Re: [gnugo-devel] corner matcher

[bump]

> Arend wrote:
> > Finally, it looks to me that your corner matchpat doesn't work for
> > joseki variations with captured stones. Is that right, or did I misread
> > the code?
> 
> hmm, what do you mean?

Here's a joseki variation where a stone has been captured. This
is the output of joseki which is then fed into mkpat -C to produce
the Joseki database. Note that a stone has been captured.

Pattern JK77

---------+
.........|
......O..|
...OXXO..|
...XOOXX.|
....*X.X.|
.....OX..|
.........|
.........|
.........|

:8,sFJ

I confirmed that this pattern gets matched in a testfile
containing this position.

(;GM[1]FF[3]
RU[Japanese]SZ[19]HA[0]KM[5.5]
PW[White]
PB[Black]
GN[White (W) vs. Black (B)]
DT[2003-02-03]
SY[Cgoban 1.9.12]TM[30:00(5x1:00)];B[qd]BL[1798];W[od]WL[1799];B[oc]
BL[1785];W[pd]WL[1798];B[pc]BL[1784];W[qe]WL[1797];B[nd]BL[1783];W[qc]
WL[1796];B[rd]BL[1782];W[qb]WL[1794];B[re]BL[1781];W[pf]WL[1793];B[qf]
BL[1780];W[nc]WL[1790];B[pe]BL[1780]
)

So I think Arend must be mistaken.

Another issue is the problem of memory corruption raised
in http://mail.gnu.org/archive/html/gnugo-devel/2003-01/msg00218.html.
This problem seems to have disappeared. I tried checking out
rel-3-3-16-pre-4 which was made close to the time of my post
but did not get these unexpected fails in nngs.tst and strategy5.tst.

I suppose what we can do is to use the patch with the modificiation
Paul suggested in:

http://mail.gnu.org/archive/html/gnugo-devel/2003-01/msg00224.html

Dan