Winboard Forum

After the first ply in FICS fischerandom games, the engine (asmFish, Stockfish) requests a draw and refuses to play after the other engine refuses the draw. The FENs also look strange:

Code: Select all

[Event "ICS Rated wild/fr match"]
[Site "freechess.org"]
[Date "2017.08.07"]
[Round "-"]
[White "GrandDroid"]
[Black "FishTest"]
[Result "*"]
[WhiteElo "2616"]
[BlackElo "2778"]
[TimeControl "180"]
[Variant "fischerandom"]
[FEN "rkrbbnqn/pppppppp/8/8/8/8/PPPPPPPP/RKRBBNQN w CAca - 161+0 0 1"]
[SetUp "1"]

{--------------
r k r b b n q n
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
R K R B B N Q N
white to play
--------------}
1. Nhg3
{Game aborted on move 1} *

[Event "ICS rated wild/fr match"]
[Site "freechess.org"]
[Date "2017.08.07"]
[Round "-"]
[White "GrandDroid"]
[Black "FishTest"]
[Result "*"]
[WhiteElo "2616"]
[BlackElo "2778"]
[TimeControl "180"]
[Variant "fischerandom"]
[FEN "nqrkbnrb/pppppppp/8/8/8/8/PPPPPPPP/NQRKBNRB w GCgc - 167+0 0 1"]
[SetUp "1"]

{--------------
n q r k b n r b
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
N Q R K B N R B
white to play
--------------}
1. Nb3
{Game aborted on move 1} *

[Event "ICS rated wild/fr match"]
[Site "freechess.org"]
[Date "2017.08.07"]
[Round "-"]
[White "GrandDroid"]
[Black "FishTest"]
[Result "*"]
[WhiteElo "2616"]
[BlackElo "2778"]
[TimeControl "180"]
[Variant "fischerandom"]
[FEN "nbrnbkqr/pppppppp/8/8/8/8/PPPPPPPP/NBRNBKQR w HChc - 167+0 0 1"]
[SetUp "1"]

{--------------
n b r n b k q r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
N B R N B K Q R
white to play
--------------}
1. Nb3
{Game aborted on move 1} *


Substituting 4.9.1 with the stable release version (Winboard 4.8.0) I obtained correct fens and correct play by the engine:

Code: Select all

[Event "ICS unrated wild/fr match"]
[Site "freechess.org"]
[Date "2017.08.08"]
[Round "-"]
[White "FishTest"]
[Black "exeComp"]
[Result "1-0"]
[WhiteElo "2772"]
[BlackElo "3172"]
[TimeControl "180"]
[Variant "fischerandom"]
[FEN "rnbnkrqb/pppppppp/8/8/8/8/PPPPPPPP/RNBNKRQB w FAfa - 0 1"]
[SetUp "1"]

{--------------
r n b n k r q b
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
R N B N K R Q B
white to play
--------------}
1. g4 g5 2. Ne3 d6 3. c3 c5 4. h4 h5 5. gxh5 gxh4 6. d3 Na6
7. h6 Nc7 8. Qh2 h3 9. Nc4 Qh7 10. Be4 f5 11. Bf3 Be6 12. Qxh3 Nc6 13. Bf4
O-O-O 14. Nbd2 d5 15. Ne3 Be5 16. Qg3 Bxf4 17. Qxf4 Rd6 18. Rh1 Bg8 19.
O-O-O e5 20. Qh4 Be6 21. Rdg1 Rd7 22. c4 e4 23. cxd5 Nxd5 24. Nxd5 Bxd5 25.
dxe4 Bxa2 26. exf5 Nb4 27. Ne4 Nc6 28. Nd6+ Rxd6 29. Rg7 Qh8 30. Bxc6 Rf7
31. Rxf7 Bxf7 32. Qe7 Rxc6 33. Qxf7 Rf6 34. Qg7 Rf8 35. f6 Rg8 36. h7 Rxg7
37. fxg7 Qxg7 38. h8=Q+ Qxh8 39. Rxh8+ Kc7 40. Re8 Kd7
{exeComp resigns} 1-0

[Event "ICS unrated wild/fr match"]
[Site "freechess.org"]
[Date "2017.08.08"]
[Round "-"]
[White "exeComp"]
[Black "FishTest"]
[Result "0-1"]
[WhiteElo "3172"]
[BlackElo "2772"]
[TimeControl "180"]
[Variant "fischerandom"]
[FEN "bbrnqnkr/pppppppp/8/8/8/8/PPPPPPPP/BBRNQNKR w HChc - 0 1"]
[SetUp "1"]

{--------------
b b r n q n k r
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
B B R N Q N K R
white to play
--------------}
1. d4 d5 {-0.13/22 8} 2. b4 b5 {-0.22/23 13} 3. Nd2 and so on


Another, smaller bug in 4.9.1 dev: Exiting from FICS with the command 'quit' hangs Winboard; exiting by File -> Quit exits normally. Winboard 4.8.0 exits normally both ways.

Because of this bug, the latest Winboard 4.9.1 cannot be used for Chess960. I use Winboard 8 instead.
I suspect the problem lies in the patch from 13 Jan 2017

Implement check-count field in FENs

A field of the form N+M between e.p. field and ply counter is now associated
with the CHECK_COUNT game state on reading and writing. It is only written
when non-zero (which again should only happen in 3check). In addition a field
of the form +N+M after the full-move number is recognized on reading as the
SCIDB check counter, which indicates checks given.