[MPlayer-users] Possible bug with libtheora support in mplayer current

[Julian Kniephoff]

Hi everyone,

I recently ran into this bug on Ubuntu (MPlayer-1.0rc2-4.2.3) and also could 
reproduce it with the current SVN version (from today: 4th of May).
If this isn't fixed in another place I overlooked, already, I could contribute 
a patch but at first some apologies and/or clarifications which may be in 
order:
This is my first post to a mailing list so if something goes wrong, sorry for 
that. If I got the threading wrong: I'm referring to this e-Mail: 
http://lists.mplayerhq.hu/pipermail/mplayer-users/2007-December/070943.html
If this is the wrong place to post a patch sorry about that, too.
And at last sorry for my possibly bad English ;).

Concerning the Bug:
If libtheora is compiled in, MPlayer uses the video stream of the ogg file to 
calculate the length (in the "-identify-phase"), whereas without the lib, it 
uses the vorbis (audio) stream - at least that is what I found out / 
interpreted from some experiments with MPlayers code.
The calculation is then: length = final_granulepos / rate
For the audio stream, rate is the bitrate, here = 44100
For the video stream, it is the framerate, here = 29.97 and granulepos's 
represent frames. So frames / (frames / time) = time seems correct but for 
theora, granule positions are encoded. Some bytes represent the frames before 
the last keyframe and the other bytes represent the frames since that 
keyframe (as far as I understood it, look up in the theora docs for furhter 
details...). These two frame numbers have to be added to get the complete 
number of frames since the start of the video (which for final_granulepos is 
the number of frames of the whole video, thus the calculation of the length 
gets correct). So what one needs to fix this is to find out how these 
information are encoded in a granulepos.

My patch basically does this but I'm not sure whether this is the way, MPlayer 
does things; it's implemented kind of naive - without great knowledge of 
MPlayers code.
With this patch, I get a length of ID_LENGTH=211.41 for the mentioned video 
which is correct (the video stream in it is a bit shorter than the audio 
stream according to ogginfo, so the 0.05 seconds difference are ok).
The patch is written against the above mentioned version of MPlayer from the 
Ubuntu repositories.

Thanks in advance,

Julian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: theora_length.patch
Type: text/x-diff
Size: 1225 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/mplayer-users/attachments/20080504/81428b00/attachment.patch>