Bug in ChessGUI

[Spock]

I've played 1,000 games with The Baron 2.23 FRC so far, and it has been totally problem-free. No time losses or anything unusual at all

[pedrox]

==============================================================================================
Eng1 (danasah416) <- 4 -354 1 1067 f2f1 a8c8 f1f2 c8c2
GetWB_InfoValues() threw a non-fatal exception.
Eng1 (danasah416) <- 5 -360 1 3611 f2f1 a8a7 e2b5 a7a5 b5e2
GetWB_InfoValues() threw a non-fatal exception.
Eng1 (danasah416) <- 6 -360 1 4231 f2f1 a8c8 e2b5 d5e5 h3g4 h5g4 f1e2
Eng1 (danasah416) <- 7 -359 1 5427 f2f1 a8c8 e2b5 d5e5 f1f2 c8c2 b5e2 c2a2
Eng1 (danasah416) <- Pipe01Time 0000001041021171
Exact move time for danasah416
Eng1 (danasah416) <- 8 -366 1 6998 f2f1 a8c8 e2b5 d5e5 b5a4 c8b8 h3g4 h5g4 f1e2
Eng1 (danasah416) <- 9 -365 1 8701 f2f1 a8c8 e2b5 d5e5 f1f2 c8c2 f2f1 e5e6 b5e8 g4h3 g2h3
Eng1 (danasah416) <- 10 -371 3 27803 f2f1 a8c8 f1f2 c8c2 f2f1 d5e6 f1f2 e6e5 f2e1 c2a2
Eng1 (danasah416) <- move f2f1
Cancelling GameMoveTimerEvent() for danasah416 at 0000001041021187.
==============================================================================================
Current position (Ply 158) : r7/8/8/3k1p1p/4p1p1/4P1PP/4BKP1/8 w - - = r7/8/8/3k1p1p/4p1p1/4P1PP/4BKP1/8 w - -
Eng1 (danasah416) time lag = 16 milliseconds, max. time lag= 94 milliseconds
Eng1 (danasah416) thought for 125 milliseconds, time left = -3 milliseconds

This game (FRC) is played at 40 moves in 6 seconds repeating. Here DanaSah lost in time by 3 milliseconds.

But there is a request from ChesGUI deep 7, however if the time taken by chessGUI not exceed 1 millisecond chessgui left to play danasah until that changes, this occurs at depth 10, danasah it took 1+1+3 milliseconds, without them possibly have passed over time control.

[pedrox]

I'm playing games without problems, I think I've played 700 games without a crash or loss of time with games of 1 minute and games 1 minute and 1 second of incr and games 40 moves en 1 minute (Now I check 32 times more if time is up).

In a game there is a loss of time for danasah, but really it is not, in the last move danasah capture a piece and both players remain with the king and bishop, the bishop is the same color, danasah sends the result draw to chessGUI, but the program continues to play and move a piece, then danasah not play and lost in time. ChessGUI then reviews the results and gives draw, but should not indicate the termination time.

In a endgamel of kings and bishops of the same color can be claimed draw.

Another little problem, sometimes when I go to play a tournament, I go to the Game menu, Engine Tournament and select Normal Chess and do nothing, I have to repeat this a second time and it works, it happens to someone else?

Pedro

Can you test DanaSah 3.97b in FRC at "40moves/1min repeating (300 games) in ChessGUI ?
Make it as gauntlet and change start positions randomly per game and alternate colours.
Maybe the rare crash is related to FRC.

If you have a ChessGUI debug of the draw, send it to me.
ChessGUI basically ignores all game results from engines except resign.
I know that the WB protocol behaves differently. Maybe I shall provide an option to behave like Winboard.

Starting a new tournament twice before it works is a bug I am aware of.
When I have enough time, I shall trace and fix it.

Matthias.

1) Make it as gauntlet and change start positions randomly per game and alternate colours. (300 games FRC)

In this case when chessgui play 3 games, it no play more, no crash of any of the 2 engines, or crash chessgui.

