Article View: gnu.chess
Article #4907Need HELP in debugging gnuchess
From: Johanes Suhardjo
Date: Fri, 17 Jun 1994 18:11
Date: Fri, 17 Jun 1994 18:11
55 lines
3000 bytes
3000 bytes
I'm just starting to take up chess programming as a hobby, which means
I'm a beginner, which means don't be too hard on me if my question is
stupid.
For some reason I pick up an older version of gnuchess (4.0 pl 17) to
start with. The gnuchessr that I compiled gives me a strange result.
These are the bestlines when it starts from the beginning board without
opening book (running on a Sun SPARC 10):
Move# 1 Target= 11243 Clock: 900000
1. 27 0 21 g1f3
2. 11 0 81 g1f3 g8f6
3& 29 0 144 g1f3 g8f6 b1c3
3. 29 1 605 g1f3 g8f6 b1c3
4& 12 1 1142 g1f3 g8f6 b1c3 b8c6
4. 12 1 1946 g1f3 g8f6 b1c3 b8c6
5& 21 2 3716 g1f3 g8f6 b1c3 b8c6 d2d4
5. 21 4 10607 g1f3 g8f6 b1c3 b8c6 d2d4
6& 13 7 23010 g1f3 g8f6 b1c3 b8c6 d2d4 d7d5
6. 13 7 23029 g1f3 g8f6 b1c3 b8c6 d2d4 d7d5
7- -9992 7 23031 g1f3 g8f6
^^^^^ why??? -> *
7& 20 7 23032 e2e3 g8f6
7. 20 8 23038 e2e3 g8f6
8& 21 8 23040 e2e3 g8f6
8. 21 8 23046 e2e3 g8f6
9& 22 8 23048 e2e3 g8f6
9. 22 8 23054 e2e3 g8f6
* Why does it return that huge negative score? If it encounters some
deadly line, it should be cut-off by the alpha-beta search, shouldn't it?
When I run it on an ancient IBM RT (the first IBM UNIX workstation),
it gives me the same thing, only that -9992 score appears at 6& move.
After that -9992 score, the computer basically chooses a move as it
likes, sometimes d2d4, sometimes d2d3, and even f2f4.
How do you guys debug such an anomality? I can't do it with a
debugger since it may take tens of thousands of search() calls before
it reaches that trouble spot. The latest I found is that all the
return statements in search() return normal values (I think), but
suddenly the line
node->score = -search (xside, ply +1, blahblahblah
returns
node->score == -9992.
How do I proceed from here? Thanks in advance.
--
Johanes Suhardjo (johanes@farida.cc.nd.edu)
--
CAUTION: Improper use of your brain may cause damages and void the
manufacturer's warranty.
Technical Details
Message-ID:
<2tsp1b$am1@news.nd.edu>Path:
arc1-ca.pugleaf.net!nntp.gmd.de!Germany.EU.net!EU.net!howland.reston.ans.net!gatech!newsxfer.itd.umich.edu!jobone!news1.oakland.edu!news.nd.edu!farida.cc.nd.edu!johanesReferences:
<2tpcf9$4s5@nuscc.nus.sg>Show Full Headers
Complete Article Headers
X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 10d71d,8eac18366caf9e5b X-Google-Attributes: gid10d71d,public X-Google-ArrivalTime: 1994-06-17 22:14:41 PST Path: arc1-ca.pugleaf.net!nntp.gmd.de!Germany.EU.net!EU.net!howland.reston.ans.net!gatech!newsxfer.itd.umich.edu!jobone!news1.oakland.edu!news.nd.edu!farida.cc.nd.edu!johanes From: johanes@farida.cc.nd.edu (Johanes Suhardjo) Newsgroups: gnu.chess Subject: Need HELP in debugging gnuchess Date: 17 Jun 1994 18:11:55 GMT Organization: Office of University Computing, U. of Notre Dame Lines: 55 Message-ID: <2tsp1b$am1@news.nd.edu> References: <2tpcf9$4s5@nuscc.nus.sg> NNTP-Posting-Host: farida.cc.nd.edu X-Newsreader: TIN [version 1.2 PL2]