[gnugo-devel] Connection tuning

[bump]

I've revised my connection patch. You can find it at:

http://www.gnu.org/software/gnugo/patches/connections_3_6.1a

The big new thing is the inhibition of C patterns where find_cuts found
something. This is contingent on !EXPERIMENTAL_CONNECTIONS because I have not
tested whether this patch works well with experimental connections enabled.

- new DEBUG_MISCELLANEOUS used only in find_eye_dragons
- connection tunning
- B patterns matched by find_cuts inhibit C patterns found by find_connections
- new connection tests

The standard CVS gives the following results.

./regress.sh . 13x13.tst 
40 unexpected PASS!
./regress.sh . strategy5.tst 
228 unexpected PASS!

Beyond this the patch produces the following results. Here connection:90
is a new test that would have failed without the patch. Excluding this
there is a net gain of 6 tests.

./regress.sh . owl.tst
18 unexpected FAIL: Correct '1 A2', got '1 B2'
./regress.sh . connection.tst
19 unexpected PASS!
90 unexpected FAIL: Correct '0', got '1'
./regress.sh . nngs1.tst
39 unexpected PASS!
./regress.sh . trevorb.tst
220 unexpected FAIL: Correct 'J3', got 'F3'
380 unexpected PASS!
390 unexpected PASS!
./regress.sh . strategy2.tst
100 unexpected PASS!
./regress.sh . nicklas4.tst
809 unexpected PASS!
./regress.sh . trevor.tst
460 unexpected FAIL: Correct '1 E14', got '0'
./regress.sh . tactics.tst
5 unexpected PASS!
./regress.sh . nngs.tst
690 unexpected PASS!
./regress.sh . strategy3.tst
101 unexpected PASS!
./regress.sh . global.tst
10 unexpected FAIL: Correct 'A5', got 'H10'
40 unexpected PASS!
./regress.sh . vie.tst
37 unexpected PASS!
50 unexpected FAIL: Correct '1 H15', got '1 G15'
./regress.sh . arend.tst
33 unexpected PASS!
./regress.sh . 13x13.tst
40 unexpected PASS!
./regress.sh . trevord.tst
180 unexpected FAIL: Correct 'S9', got 'P5'
1050 unexpected PASS!
./regress.sh . strategy4.tst
166 unexpected FAIL: Correct 'D11', got 'D8'
./regress.sh . strategy5.tst
228 unexpected PASS!

Dan