Komodo 3 generates invalid moves within 'info pv' lines?

[trojanfoe]

Hi there, I am running a test harness I have developed that can be used to play games between 2 UCI engines. However on several occasions I have seen invalid moves from Komodo 3 (under both Linux and Mac OS X).

Is this a known issue or is it commonly known?

The following log file section shows my programs interaction with Komodo (there is, of course, the possibility that my code is responsible for the error). Note that the moves from c1a3 onwards are all complete nonsense (e.g. 'h8h8b').

I attempted to run komodo interactively, entering the UCI commands shown at the top of the log file section, however komodo generated different responses to as seen below, which is strange isn't it?

Code: Select all

06/01/2012 15:54:37.931 DBG ProcessIo: komodo> 'position startpos moves d2d4 g8f6 c2c4 e7e6 b1c3 f8b4 e2e3 e8g8 f1d3 d7d5 g1f3 c7c5 e1g1 b8c6 a2a3 b4c3 b2c3 d5c4 d3c4 d8c7 c4b5 a7a6 b5c6 c7c6 a3a4 a6a5 f1e1 b7b6 f3e5 c6c7 e3e4 c8b7 f2f3 a8b8 d1e2 b8a8 a1b1 b7a6 e2a2 f8d8 e1d1 f6h5 g2g4 h5f6 c1f4 h7h6 a2b2 f6d7 e5g6 c7b7 g6e7 g8h7 d4d5 d7e5 f4e5 b7e7 c3c4 e7g5 b2c3 e6d5 c4d5 a6e2 h2h4 g5e7 d5d6 e7e8 d1d5 f7f6 e5f6 g7f6 g1f2 e2f3 c3f3 e8a4 b1b6 d8d7 f2g3 h7g8 g4g5 f6g5 h4g5 a8e8 g5h6 a4e4 f3e4 e8e4 b6c6 e4e8 c6c7 d7d8 c7g7 g8h8 d6d7 e8e6 d5c5 e6d6 c5c7 d6d3 g3f2 d3d6 g7e7 d6f6 f2e3 f6f8 c7a7 f8g8 h6h7 g8f8 a7a5 d8d7 e7d7 f8f3 e3d2 f3f2 d2c1'
06/01/2012 15:54:37.931 DBG ProcessIo: komodo> 'go wtime 94687 btime 95449 winc 0 binc 0'
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 1'
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 1 time 0 nodes 110 score cp -307 nps 1100000 pv f2f1 c1b2 f1f2 b2b3 '
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 2'
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 2 time 0 nodes 179 score cp -316 nps 1790000 pv f2f1 c1b2 f1f2 b2b3 f2b2 b3c4 '
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 3'
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 3 time 0 nodes 268 score cp -316 nps 2680000 pv f2f1 c1b2 f1f2 b2b3 f2b2 b3c4 '
06/01/2012 15&#58;54&#58;37.932 DBG ProcessIo&#58; <komodo 'info depth 4'
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 4 time 0 nodes 461 score cp -306 nps 4610000 pv f2f1 c1b2 f1b1 b2c3 b1c1 c3b4 c1c4 b4b5 '
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 5'
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 5 time 1 nodes 1023 score cp -298 nps 1022830 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 '
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 5 time 1 nodes 1061 score cp -298 nps 1060823 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 '
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 6'
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 6 time 1 nodes 1347 score cp -307 nps 1346776 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 c7b7 b3c4 b7c7 '
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 6 time 1 nodes 1442 score cp -307 nps 1441760 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 c7b7 b3c4 b7c7 '
06/01/2012 15&#58;54&#58;37.933 DBG ProcessIo&#58; <komodo 'info depth 7'
06/01/2012 15&#58;54&#58;37.934 DBG ProcessIo&#58; <komodo 'info depth 7 time 2 nodes 1890 score cp -309 nps 944955 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 c7b7 b3c4 b7c7 h5c5 '
06/01/2012 15&#58;54&#58;37.934 DBG ProcessIo&#58; <komodo 'info depth 7 time 2 nodes 2582 score cp -309 nps 1290939 pv f2f1 d7d1 f1f7 a5h5 f7c7 c1b2 c7b7 b2c3 b7c7 c3b3 c7b7 b3c4 b7c7 h5c5 '
06/01/2012 15&#58;54&#58;37.934 DBG ProcessIo&#58; <komodo 'info depth 8'
06/01/2012 15&#58;54&#58;37.936 DBG ProcessIo&#58; <komodo 'info depth 8 time 4 nodes 7099 score cp -310 nps 1774667 pv f2f1 d7d1 f1f6 a5g5 f6c6 c1b2 c6b6 b2c3 b6c6 c3b4 c6b6 b4c5 b6b7 d1h1 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 7894 score cp -293 nps 1578763 pv f2c2 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 7950 score cp -246 nps 1589963 pv f2c2 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 8006 score cp -175 nps 1601163 pv f2c2 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 8071 score cp -41 nps 1614163 pv f2c2 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 8357 score cp -5 nps 1671361 pv f2c2 c1d1 c2c1 d1e2 c1e1 e2d2 e1d1 d2c3 d1d7 a5h5 d7h7 h5h7 h8h7 c3b4 c1a3 h8h8b d2a3b a1c1 a8a1 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 8 time 5 nodes 8400 score cp -5 nps 1679961 pv f2c2 c1d1 c2c1 d1e2 c1e1 e2d2 e1d1 d2c3 d1d7 a5h5 d7h7 h5h7 h8h7 c3b4 c1a3 h8h8b d2a3b a1c1 a8a1 '
06/01/2012 15&#58;54&#58;37.937 DBG ProcessIo&#58; <komodo 'info depth 9'
06/01/2012 15&#58;54&#58;37.937 ERR Move&#58; Failed to complete move ?c1a3. Position&#58;
+---------------+
|.|.|.|.|.|.|.|.| ply=129 &#40;btm&#41;
|.|.|.|.|.|.|.|k| flags=
|.|.|.|.|.|.|.|.| ep=0
|.|.|.|.|.|.|.|.| hmclock=1
|.|K|.|.|.|.|.|.| lastMove=Kc3b4
|.|.|.|.|.|.|.|.|
|.|.|.|.|.|.|.|.|
|.|.|.|.|.|.|.|.|
+---------------+

[mar]

Are you sure that hash size was the same when you entered the commands manually? I would simply truncate the PV on the GUI side when it encounters a bad move in the PV. Perhaps Komodo uses hash table to improve PV length but doesn't check for move legality?

Martin

[trojanfoe]

I don't change the hash size from within my test program so I guess it was the same.

I have now modified my code to ignore invalid moves in the 'pv', and stop parsing the line, as they aren't as important as the 'bestmove'.

I was wondering if this was the first time this issue had been seen though?

[Dave_N]

I just noticed a similar problem - I set up a match for Houdini1.5 vs Komodo3 and Komodo eventually won the game ... At the end when both sides were printing approximately -9 scores something funny happened to the Houdini window and the Depth text box displayed some Jargon ascii characters and the program crashed ... It's a shame because it was a great game.

[trojanfoe]

What 'gui' program were you using David?

[Dave_N]

Haven't got a name for my GUI at this stage.