Fairy-Stockfish 10.3 released

[George Tsavdaris]

Cutechess GUI supports many of the SF Fairy supported variants. I recommend that with fairy stockfish.

But no analysis option which is critical.

[Fabian Fichter]

Can you please try to get the command "C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard\UCI2WB.exe 'C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard\Stockfish\fairy-stockfish-largeboard_x86-64-modern103.exe" to work on the command line (e.g., some quoting might be missing). It should stay open without printing anything, and then you can enter "protover 2". I suspect that it will then (among others) print a messed up line containing the variant names.

Doing that it says:

UCI2WB.exe fairy-stockfish-largeboard_x86-64-modern103.exe "C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard\Stockfish"
protover 2

feature variants="normal,xiangqi" setboard=1 usermove=1 debug=1 ping=1 name=1 reuse=0 exclude=1 pause=1 sigint=0 sigterm=0 done=0
feature option="UCI2WB debug output -check 0"
feature myname="Fairy-Stockfish 10.3 64 POPCNT (UCI2WB)"
feature option="Protocol -combo uci /// usi"
feature option="Debug Log File -string usi"
feature option="Contempt -spin 24 -100 100"
feature option="Analysis Contempt -combo Both /// Off /// White /// Black"
feature option="Clear Hash -button"
feature option="MultiPV -spin 1 1 500"
feature option="Skill Level -spin 20 0 20"
feature option="Move Overhead -spin 30 0 5000"
feature option="Minimum Thinking Time -spin 20 0 5000"
feature option="Slow Mover -spin 84 10 1000"
feature option="nodestime -spin 0 0 10000"
feature option="UCI_LimitStrength -check 0"
feature option="UCI_Elo -spin 1350 1350 2850"
feature option="SyzygyPath -string <empty>"
feature option="SyzygyProbeDepth -spin 1 1 100"
feature option="Syzygy50MoveRule -check 1"
feature option="SyzygyProbeLimit -spin 7 0 7"
feature option="VariantPath -string <empty>"
feature oocastle=0
feature variants="3check,5check,ai-wok,almost,amazon,antichess,asean,breakthrough,cambodian,capablanca,capahouse,caparandom,chancellor,normal,chessgi,chigorin,clobber,clobber10,codrus,courier,crazyhouse,dobutsu,embassy,euroshogi,extinction,fairy,gardner,giveaway,gorogoro,gothic,grand,hoppelpoppel,horde,janus,jesonmor,judkins,karouk,kinglet,kingofthehill,kyotoshogi,loop,losalamos,losers,makruk,micro,mini,minishogi,modern,placement,pocketknight,racingkings,seirawan,shako,shatar,shatranj,shogi,shouse,sittuyin,standard,normal,fischerandom"
feature smp=1 memory=1 done=1

It is strange that it prints "variants="normal,xiangqi"" in the first line. This should not happen and might potentially cause the problems. In any case, I now uploaded a new version of the UCI2WB adapter: https://github.com/ianfab/uci2wb/releases/tag/v10.3.

[Nordlandia]

Cutechess GUI supports many of the SF Fairy supported variants. I recommend that with fairy stockfish.

But no analysis option which is critical.

There is something. Select engine with infinite time and paste FEN. Then engine will analyze any FEN.

[George Tsavdaris]

It is strange that it prints "variants="normal,xiangqi"" in the first line. This should not happen and might potentially cause the problems. In any case, I now uploaded a new version of the UCI2WB adapter: https://github.com/ianfab/uci2wb/releases/tag/v10.3.

