Hi all,

I am using Boinc 5.10.21 on AMDx2 64 running Vista 64

I am getting client errors for nearly all of the work units that I am submitting. The stats pages for the various projects shows 'Client error' for the outcome but 'Done' for the client state. Done seems to mean done without errors which contradics the Client error message.

Looking on the World Comunity Grid website I can get a bit more detailed info on the failure but I do not understand it. Here is the log....

<core_client_version>5.10.21</core_client_version>
<![CDATA[
<message>
- exit code -1073741510 (0xc000013a)
</message>
<stderr_txt>
World Community Grid AutoDock (projects/www.worldcommunitygrid.org/wcg_dddt_autodock_5.08_windows_intelx86) version Failed to get VersionInfo size: 2

Failed to get VersionInfo size: 2
INFO: No state to restore. Start from the beginning.
ERROR: Restoring checkpoint failed. Unable to restore state!
INFO:[17:50:20] Start AutoGrid...

autogrid: autogrid4: Successful Completion.
INFO:[17:53:06] End AutoGrid...
Beginning AutoDock...
INFO: Setting num_generations: 27000
Setting maxGen to 6750
About to enter main loop...(dockings already completed: 0)
call_glss(): pop_size: 200 num_evals: 10000000 start: [17:53:27]
call_glss(): begin 2nd for loop over pop: [17:53:27]
call_glss(): end 2nd for loop over pop: [17:53:27]
call_glss(): end 1st do/while: [17:53:27]
_maxGenSeenSoFar changed: 6750
_maxGenSeenSoFar changed: 7088
call_glss(): end: [18:04:19]
_maxGenSeenSoFar changed: 7443
Updating Best Energy for WU: 0.00
call_glss(): pop_size: 200 num_evals: 10000000 start: [18:04:26]
call_glss(): begin 2nd for loop over pop: [18:04:26]
call_glss(): end 2nd for loop over pop: [18:04:26]
call_glss(): end 1st do/while: [18:04:26]
call_glss(): end: [18:15:43]
call_glss(): pop_size: 200 num_evals: 10000000 start: [18:15:45]
call_glss(): begin 2nd for loop over pop: [18:15:45]
call_glss(): end 2nd for loop over pop: [18:15:45]
call_glss(): end 1st do/while: [18:15:45]

</stderr_txt>
]]>


I hope somebody can help as I am wasting lots of CPU time at the moment.

I have just looked at the log again and think I may understand. The project is trying to restart but failing to load from its checkpoint.

Could this be because I have multiple users on the computer running Boic and when one user logs on they destroy the work done by the previous user? Just a thought!

See this FAQ. It'll give you a workaround to deal with the 0xc000013a error.

---

Could this be because I have multiple users on the computer running Boic and when one user logs on they destroy the work done by the previous user? Just a thought!

If both users run BOINC at the same time (ie. two instances of BOINC running on the same directory), they'd be destroying each other's work all the time... But I don't think that's possible, there's lock files and other mechanisms to avoid that from happening... hmm

I have added the waittokill to command to see if that helps.

I'll see how it goes.

I had this error 4 times from seti last night. I don't think it's related.

10/10/2007 02:30:14|SETI@home|Reason: Unrecoverable error for result 18mr07aa.24761.481.14.6.67_2 ( - exit code -6 (0xfffffffa))

Hello,

I'm running WinVista too and I'm having same problem as martyncarr - Client error messages. I applied FAQ on MundayWeb, but there is no reuslt - Vista kills BOINC computation so fast, that projects can't write checkpoint...
Anybody solved this problem?
Thanks in advance

10/10/2007 02:30:14|SETI@home|Reason: Unrecoverable error for result 18mr07aa.24761.481.14.6.67_2 ( - exit code -6 (0xfffffffa))

No, that's an application error, not a BOINC error. Best post that one on the Seti forums.

---

Anybody solved this problem?
Thanks in advance

The only real fix in BOINC is coming in the next major BOINC update, 6.0
But when exactly that one is released, is anyone's guess. We're not even testing it yet. End of this year/Start of next year would be my best guess.

---

Hello Ageless,

thanks for Your reply. Do You have direct link for the Seti post?
That is sad, that solution on Boinc side will be done on next year. I have already 3 dual-core PCs running Vista, so we have to wait and hope :-)
Thank You anyway :-)

I have added this to the registry also.

[HKCU\Control Panel\Desktop]
"WaitToKillAppTimeout"="20000"

As I run Boinc as a application I thought this might be more appropriate. I have tried logging on and off a few times and so far so good.

Thank you, added that to the FAQ.

---

Well that did not work either so probably best to take it from the FAQ.

I have decided to run as a service now. Unfortunatly I cannot get the screensaver to work even with the work-around. I just get a black screen.

Oh well, roll on Boinc 6

I have decided to run as a service now. Unfortunatly I cannot get the screensaver to work even with the work-around. I just get a black screen.

Graphics don't work as a service on Vista, and there is absolutely no way around it, other than waiting for BOINC 6.

Hello,
I tried both registry entries in FAQ with one interesting effect: one workunit is ended properly, second not (I'm running DualCore). Anyway it doesn't solve problem for me. So I tried Boinc Manager version 6 (6.1) with same effect. Behaviour is same as with BM 5.10.20... So right now I'm waiting for final BM6 :-)
Thanks anyway!

File - exit out of BOINC Manager (single/shared) or stop the service?

I've never lost work doing that.

---

Kathryn :o)

To KSMarksPsych - yes, this works, but I would like to install BOINC also on computers with users, which are not good in using of computers and they use it maximally for Internet and games. Therefore I would prefer automatic regular shutdown of Boinc...

I too have had this problem of Vista shutting down too quickly for BOINC.

The only sure way of not getting these client errors seems to be to shut down BOINC Manager first, WAIT 10 or so seconds and then shut down the computer.

I have tried all sorts and I now have the "Wait to kill" (see previous entries in this thread) set at 45000, but still occasionally, just using the "Wait to Kill" script results in these errors. I am now going for 60000 and I will not exit Boinc first but will try just a straight shut down for the next week and let you all know if I lose any.

I was too quick in shutting down two nights ago and lost two Einsteins !! - Not nice, especially since one of them was more than half way through.

I too have had this problem of Vista shutting down too quickly for BOINC.

The only sure way of not getting these client errors seems to be to shut down BOINC Manager first, WAIT 10 or so seconds and then shut down the computer.

Save the following batch file (.bat) and make a shortcut to it on the desktop. Use it instead of the real shutdown function on the start menu.

cd %PROGRAMFILES%\BOINC
boinccmd --quit
shutdown -s -t 20

What it does is: goes to BOINC directory, uses boinccmd to stop the core client, then shuts down after 20 seconds. shutdown with -t argument even makes it show a countdown.

Save the following batch file (.bat) and make a shortcut to it on the desktop. Use it instead of the real shutdown function on the start menu.

cd %PROGRAMFILES%BOINC
boinccmd --quit
shutdown -s -t 20

What it does is: goes to BOINC directory, uses boinccmd to stop the core client, then shuts down after 20 seconds. shutdown with -t argument even makes it show a countdown.

Thanks for that - I'm not familiar with using batch files and appreciate the help.

-t does not seem to give me a countdown on my Vista just a message saying "Windows will shut down in less than one minute".

Otherwise it seems to work well, I have used it for two shutdowns now with no client errors.

QUESTION?

In the registry there is an entry called "PreshutdownOrder" would changing the entries to include "boinccmd" cause Boinc to shut down first and fast enough?

Or would it cause more problems? I'm tempted to try it myself-but will have to clear my cache first.