Re: [gnugo-devel] Trevor's regression view

[Trevor Morris]

The problem has to do with embedded top_move_* commands by the
regress.pl viewer.  Running the regression w/out the top_move
commands inserted works correctly.  I'm working on an improved
regress.pl, but it's not ready to put up yet.  I've not seen
any problems with the regressions run the old-fashioned way.

Here's a simple way to reproduce the problem:


loadsgf games/filllib8.sgf 303
= white

27 gg_genmove white
=27 Q16

loadsgf games/filllib7.sgf 72
= black

25 gg_genmove white
=25 H1

top_moves_white
= J3 15.80 H1 10.75

26 gg_genmove black
=26 J5

top_moves_black
=

loadsgf games/filllib8.sgf 303
= white

27 gg_genmove white
=27 C19



At 09:33 PM 10/5/2001 +0200, Gunnar Farneback wrote:
>I notice that Trevor's regression view at
>http://www.public32.com/regress/ includes some occurences of FAILED
>test cases. Those are unexpected and nothing I'm able to reproduce.
>Assuming he hasn't made any private changes breaking the engine, this
>seems to indicate some portability problem.
>
>I want to focus on the test case filllib:27. The detailed view says
>GNU Go wants to play at C18, which there's really no reason for. I'd
>be interested in seeing the trace output when running the engine on
>that position with the options "-w -t -d0x4000".
>
>Has anyone else encountered unexpected regression failures with 3.1.8?
>(Current CVS should have a couple unexpected failures together with
>quite a few unexpected passes, though not the particular unexpected
>failures on Trevor's page.)
>
>/Gunnar
>
>_______________________________________________
>gnugo-devel mailing list
>address@hidden
>http://mail.gnu.org/mailman/listinfo/gnugo-devel
>