WBEC Ridderkerk new results.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 09:08:27:

Small "UCI only/PolyGlot" test result:
Problems:
List 5.12 = exits just before the games are finished (no result), lose on time.
SilkeChess = "hangs" after the games.
Piranha = ponder problem (lose on time, "hangs)".
Result of the other engines:
PolyGlot 29-04, 40/3 and ponder=on.
DUAL-P3-933, 2004.04.29 - 2004.04.30
Score Shre SOS Frui Alfi Fusc
--------------------------------------------------------
1: Shredder 7.04 14.0 / 16 XXXX 1011 0111 1111 1111
2: SOS 4.0 11.5 / 16 0100 XXXX 11=0 1111 1111
3: Fruit 1.0 10.5 / 16 1000 00=1 XXXX 1111 1111
4: Alfil 403.1 4.0 / 16 0000 0000 0000 XXXX 1111
5: Fusch 1.09 0.0 / 16 0000 0000 0000 0000 XXXX
--------------------------------------------------------
40 games: +18 =1 -21
Only SOS played with a openingbook!
Best wishes,
Leo.


WBEC Ridderkerk homepage.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 09:11:00:
Als Antwort auf:/In reply to: WBEC Ridderkerk new results. geschrieben von:/posted by: Leo Dijksman at 30 April 2004 09:08:27:

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 10:20:17:
Als Antwort auf:/In reply to: WBEC Ridderkerk new results. geschrieben von:/posted by: Leo Dijksman at 30 April 2004 09:08:27:

Small "UCI only/PolyGlot" test result:
Problems:
List 5.12 = exits just before the games are finished (no result), lose on time.
SilkeChess = "hangs" after the games.
Piranha = ponder problem (lose on time, "hangs)".

Thanks for the tests Leo!
If an engine has a problem with PolyGlot *and* the author asked you for participation in your main tournament please let me know, as we might find appropriate solutions.
Does Piranha work fine with ponder off?
Fabien.

[Günther Simon]

Geschrieben von:/Posted by: Günther Simon at 30 April 2004 11:54:17:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 10:20:17:

Small "UCI only/PolyGlot" test result:
Problems:
List 5.12 = exits just before the games are finished (no result), lose on time.
SilkeChess = "hangs" after the games.
Piranha = ponder problem (lose on time, "hangs)".

Thanks for the tests Leo!
If an engine has a problem with PolyGlot *and* the author asked you for participation in your main tournament please let me know, as we might find appropriate solutions.
Does Piranha work fine with ponder off?
Fabien.

I forgot to tell that Piranha doesnt work at all with newer versions
of Polyglot (since 2 weeks or even more...), thus I play it with an older
one in my Open Class.
(It just doesnt move, when it is its turn)
Sorry I cant provide any logs or debugs, as I am just busy with running
that tournament and leave debugging to others. (Leo experienced the
same problem AFAIK)
Best regards,
Günther

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 12:14:21:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Günther Simon at 30 April 2004 11:54:17:

Small "UCI only/PolyGlot" test result:
Problems:
List 5.12 = exits just before the games are finished (no result), lose on time.
SilkeChess = "hangs" after the games.
Piranha = ponder problem (lose on time, "hangs)".

Thanks for the tests Leo!
If an engine has a problem with PolyGlot *and* the author asked you for participation in your main tournament please let me know, as we might find appropriate solutions.
Does Piranha work fine with ponder off?
Fabien.

I forgot to tell that Piranha doesnt work at all with newer versions
of Polyglot (since 2 weeks or even more...), thus I play it with an older
one in my Open Class.
(It just doesnt move, when it is its turn)
Sorry I cant provide any logs or debugs, as I am just busy with running
that tournament and leave debugging to others. (Leo experienced the
same problem AFAIK)
Best regards,
Günther
XBOARD move c2c1
XBOARD 1-0 {White mates}
XBOARD pong 1

Thats right Günther, I had the same experience but then still Piranha "hangs" in rare cases here.
To Fabien,
Is it possible that PolyGlot "quits" an engine and itself in all cases when a game is ended to avoid this:

POLYGLOT MOVE Kc1
POLYGLOT THINKING -> WAITING
< XBOARD time 21010
< XBOARD otim 14303
< XBOARD usermove g6c6
POLYGLOT MOVE Rc6#
POLYGLOT WAITING
< XBOARD result 1-0 {White mates}
POLYGLOT GAME END
POLYGLOT WAITING
< XBOARD force
POLYGLOT WAITING
< XBOARD ping 1
this is the point where some engines "hangs" because there is no quit command, I think it must be possible to let PolyGlot always quits the engine in one or the other way?
Leo.




WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 12:32:26:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Leo Dijksman at 30 April 2004 12:14:21:

To Fabien,
Is it possible that PolyGlot "quits" an engine and itself in all cases when a game is ended to avoid this:

XBOARD move c2c1
XBOARD 1-0 {White mates}
XBOARD pong 1

POLYGLOT MOVE Kc1
POLYGLOT THINKING -> WAITING
< XBOARD time 21010
< XBOARD otim 14303
< XBOARD usermove g6c6
POLYGLOT MOVE Rc6#
POLYGLOT WAITING
< XBOARD result 1-0 {White mates}
POLYGLOT GAME END
POLYGLOT WAITING
< XBOARD force
POLYGLOT WAITING
< XBOARD ping 1
this is the point where some engines "hangs" because there is no quit command, I think it must be possible to let PolyGlot always quits the engine in one or the other way?

That's the end of the log file, right (or *** START *** after that)? If so it means PolyGlot was killed by another application (or a mysterious bug?!).
As far as I know, this never happens with WinBoard alone. I suspect that the application is WBTM (can't remember which one, I'll try to find out). I suppose the author assumed only simple engines (not combinations as we use now) would be used and would not need to do any cleanup!!!
PolyGlot always sends "quit" when WinBoard sends it as well. Unfortunately, PolyGlot is killed and can't send it.
Also as I said this can only happen if the engine *also* has an I/O bug, unfortunately extremely common.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 12:47:34:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 12:32:26:

To Fabien,
Is it possible that PolyGlot "quits" an engine and itself in all cases when a game is ended to avoid this:

XBOARD move c2c1
XBOARD 1-0 {White mates}
XBOARD pong 1

POLYGLOT MOVE Kc1
POLYGLOT THINKING -> WAITING
< XBOARD time 21010
< XBOARD otim 14303
< XBOARD usermove g6c6
POLYGLOT MOVE Rc6#
POLYGLOT WAITING
< XBOARD result 1-0 {White mates}
POLYGLOT GAME END
POLYGLOT WAITING
< XBOARD force
POLYGLOT WAITING
< XBOARD ping 1
this is the point where some engines "hangs" because there is no quit command, I think it must be possible to let PolyGlot always quits the engine in one or the other way?

That's the end of the log file, right (or *** START *** after that)?
As far as I know, this never happens with WinBoard alone. I suspect that the application is WBTM (can't remember which one, I'll try to find out).
PolyGlot always sends "quit" when WinBoard sends it as well. Unfortunately, PolyGlot is killed and can't send it.
Also as I said this can only happen if the engine *also* has an I/O bug, unfortunately extremely common.
Fabien.

The end!
If so it means PolyGlot was killed by another application (or a mysterious bug?!).
I use Jori's WBTM which do nothing else as reading the winboard.debug file afaik and nothing else except rename and save the debug file.
I suppose the author assumed only simple engines (not combinations as we use now) would be used and would not need to do any cleanup!!!
In this case I used BookThinker (with WB default protocol 2) also but the end of a log file looks always the same as above when a engine "hangs" (with or without BookThinker), the winboard debug looks fine:
otim 21010
387891 >first : usermove 387891 >first : c2c1
387907 second: force
387907 >second: ping 1
Exiting: Match Shredder 7.04 vs. Alfil 403.1: final score 1-0-0
GameEnds(0, (null), 2)
387907 >first : quit
387907 >second: quit

I will look at it further, now with protocol=1.
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 13:02:55:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Leo Dijksman at 30 April 2004 12:47:34:

I use Jori's WBTM which do nothing else as reading the winboard.debug file afaik and nothing else except rename and save the debug file.
In this case I used BookThinker (with WB default protocol 2) also but the end of a log file looks always the same as above when a engine "hangs" (with or without BookThinker), the winboard debug looks fine:
otim 21010
...
387907 >first : force
387907 >first : ping 1
387907 >second: force
387907 >second: ping 1
Exiting: Match Shredder 7.04 vs. Alfil 403.1: final score 1-0-0
GameEnds(0, (null), 2)
387907 >first : quit
387907 >second: quit

I will look at it further, now with protocol=1.

