[FFmpeg-devel] [PATCH]lavf/matroskadec: Parse stsd atom in mkv with the appropriate mov function

[Mats Peterson]

On 12/14/2015 08:06 AM, Mats Peterson wrote:
> On 12/14/2015 06:32 AM, Mats Peterson wrote:
>> On 12/14/2015 03:40 AM, Carl Eugen Hoyos wrote:
>>> Hi!
>>>
>>> Attached patch based on 3ece3e4c by Martin Storsjö fixes ticket #5071
>>> for me.
>>> I can't see how duplicating the code from mov.c could be acceptable.
>>>
>>> Please comment, Carl Eugen
>>>
>>>
>>>
>>> _______________________________________________
>>> ffmpeg-devel mailing list
>>> ffmpeg-devel at ffmpeg.org
>>> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>>>
>>
>> If you look at the patch I've made, it's not DUPLICATING the code, it
>> BORROWS some of code for the loops, but they are modified as well, if
>> you look closely, and it uses the PRIVATE DATA instead of calling some
>> superfluous stsd atom parsing function that reads from the file. I hope
>> Michael Niedermayer understands this better than you do, at least.
>>
>> Mats
>>
>
> And even if your patch will be the one to use (against all odds), it is
> still lacking the code that puts the palette in extradata, a la
> V_MS/VFW/FOURCC. Without it, MPlayer won't see the palette.
>
> Mats
>

Let me elaborate on this: The palette side data packet will be added in 
matroska_deliver_packet() in matroskadec.c alright, but the problem is 
that MPlayer will add ANOTHER palette side data packet afterwards, based 
on what's in extradata. If we don't copy the palette to extradata, like 
it's done in V_MS/VFW/FOURCC, MPlayer will use the whole contents of the 
stsd atom as the palette, which of course won't work.

Mats

-- 
Mats Peterson
http://matsp888.no-ip.org/~mats/