With this it still crashes in Winboard.

the protover 2 now is:
C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard> UCI2WB.exe fairy-stockfish-largeboard_x86-64-modern103.exe "C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard\Stockfish"
protover 2

feature setboard=1 usermove=1 debug=1 ping=1 name=1 reuse=0 exclude=1 pause=1 sigint=0 sigterm=0 done=0
feature option="UCI2WB debug output -check 0"
feature myname="Fairy-Stockfish 10.3 64 POPCNT (UCI2WB)"
feature option="Protocol -combo uci /// usi"
feature option="Debug Log File -file "
feature option="Contempt -spin 24 -100 100"
feature option="Analysis Contempt -combo Both /// Off /// White /// Black"
feature option="Clear Hash -button"
feature option="MultiPV -spin 1 1 500"
feature option="Skill Level -spin 20 0 20"
feature option="Move Overhead -spin 30 0 5000"
feature option="Minimum Thinking Time -spin 20 0 5000"
feature option="Slow Mover -spin 84 10 1000"
feature option="nodestime -spin 0 0 10000"
feature option="UCI_LimitStrength -check 0"
feature option="UCI_Elo -spin 1350 1350 2850"
feature option="SyzygyProbeDepth -spin 1 1 100"
feature option="Syzygy50MoveRule -check 1"
feature option="SyzygyProbeLimit -spin 7 0 7"
feature oocastle=0
feature variants="3check,5check,ai-wok,almost,amazon,antichess,asean,breakthrough,cambodian,capablanca,capahouse,caparandom,chancellor,normal,chessgi,chigorin,clobber,clobber10,codrus,courier,crazyhouse,dobutsu,embassy,euroshogi,extinction,fairy,gardner,giveaway,gorogoro,gothic,grand,hoppelpoppel,horde,janus,jesonmor,judkins,karouk,kinglet,kingofthehill,kyotoshogi,loop,losalamos,losers,makruk,micro,mini,minishogi,modern,placement,pocketknight,racingkings,seirawan,shako,shatar,shatranj,shogi,shouse,sittuyin,standard,normal,fischerandom"
feature egt="syzygy"
feature smp=1 memory=1 done=1

[Fabian Fichter]

It is strange that it prints "variants="normal,xiangqi"" in the first line. This should not happen and might potentially cause the problems. In any case, I now uploaded a new version of the UCI2WB adapter: https://github.com/ianfab/uci2wb/releases/tag/v10.3.

With this it still crashes in Winboard.

the protover 2 now is:
C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard> UCI2WB.exe fairy-stockfish-largeboard_x86-64-modern103.exe "C:\Program Files (x86)\WinBoard-AA\WinBoard\WinBoard\Stockfish"
protover 2

feature setboard=1 usermove=1 debug=1 ping=1 name=1 reuse=0 exclude=1 pause=1 sigint=0 sigterm=0 done=0
feature option="UCI2WB debug output -check 0"
feature myname="Fairy-Stockfish 10.3 64 POPCNT (UCI2WB)"
feature option="Protocol -combo uci /// usi"
feature option="Debug Log File -file "
feature option="Contempt -spin 24 -100 100"
feature option="Analysis Contempt -combo Both /// Off /// White /// Black"
feature option="Clear Hash -button"
feature option="MultiPV -spin 1 1 500"
feature option="Skill Level -spin 20 0 20"
feature option="Move Overhead -spin 30 0 5000"
feature option="Minimum Thinking Time -spin 20 0 5000"
feature option="Slow Mover -spin 84 10 1000"
feature option="nodestime -spin 0 0 10000"
feature option="UCI_LimitStrength -check 0"
feature option="UCI_Elo -spin 1350 1350 2850"
feature option="SyzygyProbeDepth -spin 1 1 100"
feature option="Syzygy50MoveRule -check 1"
feature option="SyzygyProbeLimit -spin 7 0 7"
feature oocastle=0
feature variants="3check,5check,ai-wok,almost,amazon,antichess,asean,breakthrough,cambodian,capablanca,capahouse,caparandom,chancellor,normal,chessgi,chigorin,clobber,clobber10,codrus,courier,crazyhouse,dobutsu,embassy,euroshogi,extinction,fairy,gardner,giveaway,gorogoro,gothic,grand,hoppelpoppel,horde,janus,jesonmor,judkins,karouk,kinglet,kingofthehill,kyotoshogi,loop,losalamos,losers,makruk,micro,mini,minishogi,modern,placement,pocketknight,racingkings,seirawan,shako,shatar,shatranj,shogi,shouse,sittuyin,standard,normal,fischerandom"
feature egt="syzygy"
feature smp=1 memory=1 done=1

