[vlc-devel] Re: (hopefully) simple question on stream clocking

[Jean-Paul Saman]

John Michael Zorko wrote:
> 
> Hello, all ...
> 
> After implementing the TS discontinuity indicator setting code, and 
> resetting the decoders on the client when this bit is detected in the TS 
> packet, and making sure I seek to an E0 packet (video ES) with a 
> sequence header, 95% of the macroblocking artifacts I see on the client 
> when the server VLC does a seek on the stream are gone.  The few that 
> are left may be due to open GOPs, but i'll worry about that later.  
> Thanks to everyone who posted hints on how to do this -- it's appreciated!
> 
Could be or you forgot to set the TS discontinuity bit on other pid's 
too! After a seek ALL PIDs in the stream need to be told that a 
discontinuity happend (video,audio,teletext,PCR packets).

> Now, i've a (hopefully) simple question ... sometimes the client VLC 
> stops updating the video after the server VLC seeks.  I've verified that 
> the server VLC seeked to where I wanted it to (the E0 PES packet with a 
> sequence header).  Sometimes it recovers after a second or two, 
> sometimes it takes far longer (10-20 seconds), sometimes never.  This is 
> sometimes accompanied by the infamous "decoder stuck" message, which 
> implies that the vout thread is the culprit i.e. perhaps on an mwait() 
> waiting to display a frame.  On the server, all it does is call 
> input_Seek() to the position I want, then set the TS discontinuity 
> indicator.  I don't do anything different with the PTS i.e. the TS muxer 
> does what it always did, it just sets the TS discontinuity indicator 
> when it should now.

Looks like a forgotten TS flag. Is there a way you can verify this?

> 
> The client, when it sees the TS discontinuity indicator on, does 
> input_EscapeDiscontinuity() -- nothing else special.  Yet sometimes 
> (after a lot of seeking on the server), the vout thread hangs.  Does the 
> server VLC need to do something special to the PTS after a seek, 
> something that it normally doesn't need to do?
> 
AFAIK you should not tinker with the PTS. If you do the DTS and PCR's 
need to be updated too, this will actually equal to a decode-encode 
cycle (reencode the audio/video).

-- 
Kind greetings,

Jean-Paul Saman


-- 
This is the vlc-devel mailing-list, see http://www.videolan.org/vlc/
To unsubscribe, please read http://developers.videolan.org/lists.html
If you are in trouble, please contact <postmaster at videolan.org>