Then reproduce the problem without WBTM, it will prove once for all it is not in fault. WinBoard /mg option launches game matches.
BookThinker 100% does not work with protocol version 2 after the first game (it does not answer "ping" so WinBoard waits forever). You should *always* restrict to version 1 when using BookThinker.
I take note that your specific problem is unrelated to BookThinker though, my point was a side note.
Notice WinBoard sends "quit" but PolyGlot claims not to receive it, that's the key here. Also notice WinBoard sends "ping" but does not wait for "pong"; it makes no sense to me.
Sure, but not with BookThinker as you could reproduce the problem without it.
Also, /xreuse "solves" some problems in specific cases.
To me the next order of business is to try to reproduce the problem without the WBTM. There is a batch file way, I heard? In that case the TM would not be running during the games; this is exactly what I want to test.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 13:34:39:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 13:02:55:

I use Jori's WBTM which do nothing else as reading the winboard.debug file afaik and nothing else except rename and save the debug file.
In this case I used BookThinker (with WB default protocol 2) also but the end of a log file looks always the same as above when a engine "hangs" (with or without BookThinker), the winboard debug looks fine:
otim 21010
...
387907 >first : force
387907 >first : ping 1
387907 >second: force
387907 >second: ping 1
Exiting: Match Shredder 7.04 vs. Alfil 403.1: final score 1-0-0
GameEnds(0, (null), 2)
387907 >first : quit
387907 >second: quit

I will look at it further, now with protocol=1.

Then reproduce the problem without WBTM, it will prove once for all it is not in fault. WinBoard /mg option launches game matches.
BookThinker 100% does not work with protocol version 2 after the first game (it does not answer "ping" so WinBoard waits forever). You should *always* restrict to version 1 when using BookThinker.
I take note that your specific problem is unrelated to BookThinker though, my point was a side note.
Notice WinBoard sends "quit" but PolyGlot claims not to receive it, that's the key here. Also notice WinBoard sends "ping" but does not wait for "pong"; it makes no sense to me.
Sure, but not with BookThinker as you could reproduce the problem without it.
Also, /xreuse "solves" some problems in specific cases.
To me the next order of business is to try to reproduce the problem without the WBTM. There is a batch file way, I heard? In that case the TM would not be running during the games; this is exactly what I want to test.
Fabien.

Why should I prove it happen/not happen with the /mg option? I dont use that!
It should work with Jori's WBTM, simply because I use that!
Again, thats what you say, sofar 2 engines work with BookThinker/protocol 2 and 2 engines dont (plays but hangs after the game!)
I am wundering why I report anything here if you know it already better anyway?
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 13:52:12:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Leo Dijksman at 30 April 2004 13:34:39:

Then reproduce the problem without WBTM, it will prove once for all it is not in fault. WinBoard /mg option launches game matches.
BookThinker 100% does not work with protocol version 2 after the first game (it does not answer "ping" so WinBoard waits forever). You should *always* restrict to version 1 when using BookThinker.

Why should I prove it happen/not happen with the /mg option? I dont use that!
It should work with Jori's WBTM, simply because I use that!
Again, thats what you say, sofar 2 engines work with BookThinker/protocol 2 and 2 engines dont (plays but hangs after the game!)
I am wundering why I report anything here if you know it already better anyway?

You claim that WBTM is fine, whereas I claim the opposit. Reproducing the problem without WBTM would prove you are right. It also works if you reproduce the problem without /mg of course.
Because you rely on looks (hang, crash, loss on time, etc ...), whereas I study the log files. I can detect problems you can't see like time control defects. For one-game matches the problem might not show up, it depends on WinBoard behaviour to an error from BookThinker.
My advice that fixes one general problem is to always use version 1 when using BookThinker. There are many other problems, but this one is easy to avoid. I did not claim that this is the problem you are now reporting.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 14:41:31:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 13:52:12:

Then reproduce the problem without WBTM, it will prove once for all it is not in fault. WinBoard /mg option launches game matches.
BookThinker 100% does not work with protocol version 2 after the first game (it does not answer "ping" so WinBoard waits forever). You should *always* restrict to version 1 when using BookThinker.

Why should I prove it happen/not happen with the /mg option? I dont use that!
It should work with Jori's WBTM, simply because I use that!
Again, thats what you say, sofar 2 engines work with BookThinker/protocol 2 and 2 engines dont (plays but hangs after the game!)
I am wundering why I report anything here if you know it already better anyway?

You claim that WBTM is fine, whereas I claim the opposit. Reproducing the problem without WBTM would prove you are right. It also works if you reproduce the problem without /mg of course.
Because you rely on looks (hang, crash, loss on time, etc ...), whereas I study the log files. I can detect problems you can't see like time control defects. For one-game matches the problem might not show up, it depends on WinBoard behaviour to an error from BookThinker.
My advice that fixes one general problem is to always use version 1 when using BookThinker. There are many other problems, but this one is easy to avoid. I did not claim that this is the problem you are now reporting.
Fabien.

