Re: [gnugo-devel] matcher_status question

[bump]

> First, I think the result of the discussion about my matcher_status patch
> was that, as implemented, it would only ever return states of ALIVE, DEAD
> or CRITICAL, and never UNKNOWN.  however, I am getting all four.  Is this
> a big with my code, or is something else going wrong?  it's fairly
> repeatable -- just load up an early stage small game, and it'll show
> UNKNOWNs.

The matcher status is less likely to be UNKNOWN than the owl
status is to be UNCHECKED, which is why I recommended using
the matcher status than the owl status. In dragon.c around line 459
we learn that if the owl status is UNCHECKED and the matcher
status is DEAD or CRITICAL then the matcher status is made
UNKNOWN.

> Second, for some unknown reason GNU Go isn't finishing it's games, I don't
> think.  the sgf file attached is an example of such.  Any idea what's
> going on?

If I load this game, GNU Go generates a move at D8, which is the
only move to make 2 eyes. If I load it with --color black then
the move at E8 is generated. (Actually this lets W get a ko
so F6 is a better move.)

The SGF file ends with passes by both players, with W passing first.

Which color is GNU Go?

Dan