[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] two BUGs
From: |
Arend Bayer |
Subject: |
[gnugo-devel] two BUGs |
Date: |
Mon, 17 Dec 2001 00:06:12 +0100 (CET) |
Here are two bugs that I stepped on while I was debugging my
move_reasons-patch (which should come with the next e-mail). The first
seems to be due to the reading cache; as I don't know that part of gnugo
yet, maybe someone else wants to have a look at it? The second one
is in move_reasons itself and should be easier to detect.
1. BUG:
Am I right that the cache size should only affect the time consumed,
but not the move valuation? If so, here is a bug:
The SGF file is the recent NNGS game looknglass-gnugo-3.1.16-200112142136.sgf,
the problem is with move N13 at move number 70.
With
gnugo -l looknglass-gnugo-3.1.16-200112142136.sgf -L 70 -t -d0x800
(which runs with default cache size 16) I get as output by move_reasons:
(...)
Move at N13 cuts L16 and Q15
Move at N13 strategically or tactically unsafe
(...)
N13: 0.83 - change in territory
N13: 1.79 - strategic effect on L16
N13: 2.44 - strategic effect on Q15
N13: 1.50 - added due to followup (3.00) and reverse followup values (0.00)
Move generation values N13 to 6.56
Instead, using
gnugo -l looknglass-gnugo-3.1.16-200112142136.sgf -L 70 -t -d0x800 --cache-size
150
the output is:
(...)
Move at N13 cuts L16 and Q15
Move at N13 defends N19
(...)
N13: 0.0 - defense of N19 (inessential)
N13: 5.29 - change in territory
N13: 1.79 - strategic effect on L16
N13: 2.44 - strategic effect on Q15
N13: 1.50 - added due to followup (3.00) and reverse followup values (0.00)
Move generation values N13 to 11.02
(I found this to be completely reproducible, the CVS version I used
had date -D 15 Dec 2001 23:05.)
2. BUG
This is directly visible in the DEBUG_MOVE_REASONS output. Either the
worms array or some move reason field gets corrupted resulting in output
like
Move at R19 attacks R17 (defenseless)
by list_move_reasons and later:
R19: 0.00 (secondary) - attack on [138199048] (defenseless)
This happens always in the case of an attack on a defenseless worm
(e.g. the same NNGS game looknglass-gnugo-3.1.16-200112142136.sgf, move 56).
I hope to find time tomorrow myself to have a look at this.
-Arend
- [gnugo-devel] two BUGs,
Arend Bayer <=