If you know it so well, why this "WBTM problem" happen consequent with the same engines using protocol 2 and not with the others?
Also I should like to know which of the two "polyglot.exe"s is quit and which one is still loaded when a engine hangs?
With WBTM, BookThinker and ponder=off (single CPU):
Rank|Name |1|2|3|4|5|Pts |S-B |
----|-------------|-|-|-|-|-|--------|-----|
1.|Shredder 7.04|X|1|1|1|1| 4.0 / 4| 6.00||####
2.|SOS 4.0 |0|X|1|1|1| 3.0 / 4| 3.00||###
3.|Fruit 1.0 |0|0|X|1|1| 2.0 / 4| 1.00||##
4.|Alfil 403.1 |0|0|0|X|1| 1.0 / 4| 0.00||#
5.|Fusch 0.1.09 |0|0|0|0|X| 0.0 / 4| 0.00||
(SOS played with his own book!)
no timeloses, two engines hangs when using protocol 2 (seems "fixed" when using protocol 1).
Fusch seems to have a "runtime" problem when it runs at WindowsXP!?
Leo.




WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 15:18:15:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Leo Dijksman at 30 April 2004 14:41:31:

If you know it so well, why this "WBTM problem" happen consequent with the same engines using protocol 2 and not with the others?
Also I should like to know which of the two "polyglot.exe"s is quit and which one is still loaded when a engine hangs?
With WBTM, BookThinker and ponder=off (single CPU):
Rank|Name |1|2|3|4|5|Pts |S-B |
----|-------------|-|-|-|-|-|--------|-----|
1.|Shredder 7.04|X|1|1|1|1| 4.0 / 4| 6.00||####
2.|SOS 4.0 |0|X|1|1|1| 3.0 / 4| 3.00||###
3.|Fruit 1.0 |0|0|X|1|1| 2.0 / 4| 1.00||##
4.|Alfil 403.1 |0|0|0|X|1| 1.0 / 4| 0.00||#
5.|Fusch 0.1.09 |0|0|0|0|X| 0.0 / 4| 0.00||
(SOS played with his own book!)
no timeloses, two engines hangs when using protocol 2 (seems "fixed" when using protocol 1).

I am afraid I don't understand this question. Do you mean that protover 1 seems to solve the problem?
Also a small precision: what I claim to know well was one problem due to BookThinker and the solution to it is to use protover 1.
About WBTM (I don't remember which one), I only *suspect* a problem. The log files are "cut" so there is no definite proof of what is really going on. I can only guess here.
I have asked many times in this forum (not to you specifically) whether the problem can happen with WinBoard alone, and am still waiting for an answer (that's about one month now). So I guess it works fine with WinBoard alone. Also I see nothing in WinBoard's source code that could cause that problem.
There are also two ways I could interpret this question. Two PolyGlots as in engine1 and engine2? Or the two PolyGlots that seem to be running for each engine, due to CYGWIN1.DLL internals?
If I guess correctly what CYGWIN1.DLL is doing, when an engine hangs there will be a PolyGlot ghost as well. That is not the reason why the engine hangs in the first place.
Do I understand correctly that it works fine?
Or is it different with ponder on?
Let's discuss only existing problems, and try to fix them in order.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 16:11:19:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 15:18:15:

If you know it so well, why this "WBTM problem" happen consequent with the same engines using protocol 2 and not with the others?
Also I should like to know which of the two "polyglot.exe"s is quit and which one is still loaded when a engine hangs?
With WBTM, BookThinker and ponder=off (single CPU):
Rank|Name |1|2|3|4|5|Pts |S-B |
----|-------------|-|-|-|-|-|--------|-----|
1.|Shredder 7.04|X|1|1|1|1| 4.0 / 4| 6.00||####
2.|SOS 4.0 |0|X|1|1|1| 3.0 / 4| 3.00||###
3.|Fruit 1.0 |0|0|X|1|1| 2.0 / 4| 1.00||##
4.|Alfil 403.1 |0|0|0|X|1| 1.0 / 4| 0.00||#
5.|Fusch 0.1.09 |0|0|0|0|X| 0.0 / 4| 0.00||
(SOS played with his own book!)
no timeloses, two engines hangs when using protocol 2 (seems "fixed" when using protocol 1).

