[mythtv-commits] Ticket #9410: Positioning problems with BBC HD and BBC ONE HD recordings
MythTV
noreply at mythtv.org
Mon Jan 17 13:00:47 UTC 2011
#9410: Positioning problems with BBC HD and BBC ONE HD recordings
-------------------------------------+-----------------------------
Reporter: mythtv@… | Owner: jpoet
Type: Bug Report | Status: accepted
Priority: critical | Milestone: 0.24.1
Component: MythTV - Video Playback | Version: Unspecified
Severity: medium | Resolution:
Keywords: | Ticket locked: 0
-------------------------------------+-----------------------------
Comment (by mythtv@…):
I've prepared a new version which removes all assumptions of alignment
between NALs and the packets passed into the parser (for the entire
change, see the four commits between fixes/0.24 andfix9410-bbc-hd in my
fork at https://github.com/Glidos/mythtv, or the patch I'll upload in the
next comment). I've added debug to both this version and the previous
version, and they give identical results on a BBC HD sample, when run with
John's test program, passing 128KB chunks. With this new version I've also
tried with 128B chunks passed in, and that still gives identical results
(which was the whole point of the change).
With the previous version, I was seeing the recordedmarkup table filling
up with huge numbers of width/height/aspect records during recording
(eventually leading to mythbackend stalling). I guessed this was because
the SPS was not being passed in a single continuous chunk. The guess seems
to be correct, because now I see a single record added with sensible
values.
In any case, the dtvrecorder class looks to pass in a TS packet at a time,
rather than building up PES packets; avformatdecoder, on the other hand,
looks to build up packets, which would explain why "mythcommflag
--rebuild" would work and recording not. Also, H.222 specifies that there
are no assumptions of alignment between NALs and PES packets unless a
certain flag is set.
I haven't had time yet to check a sample of John Poet's problematic HD-PVR
recording, but I'm hoping that this new version might be a fix. I'm
possibly now seeing problems with ITV HD recordings, but I need to do more
testing. I have a past recording that now wont play. I used the test
program to pass it through the parser, and it found frames about every
50KB, with every 32nd being a key frame, so it looks to be working
correctly. I hope it's not a nasty side-effect of the fact that now we now
often leave parsing a NAL until finding the start code of the next.
--
Ticket URL: <http://code.mythtv.org/trac/ticket/9410#comment:7>
MythTV <http://code.mythtv.org/trac>
MythTV Media Center
More information about the mythtv-commits
mailing list