[FFmpeg-devel] [PATCH] dts parser fails to parse DTS-HD Master streams due to changing frame size

[Kostya]

On Thu, Jul 24, 2008 at 07:26:59PM +0200, elupus wrote:
> On Wed, 11 Jun 2008 17:38:15 +0200, elupus wrote:
> 
> > On Wed, 11 Jun 2008 11:19:22 +0300, Kostya wrote:
> > 
> >> On Wed, Jun 11, 2008 at 09:23:03AM +0200, Benjamin Larsson wrote:
> >>> [12 quoted lines suppressed]
> >> 
> >> The reason is called DCA specification where it is stated that all frame
> >> sizes should be equal.
> >> 
> > 
> > What I see wrong in the parser is that you consider framesize to be from
> > one startcode to the next startcode, not the encoded framesize as given by
> > the dts header. 
> > 
> > This means that like in the case of DTS HD, where there is extra "junk"
> > after the full core dts frame (which btw, will have same framesize as
> > previous frame), the parser will never find a valid end code.
> > 
> > Another error is that if this even happens once in the stream due to bit
> > errors, the parser will then never resync, as pc1->size is unconditionally
> > increased, and compared to the fixed pc1->framesize.
> > 
> > Also what happens with a TS stream transmitted over tv network, when they
> > for example change bitrate or channel count? wouldn't that change
> > framesize?
> > 
> > Joakim
> 
> Kostya, could you comment on this?

I can.
The reason that core DCA frames should have equal size stays, we can't disregard it
and there's at least one stream that has marker inside frame.
As for DTS HD, I don't know a way to determine fixed frame size (for now) neither
have samples to test. I'm not against supporting it, not at all.
DTS stream with errors, I think it may be easily fixed by replacing size == framesize
with size >= framesize. I am just too lazy to create and test such sample.

You can also try your luck with Benjamin Larsson, he's DCA co-maintainer.
 
> Regards
> Joakim