AlvaroBegue wrote:I modified the code a tiny bit to report whatever the fathom interface says has gone wrong. Unfortunately, it's not much. It will probably just tell you "ERROR: No tablebases found", which isn't very informative.
I don't use Windows, but my memory of programming on MS-DOS around 20 years ago is that both "C:/blah" and "C:\blah" should work.
Does anyone have an idea of how to debug this?
Just an idea... Does your implementation perhaps expect up to 6men?
My folder (as the name says) is the complete set for 5men!
No, I also have just the 5-men tablebases, and it works fine for me.
Anyhow I will download your new attempt now and check what it says.
BTW the download size of 'master' is lying always
Guenther
Oh, I didn't notice that. I also don't know how to make older versions available for download on BitBucket. Perhaps I should just get another website to host direct downloads.
AlvaroBegue wrote:I modified the code a tiny bit to report whatever the fathom interface says has gone wrong. Unfortunately, it's not much. It will probably just tell you "ERROR: No tablebases found", which isn't very informative.
I don't use Windows, but my memory of programming on MS-DOS around 20 years ago is that both "C:/blah" and "C:\blah" should work.
Does anyone have an idea of how to debug this?
Just an idea... Does your implementation perhaps expect up to 6men?
My folder (as the name says) is the complete set for 5men!
No, I also have just the 5-men tablebases, and it works fine for me.
Anyhow I will download your new attempt now and check what it says.
BTW the download size of 'master' is lying always ;-)
Guenther
Oh, I didn't notice that. I also don't know how to make older versions available for download on BitBucket. Perhaps I should just get another website to host direct downloads.
I don't think that is necessary. ZuriChess e.g. has older downloads in the same folder as the master repository at bitbucket.
(from looking at the difference it seems you can add branches for different releases?) https://bitbucket.org/zurichess/zuriche ... =downloads
For the issue with the syzygy crash I am running out of ideas though.
Probably a debug version would be needed?
First the error should be reproduced on another windows machine,
hopefully someone reports here too.
For the specs: I am running it on Win7-64 Ultimate on an older quadcore, which means I use the default ruydos compile,
because there are no popcount instructions available. (this shouldn't be the reason, because it runs w/o the syzygy path given
and I already had played a lot of games with version 1.02 on that machine.)
For the issue with the syzygy crash I am running out of ideas though.
Probably a debug version would be needed?
First the error should be reproduced on another windows machine,
hopefully someone reports here too.
For the specs: I am running it on Win7-64 Ultimate on an older quadcore, which means I use the default ruydos compile,
because there are no popcount instructions available. (this shouldn't be the reason, because it runs w/o the syzygy path given
and I already had played a lot of games with version 1.02 on that machine.)
Ah! That's probably [at least part of] it! I have a separate compilation command for fathom and I wasn't careful, so it uses popcount for that part of the code.
Bester Zug: Se2xd4 Zeit: 0:10.266 min K/s: 6.193.021 Knoten: 63.013.981
### 6.: to RuyDos 1.0.11 x64 (0): position fen 4k3/4p3/8/8/3p4/8/4N3/4KB2 w - - moves e2d4 e8d7
### 7.: to RuyDos 1.0.11 x64 (0): go movestogo 39 wtime 469734 btime 472890
### from RuyDos 1.0.11 x64 (0): ERROR: Wrong position command (`moves' expected)
##### Unknown token: ERROR:
Oh, this is a completely different problem. RuyDos expects two numbers before the "moves" part of the "position" command, which are supposed to be move counters of some sort. Those numbers are not very important, but I thought they were part of the UCI interface.
Ah! That's probably [at least part of] it! I have a separate compilation command for fathom and I wasn't careful, so it uses popcount for that part of the code.
Please download again and give it a try.
Success!
I have found a little glitch though. It seems it does something different
in analyze mode when a position appears which is covered by syzygy tablebases.
e.g. if I want to analyse a position which is already 5men and 5men syzygy are used somehow there is no visible output in WB via polyglot.
(I guess it shouldn't answer with bestmove immediately and no score? - nothing appears in the output window of WB)
BTW the position really happened in that strange game posted by Dimitri Gusev. http://talkchess.com/forum/viewtopic.php?p=719973
(It seems Stockfish did not use the 5men syzygy there and I was curious what they would say here and if Bobcat had a draw or not in that ending)
Ah! That's probably [at least part of] it! I have a separate compilation command for fathom and I wasn't careful, so it uses popcount for that part of the code.
Please download again and give it a try.
Success!
I have found a little glitch though. It seems it does something different
in analyze mode when a position appears which is covered by syzygy tablebases.
e.g. if I want to analyse a position which is already 5men and 5men syzygy are used somehow there is no visible output in WB via polyglot.
(I guess it shouldn't answer with bestmove immediately and no score? - nothing appears in the output window of WB)
Yes, that's what I did. I am not sure what the desired behavior is, but this is good enough for winning games, which is my primary focus.
I think if the root of the tree is already in the tablebases, I could prune all the moves that are not in the optimal class and then search normally, or perhaps with the bitbases disabled. Or maybe I could just print an info message with the score before I return a best move.
