QuickTopic free message boards logo
Skip to Messages



^     All messages            824-839 of 839  808-823 >>
Jukka Santala
01:32 PM ET (US)
By the way, if you don't wan't/can't visit the web-site at http://www.egroups.com/group/folding-at-home , you can also conveniently subscribe by sending e-mail to folding-at-home-subscribe@egroups.com

The list has been set publically readable some time ago, so you can read the messages through the web-page without registering. You need to be a member to post messages currently, though, in hopes of avoiding future spam.
Scott Kroyer (Kroyer)
12:00 AM ET (US)
To all newcomers:
This list/discussion board is dead as of December 19th, 2000. Instead, please visit

New posts to this board will most likely go unanswered. See you at http://www.egroups.com/group/folding-at-home!
Scott T. Jensen
11:12 PM ET (US)
OK, I guess move to eGroup is done. To all that have posted to me, I'll post replies to you from the eGroup list.

VIJAY - Please can "Discuss" direction from here to the eGroup list. It would be kind of mean to have newbies post to a dead list.
Vijay Pande
09:37 PM ET (US)

ok, since there has only been praise for egroups and nothing against it, let's move there. I won't be checking out the QuickTopic stuff anymore then and will change the link on the web site.

09:35 PM ET (US)
Poof: What can I say... (starts with a B and ends with a astard ;D
09:21 PM ET (US)
Voluspa - you don't have any betahairpins 'cause I've been getting them finally!!! A whole lot too in the past couple of days!! Just downloaded another one a few minutes ago...

09:18 PM ET (US)
Jérôme: My 486/DX33 is working on a zinc finger (previous was a villin), the P120@133 on a villin and the P166@200 on another villin ;-) Haven't seen a betahairpin for over a month... (still going at Server 1 RUN 2, and others?).

Scott T. J.: Would it be considered cheating if people here popped in at regular intervals to mumble about folding on some existing newsgroup?

Mike: I installed the 1.31 lin client and captured (through Ethereal) the very first communication with stanford. No problems. My distro uses glibc-2.1.3

Vijay: Wasn't an inhouse Squishdot where you finally landed in the first migration discussion... I'd be inclined to say "don't load an already burdened donkey" (no personal comparison :) I'm in favour of _any_ move. Just cut this QuickTopic navelsträng ('tummy-cord') quickly and painlessly. But it is you who must wield the scissors.
09:08 PM ET (US)
Locabinx - I've been over the for the past couple hours. I even posted a reply to Scott Jensen's message to test the threading.

Havent' seen you post over there yet...

08:53 PM ET (US)
Head 'em up. Move 'em out.
Write 'em in, Raw-mail.
check out message # 822 and
hypertext link. Let's meet
at e-group. It's THE place
for folding!
Jerome Landgräfe
08:50 PM ET (US)
Hi Mike,
I've run strace fahclient (although for performance reasons, I mainly run f@h under Windows). Concerning your questions:
#1: I can verify this.Nevertheless, f@h runs without a glitch (even doing a fresh install). AFAIK, sd.dat is used to verify that the client has been shut down properly. If present, f@h starts in normal mode and resumes work from checkpointed files or requests a new WU if necessary. If sd.dat isn't present, then f@h starts in recovery mode, deletes partial WU results and requests a new WU. See my post #754 (btw, an answer to my post #756 regarding the safety of this workaround for the scientific results would be greatly appreciated ;-) ) After testing if present, sd.dat is deleted (recreated after ctrl-c, so after a crash there is no sd.dat)
#2:serverip.dat is sent when successfully connected to the server
#3 see #1
#4 there is no (unknown) signal 32 or any other unknown signal mentioned in my output of strace -omyfile fahclient
#5 got the same problem with strace (I'm not familiar with it, though): how does one capture different threads of a process with strace? strace -f or -ff only works for child processes and generates no results in this case.

Btw, I am running SuSE Linux 7.0 (Kernel 2.2.16,glibc 2.1.2-..) on a dialup connection.
If there is something else I can look into, just ask ;-)
We've got to squash the bug :o)...
But maybe it really simply is a library incompatibility problem. As strace doesn't catch the network calls (at least the way I have set it up...), there is no need posting it here. If you're nevertheless interested in the outputfile, no problems.
Sorry for the long post, but there where quite a lot of questions...
08:45 PM ET (US)
Mike - I don't know how helpful this is but my scrlog.txt file lists an IP... so somewhere I'm guessing it does the lookup and then writes the numerical IP in the file?

