22 July 03 Changelog

[djp]

I got the update, but it doesn't seem to work yet. Here's the What's New.txt file's listing of changes:

07/22/2002

- Algorithm updated - added back energy minimization
between generations, side chains no longer held
fixed in core, and individual structure timeout
decreased by a factor of 2.5 (thus it should get
stuck for less absolute time on average)
- Windows version now supports up to 8 service
installs (each must be in its own directory) for
users with octopi machines
- Added 'purgeuploadlist' function to allow removal
of files from filelist.txt - for Advanced users
only; see the readme for details
- Made several error messages clearer
- Fixed some potential problems with temporary files
from different processes overwriting each other
(when running multiple copies of the client)
- Fixed bug preventing uploading when filelist.txt >
32767 bytes (usually near 337 buffered gen.)
- Fixed FPE on Tru64
- Fixed sin(phi)=0 fatal error message on IRIX
- Fixed handle leak on Windows
- filelist.txt is never erased until a new one is first
created, so it should never be missing if the program
should crash
- On Windows, the file foldtraj.ini (created and used
automatically by the client) goes in the client
directory now
- Logo should show up properly on screensaver now
- Fixed crash if try to upload data with Windows
filelist.txt on UNIX, or vice-versa

[bwkaz]

Originally posted by djp
- Added 'purgeuploadlist' function to allow removal
of files from filelist.txt - for Advanced users
only; see the readme for details

OooOoo! Nifty!

- Fixed bug preventing uploading when filelist.txt >
32767 bytes (usually near 337 buffered gen.)

Ahh, so that's what that bug was. Good to know they aren't using fixed-size buffers... anymore.

[m0ti]

Great Work, Howard!

Now with the purgeuploadlist I should be able to get out that next version of dfQ.

Should probably be next week sometime as I am mucho busy; but ( ) who knows?

[showdown]

- Fixed sin(phi)=0 fatal error message on IRIX

Yay! Works great now!

[DocWardo]

The question is: what is different in this win32 executable that wasn't in the last mini update that we could download to fix the handles issue. Cause something has severly borked the speed on windows. All my machines are crunching full cpu, but I have gone from 2-3K points/hour to 2-3 hundred points/hour. and the linux clinets seem to be the only ones running (too bad I only have two old linux boxes)

[Brian the Fist]

Originally posted by bwkaz
OooOoo! Nifty!

Ahh, so that's what that bug was. Good to know they aren't using fixed-size buffers... anymore.

No no, not using fixed buffer sizes, we were using fixed-size 2-byte int's!!! Those silly C hard limits of data types, eh?

[bwkaz]

You were intentionally using a short? Whoa...

Any 32-bit compiler should, AFAIK, interpret "int" as a 32-bit type... you need to (or at least, I've always needed to, with gcc) specifically request "short" to get a 16-bit type.

Well, whatever. At least it's working now.