 0;
>  }
> +
> +int ff_id3v2_tag_len(const uint8_t * buf)
> +{
> +    int len = ((buf[6] & 0x7f) << 21) +
> +        ((buf[7] & 0x7f) << 14) +
> +        ((buf[8] & 0x7f) << 7) +
> +        (buf[9] & 0x7f) +
> +        ID3v2_HEADER_SIZE;
> +    if (buf[5] & 0x10)
> +        len += ID3v2_HEADER_SIZE;
> +    return len;
> +}

is this correct?
i mean this does either X + ID3v2_HEADER_SIZE or X + 2*ID3v2_HEADER_SIZE


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

I know you won't believe me, but the highest form of Human Excellence is
to question oneself and others. -- 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/20090115/e0d76cd0/attachment.pgp>