[Live-devel] dump h264 video problem

[jacopo]

On 07/29/2010 10:14 PM, Ross Finlayson wrote:
>> ]I've found an h264 video stream that I'm able to dump and play
>> correctly
>> rtsp://media.lscube.org/tests/tc.mov
>>
>> This one suffers the problems I was talking about:
>>  rtsp://streaming1.osu.edu/media2/ufsap/ufsap.mov
>>
>> Both have been dumped with
>> ./openRTSP -t -d 20 -q
>> command line (no fps and w/h has to be specified for firts one, for
>> second none worked)
>
> The second stream's video has width 320 pixels, and height 240 pixels,
> so you should specify those if you're trying to record it using
> "openRTSP -q".  (I'm not sure about the frame rate.)
>
>
>> I'm not sure about one thing. H264VideoFileSink class that inserts 4
>> bytes long delimiter between NAL units, is used only when streaming not
>> when receiving a strem, right?
>
> Wrong.  Because it's a 'FileSink', it's used only when *receiving* a
> stream.  If "openRTSP" is used to receive a H.264/RTP stream normally
> - without using the "-q" or "-4"  options - then the
> "H264VideoFileSink" class is used for the output.  If, however,
> "openRTSP" is used with the "-q" or "-4" options, then a different
> 'FileSink' class - "QuickTimeFileSink" - is used.  Because you're
> using the "-q" option, this is what you are using.
Yes, I saw that into sources..
>
>
>> If so, the problem must be container related (QuickTime or mp4 does't
>> matter for me) not codec related, since libraries are just dumping an
>> encoded stream that is not touched or modified when saved to a file (or
>> printed to stdout) inside the choosen container...
>> Am I totaly missing the point?
>
> I'm not sure.  But the basic problem here is that the ".mov" (or
> ".mp4") file format was very badly designed, and is especially
> ill-suited for what we are trying to do: Record incoming media streams
> into a file.  The "QuickTimeFileSink" class tries to do this as well
> as it can, but the basic flaws of the file format makes it difficult. 
> Nonethess, it's possible that the "QuickTimeFileSink" implementation
> can be improved, so feel free to take a look at it, and let us know if
> you have any suggestions for improvements/bug fixes.  Remember, You
> Have Complete Source Code.
Thanks a lot Ross.
I'm digging deep into code from a couple of days... If I'll be able to
do something usefull I'll write there ASAP.

thanks again
    j