Pls look at RID=11253076, WUID=4128572, name=05ap10al.22292.6616.140733193388035.14.174. stderr{
...
Unhandled Exception Detected...

- Unhandled Exception Record -

Reason: Breakpoint Encountered (0x80000003) at address 0x74EA3219
...
}

Also, setiathome_6.98_windows_intelx86__cuda42.exe suddenly takes about 3 HRs to process a WU, whereas before it took about 17 min, when two WUs per GPU are being processed (anonymously). Seems to be working normally when only one WU per GPU is scheduled.

---

Task 11253076 ended with

Exit status 197 (0xc5) EXIT_TIME_LIMIT_EXCEEDED

When BOINC kills a task like that, it deliberately trips the 'Breakpoint Encountered' trigger, so that you (or the developer) gets a full debugger dump, in case the unexpectedly long runtime is caused by an infinite loop or other programming error. So no problem there.

It looks as if both issues are one and the same.

You have two GPUs, both GTX 570, and you're running two tasks on each? That doesn't seem excessive - plenty of people run that many or more with the same application on the main project, and I run four at a time there on a single GTX 670 - no out-of-proportion runtime increase there.

Obviously, we have to test the same configuration with the new v7 processing we're here to test, and it would also be wise to re-test, both singly and in multiples, with the new x41zb application which Eric is planning to deploy today. But provisionally, I'd be inclined to look for hardware issues affecting your host only - it's not a reported problem with the application in general. How do your other hosts behave under similar conditions?

BTW, could you (and anyone else running Anonymous Platform) please ensure that all traces of the original x41z are removed from app_info.xml and replaced by x41zb, as soon as possible after the new versions appear on the applications page? That'll help us monitor for any remaining validation problems which might inhibit transfer to Main.

That could also be the problem which I reported some days ago at the Alpha mailing list.
I didn't get the time limit exceeded because my hardware isn't that fast. If it is overrunning one or two hours it is still inside the limits, I see it and reboot the machine.

On faster machine it could be of course that BOINC hit simply the time limit.

EDIT: Can you enable task_debug and wait if it happens again? Then check if there is something in the log around the time when the task hit 100%?

---

Christoph