Re: [gnugo-devel] atari_atari problems

[Trevor Morris]

>Yes, I noticed those when looking through the test cases. It would be
>useful to have a list of all failed test cases involving atari_atari.
>It seems to me like these could without too many false hits be
>identified by matching the string "combination attack" in the trace
>outputs presented in the regression view. Trevor, would it be easy for
>you to do this string matching? Similarly it would be interesting to
>know which of the failures might be due to incorrectly detected
>blunders, which could be identified by matching the string "would be a
>blunder".

I created a listing that may help:
http://www.public32.com/games/go/special

You can view the problems on-line at:
http://www.public32.com/regress/?special=1

Anyone should be able to do this at home.  I suspect no one else
has tried regress.pl yet, due to extreme lack of documentation
and no bug reports!  If anyone's interested in trying, please let
me know, and I'll write something up.

-Trevor


For (my!) reference, here's the script I used to create the
listing:

#!/usr/bin/perl

use strict;
use warnings;

foreach my $file (<html/*.tst/*.trace>) {
  open (F, $file) or die "Couldn't open $file: $!";
  $file =~ /^html.(.*).tst.(.*).trace$/ or die "Funny trace file name: $file";
  my $prob = "$1:$2";
  my $printed = 0;
  while (<F>) {
    if (/would be a blunder/ || /defends against combination attack/) {
      print "$prob\n" unless $printed++;
      print "#   $_";
    }
  }
  print "#\n" if $printed;
  close(F);
}