Hi, first time posting!

[Blackted]

I have noticed a problem with Winboard, both 4.2.x and 4.3.x, when I am using it as an interface to a variant playing chess engine on ICC.

While playing the W26 variant, giveaway chess, on ICC, if the opponent castles I get the following error in the winboard window:

Couldn't parse move "O-O"

Does anyone have any clues as to what is cauing this? Also i can't seem to get the debug parameter to work.

Ted Black

[Blackted]

Just to clarify, the error message is:

Couldn't parse the move "O-O" from ICS.

I'm not sure if its aproblem with Winboard or the engine, but is does this with both engines I am running, Pulsar and Sjeng.

[Blackted]

Finally worked out debug mode for Winboard, here is the output I am getting from the debug file:

load 8x8 board
parseboard 7, castling = -1 -1 6 7 0 4
accepted move O-O from ICS, parse it.
moveNum = 7
board = 0-8 x 8
move to parse: O-O
short castling 4 6
7 0 4 7 0 4 Legality test? e1g1
Illegal move from ICS 'O-O'
board L=0, R=8, H=8, holdings=0
Move parsed to 'O-O (0:02)'
nps: w=-1, b=-1

[H.G.Muller]

Hmm, from the debug file it seems that WinBoard thinks the white Rooks do not have castling rights (castling -1 -1 ...). I woud have to see the entire debug file to see where exactly they get lost. The castling rights do look fishy indeed, because the file of the white King (6) is different from that of the black King (4). It seems like WinBoard might derive the castling rights fom the board position as it is after castling, which it is not supposed to do, and certainly not in this variant. (For FRC some tricks are needed, as ICC sends you wrong castling rights there.)

Can you mail the entire debug to h.g.muller AT SIGN hccnet (DOT) nl?

[Blackted]

Thanks for your reply. The Debug file has been sent as an email attachment.

[PulsarMike]

HI,

It took me a minute to think about this as i have not run giveaway on icc in over a year. But here is what i recall. Winboard is more set up for fics rules or suicicide not giveaway. there is no castling in fics rules or suicide. when icc implemented suicide they left the castle move intact which was not apart of the game before. the work around i found was when running giveaway on icc, turn off move legality. then it works to let you castle as long as the engine supports it, which pulsar does.

Mike

(ps took me a bit to realize i had a message)

[Blackted]

Thanks Mike for your reply. Turning legality check off in Winboard seems to do the trick.
Castling is allowed in Giveaway!

[H.G.Muller]

OK, I was confused first by the debug output. It says castling rights = -1 -1 6,, meaning that white castling is disallowed, but this was only for the position of board 7, which is send together with the move that was done from board 6. Indeed, on the line with "Legality test" it does print 7 0 4 7 0 4, which means full castling rights for both Rooks (file 7 and 0) and King (in file 4). It is these rights that wer used in the Legality test.

The problem is indeed as Mike remarks: it considers castling illegal in Giveaway. Apart from the castling rights that I programmed in (which are updated move to move) there is a now-obsolete mechanism, that defined flags for all castling rights that stayed constant during the entire game. (Only useful to suppress castling in games that do not have it, like Shatranj). It is on testing these flags that WB decided the castling was illegal, although the rights I defined (essentially keeping track of which pieces moved) formed no obstacle.

The question is what to do about this? Currently WB treats Suicide and Giveaway exactly the same, setting the IGNORE_CHECK flag and clearing all casling-right flags. Would it break anything if I enabled castling in Giveaway? I guess it never hurts to be too liberal; in the worst case the ICS would refuse the move, and WB would take it back, and (playing as a Human) you would have lost a bit of time compared to WinBoard not allowing you to make the move in the first place. I guess no FENs with castling rights are send to the engine that could confuse it, unless you would try to play Giveaway from a non-standard opening position. And on servers that implement Giveaway without castling, I suppose they would also send boards specifying 'no castling rights', and WB would simply follow that.

So in the next release, I will remove the clearing of castling rights in Giveaway.

[PulsarMike]

I don't know if this is also an issue but from the icc help file http://www.chessclub.com/help/Giveaway

Castling is sometimes legal; just ignore the check considerations, since there is no concept of "check" in Giveaway.

that means in giveaway you can castle while in check, like icc atomic, I dont know if not clearing the flags would deal with the special check rule. If you allowed castling but prohibited it in check, it would not completely follow the giveaway rules. keep in mind that other than the castle move being allowed in giveaway, in giveaway and suicide the king is just another piece no different than a pawn or knight and can be captured, so it makes sense check wouldnt matter. check never matters in giveaway.

Also another differece is when you have no more moves, which happens when pawns meet each other and cant progress. like all that is left is e4e5 pawns. on icc if you have no moves you win ( i think its assumed by the geometry since there is no check rule that both sides must be out of moves.) on fics if there are no more moves a count is done of the remaining pieces. he with the fewest pieces wins. Imagine a white pawn on e3. black has two pieces left and they are on e5 and e6. white moves e3e4 locking pawns now at e6e5 ( 2 black pawns ) and white at e4. black to move, he has no moves. white had last move. on icc black wins cause he ran out of moves first. in suicide white wins becaue he has only one piece at the time moves run out and black has two. who ran out of moves first doesnt matter in suicide. I dont recall if a tie is a draw or it goes to the player who ran out first.

In ics play this is not an issue i've found and the server handles game end. I dont know if its an issue in at home play. Its also not a big issue though. I've met fics players who say they dont care if the fics or icc rules are used. Some suicide players prefer the icc rules actually of giveaway on this. but I wanted to be complete on the differences.

Mike

[PulsarMike]

with regard to the game end with no moves, it occurs to me that 2 pawns could lock and there is one knight on board so one guy runs out of moves. he wins in both giveaway and suicide. its only when both run out of moves at the same time that the giveaway and suicide rules diverge with icc giveaway giving the win to the first guy to run out of moves and fics suicide ( and bear in mind servers doing fics rules exist beyond fics and can use winboard ) the remaining pieces are counted with the fewest pieces winning if both run out of moves on consecutive turns.

Mike

[Blackted]

Ok, thanks guys for taking the time over this.

I noticed that turning legality check off in WB sometimes causes problems with Crazyhouse when placing a piece on the board, so i guess I have to be careful about what variants my comp account is playing on ICC when setting legality on or off.

I have used WB with Sjeng playing Losers and Crazyhouse for ages without problem on ICC, and Pulsar too. Maybe I will stick to those variants for now.

Cheers

Ted Black