The output looks perfectly fine, so apparently Fairy-Stockfish is working well with the adapter. Please make sure that the command you are using in Winboard calls the adapter in the same way as you do it via the command line, because my only explanation is that this is not the case, otherwise WinBoard would not crash when loading the engine. You can e.g. let WinBoard write a debug log file (there should be documentation on how to do this) where you can see the communication between engine and GUI to identify what is wrong.

[George Tsavdaris]

The output looks perfectly fine, so apparently Fairy-Stockfish is working well with the adapter. Please make sure that the command you are using in Winboard calls the adapter in the same way as you do it via the command line, because my only explanation is that this is not the case, otherwise WinBoard would not crash when loading the engine.

That explanation does not explain why the previous Fairy-SF largeboard versions did work perfectly nor why the current Fairy-SF 10.3 non-largeboard works also perfectly.
There is something wrong with largeboard 10.3 is my explanation.

You can e.g. let WinBoard write a debug log file (there should be documentation on how to do this) where you can see the communication between engine and GUI to identify what is wrong.

Maybe HGM can help on this....

[hgm]

It seems that WinBoard crashes, which almost by definition is a WinBoard bug. String handling in WinBoard unfortunately is not very robust; it mostly doesn't test for buffer overrun, and just relies on all text buffers being large enough. (Currently set to 512 bytes through the macro MSG_SIZ.) Perhaps the list of supported variants is simply too long, and causes a buffer overflow somewhere down the line.

[George Tsavdaris]

It seems that WinBoard crashes, which almost by definition is a WinBoard bug. String handling in WinBoard unfortunately is not very robust; it mostly doesn't test for buffer overrun, and just relies on all text buffers being large enough. (Currently set to 512 bytes through the macro MSG_SIZ.) Perhaps the list of supported variants is simply too long, and causes a buffer overflow somewhere down the line.

Oh, but why previous Fairy-SF largeboard worked or current Fairy-SF 10.3 non-largeboard works?
Can it be without a bug introduction into the largeboard 10.3?
I guess it can as your last sentence suggests.
Damn. Too bad, i really wanted this new 10.3 SF.

[Fabian Fichter]

It seems that WinBoard crashes, which almost by definition is a WinBoard bug. String handling in WinBoard unfortunately is not very robust; it mostly doesn't test for buffer overrun, and just relies on all text buffers being large enough. (Currently set to 512 bytes through the macro MSG_SIZ.) Perhaps the list of supported variants is simply too long, and causes a buffer overflow somewhere down the line.

Oh, but why previous Fairy-SF largeboard worked or current Fairy-SF 10.3 non-largeboard works?
Can it be without a bug introduction into the largeboard 10.3?

Of course it can, because the list of supported variants of course got longer. That is why I asked you to try the non-large-board version of 10.3, because that one roughly has the same number of supported variants as the large-board version of 10.2. I just wanted to make sure that the bug definitely is not in Fairy-Stockfish or the UCI2WB adapter before claiming that it is on the GUI side.

[George Tsavdaris]

Of course it can, because the list of supported variants of course got longer. That is why I asked you to try the non-large-board version of 10.3, because that one roughly has the same number of supported variants as the large-board version of 10.2. I just wanted to make sure that the bug definitely is not in Fairy-Stockfish or the UCI2WB adapter before claiming that it is on the GUI side.

So when new Winboard 5.0 with unlimited buffer? We live in 2019 so these limits are bad.
It's amazing there is no single GUI except Winboard that can run Fairy-SF with Winboard's abilities(to run engine matches and analysis)!!
Having such a good engine and no way to run it! Command line sucks big time and it's impractical.