[mythtv] Bug? HD-3000 DVB driver timestamps

[William Uther]

On Mon, 18 Apr 2005 23:16:30 -0600, John Patrick Poet 
<john at BlueSkyTours.com> wrote:

> I switching from the V4L2 HD-3000 drivers to the DVB HD-3000 drivers on
> Wednesday, 13 Apr 05.  While they seem to work fine, commercial 
> flagging
> suddenly seemed broken.

[snip]

> To give you an example of what I am talking about, here is a list of
> detected comercial markers compared to the actual position:
>
> detected        actual      difference
> 0:07:51.09  --  0:08:20.09     0:29
> 0:17:21.22  --  0:18:25.22     1:04
> 0:20:26.16  --  0:21:42.16     1:16
> 0:30:38.29  --  0:32:32.29     1:54
> 0:33:55.23  --  0:36:01.23     2:06
> 0:45:33.24  --  0:48:03.24     2:30
> 0:49:25.05  --  0:52:27.05     3:02
> 0:58:41.01  --  1:02:01.01     3:20
>
> It kinda looks to me like the commercial flagger is using the "real"
> keyframe positions, but the seektable is off, so when played the wrong
> part of the show is skipped.
>
> Any ideas why the seektable slowly gets more and more off with the DVB
> driver?

Interesting....

This may be related to the recent fix for this:

http://www.mythtv.org/bugs/show_bug.cgi?id=221

discussed here:

http://www.gossamer-threads.com/lists/mythtv/dev/121688

Assuming I'm reading your 'actual' vs 'detected' the right way around, 
then if the system was detecting bogus picture start codes in 
DTVRecorder::FindKeyframes() then you'd see what you're seeing.

There are a few ways to start debugging it:

a)

If you have a look at FindKeyframes() in libs/libmythtv/dtvrecorder.cpp 
you'll see a debug define:

#define DEBUG_FIND_KEY_FRAMES 0 /* set to 1 to debug */

Turn on that define and get a backend log as you record something.

b)

Apply the patch attached to the bug report above in reverse (patch 
--reverse) to current CVS.  See if the problem with DVB recording goes 
away.

c)

Try re-building the seek table with `mythcommflag --rebuild`.  While 
debugging bug #221, I found that mythcommflag did not fix the seek 
table correctly.  I haven't looked into why yet.

Be well,

Will         :-}

--
Dr William Uther                           National ICT Australia
Phone: +61 2 9385 6357               Computer Science and Engineering
Email: willu at cse.unsw.edu.au          University of New South Wales
Web: http://www.cse.unsw.edu.au/~willu/     Sydney, Australia