Polyglot + Alfil

[Volker Pittlik]

Geschrieben von:/Posted by: Volker Pittlik at 31 March 2004 12:24:52:

[Event "Computer chess game"]
[Site "VPITTLIK"]
[Date "2004.03.31"]
[Round "1"]
[White "Bookthinker-Alfil-403-1"]
[Black "Movei00_8_178"]
[Result "0-1"]
[TimeControl "120+2"]
1. d4 f5 2. g3 e6 3. Nf3 Nf6 4. Bf4 Nc6 5. h4 d5 6. c3 Bd6 7. Bxd6 Qxd6 8.
Nbd2 O-O 9. Ng5 e5 10. dxe5 Nxe5 11. Qb3 a5 12. Bg2 a4 13. Qc2 Bd7 14. Rd1
a3 15. b3 h6 16. Ngf3 f4 17. Nxe5 Qxe5 18. Nf3 Qd6 19. gxf4 Qxf4 20. Rd4
Qd6 21. Qg6 Ra6 22. Nd2 c5 23. Rd3 Qe7 24. Qg3 Re6 25. Bf3 Ne4 26. Qg2 Bc6
27. Bxe4 dxe4 28. Re3 Ref6 29. Rf1 Rd8 30. Nxe4 Re6 31. f3 Qxh4+ 32. Qf2
Qf4 33. c4 g5 34. Qg1 b6 35. Rd3 Rxd3 36. exd3 Qc1+ 37. Kf2 Qb2+ 38. Ke3
Qxa2
{Black wins on time} 0-1

Winboard.debug:


307531 >second: f2e3
307640 <second:  1 462 0 249 b2a2
307640 <second:  2 388 0 390 b2a2 g1g4
307640 <second:  2 388 0 1653 b2a2 g1g4
307640 <second:  3 336 1 3140 b2a2 g1g4 c6d7
307640 <second:  3 336 3 7937 b2a2 g1g4 c6d7
307640 <second:  4 318 3 9378 b2a2 g1g3 a2b3 g3c7
307703 <second:  4 318 11 26759 b2a2 g1g3 a2b3 g3c7
307718 <second:  5 319 12 32028 b2a2 g1g3 a2b3 g3c7 b3b2
307890 <second:  5 319 29 75424 b2a2 g1g3 a2b3 g3c7 b3b2
307968 <second:  6 302 37 93576 b2a2 g1g4 c6d7 g4f5 a2b3 e3d2 b3b2 d2e1
308406 <second:  6 302 81 214387 b2a2 g1g4 c6d7 g4f5 a2b3 e3d2 b3b2 d2e1
308656 <second:  7 312 106 272480 b2a2 g1g3 a2b3 g3b8 g8f7 b8c7 f7g6 f1h1 a3a2
309703 <second:  7 312 211 558666 b2a2 g1g3 a2b3 g3b8 g8f7 b8c7 f7g6 f1h1 a3a2
312500 <second:  8 254 489 1230824 b2a2 g1g4 c6d7 g4f5 a2b3 f3f4 g5f4 f1f4 b3a4 f5f8 g8h7 f4f7 h7g6
317296 <second:  8 254 970 2492497 b2a2 g1g4 c6d7 g4f5 a2b3 f3f4 g5f4 f1f4 b3a4 f5f8 g8h7 f4f7 h7g6
317296 <second: tellics kibitz   2.54/8 time=   9.70 nps= 256879 b2a2 g1g4 c6d7 g4f5 a2b3 f3f4 g5f4 f1f4 b3a4 f5f8 g8h7 f4f7 h7g6
317296 <second: move b2a2
317312 >first : time 4854
otim 3725
317312 >first : usermove 317312 >first : b2a2
320343 <first : 4 -242 6 0 Ra1 Qb2 Rb1 Qc2
320531 <first : 5 -217 33 21620 Ra1 Qb2 Rb1 Qc2 Qc1
320546 <first : file "xboard.cpp", line 996, assertion "RootMovePos>=0&&RootMovePos<RootMoveNb" failed
GameEnds(32, Black wins on time, 4)
365843 >first : result 0-1 {Black wins on time}
Interrupting second
365843 >second: result 0-1 {Black wins on time}
365859 >first : quit
365859 >second: quit


Alfil-log:


POLYGLOT MOVE Qxa2
POLYGLOT WAITING -> THINKING
POLYGLOT START SEARCH
> ENGINE position startpos moves d2d4 f7f5 g2g3 e7e6 g1f3 g8f6 c1f4 b8c6 h2h4 d7d5 c2c3 f8d6 f4d6 d8d6 b1d2 e8g8 f3g5 e6e5 d4e5 c6e5 d1b3 a7a5 f1g2 a5a4 b3c2 c8d7 a1d1 a4a3 b2b3 h7h6 g5f3 f5f4 f3e5 d6e5 d2f3 e5d6 g3f4 d6f4 d1d4 f4d6 c2g6 a8a6 f3d2 c7c5 d4d3 d6e7 g6g3 a6e6 g2f3 f6e4 g3g2 d7c6 f3e4 d5e4 d3e3 e6f6 h1f1 f8d8 d2e4 f6e6 f2f3 e7h4 g2f2 h4f4 c3c4 g7g5 f2g1 b7b6 e3d3 d8d3 e2d3 f4c1 e1f2 c1b2 f2e3 b2a2
> ENGINE go wtime 48540 btime 37250 winc 2000 binc 2000
< ENGINE info score cp -242 depth 4 seldepth 5 time 62 pv f1a1 a2b2 a1b1 b2c2
> XBOARD 4 -242 6 0 Ra1 Qb2 Rb1 Qc2
< ENGINE info depth 4 nodes 21620
< ENGINE 
< ENGINE info nodes 21620
< ENGINE 
< ENGINE info score cp -217 depth 5 seldepth 6 time 328 pv f1a1 a2b2 a1b1 b2c2 g1c1
> XBOARD 5 -217 33 21620 Ra1 Qb2 Rb1 Qc2 Qc1
< ENGINE info depth 5 nodes 138688
< ENGINE 
< ENGINE info nodes 138688
< ENGINE 
< ENGINE info currmove f1a1 currmovenumber 1
< ENGINE 
< ENGINE info currmove f1b1 currmovenumber 2
< ENGINE 
< ENGINE info currmove g1g4 currmovenumber 3
< ENGINE 
< ENGINE info currmove g1g3 currmovenumber 4
< ENGINE 
< ENGINE info currmove f1e1 currmovenumber 5
< ENGINE 
< ENGINE info currmove g1h1 currmovenumber 14
< ENGINE 
< ENGINE info currmove g1g2 currmovenumber 15
< ENGINE 
< ENGINE info currmove g1g5 currmovenumber 18
file "xboard.cpp", line 996, assertion "RootMovePos>=0&&RootMovePos<RootMoveNb" failed


Maybe an Alfil error. Testing will continue.
Volker

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 12:30:09:
Als Antwort auf:/In reply to: Polyglot + Alfil geschrieben von:/posted by: Volker Pittlik at 31 March 2004 12:24:52:

Maybe an Alfil error. Testing will continue.
Volker
< ENGINE info currmove g1g5 currmovenumber 18
file "xboard.cpp", line 996, assertion "RootMovePos>=0&&RootMovePos

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 13:16:49:
Als Antwort auf:/In reply to: Polyglot + Alfil geschrieben von:/posted by: Volker Pittlik at 31 March 2004 12:24:52:

Maybe an Alfil error. Testing will continue.
Volker

I think it is.
My reasoning is the following:
UCI specifies that the engine can send the current-move-searched number.
Unfortunately, in UCI the engine cannot send the total move number (contrary to xboard).
So that means the interface has to count them.
Which in turn means it can only consider strictly-legal moves.
So the engine must count them in the same way, for coherency.
So to sum up:
1) bad design in UCI
2) all the same, compliant engines must consider only strictly-legal moves at the root of the search (first ply from the current position)
So IMO this is a bug in AlFil, and should be reported accordingly. I have already reported the same bug in Gothmog (only for en-passant it seems, though).
Note that the release version of PolyGlot will not complain about this anymore, as it is a harmless bug.
If you want, you can ignore the problem. I believe however that finding bugs is the hardest part, fixing them is usually much easier.
Fabien.

[Volker Pittlik]

Geschrieben von:/Posted by: Volker Pittlik at 31 March 2004 14:02:30:
Als Antwort auf:/In reply to: Re: Polyglot + Alfil geschrieben von:/posted by: Fabien Letouzey at 31 March 2004 13:16:49:

...

So IMO this is a bug in AlFil, and should be reported accordingly.
I have already reported the same bug in Gothmog (only for en-passant it seems, though).
Note that the release version of PolyGlot will not complain about this anymore, as it is a harmless bug.

Maybe I do, but it is not my intention to find bugs in engines.
For further testing I suggest to use something as a priority list what to test/debug first. I think we have some
Class A phanomenoms:
- Something what have to be solved first
Class B
- Errors which are caused by bad engine designed and can be ignored
Class C
- Something "nice to have" as the Bookthinker issue
Regards
Volker

[Fabien Letouzey]

Geschrieben von:/Posted by: Fabien Letouzey at 31 March 2004 14:13:20:
Als Antwort auf:/In reply to: Re: Polyglot + Alfil geschrieben von:/posted by: Volker Pittlik at 31 March 2004 14:02:30:

So IMO this is a bug in AlFil, and should be reported accordingly.
I have already reported the same bug in Gothmog (only for en-passant it seems, though).
Note that the release version of PolyGlot will not complain about this anymore, as it is a harmless bug.

Maybe I do, but it is not my intention to find bugs in engines.
For further testing I suggest to use something as a priority list what to test/debug first. I think we have some
Class A phanomenoms:
- Something what have to be solved first
Class B
- Errors which are caused by bad engine designed and can be ignored
Class C
- Something "nice to have" as the Bookthinker issue

I don't understand, I consider this thread not to be a PolyGlot issue at all.
In PolyGlot's case, that's the engine-unloading problem.
Dann sent me a log file (Shredder), but it is unclear what really happened.
I suspect WinBoard does strange things at exit but the log files don't show anything at all.
Unfortunately I have just found a problem in my error-reporting code, and it is possible that some OS error does not appear in the log file. At the same time I have no proof that it wasn't working fine. I hope that this becomes clearer in the next version (in about 10 hours).
The problem is that I can't guarantee beforehand if they are from the engine or from PolyGlot. For each problem, it must be checked that PolyGlot was not in fault. In programming it is a good habit to always suspect the most recent software.
So far all the "RootMove" stuff enters that category, and most other assertions although I relaxed some of them already (depth/seldepth). I can specify which could be further ignored when answering specific threads.
Yes, we are not yet at this stage; I will start a thread when the "unload" problem is clearly identified.
Fabien.