[mythtv] Re: [PATCH] Soft padding proof of concept

[Bruce Markey]

David Shirley wrote:
> Hi David,
> 
> actually with the current SVN (no soft3.patch) there are no conflicts
> for the back-to-back recordings.

That is a matter of circumstance. What happens when there are
two tuners and four shows that overlap each other has been reported
five times before and this sixth time is has nothing to do with
padding time per se.

> On 10/16/05, David Engel <gigem at comcast.net> wrote:
> 
>>On Sat, Oct 15, 2005 at 08:38:16PM +1000, David Shirley wrote:
>>
>>>i would have assumed that the scheduler would have picked card 2 for
>>>#2, then card 1 for #3 as it did, but when i add #4 it should just use
>>>card 2, instead it conflicts #3 and uses card 1.

1) Mysterious Island         19:40-23:00 --+
2) Payback                   21:39-23:30 --|--+
3) Executive Command         23:00-00:40 --+  |
4) Tri Nations Rugby League  23:30-01:30 -----+

1 overlaps 2 which overlaps 3 which overlaps 4. 1 and 3 will fit on
the same card as will 2 and 4. However. all four have the same priority
so it's up to the tie breaking to determine which will be placed first
then next. See http://www.mythtv.org/docs/mythtv-HOWTO-11.html#ss11.7
under "Scheduling decisions".

It is a matter of circumstance whether 3 or 4 will get the same
card as 1. If 3, they fit. If 4 is placed first, 2 and 3 can't
fit on the same card. MoveHigherRecords() wasn't (past tense)
able to handle this specific case.

>>You're being bitten because the scheduler only tries one level of
>>redo.  This is most apparent when there are tightly packed programs
>>with no later showings available and happens even without soft
>>padding.

Correct. This is not an issue introduced by soft padding and is
entirely a diversion.

>>  You'll need to tweak priorities and/or use overrides to fix
>>these cases manually.

The solution to this has always been to tweak the priority of one
of them to force 1 and 3 or 2 and 4 to pair up rather than leaving
it to chance. It's 50-50 for 3 or 4 to be paired with 1 and any
number of changes could reverse it. Note that this is a very
specific case. This isn't an issue with one card, or three or more
cards, or three overlaps, or if 1 and 3 overlap or 2 and 4 overlap,
or any of the consecutive shows don't actually overlap.

Changeset 7490 "Fixed a bug where TryAnotherShowing might not consider
other programs at the same time on another input" does appear to
fix this.

One card:
--- print list start ---
Title - Subtitle                    Chan ChID Day Start  End   S C I  T N Pri
One (Manual Record) - "Mon Oct 17 0   10 1010  17 01:00-03:00  1 1 1  S 1   1
Two (Manual Record) - "Mon Oct 17 0   20 1020  17 02:00-04:00  1 0 0  S C   0
Three (Manual Record) - "Mon Oct 17   30 1030  17 03:00-05:00  1 0 0  S C   0
Four (Manual Record) - "Mon Oct 17    40 1040  17 04:00-06:00  1 1 1  S 1   1

One and Four win with the higher priorities.

Two cards:
One (Manual Record) - "Mon Oct 17 0   10 1010  17 01:00-03:00  1 1 1  S 1   1
Two (Manual Record) - "Mon Oct 17 0   20 1020  17 02:00-04:00  1 2 2  S 2   0
Three (Manual Record) - "Mon Oct 17   30 1030  17 03:00-05:00  1 1 1  S 1   0
Four (Manual Record) - "Mon Oct 17    40 1040  17 04:00-06:00  1 2 2  S 2   1

Three and Four are flipped. Three gets card 1 rather than Four so
that Three will not be marked as a conflict.

--  bjm