Re: [gnugo-devel] snapback, connect.tst 23, 39

[Daniel Bump]

Teun wrote:

> > No, it has nothing to do in board.c. It has passed the boundary
> > between basic board manipulation and tactical reading. Should it be
> > moved anywhere it would be to reading.c, but there we already have
> > functionality to do this. To be more precise, if you want to do
> > snapback(str) you should instead do attack(str, NULL) and reverse the
> > result.
> 
> Then it would make sense to rewrite snapback in terms of already
> existing functions, add it to reading.c and reuse that in readconnect.c

The issue is that the readconnect.c algorithms require a more
precise version of the reading shadow than is used by the
reading code. See:

http://mail.gnu.org/pipermail/gnugo-devel/2001-November/000407.html

For this reason we can't use the tactical reading code in
readconnect.c as we would like to. Gunnar's patch in the cited
message came with a caveat "not for CVS", because of this
exact reason.

Dan