Unrelated, though my 64 bit client produced again a perfect result, the funny with the negative credits has apparently something to do with... mostly 64 bit clients generating corrupt output files. Noted before by the tech, the 2 minute issue has the attention, but the focus remains first on getting 2 projects launched.

Personally, I'd go for a project mix also because HPF2 actually belongs to the more cpu heavier. [ot]Currently have snug up to 3 badge simultaneous upgrades (now within 10 days, 10,8,4 days to be exact), trying to time this so they fall on the same day (an outside chance I even get a fourth, virtual, to happen for an anticipated level, 11 CPU days out). Why it's done like that... because I can :D [/ot]

Hypernova,

If the 2 minute fails grow too large in numbers within 24 hours (UTC timekeeping) you'll be facing quota dry-up, so be sure to continue to return valid results in-between.

Tonnes of errors indeed but what you have there looks like already one-third towards Ruby.

The rule works like each error counts as quota 1 reduction and eventually even if good results will double the quota again there the hard daily of 80 per core or whatever the number is. You're on about > 50% error out rate. Having a buffer in this particular case helps to ensure you keep crunching.

Hypernova, count the PV as valid, near guaranteed. Then 490:422 makes it presently a ratio of 54:46, better than 50% valid. Good enough to keep running.

Watch out for those notorious and rare HPF2 jobs that loop endlessly. See a task with a run time / % progress ratio that's abnormal, suspend the client with LAIM (Leave application in Memory) OFF and resume after 30 seconds so the job unloads from memory and resumes from last checkpoint. 99.9999999% it finishes then properly and faultless. BOINCview used to have a warning mechanism for low progress jobs and think BOINCTasks has it too but have not toyed with it.

I am wondering if we have ever seen such errors in Linux or Mac machines? Indeed my question tends to give my answer: I have the feeling that all these errors have been reported by Windows (all flavors) users.

I have run about 50 CPU days of HPF2 WUs recently in order to get the emerald badge, most under Ubuntu 64 and the rest (<20 %) under XP 32 and the only error I have seen was under XP 32 in the quad.

Could this be just another general Windows flaw, or a not-so-good library used only by Windows jobs?