Re: [gnugo-devel] Another reading patch

[Arend Bayer]

Evan wrote:

> This patch applies on top of evan_3_10.1.
>
> - defend_both now returns ko results properly.
> - defend_both attempts to upgrade ko results.
> - defend_both uses attack_either() instead of !attack().
> - comments in attack_either and defend_both updated.
>
> I mentioned on nngs that the defend_both patch didn't do much, and that
> attempting to upgrade ko results in attack_either helped.  Turns out I had
> it backwards.
>
> Regression delta is
>
> 13x13:40 PASS
> 13x13:42 PASS
>
> Thanks

Evan, the important part (and the hard work) is not writing such a
patch, it is evaluating it. In those positions above, it is not at all
clear where a modified defend_both/attack_either would make a
difference. Well, actually, for 13x13:42 I could guess where
attack_either might find s.th., but there it would be wrong.

Modifying attack_either/defend_both to be recursive (this is what you
actually do) _might_ be a good idea, but it needs careful evaluation.
You should at the very least be able to point out some positions where
the old defend_both resp. attack_either misses defenses resp. attacks,
and your revised version finds them.

As it is there is no telling whether your patch might contain stupid
bugs. Of course I could look at those positions and find out where the
difference comes from, but I don't see why I should do that work. And
you would have to do that anyway, as it might tell you where your patch
needs improvement.


Arend