RE: [gnugo-devel] nando_3_15.1a

[Portela Fernand]

Dan wrote:

> Are we really confident the failures are accidental?

I think it must be related to the fact that fast_defense() sometimes finds
different defense points than previously returned by defendN(). And since
attack_and_defend() is called in many places and returns only one defense
point, it might very well change a couple other results elsewhere.

This is just a wild guess, do you want me to get the proof of this ? I'm
not too familiar with debugging tactical reading cases, but if you think
it's important, I'll try my best.

/nando