I am afraid I don't understand this question. Do you mean that protover 1 seems to solve the problem?
Also a small precision: what I claim to know well was one problem due to BookThinker and the solution to it is to use protover 1.
About WBTM (I don't remember which one), I only *suspect* a problem. The log files are "cut" so there is no definite proof of what is really going on. I can only guess here.
I have asked many times in this forum (not to you specifically) whether the problem can happen with WinBoard alone, and am still waiting for an answer (that's about one month now). So I guess it works fine with WinBoard alone.
Also I see nothing in WinBoard's source code that could cause that problem.
There are also two ways I could interpret this question. Two PolyGlots as in engine1 and engine2? Or the two PolyGlots that seem to be running for each engine, due to CYGWIN1.DLL internals?
If I guess correctly what CYGWIN1.DLL is doing, when an engine hangs there will be a PolyGlot ghost as well. That is not the reason why the engine hangs in the first place.
Do I understand correctly that it works fine?
Or is it different with ponder on?
Let's discuss only existing problems, and try to fix them in order.
Fabien.

Right, IF an engine hangs I see also one polyglot.exe in the taskmanager, must be the "ghost" then.
Seems so but it are only 4 games each!
Shredder and Alfil with protocol 1 and Fruit with protocol 2 (default), only Fuschs I am not sure whats the problem because I get an "runtime library" error when a game is finished. (can be WindowsXP/Fusch, I will test later)
No idee, I have no dual free at the moment to test with ponder=on.
The tourney I reported this morning was with ponder=on but without BookThinker!
For the moment it should be nice to have the possibillity to "quit" a engine in cases it "hangs" when PolyGlot is killed for some "unknown" reason and can't send it, for example with a option in the polyglot.ini which can be set on/off.
By my opinion it must be possible to let polyglot kill an "problem" engine at this point: "POLYGLOT GAME END", the game is finished anyway only the "hanging" engine stops the tourney(s) until it is stopped by hand.
Leo.



WBEC Ridderkerk homepage.

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 30 April 2004 17:02:47:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Leo Dijksman at 30 April 2004 16:11:19:

Right, IF an engine hangs I see also one polyglot.exe in the taskmanager, must be the "ghost" then.

Or is it different with ponder on?
Let's discuss only existing problems, and try to fix them in order.

No idee, I have no dual free at the moment to test with ponder=on.
The tourney I reported this morning was with ponder=on but without BookThinker!
For the moment it should be nice to have the possibillity to "quit" a engine in cases it "hangs" when PolyGlot is killed for some "unknown" reason and can't send it, for example with a option in the polyglot.ini which can be set on/off.
By my opinion it must be possible to let polyglot kill an "problem" engine at this point: "POLYGLOT GAME END", the game is finished anyway only the "hanging" engine stops the tourney(s) until it is stopped by hand.

Yes, this is "normal".
Frankly I don't think ponder on could bring new problems. BookThinker is a different story.
This is not possible, by definition of beeing killed.
I see. "POLYGLOT GAME END" means WinBoard has sent a game result.
This would be a dirty trick of course, but it could "fix" some problems for one-game matches (no /mg option).
I'll see what I can do.
Fabien.

[Leo Dijksman]

Geschrieben von:/Posted by: Leo Dijksman at 30 April 2004 17:56:39:
Als Antwort auf:/In reply to: Re: Small polyglot test geschrieben von:/posted by: Fabien Letouzey at 30 April 2004 17:02:47:

Right, IF an engine hangs I see also one polyglot.exe in the taskmanager, must be the "ghost" then.

Or is it different with ponder on?
Let's discuss only existing problems, and try to fix them in order.

No idee, I have no dual free at the moment to test with ponder=on.
The tourney I reported this morning was with ponder=on but without BookThinker!
For the moment it should be nice to have the possibillity to "quit" a engine in cases it "hangs" when PolyGlot is killed for some "unknown" reason and can't send it, for example with a option in the polyglot.ini which can be set on/off.
By my opinion it must be possible to let polyglot kill an "problem" engine at this point: "POLYGLOT GAME END", the game is finished anyway only the "hanging" engine stops the tourney(s) until it is stopped by hand.

Yes, this is "normal".
Frankly I don't think ponder on could bring new problems. BookThinker is a different story.
This is not possible, by definition of beeing killed.
I see. "POLYGLOT GAME END" means WinBoard has sent a game result.
This would be a dirty trick of course, but it could "fix" some problems for one-game matches (no /mg option).
I'll see what I can do.
Fabien.

We will see, I am now testing with ponder=on without BookThinker to be 'sure' there is no problem there and tomorrow it will be both, ponder=on and BookThinker.
Thanks, I hope you can do the trick!
Leo.



WBEC Ridderkerk homepage.