[FFmpeg-devel] [PATCH] Lost VBR header

Peter Belkner pbelkner at snafu.de
Fri May 6 20:53:19 CEST 2011


On 29.04.2011 21:08, Peter Belkner wrote:
> Hi,
> On 28.04.2011 21:43, Michael Niedermayer wrote:
>> This patch results in differnt output when the resulting mp3 is decoded
>> also there are error messages and 3456 samples (6*576) missing at the
>> start
> Unfortunately I'm not able to exactly reproduce this error. In some 
> cases I found some frame headers corrupted.
> I changed the following:
>    * Dynamic bitrate index determination to really make sure that the
>      TOC fits into the XING frame.
>    * In mp3_write_trailer() changed the order of writing ID3v1 and
>      fixing the XING header. (Fixing the XING header is now the final
>      operation. This makes the corrupted frame error vanishing in my
>      tests. Is there someting I'm missing with AVIO?)
> In order to avoid overflow in TOC statistics (wrapping to negative 
> values):
>    * Diveded the bitrate by 1000.
>    * Changed the statistics data types from "int" to "uint64_t".
> Please find a revised patch below.
> With the new patch I could not find any difference the following way:
>    unpatched/ffmpeg -i a.mp3 -acodec copy -y b.mp3
>    unpatched/ffmpeg -i b.mp3 -y b.wav
>    patched/ffmpeg -i a.mp3 -acodec copy -y c.mp3
>    patched/ffmpeg -i c.mp3 -y c.wav
>    diff --binary b.wav c.wav
> Could you please share the test case (MP3) in case the error is not 
> vanishing in your environment?
> Peter
Is there any chance to get this into FFmpeg?

Please let me know how I can contribute.

Thank you for your effort.


More information about the ffmpeg-devel mailing list