Michael Niedermayer <[email protected]> added the comment:

On Thu, Sep 10, 2009 at 06:37:21PM -0700, Baptiste Coudurier wrote:
> On 09/10/2009 06:57 PM, Michael Niedermayer wrote:
>> Michael Niedermayer<[email protected]>  added the comment:
>>
>> On Thu, Sep 10, 2009 at 10:34:13PM +0000, ndrasmussen wrote:
>>> When adding a data stream to an MPEGTS, the packets in the stream are 
>>> buffered
>>> up as if they were audio packets. I am not sure if this is per the 
>>> specification
>>> as I have not been able to find any information on it. I put this as a 
>>> bug
>>> because I have been given other MPEGTS files that do not have the data 
>>> stream
>>> buffered as I am seeing using ffmpeg. It can be fixed by simply adding
>>> STREAM_TYPE_DATA to the "if" statement on line 736 of file mpegtsenc.c.
>>
>> what you write is all very vague, its not clear at all what kind of data
>> stream you use nor what problem you have. Its hardly the way packets are
>> buffered as such ...
>>
>> Also just to clarify the thing you call "buffering up" is not
>> audio specific, it can be done for video as well its just that our ts 
>> muxer
>> isnt doing it. I think it was disabled because it caused some problem with
>> some decoder, i think i was quite unhappy about this solution, though iam
>> not sure if i mentioned that in public ...
>
> Apparently, many set-top boxes only support one video frame per PES packet, 
> it's better for timestamps anyway ;)
>
> Also I found out that the cisco mpeg analyzer do not support audio frames 
> spanned accross multiple PES packets. It supports more than one audio frame 
> though.

i really cant belive that something calling itself mpeg analyzer would not
support half of the spec ...

have the decoder & analyzer been tested with streams known to be fully
compliant or just the output of our mpeg ts muxer?
The reason why iam asking is to exclude the case that another bug in
the muxer is causing the problem and making frame==PES could just be
covering that issue up.

Someone should take a known to be fully compliant mpeg ts muxer that supports
spanning video frames accorss PES and try its output if this has not been
done.

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Thouse who are best at talking, realize last or never when they are wrong.

_____________________________________________________
FFmpeg issue tracker <[email protected]>
<https://roundup.ffmpeg.org/roundup/ffmpeg/issue1374>
_____________________________________________________

Reply via email to