Re: [gnugo-devel] how to patch regress 13x13:6

[kevin yong]

 --- Evan Berggren Daniel <address@hidden>
wrote: > Any chance you have patches for 13x13b:6 or
> 13x13b:8?

Hi, Evan:

  1. regard 13x13:6

  we identified B3 was suggested by 2 patterns D301 &
D1383. and you suggested solution:

D1383 shouldn't be matching because of the constraint.
 D301 appears to be the culprit.  Looking at the
pattern, it appears that the only way the stone could
be disconnected is a shortage of liberties. therefore,
adding a constraint like !oplay_attack(*,B) with B
being the two stones on the left would seem workable.
Adding a connection reading constraint or a b
classification to the pattern would work, but this
constraint should be faster to check than either of
those.

  originally, D301 looks like:

Pattern D301

X..?
O.*.
O..?

:8,E,value(80)

X..?
O.*a
O..?

;owl_escape_value(a) > 1

  then, i try to just add:



Pattern D301

X..?
O.*.
O..?

:8,E,value(80)

X..?
B.*a
B..?

;!oplay_attack(*,B)

X..?
O.*a
O..?

;owl_escape_value(a) > 1

  and compiler reports error.

  so, i comment out the old constraint:


Pattern D301

X..?
O.*.
O..?

:8,E,value(80)

X..?
B.*a
B..?

;!oplay_attack(*,B)

#X..?
#O.*a
#O..?
#
#;owl_escape_value(a) > 1

  compile ok. but when i run it, D301 still matched
with B3 for B5. i m not sure i have followed your
instructions correctly.

  i appreciate your help and like to hear more
suggestions from you.

kevin.



______________________________________________________________________ 
Post your free ad now! http://personals.yahoo.ca