[MPlayer-dev-eng] Re: [PATCH] -identify vids, aids, sids, alangs and slangs

[Nico Sabbi]

Moritz Bunkus wrote:

>Hey,
>
>  
>
>>Hmm... I don't think I understand. The only place I found where the
>>language is set is in demux_mkv_read_trackentry, which is called from
>>demux_mkv_read_tracks. And display_tracks is called only after
>>demux_mkv_open is done calling demux_mkv_read_tracks. So why is it
>>too early? Did I miss something?
>>    
>>
>
>Ups sorry, I mixed up places :) I thought the part I was complaining
>about was to be inserted in demux_mkv_read_trackentry. No,
>display_tracks is OK of course.
>
>  
>
>>The only easy way to fix this mess I could think of was to change the
>>accepted aids and vids to match the internal ones,
>>    
>>
>
>Or fix the demuxers to all use zero-based IDs (which would be MUCH more
>consistent than the mess we have now...).
>
>  
>

think of mpeg-ts where there can be almost 8192 pids; each of them may 
or may not
have an entry in the PMT, and when use -alang/slang eng you have to find 
the corresponding
numeric pid; and what if a pid shows 5GB after the first one ?
simply put it's not feasible to 0-base the pids in the case of TS

>>but I understand you (and Richard) object to that,
>>    
>>
>
>Strongly. It's a matter of usability (which mplayer lacks in too many
>ways already). Having to call mplayer once just to find out the correct
>IDs to use for the first audio track or so is simply not acceptable.
>
>  
>
>>So the demux_aid_vid_mismatch is the workaround that tells new_sh_*
>>not to print the ids it gets, in which case it is the demux module
>>that prints them.
>>    
>>
>
>In fact this could be changed. For demux_ogg.c you've already introduced
>a new member "id" which stores the -Xid ID. How about you extend the
>patch so that demux_ogg.c always uses those? :)
>
>  
>
>>>-alang doesn't work for Ogg (yet), just so you know, but this would be
>>>the right place to output it. So let's keep it.
>>>      
>>>
>>I understand, but a language name is useful for displaying in a GUI.
>>    
>>
>
>Sure. I just wanted to point it out.
>
>  
>
>>The patch will be one line shorter, but the resulting code one line
>>longer. Whatever your preference. :-)
>>    
>>
>
>Shorter patches. Makes reviewing easier.
>
>  
>
>>Another minor problem in demux_ogg is that demux_ogg_check_comments
>>only gets called for streams that are actually going to be used, and
>>so the other stream languages are not known until they are actually
>>played. But I am guessing there is not easy fix for that.
>>    
>>
>
>Well, the "best" way IMHO would be to have demux_ogg_open parse ALL
>header packets of ALL streams before returning. That way we could also
>enable -alang.
>
>  
>
>>Thanks for looking at the patch.
>>    
>>
>
>You're welcome. I'm sorry I cannot really comment on the other
>parts. demux_ogg.c and demux_mkv.c are the ones I know best, so I
>preferred to talk only about those.
>
>Mosu
>
>  
>