[mythtv] Bug in 0.16 scheduler?

[Rob Walker]

On Thursday 14 October 2004 02:05 pm, David Engel wrote:
> On Thu, Oct 14, 2004 at 07:33:38PM +0200, Thomas B?rkel wrote:
> > David Engel wrote:
> > What I propose is, that it first should ignore all priorities and try to
> > schedule all shows sorted by starting time. And only if that causes a
> > conflict, it should go the full approach, as it does now (as explained
> > by Bruce). OK, maybe the tuner priority must be taken in account, but
> > not the show, recording type or channel priorities.
>
> Priorities, especially the implied ones, have to be taken into
> account.  Remember, one of the goals was to make the scheduler
> deterministic.  The implied priorities were chosen so the scheduler
> would generally do the right thing.  In your case, it didn't work out.

When I was reading the initial post, the first question I thought of was, 
"where do the tuner priorities get set?  It doesn't seem that the poster 
knows that he has one tuner at a higher priority than the other, nor that he 
wants it this way."

If it were possible to have each tuner set to equal value (what if they were 
both the same type of card, would it make sense then?), would that have 
solved the original problem with the current scheduler code?

If this problem was caused by a default setting of "lower numbered tuner cards 
have higher priority by default", then what about the following:

A new algorithm of "straightforward" could be used if a setting of "all tuner 
cards are same priority" was selected.  Would that solve the problem for the 
original poster?  Is is possible that there are other people who have this 
same problem, but don't know about it?

> > >sure the scheduler doesn't get into an infinite loop or take an
> > >inordinately long time when exhaustively trying every combination of
> > >program and tuner.

Yes, that's a problem, unless the tuner were re-written in lisp.  :-)  Maybe 
this would only take into account if there were identical tuners and 
identical recording settings.

> No, you're not talking nonsense.  What you are doing, however, is
> reinforcing my assertion that no matter how good the scheduler is,
> somebody will always be able to find a case where it doesn't do what
> they want.

Agreed.  I am also a fan of the bounty method of getting bugs fixed.  Make a 
list for people to commit some $$$ to their favorite bugs.

rob