Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Comet captures its own King at e1 of 5th game in Comet-Crafty match

Author: Robert Hyatt

Date: 11:20:27 05/28/98

Go up one level in this thread


On May 28, 1998 at 11:38:29, Ulrich Tuerke wrote:

>On May 28, 1998 at 01:56:39, Komputer Korner wrote:
>
>>I don't get any Crafty flagging any more using WINBOARD 3.6.9  but it
>>still
>>doesn't use it's opening book in WINBOARD. However the subject is a new
>>phenomenon that I haven't seen since the missing a1 rooks that Crafty
>>caused
>>by the ..... info that Tim has now fixed when both programs ponder. I
>>will
>>send the log and debug files to Tim. After 1.f3xe1?????????????????e5
>>2.Nf3e4  Comet let it's flag fall by thinking for 10 minutes.  I guess
>>it
>>didn't know what to do without a king:)))))))).  Not only was the self
>>mate
>>king capture illegal but the Knight started the game at f3!!!!!!!!!!!.
>>--
>
>Meanwhile, Tim Mann was so kind to send me the debug file and his
>conclusions/assumptions about the above problem. (Of course, the move
>f3e1 referred to the previous game, and not to the new one).
>
>There seems to be a problem in communication between the new winboard
>and comet when the multi-game mode in winboard is used. In fact, this is
>completely untested in Comet because this feature comes with the new
>winboard version, just going to be released.
>
>As far as I understood from Tim, winboard uses a timeout of 5 seconds
>between interrupting the chess engine and the subsequent start of the
>new game. Tim's conclusion is that there seem to be cases where Comet
>needs more than 5 seconds to interrupt its search. The above problem is
>of this kind. Unfortunately, I am not able to produce situations where
>Comet can't be quickly interrupted. I can't reproduce the problem at
>all. When pondering or searching, Comet looks about twice per second if
>new input is present (using the routines suggested in the Crafty and
>Gnuchesx source code "PeekNamedPipe"). So 5 seconds should be more than
>sufficient.
>
>May be someone knows if it is possible that the windows pipes can have
>some peculiar timing behaviour ?
>
>Any hint is welcome.
>
>Regards, Uli


two issues.  #1 is to steal the peeknamedpipe() code from Crafty if you
haven't already. Tim wrote that and sent it to me and it has worked
without
glitch since installing.  #2 is that you *must* accept a "quit" command
whether you are pondering or searching, and you must quit *now*... not a
few seconds later.  If you don't quit now, you will break things, of
course,
as Tim explained...

I once had a bug where quit was accepted while pondering, but not while
searching...  and got the same kinds of problems..



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.