[FFmpeg-soc] SVQ3 depacketizer

[Martin Storsjö]

On Thu, 1 Jul 2010, Luca Barbato wrote:

> On 6/30/10 6:38 PM, Ronald S. Bultje wrote:
> > Hi,
> > 
> > On Wed, Jun 30, 2010 at 4:19 AM, Martin Storsjö<martin at martin.st>  wrote:
> > > On Wed, 30 Jun 2010, Josh Allmann wrote:
> > > > +/**
> > > > + * @file libavformat/rtpdec_svq3.c
> > > > + * @brief RTP support for the SV3V (SVQ3) payload
> > > > + * @author Ronald S. Bultje<rbultje at ronald.bitfreak.net>
> > > > + */
> > > 
> > > Lately, we've been instructed to remove the file names from these @file
> > > directives, and just keep "@file". Also, it would be good to mention which
> > > RFC this implements, to make it easier for the reader to look up the
> > > details.
> > 
> > No RFC, this is RE'ed stuff. :-(.
> 
> then it should be put in the multimedia wiki and we should reference it
> 
> > 
> > > > +/** return 0 on packet,<1 on partial packet or error... */
> > > 
> > > Isn't this comment a bit wrong? Iirc, return 0 on packet, 1 on partial
> > > packets (that is, more packets can be returned),<0 on error. When
> > > checking other depacketizers, others seem to have some kind of confusion,
> > > too.
> > 
> > Depends on spreading. We return 1 if the RTP packet contained multiple
> > demuxer packets, e.g. two video frames. What happens here is that>=2
> > RTP packets contain 1 video frame, so we don't have any data after 1
> > RTP packet (partial packet) and thus return -1.
> 
> Doesn't sound nice if <0 is also used for errors.

It's no problem actually, it uses AVERROR(EAGAIN) for these cases, which 
most calling code should handle just fine.

// Martin