[mythtv] problem with eit-cache

Janne Grunau janne-mythtv at grunau.be
Tue May 5 00:34:46 UTC 2009


On Mon, May 04, 2009 at 09:10:37PM +0300, Otto Kolsi wrote:
> Janne Grunau wrote:
> > On Tue, Apr 28, 2009 at 08:57:42PM +0200, Markus Schulz wrote:
> >> I'm searching for a potential bug inside the eit-parser for the program 
> >> guide.
> > 
> > ...
> > 
> >> These two shows are running consecutively but the first has the _wrong_ 
> >> title but correct description.
> > 
> > I'm seeing this error too.
> 
> This is exactly what I've been seeing for some time now (months?).
> 
> What I noticed was that titles/descriptions got mixed when programs had 
> overlap. So the endtime of the first program could be e.g. 2sec - 45sec 
> later than the starttime of the second program. Can you check if you 
> have the same situation?

That is/was at least in some cases a trigger for this bug. A EIT fixup
might be actually to blame for introducing short overlaps (up to 10
seconds). If the seconds of the starttime is in the range :55 - :05 the
starttime is rounded to :00. The duration of the event is not changed.
This introduces short overlaps. Which shouldn't be a problem. A overlap
of a couple of seconds should be easily fixable.

> I verified this with myth and outside Myth with several different 
> applications (dvbsnoop etc). I also wrote some quick hackish code to fix 
> some of these overlaps. That fixed the problem for me and I haven't 
> actually checked whether overlaps or title/description corruption still 
> occurs.
> 
> I'd like to figure out what really is causing this and of cource to find 
> a proper fix for it :)

See [20485] or [20486].

> Janne, I think you mentioned briefly in IRC that 
> one possibility would be to allow disabling of the "overlap fixing 
> algorithm" in eit.cpp because we now have multirec and it doesn't hurt 
> so much if programs have small overlapps..

Well I thought the overlaps weree in the EIT data. I hope that the
fixing code is now mostly correct. I still have to think what to do with
EITFixup::TimeFix(). I'm leaning towards removing it.

Janne


More information about the mythtv-dev mailing list