Perhaps Rickjb's wingman was a new system, went into low power mode or was restarted frequently, but it would still need to have been using an old client; with runtime instead of CPU time being reported, assuming the 12h cut-off exclusively goes by cpu time and does not read run time.

3.22h was the shortest valid I have seen on a SB (HT on) and 8.22 the longest. i7-920 (HT on) is 5.41 to 9.77h.
These complete all 16 sub-jobs.

Only one i7-920 PV listed with 12h time:

<stderr_txt>
INFO: No state to restore. Start from the beginning.
[22:31:07] Number of jobs = 16
[22:31:07] Starting job 0,CPU time has been restored to 0.000000.
[22:33:46] Finished Job #0
...
[05:08:30] Finished Job #13
[05:08:30] Starting job 14,CPU time has been restored to 23543.234375.
[06:46:08] Finished Job #14
[06:46:08] Starting job 15,CPU time has been restored to 29379.062500.
Killing job because cpu time has been exceeded. Subjob start time = 0, Subjob current time = 1088204996
[10:37:23] Finished Job #15
10:37:33 (3096): called boinc_finish

The last sub-job was pushing 4h, but still got cut off.
Unless it was the case that this last sub-job was allowed to run to completion, and this was recently changed. But doubt it, this was also sent on the 20th and returned on 21st.
I can see 4 pages of CEP2 tasks and nothing similar. Pity we can't see some Boinc and system info with the tasks at WCG. It would put pay to a lot of guesswork.

Was not sure about CPU time or runtime being only used by Boinc in the past; it’s been a while since I used such an old version.
Sounds like Use CPU, While computer is in use was set to Off. Still don’t understand how it went past the first job though?

[OT Winge]
I think these sort of problems stem from one of the fundamental problems with Boinc - Boinc cannot distinguish between different task types from the same Project. Therefore you cannot have independent default settings on a project-by-project basis.
As a result, if I want to run 4 CEP2 tasks, I cannot. If I want to have CEP2 use the CPU when system is in use but not HCC, I cannot. I cannot configure Boinc to crunch one project but not another when the system is in use. When the system is using more than 25% of the CPU, or at different times of the day (office hours), I can’t continue running one project, but stop using cores for another project, not even after say after 2h (to reduce power usage), or run different numbers of tasks, or to allow a GPU task to continue but not CPU tasks (at peak power times, when a specific program starts, or time of day), or continue to completion if completion will be within x minutes…

The more cores/threads and GPU's and the more applications (HCC, HPF…) from the one Project (WCG) there are, the more users will want to explicitly configure individual projects (apps). Boinc as is, has too many problems for multiple application Projects, doesn’t cater well for mixed Project crunchers, and when it comes to GPU crunchers, well that was just an afterthought. So these problems are not just limited to WCG, it affects dozens of projects, including GPU projects, and projects that use CPU’s, ATI GPU’s and Nvidia GPU’s.

If WCG started using their own seperate Boinc version it would be useful, so long as the Berkeley one could run along side (for other projects).
[/OT Winge]

Sorry, yes, forgot about the Runtime added in need for GPU tasks.

Quick look, Win98 still contributes 1346 RAC and ME 366... amazing.

Thanks

--//--

Thanks, Rickjb,

The snip you posted, also showing the wallclock time suggests there was like 2.25 hours at most since the restore at 30.031 seconds, so it's odd how 10.6 hours at most slipped in as 34.25 hours.

I wrote platform, but meant client version which is printed at top of log.

Anyway, I've polled the techs on this oddball and hope they'll get to it before the task is assimilated out of the live db.

cheers

--//--

Thanks for sharing that. Been following the reports and discussion on the Alpha mail list since start. So far unable to reproduce with .26 or .33, both quad and duo [Laptop] on W7[SP1] or Linux [quad] 11.04. Whilst running CEP2, mostly 3 concurrent on quad, I'd recommend to give it real serious time before unplugging power. At any rate, powering up properly resumes BOINC, again it needing serious time to fetch the state-image from disk.

BTW, Did not know .33 had advanced to ''recommended'' and to this moment showing (MAY BE UNSTABLE - USE ONLY FOR TESTING) . Though there's pushing, it'd be [fill in the blanks] to promote this version via a message in the clients, knowing this [to me] serious issue is loitering. They can do the aggressive developing all they like, but it's time for them to not interrupt production grids until there's a solid, proven release.

--//--