EXchess castling bug?

[Volker Pittlik]

EXchess (v5.01 beta) sometimes stops playing and loses on time.

Can someone confirm or deny my observation? All games played under Linux with xboard.

I found out one condition when it happens. In these games:

Code: Select all

[Event "Computer chess game"]
[Site "vlk-ubuntu"]
[Date "2006.08.10"]
[Round "2"]
[White "exchess"]
[Black "Arasan 9.3.1"]
[Result "*"]
[TimeControl "60+1"]
[FEN "r1bqk2r/1p2bpp1/p1nppn1p/8/3NP3/2N1B3/PPPQ1PPP/2KR1B1R w kq - 0 1"]
[SetUp "1"]

{--------------
r . b q k . . r
. p . . b p p .
p . n p p n . p
. . . . . . . .
. . . N P . . .
. . N . B . . .
P P P Q . P P P
. . K R . B . R
white to play
--------------}
1. Nxc6 bxc6 2. Qd4 O-O
*

[Event "Computer chess game"]
[Site "vlk-ubuntu"]
[Date "2006.08.10"]
[Round "1"]
[White "Arasan 9.3.1"]
[Black "exchess"]
[Result "*"]
[TimeControl "60+1"]
[FEN "rnbqkb1r/1p3ppp/p2ppn2/8/3NP3/2N1BP2/PPP3PP/R2QKB1R b KQkq - 0 1"]
[SetUp "1"]

{--------------
r n b q k b . r
. p . . . p p p
p . . p p n . .
. . . . . . . .
. . . N P . . .
. . N . B P . .
P P P . . . P P
R . . Q K B . R
black to play
--------------}
1... e5 2. Nb3 Be6 3. Qd3 Nc6 4. O-O-O
*

EXchess rejects the castling move and is losing on time.

from the debug:

Code: Select all

38833 <first : (debug) done pondering
38833 <first : (debug) ponder move = e1-c1
38833 <first : (debug) out of ponder()
38833 <first : (debug) sending ponder move
38834 <first : move o-o-o
Interrupting second
38835 >second: time 5297
otim 4517
38835 >second: e1c1
AnimateMove: piece 5 slides from 4,0 to 2,0
38971 <second:
38972 <second:
38972 <second:
38972 <second: Error(e1c1)
38972 <first : 10      18    296   850255 O-O-O Be7 Bg5 Nb4 Qd2 Bxb3 axb3 O-O Bxf6
38972 <first : (debug) predicted reply = f8-e7
38972 <first : (debug) starting to ponder
GameEnds(0, (null), 2)
Interrupting first
97521 >first : force
97521 >first : ping 2
97521 >second: quit
97543 >first : quit

If games are played manually under xboard this does not happen. Usually castling is no problem but under certain circumstances I don't know the error occurs.

I tried it with the precompiled version as well as with one compiled from the sources. There is no difference. The author is informed already.

regards

Volker

[Guenther Simon]

Hi Volker,


Did you use "setup board"? Or you made the moves from the initial position?

If the former then ExChess may not have interpreted the castling rights correctly.

I'm guessing of course.


Andrew.

My memory could be wrong, but IIRC once I tried to play with fixed
positions from collected FENs instead of a PGN(/lpf vs. /lgf), I think
I noticed that WB(xboard?) doesn't send the castling rights at all?

This should be easy to verify though from Volkers full debug,
from the position where xboard sends the start FEN.

Guenther

[Volker Pittlik]

... I think
I noticed that WB(xboard?) doesn't send the castling rights at all?

This should be easy to verify though from Volkers full debug,
from the position where xboard sends the start FEN.

Guenther

That's almost correct. EXchess does not use the "feature" command to tell xboard if it understands the setboard command.

Therefore xboard uses the "edit" command to set up the position. AFAIK it is not possible let the engine know about the castling rights that way. That explains the engine's behaviour.

Trying to hack it.

Volker

[Volker Pittlik]

...That explains the engine's behaviour.
...

Of course the engine understands the "setboard" command in the command line mode.

Volker

[Volker Pittlik]

To avoid this problem I have added the following lines in main.cpp:

Code: Select all

  else if(!strcmp(response, "xboard")) { xboard = 1; }
  //beginn feature command VP
  else if(!strcmp(response, "protover")) {
      cout << "\n";
      cout << "feature setboard=1 draw=0\n";
      cout << "feature variants=\"normal\"\n";
      cout << "feature myname=\"EXChess 5.01-beta(hacked)\"\n";
  }
  //end feature command VP
  else if(!strcmp(response, "build")) { build_book(game.pos); }


So far no new trouble related to that hack has occured here .

The castling moves in the setup position are now accepted by EXchess.

Volker