2) Make it as gauntlet and change start positions randomly per cycle and alternate colours. (300 games FRC)

In this case all games are played. But I have 3 games to crash, the game ends in 15, 18 moves, etc, away from the control of time, a windows screen indicates an error in danasah.exe

The file EngInfo1 = EngInfo2 and EnOupt1 = EngOupt2. I send this files by e-mail.

Pedro

[Matthias Gemuh]

==============================================================================================
Eng1 (danasah416) <- 4 -354 1 1067 f2f1 a8c8 f1f2 c8c2
GetWB_InfoValues() threw a non-fatal exception.
Eng1 (danasah416) <- 5 -360 1 3611 f2f1 a8a7 e2b5 a7a5 b5e2
GetWB_InfoValues() threw a non-fatal exception.
Eng1 (danasah416) <- 6 -360 1 4231 f2f1 a8c8 e2b5 d5e5 h3g4 h5g4 f1e2
Eng1 (danasah416) <- 7 -359 1 5427 f2f1 a8c8 e2b5 d5e5 f1f2 c8c2 b5e2 c2a2
Eng1 (danasah416) <- Pipe01Time 0000001041021171
Exact move time for danasah416
Eng1 (danasah416) <- 8 -366 1 6998 f2f1 a8c8 e2b5 d5e5 b5a4 c8b8 h3g4 h5g4 f1e2
Eng1 (danasah416) <- 9 -365 1 8701 f2f1 a8c8 e2b5 d5e5 f1f2 c8c2 f2f1 e5e6 b5e8 g4h3 g2h3
Eng1 (danasah416) <- 10 -371 3 27803 f2f1 a8c8 f1f2 c8c2 f2f1 d5e6 f1f2 e6e5 f2e1 c2a2
Eng1 (danasah416) <- move f2f1
Cancelling GameMoveTimerEvent() for danasah416 at 0000001041021187.
==============================================================================================
Current position (Ply 158) : r7/8/8/3k1p1p/4p1p1/4P1PP/4BKP1/8 w - - = r7/8/8/3k1p1p/4p1p1/4P1PP/4BKP1/8 w - -
Eng1 (danasah416) time lag = 16 milliseconds, max. time lag= 94 milliseconds
Eng1 (danasah416) thought for 125 milliseconds, time left = -3 milliseconds

This game (FRC) is played at 40 moves in 6 seconds repeating. Here DanaSah lost in time by 3 milliseconds.

But there is a request from ChesGUI deep 7, however if the time taken by chessGUI not exceed 1 millisecond chessgui left to play danasah until that changes, this occurs at depth 10, danasah it took 1+1+3 milliseconds, without them possibly have passed over time control.

I think you don't need to worry about time losses in this version of DanaSah.
I consider it normal for an engine to overstep time by 3 ms at move 80 and
time control 40moves/6sec.

The debug shows that:
- at time 0000001041021171, ChessGUI saw "move f2f1" in input pipe.
- ChessGUI processed the move at time 0000001041021187 (16 ms later).
- it is the 0000001041021171 that ChessGUI uses to determine whether game is
lost on time.

[Matthias Gemuh]

1) Make it as gauntlet and change start positions randomly per game and alternate colours. (300 games FRC)

In this case when chessgui play 3 games, it no play more, no crash of any of the 2 engines, or crash chessgui.

2) Make it as gauntlet and change start positions randomly per cycle and alternate colours. (300 games FRC)

In this case all games are played. But I have 3 games to crash, the game ends in 15, 18 moves, etc, away from the control of time, a windows screen indicates an error in danasah.exe

The file EngInfo1 = EngInfo2 and EnOupt1 = EngOupt2. I send this files by e-mail.

Pedro

1) that is the issue that started this thread, relevant only after a loss on time. I still need to investigate whether the bug is 100% fixed.

2) there is definitely a rare bug in latest DanaSah version.
The question is whether this affects only FRC or normal chess also.
I thought I once saw it in normal chess also, but I am not sure anymore.
The debug you sent does not help because the bug happens deep inside DanaSah's search.

Best,
Matthias.

[Matthias Gemuh]

This link will be valid only till tomorrow evening:

http://www.hugedrive.com/published/WG/s ... =-5d300fe1

it's a current bugfixed ChessGUI.exe !
Official next ChessGUI package shall follow, after a long test.

