Some Windows bugs fixed...

We have fixed a number of minor bugs which mostly (if not exclusively) affect Windows users. The handle leak should be fixed. It is unclear if this was also what was causing the memory leak (ie the handles using memory) although I doubt it... The handle leak was definitely related to the -g frequency option.

Also some bugs or potential bugs with the TMP / DFPTEMP paths and files are fixed. Specifically, those few people who observed a "Insert duplicate residue" error should hopefully never see this message again. There should be no need to have separate TMP directories for multiple processes on multiple CPU machines.

If you are feeling lucky, the updated Windows EXE is here:

ftp://ftp.mshri.on.ca/pub/distribfol...ldtrajlite.zip

You can just overwrite your current executable with this one (assuming you've got the Phase II client already installed that is). I suggesting backing up the old EXE first just in case.

So if you are experiencing troubles please try out this one and let us know what happens, esp. re the handle leak and memory leak.

In the mean time we will search for the remaining 2-3 problems which have yet to be reproduced on our end...

just a quick comment - the new EXE doesn't want to run as when installed as a service (I get the below message when starting via Administrative Tools|Services, and dfGUI won't start it either)...



not sure if this is mean to be or what but I'd thought I'd mention it...

Two more things I forget to mention:

:drums: IMPORTANT! :|party|:

1. Please remember it is normal for memory usage to go up DURING a generation - but it should get 'flushed' at the end of a generation (if you keep it going)

2. The new executable also fixed the Windows directory permission issue. So you shouldn't need write permission in the WINNT or WINDOWS directory any longer :looney:

Ok I lied there's 3:

3. I've been messing around with the service stuff so Ill fix that tomorrow - for now command line only...

Looking great so far!
I put it on 8 boxen, all w2k except for 1 XP boxen :thumbs:
Disabled the scheduler from restarting the client every night, let ya know tomorrow :D
:cheers:

I also am unable to run this version as services on my dually xp pro rig. Same kind of error. I'm back on the original file.
I did try to uninstall it as a service then reinstall. :confused:

Command line only - NO SERVICE - yet ;)

It is doing so well, I added it to 21 more boxen, all w2k, 3 of em Duals. All AMD rigs of course. ;) :D
Nice work Doc! :smoking:

Steady as she goes!!

Almost 4 hours now and the file handles are only 2 above where they initially started.

Memory is steady at around 90 MB. :cheers:

---------

Brian, you do know that the linux client also suffers from a similar or the same memory leak ??

workin well for me too.
over 10 hrs of cpu time -g1 47 handles and 92 megs of RAM on xp pro.
well done, thank you.

:notworthy Howard :hifi:
Working Great on win98/w2k/XP :D :thumbs:

:|party|: Steady at 90 Megs Cpu1, Cpu 2 on old version ( started together) 107 Megs

Note that I did NOT find or fix any memory leaks so I still expect that to be present, but we'll see..

Well Howard, the Universe has a sense of humour. You may not have found the leak, you may not have put a patch on the leak, but whatever you fixed is so happy to be working it went out and fixed it for you. I can tell you the updated foldtrajlite does not leak memory like the old version, always sits on 90 megs while my Duallies 2nd Client with the old version is now up to 120 Megs.

:hifi:

I went back to your new trajlite file after overnight finding that nearly my whole 512 meg of memory was taken over by the clients on my dually.
I started very early this morning at 88 to 89 meg per client...I'm still at 88 to 89 meg per client as of 5 hours runtime.
For now I'll use dfgui to hid the apps windows but I look forward to your next test release if it offers 'service' operation again.

Ok, I've updated the binary again, same link as above (make sure your browser isnt just caching it...). The service thingy should work now. As a bonus you can now have up to 8 services (!) since someone asked for that. They should all work in harmony and with no TMP file collisions anymore I hope. I think I also may have inadvertantly found and fixed a bug that prevented the 2nd service from working properly (who knew?). Specifically it would explain why the 2nd service never restarted properly after an auto-update. We will see if this changes now I guess..

Still a few more bug fixes to come esp. the gen 337 one and the sin(phi)=0 on IRIX. Keep us posted on the memory leak if its still an issue and any similar old problems.

Thanks Howard!

Yay! It runs as a service AND it doesn't leak :|party|:

If most/all of these bugs are corrected, will you release updates for the affected clients, or wait for the next protein? Or, is a memory leak-fixing Linux client on the way out, too? :D

I've updated and been running it as a service for ~4 hours now and so far it looks good - no problems, no leak :thumbs:

i installed as a service last night (win xp pro, service install did not work with the previous foldtrajlite.exe files)
it ran fine all night, until my dsl went down, but other than that, everything is A-OK on my end. thanks howard.

Has the handle leak issue been fixed for all clients or just the Windows one? Would it be possible to get a clarification on this point?

Windows - just like it says in the subject ;)

Nary a word has been spoken about any other OS.

I thought the memory leak was somehow related to the 'gen337'-failure? After reading this thread, I understand it isn't?

Can somebody clarify this?

First, this has only been released for windows so far.

It looks like Howard fixed the memory leak (without really knowing it :D ). At least no one that has upgraded to the new release, has seen any memory leaks, me included. I've had the new client running as a service on this machine (P4 2Ghz, WinXP Pro) for close to a day and there's no signs of any memory leak :smoking:

Quote:

---

Originally posted by [DPC]Mobster
I thought the memory leak was somehow related to the 'gen337'-failure? After reading this thread, I understand it isn't?

Can somebody clarify this?

---

The two don't seem to be related, no. I've seen memory leaking long, long before gen 337 -- and I don't run offline, either, so I'd never see the 337 bug anyway.

Indeed, the memory leak has no connection to the Generation 337 error.

On the bright side, we have discovered what is causing Gen 337 errors and it should be fixed by next week - please see the News page on the website for more information.

Quote:

---

On the bright side, we have discovered what is causing Gen 337 errors and it should be fixed by next week - please see the News page on the website for more information.

---

So, do you still need 337-bug samples, or should I just upload the results before my test boxen reach that point?

We already got what we needed - thanks

Will the update be an auto update, or have to be manually installed ?