[FFmpeg-devel] [PATCH 3/3] lavf/mp3dec: read encoder delay/padding from Info tag

[Michael Niedermayer]

On Mon, Oct 03, 2016 at 05:45:08PM -0700, Jon Toohill wrote:
> Muxers can check AVCodecParameters.initial_padding for the
> encoder+decoder delay, and read the AV_PKT_DATA_SKIP_SAMPLES
> side data from the last packet for the encoder padding.
> 
> This change also fixes the first_discard_sample calculation
> which erroneously included the decoder delay. Decoder delay
> is already accounted for in st->skip_samples. The affected
> FATE tests have been updated accordingly.
> ---
>  libavformat/mp3dec.c                 |  3 ++-
>  tests/ref/fate/audiomatch-square-mp3 |  2 +-
>  tests/ref/fate/gapless-mp3           | 10 +++++-----
>  3 files changed, 8 insertions(+), 7 deletions(-)
> 
> diff --git a/libavformat/mp3dec.c b/libavformat/mp3dec.c
> index 56c7f8c..e8b2428 100644
> --- a/libavformat/mp3dec.c
> +++ b/libavformat/mp3dec.c
> @@ -239,9 +239,10 @@ static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
>  
>          mp3->start_pad = v>>12;
>          mp3->  end_pad = v&4095;
> +        st->codecpar->initial_padding = mp3->start_pad + 528 + 1;
>          st->start_skip_samples = mp3->start_pad + 528 + 1;
>          if (mp3->frames) {
> -            st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf;
> +            st->first_discard_sample = -mp3->end_pad + mp3->frames * (int64_t)spf;
>              st->last_discard_sample = mp3->frames * (int64_t)spf;
>          }
>          if (!st->start_time)
> diff --git a/tests/ref/fate/audiomatch-square-mp3 b/tests/ref/fate/audiomatch-square-mp3
> index 8de55c2..05176a0 100644
> --- a/tests/ref/fate/audiomatch-square-mp3
> +++ b/tests/ref/fate/audiomatch-square-mp3
> @@ -1 +1 @@
> -presig: 0 postsig:0 c: 0.9447 lenerr:0
> +presig: 0 postsig:-529 c: 0.9334 lenerr:-529

isnt this exactly correct before and wrong after this change ?

zero signal before and zero signal after the original is what one would
expect and equal length and high correlation

every number that changes gets worse

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Never trust a computer, one day, it may think you are the virus. -- Compn
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: Digital signature
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20161004/496de3bb/attachment.sig>