Trouble uploading work with linux client

[AMD_is_logical]

I'm using the Linux client (Intel Compiler) and tried uploading work for the last protein using the -ut switch. This was from a batch file that had always worked fine. The program got to where it was trying to upload, then announced that it had crashed, then exited.

It didn't crash if there was no work to upload, but always crashed if it tried to upload work.

After much playing around, I found that it would upload if I added the -qt switch.

I haven't tried it with new work yet, as none of my nodes have finished gen 0 yet.

[Keller]

Thanks ... I noticed the same problem and thought my work was lost.

[Brian the Fist]

Sound like it could be a bug. Could you provide clear, concise steps to reproduce this problem so I can test it (on Linux)? Thanks.

[Keller]

I had some buffered generations from the previous Client/Protein .
Then I started the client and the autoupdate proceeded successfully. I restartet the client and the client crashed (it said caugh sig 11). After a view in the error.log there is shown a Prev Gen Missing error ... I dunno why it is there because evrything went well with my connection but i think this error is not the reason why the client crashed completely ...
(I guess "caught sig 11" means a prev gen error but I dont know it exactly, cause this is the first error since im using Linux ).
Well ... I think its important to say that all points are shown in the stats ...
Now the client tries to go through generation 0 of the new protein

[Brian the Fist]

I believe this is just from a misplaced ncurses call and nothing serious - you can run with quiet=true (-qt flag) to avoid it for now. We'll try to get fixed binaries posted tomorrow with a clear explanation of what the problem was.

[Mikus]

Originally posted by Brian the Fist
I believe this is just from a misplaced ncurses call and nothing serious - you can run with quiet=true (-qt flag) to avoid it for now. We'll try to get fixed binaries posted tomorrow with a clear explanation of what the problem was.

The trouble with using the -qt flag is that then I DO NOT SEE what the client does. In particular, I do not see how many filesets it uploads. To me, -qt is not an acceptable solution.

You wanted directions as to how to recreate this on Linux: (1) have at least one fileset completed and ready to upload, but have the client halted. (2) be connected to the internet. (3) issue ./foldit -- simplest way is with the -u t flag. Doing this (with a client which I downloaded TODAY from the DF download server), I get the message that the client has crashed. Error.log shows:
========================[ Nov 1, 2003 3:59 PM ]========================
Starting foldtrajlite built 2003.10.26
Sat Nov 1 15:59:19 2003 ERROR: [001.001] {foldtrajlite2.c, line 2073} Caught sig 11

mikus

[AMD_is_logical]

Originally posted by Mikus
The trouble with using the -qt flag is that then I DO NOT SEE what the client does. In particular, I do not see how many filesets it uploads. To me, -qt is not an acceptable solution.

Hmmm... The client creates files of the form <handle>*.bz2 as it uploads, so maybe you could count the number of such files (with the help of the wc command), and then you would know how many units had been uploaded. You could compare that with the number of fold_* files.

[Brian the Fist]

I did say this was only until we posted a fix. Anyhow it has been corrected and the new binary packages will be posted for Download in a few hours on the Download page of the web stie.

[Ned]

Anyhow it has been corrected and the new binary packages will be posted for Download in a few hours on the Download page of the web stie.

Did this upload ever occur??? I just got the error

Ned