On Sun, Sep 01, 2013 at 06:47:33PM +0200, Christian Marillat wrote:
> Craig Sanders <cas@???> writes:
>
> > On Mon, Sep 02, 2013 at 12:33:20AM +1000, Craig Sanders wrote:
> >> at first i though it was a problem with keyframe indexes, because
> >> moving
> >
> > looks like it is some problem with the index generation.
> >
> > if i rebuild the index for a recording with e.g.:
> >
> > mythtranscode --chanid 1021 --starttime 20130901093900 --showprogress --buildindex
> >
> > then editing and cut/transcode work properly again.
>
> May be this patch is the guilty ?
>
> https://github.com/MythTV/mythtv/commit/5b306f56a310b3358c75d372e8cbec9d84829189
i'm no expert on the myth source but it doesn't seem likely - there
isn't a cut list when the show is first recorded (and ad-scanned and
transcoded), so that "if (useCutList)" code doesn't get executed - and
the "Honoring the cutlist..." message is not logged.
it might be in mythcommflag. it's a shame that myth logging is so
difficult to read and analyse (putting the chanid_starttime in every
single log entry would help a lot for tracing what happens with a
particular recording), but the only errors i can see associated with the
recording above are lots of lines like:
Sep 1 19:54:16 indra mythlogserver: mythcommflag[13688]: E Decoder fileringbuffer.cpp:806 (Seek) FileRingBuf(/video/myth/1021_20130901092900.mpg): Seek(-1, SEEK_SET) Faile
d
the record worked OK, the initial transcode and keyframe generation
seemed to work OK, but then there's lots of errors on the commflagging -
1334 lines in the log for this recording.
then later when I rebuilt the keyframe index and transcoded again, there
were no errors.
i'll try changing my recording jobs so that they don't use commflag (as
i mentioned earlier, it hasn't worked reliably for ages so i've been
cutting out ads without the aid of mythcommflag anyway - most of my
recordings are on the commercial-free public broadcaster ABC anyway so
even when it worked i was mostly using it to detect the actual beginning
and end of the show) - and i'll have to write a script to rebuild the
keyframe indexes for all recordings for the last week or so.
it could be
http://code.mythtv.org/trac/ticket/11141
which fixes an off-by-one error in experimental commflagging, which
was closed about 4 weeks ago. I had the experimental methods enabled,
so i'll also try with them disabled.
running mythcommflag from the command line - with method=7 (all except
experimental methods) seemed to work OK.
so, looks like the lesson here is "don't use commflag methods marked as
experimental" :)
i can't remember when i turned that option on - probably several years ago
when i first started running mythtv.
craig
--
craig sanders <cas@???>
