🚀 go-pugleaf

In article <3a0144f2.54852587@news>, Henrik Gram <henrikg@funcom.com> wrote:
>mann@milwaukee.pa.dec.com (Tim Mann) wrote:
>
>>Simon Waters wrote:
>>> Looks to me like perhaps it tried to play 1. ... c6 and got
>>> misunderstood
>>
>>I don't see how xboard could misunderstand GNU Chess 5's attempt to play
>>c6 (or c7c6) as Nc6 (or b8c6).  What seems a bit more likely is that
>>GNU Chess 5 played c6 internally but sent something different to xboard.
>
>I suspect that too. Btw, this reminds me of a mistake I made myself
>when making the lan/san move parser for my engine.  The move: b8d7 (as
>sent by xboard) to me first went through the various san cases before
>lan, which made the engine think that b8d7 meant Bishop on 8th rank to
>d7, which clearly isn't the case :)   With olithink, it was later the
>night at d7 that moved diagonally to b5, so I think the engine
>misinterpreted the move b8d7 just as mine once would. Perhaps this too
>is a problem with gnuchess 5 ?

Interesting point, but I don't think you can make a theory out of it that
explains what happened in this game.  I don't see a place where gnuchess could
have misinterpreted a move of the user's to cause this, do you?

By the way, xboard handles the "b" vs. "B" cases basically by insisting that
you have to get the case right.  So "1. b4 e6 2. B1a3" will move the bishop,
while "1. b4 e6 2. b1a3" will move the knight, and "1. b4 e6 2. Nc3 Nf6 
3. b1a3" will call b1a3 an illegal move (trying to move an empty square).
I'm not sure this is the ideal solution, but so far I've had few if any
complaints from people who want to be able to use "b" for a bishop or "B"
for the b-file.

It's not obvious at all from the xboard code that this is what is happening.
The patterns in parser.l allow lower case piece names, though not upper case
coordinates, so b1a3 matches both the pattern for coordinate notation and the
pattern for a SAN bishop move with rank disambiguator.  But the coordinate
notation pattern appears first in the .l file, so that's the one that actually
fires.  Ick...

	--Tim
  

-- 
Tim Mann  tim.mann@compaq.com  http://www.tim-mann.org
Compaq Computer Corporation, Systems Research Center, Palo Alto, CA