Matthias.

ChessGUI 0.135 is now available.
Everybody is invited to hunt for bugs in it.
You still have to start new tournaments twice sometimes.
GUI will automatically hunt for engine logos.

Matthias.

[pedrox]

200 games FRC, gaunlent, alternate colours, random per Game.

The tournament stop in game 3, game 3 is ok.

Code: Select all

Point 4 in NextGame().
Point 7 in NextGame().
Now in GetGauntletPairing() at Time0000000013036984


Game Nr. 4  (Hermann 2.4  <==*==> danasah416) in Tournament "chess960-1"
Hermann 2.4
danasah416
Cycle = 2
Now in DisplayGameSchedule() at Time0000000013036984

Now in DisplayRoundRobinTable() at Time0000000013037015

GetPrivateProfileString() no value for StartPositionFile in chess960-1 (D:\Documents and Settings\pedrox\Escritorio\ChessGUI_net\TOURNAMENTS\chess960-1.e2e), nResult = 0, OutStr = 

[Matthias Gemuh]

200 games FRC, gaunlent, alternate colours, random per Game.

The tournament stop in game 3, game 3 is ok.

Code: Select all

Point 4 in NextGame().
Point 7 in NextGame().
Now in GetGauntletPairing() at Time0000000013036984


Game Nr. 4  (Hermann 2.4  <==*==> danasah416) in Tournament "chess960-1"
Hermann 2.4
danasah416
Cycle = 2
Now in DisplayGameSchedule() at Time0000000013036984

Now in DisplayRoundRobinTable() at Time0000000013037015

GetPrivateProfileString() no value for StartPositionFile in chess960-1 (D:\Documents and Settings\pedrox\Escritorio\ChessGUI_net\TOURNAMENTS\chess960-1.e2e), nResult = 0, OutStr = 

Hi Pedro,

Was game number 2 a lose on time or was result of game 2 "unknown"?
Official ChessGUI 0.135 will not stop even after an engine loses on time, all those time loss problems are now fixed.
No "unknown" results anymore.

If you want to continue searching for the crash bug in DanaSah, I suggest that you try using not ChessGUI but Arena instead.
After Arena, come back to ChessGUI

Best,
Matthias.

[pedrox]

200 games FRC, gaunlent, alternate colours, random per Game.

The tournament stop in game 3, game 3 is ok.

Code: Select all

Point 4 in NextGame().
Point 7 in NextGame().
Now in GetGauntletPairing() at Time0000000013036984


Game Nr. 4  (Hermann 2.4  <==*==> danasah416) in Tournament "chess960-1"
Hermann 2.4
danasah416
Cycle = 2
Now in DisplayGameSchedule() at Time0000000013036984

Now in DisplayRoundRobinTable() at Time0000000013037015

GetPrivateProfileString() no value for StartPositionFile in chess960-1 (D:\Documents and Settings\pedrox\Escritorio\ChessGUI_net\TOURNAMENTS\chess960-1.e2e), nResult = 0, OutStr = 

Hi Pedro,

Was game number 2 a lose on time or was result of game 2 "unknown"?
Official ChessGUI 0.135 will not stop even after an engine loses on time, all those time loss problems are now fixed.
No "unknown" results anymore.

If you want to continue searching for the crash bug in DanaSah, I suggest that you try using not ChessGUI but Arena instead.
After Arena, come back to ChessGUI

Best,
Matthias.

No.

The three games were won by danasah with termination checkmate. If the game 3 had ended with a loss of time or crash could not have put the message.

The bug is not in danasah, as it also happens when not playing danasah.

But this not only happens once, it is always reproducible on my computer each time I select an gaunlent FRC, alternate colors and random per game. With random per cycle all is well.

Perhaps the bug is because in the gaunlent just 2 engines, per example Hermann 2.4 y Ufim 8.02?

In Arena I can not find the bug nor crash, as I've played thousands of games in the normal and FRC variant.

Pedro

[pedrox]

Gauntlet FRC, Alternate colours, random per game, now 3 engines, 100 Cycles, 400 games.

Stop after game 5.

Best,

Pedro