This may all be obvious though... ;-)

Here's part of the lin 1.31 scrlog.txt output on my SuSE 6.4 when it completes the final frame of a WU, makes the attempt to connect and upload, and then download new work:



Writing evd data
Finished a frame (100)
TINKER is Exiting following Normal Termination
Returned from dynamic.
Leaving Run()
Finished running, sending data
Connecting to server
IP =
Sent result block.
start 00000000 / end 00000001
About to delete files
Deleting files: prefix project, suffix arc,dyn,xyz,key,bed,evd & work.dat
Finished deleting files
Starting loop
Deleting files: prefix project, suffix arc,dyn,xyz,key,bed,evd & work.dat
Finished deleting files
Getting work
Connecting to assignment server
IP =
Getting server from AS
Connecting to server
IP =
Connection made
Work request made
Waiting to receive work assignment
Work assignment received
Received assignment block:
current 00000000 / end 00000001
nsteps: 50000 dt: 2.000000 dt_dump: 10.000000 temperature: 300.000000
" 177 betahairpin
     1 CH3 -28.616882 -11.586716 45.773370 7 2
"parameters ./opls.prm

cutoff 16.0
taper 12.0

Got work, writing checkpoint file
Starting run
Writing local files: project.xyz project.key.
Attempting to write file "project.key"...Successfully wrote file "project.key".
Attempting to write file "project.xyz"...Successfully wrote file "project.xyz".
PROJECT="project", NSTEPS=50000, DT=2.00, DTDUMP=1.00, TEMP=300.00
TINKER: Software Tools for Molecular Design
Version 3.8 October 2000
Copyright (c) Jay William Ponder 1990-2000
08:26 PM ET (US)
see you all at e-group :)
06:58 PM ET (US)
Jerome and siraj (#823): I'm running "strace fahclient" to
do a system call trace of the linclient 131. I've noticed one weird
thing in the following order: 1) after opening sd.dat and writing a one it closes this file, then 2) it tries to a do a rmdir(sd.dat) which
generates ENOTDIR error since it isn't a directory, then followed by a unlink(sd.dat).
Question: Why the rmdir and unlink? Jerome can you verify this too.
Question#2: I see that serverip.dat is missing. The binary
has it and the winsaver has the actual file.
Question#3: After you open the pipe() your write sd.d with
length 5. Whats this?
Question#4: Jerome are you getting unknown signal 32? This may be user defined. I'll have to try the books. Question #5: I'm having problems being able to trace which process does the actual network lookup or reads the client.cfg file. I'm stuck reading forever on reading scrlog.txt. Anybody have better luck?
Jérôme Landgräfe
06:55 PM ET (US)
just a question:
I have noticed today, that only the top 25 users are still mentioned in the server x run y stats. Is this permanent (problems with server load my tables getting too big?) , will we see the old listing again or are there maybe plans to add an input form on each run-page to query for the number of WU one contributed to this particular run? Or has the information been moved elsewhere?
Just curious,
Jérôme Landgräfe
06:06 PM ET (US)

sorry: Siraj already posted some of this while I was typing ....

Re: #819
Hi Mike,
as far as I understood, f@h first contacts the assignment server (AS) vsp20, which then attributes a server (currently vsp21-vsp24) from which to retrieve a WU according to client speed (fast computers are more likely to get villins, if my memory serves me right) & according to load-balancing needs. By first using the AS, vijay can add new servers to handle the load without releasing new clients. Correct me if I'm wrong.
Your problem really seems to be a tricky one :-( .
squishdot: would be nice, as long as, like Mike & Poof already mentioned, performance issues aren't (an issue ;-) )
Edited 12-19-2000 06:09 PM
Scott T. Jensen
06:03 PM ET (US)
Vijay - I think regardless what we do about this list, we should still set up a newsgroup to hopefully bring in usenet users to this project.
^     All messages            824-839 of 839  808-823 >>

Print | RSS Views: 13574 (Unique: 3110 ) / Subscribers: 0 | What's this?