Author: Robert Hyatt
Date: 17:11:01 11/23/97
Go up one level in this thread
On November 23, 1997 at 13:44:57, Keith Ian Price wrote: >On November 23, 1997 at 09:20:08, Robert Hyatt wrote: > >>On November 22, 1997 at 23:44:51, Tim Mirabile wrote: >> >>>On November 22, 1997 at 17:14:03, Robert Hyatt wrote: >>> >>>>On November 22, 1997 at 16:18:00, Darrin Moon wrote: >>>> >>>>the problem is obvious looking at the log files. You have corrupted >>>>endgame databases. There are *no* mate in 44's in the 3-4 piece >>>>endings, so you have something wrong. Most likely one or more (or all) >>>>are corrupted. >>>> >>>>Here is a complete list of the 3-4 piece files and their sizes. There >>>>must be *no* deviations in file size: >>> >>>Perhaps he FTP'd them in ASCII mode instead of BINARY? >> >> >>possibly. Also these things can fill the drive when you unzip 'em, >>which >>can truncate one here or there. But they *must* be correct, because >>Crafty >>probes them in a completely trusted manner. If the file is corrupted, >>bad >>results will show up... > >Perhaps a quick checksum on the available tablebases, upon firing up >Crafty would be a way to avoid unnecessary losses due to corrupted >tablebases. Also, I believe I recall you saying that all of the 3 and 4 >man bases must be present. If this is true, then a count should be added >also. Just a suggestion, in case Crafty were to be tested on SSDF for >example. > >kp the problem is that "quick checksum" is an oxymoron when you have the 5 piece files as well. try reading thru 2.5 gigs and checksumming each file... and the flag will fall before the first move is made. :) simplest solution is to watch the sizes... I will try to dream up some sort of "certificaton maybe... such as reading the first/last byte of each file and checking it or some such thing...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.