We did have a report of this earlier, but we were unable to pin down the exact cause of the problem. As I remember, it only affected FAAH according to that report. Since a new version of the FAAH code is about to be released, I advised waiting for that to see if the problem existed in the new code.

This is the first I've heard of this particular behaviour with HPF2, although there is a similar (very rare) bug that sometimes stalls HPF2 on any platform (Windows, definitely).

If/when this happens again, please will you try suspending and resuming computation, instead of the more brute-force methods you have tried successfully?

I will bring this to the attention of the techs, and perhaps they will be able to do an extended beta test of the new FAAH code on the Mac platform (normal beta tests only involve a few hundred work units, and very rare problems aren't always picked up on).

The big difference between GC on one side and HPF2/FAAH on the other is that GC uses a comparison algorithm (SSEARCH) - it is doing memory manipulation and integer arithmetic. HPF2 and FAAH both do extraordinary amounts of floating point mathematics. Read this: http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.18 for a particularly scary example of the pitfalls awaiting the WCG techs.

Works for me in Firefox 2.0. But just in case IE or some other browser is screwing it up, here's a http://www.parashift.com/c++-faq-lite/newbie.html#faq-29.18 different way of making a link in BBCode.

edit: the site may be having DNS problems. Try again later.

Hey Sek, isn't this regarding this thread?
http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=15380
I'm sure just a copy and paste error

As to the OSX problem. I have this same problem. Suspending and resuming the computation does not help. You have to shut boinc down and re-start it to get it working again. I am trying one thing, under my processor preferences I unchecked "Allow Nap". I've noticed it gets checked without my doing. Wondering if that might be the problem.
Edit: To get the processor prefs you need to have the developer tools loaded. I think the tools are called CHUD or something like that. It's on the install dvd.

We've tried suspending/resuming the task and project and boinc from the top menu to no avail. The only method that appears to work are the aforementioned brute force methods

TBH, Waiting 30 seconds doesnt make things much better from the user side. I run 16 dedicated macs of various flavors and its easier to just restart the herd than go to each machine to see if its stalled.

Not to be combative, but based on the small number of the people in my team that run OSX (they all report the problem), and my experience, I wouldn't characterize the problem as "rare" and is present in both the remaining projects (FAAH, and HPF2) though they might be different problems (with the same symptoms). The strange part is it rarely if ever, affects more than 50% of the cores available - often its only one. Some also have the impression that HPF2 has this problem "to a lesser degree" than FAAH, though I personally cant verify this - Its still a royal PITA with 16 machines.

If I can be of any assistance in hunting this thing down please LMK here. In the meantime, I suppose my only option is to bootcamp windows on these things and get them back to work as things are quite untenable as they currently stand. <sigh> More licensing revenue for MSFT...

-SS

P.S. This CHUD is on the OSX install DVD?

Kudos to all Techies everywhere who deal with this stuff!