Re: [gnugo-devel] Extremely bad mistake

[Gunnar Farneback]

SP Lee wrote:
> I found the root of the problem. However, I can't provide a patch
> because I'm not familiar with the program structures yet.

Maybe you could describe how you came to your conclusion?

> In the reading process, attack and defend is carried out virtually

I suppose you are talking about the temporary and quickly reversible
moves performed by trymove()/tryko()/komaster_trymove() and undone by
popgo(). 

> and it seems the actual board position is unchanged.

If you by board position mean the contents of the board[] array you
are wrong.

> However when gnugo checks if A16 is a suicide move, it doesn't know
> there is already a stone at A18 and gets the conclusion A16 is a
> suicide move. This in turn makes the defence of C19 failed on the
> attack of A18.

I'm afraid you have misunderstood something very basic about the board
and tactical reading code in GNU Go. At least this description doesn't
make much sense to me.

> I think when checking suicide moves, gnugo must take into account of
> any virtual moves already played.

Of course it does. There's no way it could get any analysis correct if
it didn't.

/Gunnar