TalkChess.com

I had this problem today. EXchess 7.96 failed to move and lost on time, but the position was drawn by 3 fold repetition, according to Stockfish analysis engine (and myself if I read the rules correctly)
It was 40 moves in 2 minutes repeating in Arena, so plenty of time. Several other games in the 20 game series between the 2 engines ended correctly due to 3 fold repetition.
[pgn]
[Event "My Start Positions"]
[Site "A3700"]
[Date "2017.02.03"]
[Round "6.2"]
[White "EXchess 7.96"]
[Black "EXchess 7.31b"]
[Result "0-1"]
[ECO "B07y"]
[TimeControl "40/120:40/120:40/120"]
[PlyCount "34"]

1.e4 d6 2.d4 Nf6 3.Nc3 g6 4.Be2 Bg7 5.g4 Na6 6.g5 Nd7 7.h4 c5 8.d5 c4 9.h5 Nac5 10.h6 Bxc3+ 11.bxc3 f6 12.Qd4 Qa5 13.f3 b5 14.Bd2 Qa3 15.Bc1 Qa5 16.Bd2 Qa3 17.Bc1 Qa5
White forfeits on time
0-1
[/pgn]
After correcting the above result, the contest finished 13.5 : 6.5 in favour of v7.96

Hi Dan,

First I deleted the search par file from 7.96. Suddenly the program was running on 1 CPu.

Then I copied the search file from version 7.88 in the 7.96 folder. The program crashed again.

Without search par file 7.96 is running but with 1 cpu. Not with 4 or 8.
With search par file it crashes immediately.

Colin-G wrote:I had this problem today. EXchess 7.96 failed to move and lost on time, but the position was drawn by 3 fold repetition, according to Stockfish analysis engine (and myself if I read the rules correctly)

Looks like Arena does not know that a Pawn double-push does not create e.p. rights if there are no Pawns that can actually perform the capture.

Of course EXchess claims a draw, but when you run it through Wb2Uci, this claim cannot be passed to the GUI.

Of course using Uci2Wb to run WB engines on Arena is unwise. Arena supports WB protocol directly, and channeling the communication through the less capable UCI protocol only loses functionality.

hgm wrote:...

Of course using Uci2Wb to run WB engines on Arena is unwise. Arena supports WB protocol directly, and channeling the communication through the less capable UCI protocol only loses functionality.

You're right H.G.
It makes no sense (at least it is unnecessary...) to run a WB-Engine as UCI under Arena with an adapter.

I tried 32bit as WB-Engine under Arena 3.5.1 with search.par in the same directory as the .exe. No problems, running perfect with 256 MB Hash (or more) and as much cores as I can set in Arena with F11/Winboard/Max CPU Cores (my laptop is dual core).

I only have 32bit OS (Win7) on my Laptop here, will try 64bit next week when at home again.

hgm wrote:

Colin-G wrote:I had this problem today. EXchess 7.96 failed to move and lost on time, but the position was drawn by 3 fold repetition, according to Stockfish analysis engine (and myself if I read the rules correctly)

Looks like Arena does not know that a Pawn double-push does not create e.p. rights if there are no Pawns that can actually perform the capture.

Of course EXchess claims a draw, but when you run it through Wb2Uci, this claim cannot be passed to the GUI.

Of course using Uci2Wb to run WB engines on Arena is unwise. Arena supports WB protocol directly, and channeling the communication through the less capable UCI protocol only loses functionality.

Why did you think I did not have EXchess installed directly as a Winboard engine?
Both engines were installed in Arena as Winboard engines when the problem occured.
No adapters were used.

I am running EXchess 7.92b 64-bit 1CPU as a winboard engine using the search.par file.
No issues at all so far.

Earlier in this thread people were talking about using Uci2Wb with Arena, but I see now that was not you.

I guess Arena then simply ignores the engine's draw claim, because it (erroneously) did not think this should be a draw itself.

attakinski wrote:Hi Dan,

First I deleted the search par file from 7.96. Suddenly the program was running on 1 CPu.

Then I copied the search file from version 7.88 in the 7.96 folder. The program crashed again.

Without search par file 7.96 is running but with 1 cpu. Not with 4 or 8.
With search par file it crashes immediately.

I have had the same problem. The solution is:

in search.par: THREADS 1

but in Arena use the command line parameters to modify the number of cores. For example for 4 cores:

xb cores 4

Dokterchen wrote:

attakinski wrote:Hi Dan,

First I deleted the search par file from 7.96. Suddenly the program was running on 1 CPu.

Then I copied the search file from version 7.88 in the 7.96 folder. The program crashed again.

Without search par file 7.96 is running but with 1 cpu. Not with 4 or 8.
With search par file it crashes immediately.

I have had the same problem. The solution is:

in search.par: THREADS 1

but in Arena use the command line parameters to modify the number of cores. For example for 4 cores:

xb cores 4

Thank you! This is the final clue that I needed. It turns out that a logging file lock was not initialized until after search.par was read; however, if more than one thread is started in search.par, the program tries to note that in the log (if logging is turned off, it still tries to grab the lock). This did not cause a problem on my Mac, but it did cause a problem on Windows. I fixed the bug and put a new version v7.97 on my website. It is otherwise identical to v7.96.

My thanks to everyone for trying various combinations to help for this out!

- Dan