Re: [gnugo-devel] Owl threats

[Daniel Bump]

> There's a bug in the patch giving the effect that also worm threats
> have been disabled, which I don't think was the purpose. This breaks
> e.g. Arend:36. Trevor later added a pattern CF3 in endgame.db to solve
> this test case, but with worm threats working the pattern is
> redundant.

About the worm threats, this isn't supposed to happen! 

The variables owl_threats and disable_threat_computation have different
meanings because once set, owl_threats doesn't change, but
disable_threat_computation gets turned on and off by aftermath.

> The patch below fixes the bug and makes the
> disable_threat_computations variable (needed by aftermath) independent
> from the owl_threats variable.

It might simplfy things if you don't put this in the CVS yet.
I'm in the process of breaking the owl loop in dragon.c into
two, separating the owl threat loop. This simplifies the code.
Also I'd like to run the owl threat loop after semeai(). The
proximate reason for this change is that we'd like to run
the owl threats for a thrashing dragon, that is, a dead
stone just played by the opponent.

After breaking this loop into two we can fix the problem you found.

At the moment if I compile with --owl-threats there is a
crash in trevorb:280.

Dan