Windows S5R3 App 4.11 available for Beta Test
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250655727
RAC: 34274
18 Oct 2007 10:26:50 UTC
Topic 193237
(moderation:
A new Windows App is available from our Beta Test Page.
This is the one that will throw a floating-point exception in case of a serious numerical error instead of the time-consuming explicit checking we did before (I never did this on Windows, prepare for the unexpected...).
I think that the 4.11 will be able to do the workunits on which the 4.07 fails (se thread "validate errors").
With the app_info.xml in the package the 4.11 should pick up tasks where 4.07 left them.
Please test & report!
BM
BM
Windows S5R3 App 4.11 available for Beta Test
)
Yup, this time no problems:
18-Oct-07 12:39:05|Einstein@Home|Restarting task h1_0520.20_S5R2__153_S5R3a_0 using einstein_S5R3 version 407
I've switched host 1001562
)
I've switched host 1001562 (twin sister of the one I'm doing the timing charts on). Like Jord, no problems - picked straight up from 4.07. The matched pairing will give a performance comparison, if anyone's interested.
RE: With the app_info.xml
)
I've switched four hosts, all of which were running 4.07 beta by ap_info. The three Windows XP machines (Banias, Core 2 Duo E6600, and Core 2 Quad Q6600) appear to have resumed a total of six in-process results without complaint, and one such has already completed. As the confirmation result remains unsent (over five days after issue to me!) validation may have to wait a long time.
However the fourth host, a Coppermine Pentium III running Windows 98SE errored out instantly on restarting.
As the stderr out is available at the linked result page I won't transcribe it here, but in case it is of any interest, here is the message traffic for the event.
10/18/2007 9:29:23 AM Output file h1_0295.80_S5R2__36_S5R3a_2_0 for task h1_0295.80_S5R2__36_S5R3a_2 absent
10/18/2007 9:29:23 AM Computation for task h1_0295.80_S5R2__36_S5R3a_2 finished
10/18/2007 9:29:23 AM Reason: Unrecoverable error for result h1_0295.80_S5R2__36_S5R3a_2 ( - exit code 99 (0x63))
That same host without tampering has subsequently downloaded three fresh results in rapid succession and errored each of them as well. So the issue here appears not to be with the in-process resume but rather with either my installation or with compatibility of the beta ap with my system.
I've left the offending Windows 98SE machine alone, just suspending Einstein, in case there is any useful information its current state might yield. If I don't get requests or suggestions within a couple of hours, I'll try just rebooting the machine, restart BOINC, resume Einstein, and check if the next result does the same thing, and if so I'll go back to 4.07.
[Edited for multiple typos, then later actually to put in the link to the first error which I'd claimed, but failed to put in the original post]
Thanks for the report. I need
)
Thanks for the report. I need to think and investigate.
I had this a few hours ago on a Mac after installing an App built from the same code for internal testing. After four client errors, however, it now seem to be working fine. Reminds me to what I saw with som 5.10 BOINC clients that apparently started working properly only after a few client errors. Maybe something is broken in the BOINC version we are linking (also see the thread for the Linux Beta).
Better wait until the tasks in progress have finished before switching to the new App version, just to be on the safe side.
BM
BM
If we were not having
)
If we were not having problems with 4.07, then is there anything other than just making sure 4.11 doesn't crash?
I'm sorry. I think I know
)
I'm sorry. I think I know where this error comes from, and it's not BOINC; it's actually my code. I'll probably not be able to fix it before tomorrow. Glad to have this Beta Tests...
BM
BM
RE: I'm sorry. I think I
)
I can't test any fix on my erroring Windows 98 host until tomorrow anyway. I thought I'd check to see if it would heal itself as you reported on the Mac while I de-scaled a humidifier. Hit my daily limit of 35, so no new results for that host until tomorrow--though maybe that is only six hours away in UTC. In any case I'll remove the ap_info from that host and restart.
Actually, it feels like there is more point in participating in the beta tests when something useful turns up. Glad to be of service.
http://einstein.phys.uwm.edu/
)
http://einsteinathome.org/task/87617346
This one was mostly done by v4.07 (80+%) but finished up with v4.11. And, it validated OK.
A new App that has this bug
)
A new App that has this bug fixed is available. Transition from 4.11 should be smooth.
BM
BM
Never mind... goes in the
)
Never mind... goes in the other beta thread.
Kathryn :o)
Einstein@Home Moderator