[MPlayer-dev-eng] [PATCH] fix mpg123 resync (for massive seeking)

[Ivan Kalvachev]

On 5/3/12, Thomas Orgis <thomas-forum at orgis.org> wrote:
> Am Thu, 3 May 2012 00:22:52 +0300
> schrieb Ivan Kalvachev <ikalvachev at gmail.com>:
>
>> However sometimes I see some debug messages and parser errors. (no
>> audible artifacts or skips, it resumes normal play as it should)
>
> Do you mean
>
> 	trying to decode obviously invalid header
> 	Note: missing %d bytes in bit reservoir for frame %li
>
> ? The first is a harmless message that trunk mg123 printed for a while,
> but which is gone in released 1.14.0, the second just means that one
> started decoding in the middle of the stream where a frame depends on
> bits from previous frames and cannot be properly decoded if it's the
> first in line.
>
>> We talked about the possibility of avi demuxer returning garbage,
>> that could confuse the mpg123 parser. In your investigation have you
>> found any clues if this might be true?
>
> Not really, mpg123 1.14.0 was wholly silent with the test video.
>
>> Was mpg123_getformat() confused
>> by completely normal packages?
>
> The confusion was not about the packages themselves, but about the
> question if they were feed to the decoder when needed /
> mpg123_getformat() expected more data without reason.
>
>> Is it possible that e.g. packets cut mp3 frames at strange places
>> (e.g. at half) and this causes the problems.
>
> That would not matter to mpg123: You can feed it 1-byte blocks, if you
> want. It is a full bitstream in, audio byte stream out decoder.
>
> Now, what errors do you see? On which mpg123 version?

You got me, I'm with 1.13.8 version.
After all, I had to test the patch with "broken" mpg123 versions.

I was just curious and wanted more details. Thank you for detailed reply.

If I find any other problem, I'll inform you.

Best Regards.