[vlc-devel] commit: Remove non-sensical NPT computation code from c9569b35.
Ross Finlayson
finlayson at live555.com
Fri Jul 25 01:52:13 CEST 2008
>This is an entire session for that URL. it starts playing, and after a
>couple of secs I execute a seek. Obviously some of those PTS values
>are out of whack.
Yes, this stream is real weird. Looking at the PTSs for the audio
stream only, we see that all of the diffs are 96000 microseconds,
except for the following:
[00000403] live555 demux debug: PTS 1216933811532764, diff for track
mp4a, is now: -7004189878605393497
This is just initialization => OK
[00000403] live555 demux debug: PTS 1216933810096000, diff for track
mp4a, is now: -1436764
This is the transition between unsynched and (RTCP)synced PTSs => OK
[00000403] live555 demux debug: PTS 1216933817096000, diff for track
mp4a, is now: -7400000
[00000403] live555 demux debug: PTS 1216933824096000, diff for track
mp4a, is now: -7400000
[00000403] live555 demux debug: PTS 1216933831096000, diff for track
mp4a, is now: -7112000
[00000403] live555 demux debug: PTS 1216933838096000, diff for track
mp4a, is now: -2696000
WTF? Why is the server reducing the PTS by several seconds at a
time? I have no idea, but the server seems to be broken.
[00000403] live555 demux debug: PTS 1216933845096000, diff for track
mp4a, is now: 88000
[00000403] live555 demux debug: PTS 1216933852096000, diff for track
mp4a, is now: 88000
These are probably just the result of minor clock sync adjustments at
the server end => OK
[00000403] live555 demux debug: PTS 1216934815648000, diff for track
mp4a, is now: 960672000
This is the jump that occurred after you did the seek. You had
played the stream for about 44 seconds, and then requested a seek to
start at point (roughly) 1030 seconds, so that would be a jump of 986
seconds. That's a bit different from the PTS jump that we actually
saw (960 seconds), but it kinda makes sense...
[00000403] live555 demux debug: PTS 1216934815648000, diff for track
mp4a, is now: -1344000
Again, WTF? Why is the server reducing the PTS by more than 1 second??
[00000403] live555 demux debug: PTS 1216933859096000, diff for track
mp4a, is now: -963656000
Arggh!! WTF is this server doing?? This makes no sense to me!!
The bottom line is that this server is broken. I don't know why, but
it might have something to do with the fact that it's part of an
Akamai content distribution network.
In any case, it seems that - to handle such streams - VLC (and other
RTSP/RTP clients) need to watch for unusually large changes in PTS,
and reset the PTS->PCR mapping accordngly.
What's strange, though, is that when I tried playing the same stream
recently (using the "openRTSP" client, without seeking), I didn't see
any unusual PTS changes (for either the audio or video stream). So I
don't know why you saw those strange PTS changes (before you did the
seek)...
--
Ross Finlayson
Live Networks, Inc.
http://www.live555.com/
More information about the vlc-devel
mailing list