The case that the App is interrupted between writing the last checkpoint and writing the final result file is rather rare....
Possibly less rare than you might think.
I've noticed a distinct tendency for BOINC to call for new work during the last couple of minutes work on a running task.
Scheduler contact --> debt recalculation, which in turn may --> task switch.
If a task switch is indicated, BOINC will wait until the next (final?) checkpoint before switching.
If the user doesn't have 'leave tasks in memory' checked, or switches their machine off while the task is waiting to run for the final time, the task switch could be fatal to the WU. Well worth fixing.
And a particularly opportune time to be fixing it, prior to the release of S5R5 - if that's still the plan. Any news on that front, Bernd?
The case that the App is interrupted between writing the last checkpoint and writing the final result file is rather rare....
Possibly less rare than you might think.
Actually with the features of newer core clients like throtteling and EDF mode these effects have become more frequent. The bug was in the original HierarchicalSearch code since we began to use it for S5R2 in May last year (IIRC), I've never noticed this type of crash as a problem back then.
RE: The case that the App
)
Possibly less rare than you might think.
I've noticed a distinct tendency for BOINC to call for new work during the last couple of minutes work on a running task.
Scheduler contact --> debt recalculation, which in turn may --> task switch.
If a task switch is indicated, BOINC will wait until the next (final?) checkpoint before switching.
If the user doesn't have 'leave tasks in memory' checked, or switches their machine off while the task is waiting to run for the final time, the task switch could be fatal to the WU. Well worth fixing.
And a particularly opportune time to be fixing it, prior to the release of S5R5 - if that's still the plan. Any news on that front, Bernd?
RE: And a particularly
)
The bug is fixed in the code now. WRT S5R5 - Badri who intended to run some simulations fell sick. Hopefully will be back at work soon.
BM
BM
Thanks for the update,
)
Thanks for the update, appreciate it.
Edit: just wondering if this is another example of it, since this person's WU both appear to have crashed at the same point in the task.
RE: RE: The case that the
)
Actually with the features of newer core clients like throtteling and EDF mode these effects have become more frequent. The bug was in the original HierarchicalSearch code since we began to use it for S5R2 in May last year (IIRC), I've never noticed this type of crash as a problem back then.
BM
BM
- exit code -1073741131
)
- exit code -1073741131 (0xc00002b5)
task1
task2
CPU QX6700 oc. 3517(11x320), RAM 2x1GB oc. 960(4-4-4-12)
Running 2 Einstein 6.05 and 2 SETI AK_v8_win_SSSE3x. Core temps are below 60C.
Pc is PRIME and MEMTEST stable and running only BOINC 24/7 more than year with no
errors until now.
Running now with lower OC
RE: - exit code -1073741131
)
Hi
Reason: "Unknown exception"
Could be hardware or the code it appers to me. The OC seems like a good place to start.
Thanks for reply. I run
)
Thanks for reply. I run optimised 6.05 applications near month with no problem
but week ago I raised OC and after six days I get errors. Like Richard says
optimised applications put more stress on the CPU so the OC is possible problem.
Will try a few weeks with lower OC.
RE: RE: And a
)
I built a Windows App from this code that should fix the problem.
BM
BM