Stockfish problem in Scid

[Marek Soszynski]

I've been happily compiling most each new 64-bit Linux Stockfish developmental version to use in Scid. Unfortunately something has gone wrong with recent versions. They compile okay (and run the benchmark) but when I change the UCI parameters within Scid they go on to crash. Typically I change only the hash and switch off the book. I can't remember the exact version when this started to happen. Currently I'm on 120311.

[mcostalba]

They compile okay (and run the benchmark) but when I change the UCI parameters within Scid they go on to crash. Typically I change only the hash and switch off the book.

Do you change the hash while searching? In this case you should not. A recent change made option setting effect immediate, so if you change the hash size it will be changed immediately and if the engine is searching a crash is assured.

[Marek Soszynski]

The hash wasn't changed while searching. I've now restarted the comp and Scid doesn't show the Stockfish analysis at all nor nodes or depth etc., though System Monitor confirms SF is running.

[Justin Sherron]

I've been happily compiling most each new 64-bit Linux Stockfish developmental version to use in Scid. Unfortunately something has gone wrong with recent versions....

...Scid doesn't show the Stockfish analysis at all nor nodes or depth etc., though System Monitor confirms SF is running...

I don't know if this is helpful, but a few days ago I observed these same symptoms after compiling a 32-bit version of Stockfish and using the X-Board GUI. It seems not to be a 64-bit/Scid specific issue.

[IGarcia]

Currently I'm on 120311.

And tomorrow you will be in 120312 version

..its current day, unless you set

[Marek Soszynski]

I've been running Stockfish 64-bit 9934b8e in Scid. It runs normally with the default UCI settings. However, after I increase the hash, the analysis window is blank. System Monitor shows Stockfish is running, taking up the increased hash, though not using the second processor.

Stockfish 2.2.2 JA is okay.

[Justin Sherron]

I've been running Stockfish 64-bit 9934b8e in Scid. It runs normally with the default UCI settings. However, after I increase the hash, the analysis window is blank. System Monitor shows Stockfish is running, taking up the increased hash, though not using the second processor.

Stockfish 2.2.2 JA is okay.

Are you making PGO-optimized builds when you compile? When I make PGO builds, the program malfunctions in a manner similar to your description (as well as making X-Board unresponsive). It appears that non-PGO builds work perfectly fine though, at least on my machine.

The earlier Stockfish 2.2.2 code compiles and runs fine regardless of its PGO status when compiling.

[Marek Soszynski]

I've been running Stockfish 64-bit 9934b8e in Scid. It runs normally with the default UCI settings. However, after I increase the hash, the analysis window is blank. System Monitor shows Stockfish is running, taking up the increased hash, though not using the second processor.

Stockfish 2.2.2 JA is okay.

Are you making PGO-optimized builds when you compile? When I make PGO builds, the program malfunctions in a manner similar to your description (as well as making X-Board unresponsive). It appears that non-PGO builds work perfectly fine though, at least on my machine.

The earlier Stockfish 2.2.2 code compiles and runs fine regardless of its PGO status when compiling.

Thank for the suggestion, Justin. Unfortunately, on my machine the non-PGO builds suffer the same fate: default UCI settings fine, changed settings crash.

[stevenaaus]

I've been happily compiling most each new 64-bit Linux Stockfish developmental version to use in Scid. Unfortunately something has gone wrong with recent versions. They compile okay (and run the benchmark) but when I change the UCI parameters within Scid they go on to crash. Typically I change only the hash and switch off the book. I can't remember the exact version when this started to happen. Currently I'm on 120311.

I can verify this on 32bit Fedora 7. The engine becomes unresponsive after these commands

Code: Select all

Scid  : setoption name Clear Hash value false
Scid  : isready

If one edits ~/.scid/config/engines.dat and removes the
{{Clear Hash} {VALUE}} tuplet, the engine works ok.

[Marek Soszynski]

Thanks for the pointer, Steven.

By editing the DAT file directly and inserting...

Code: Select all

UCIoptions {{Hash 512} {OwnBook false}}

... I was able to achieve what I wanted. On first installing the 9934b8e engine in Scid, UCIoptions is blank otherwise. That doesn't seem to be the case when other engines are installed, including stockfish-222-64-ja.