BOINC Client cannot detect that computer is being used.
5 Jun 2010 2:20:46 UTC
Topic 194959
(moderation:
Unfortunately I was convinced to upgrade my BOINC client from the stable version 5.4.11 to the unstable version 5.10.28 .
This is the highest version which can be made to run on my Linux-based computer.
The newer version can't be controlled. It simply ignores the fact that I'm typing away on my keyboard, even though in all the preferences I've forbidden it to run until the computer has been idle for 10 minutes.
What can I do?
BOINC Client cannot detect that computer is being used.
)
I am running BOINC 6.10.56 on my SuSE Linux 11.1 and it works OK. I am also running 6.2.19 on a Solaris virtual machine running as a guest OS on my Linux box via VirtualBox.
Tullio
I suppose that I should
)
I suppose that I should mention that I found it necessary to custom-compile a version of BOINC, for versions mentioned in thread 8010.
Basically, with Debian Etch, the highest version in the repository is 5.4.11 , and when I tried just to run the binary which is currently popular, I got the message that it needs GLIBC 2.4 , while I know that I only have GLIBC 2.2 .
And version 5.4.11 was giving me problems.
Yet it would still be possible for me to continue using 5.10.28 . It's just that I need to suspend it by hand, by selecting Suspend in the command menu. And yes, the core client is responding to some of the other settings which I give it.
BUT, does anybody know whether it's necessary to give these source trees a special compiler flag, so that it will interface with the X server? I know that it's using wxWidgets successfully, and that the Makefile should be recognizing the X anyway. But what should be, is sometimes just different from what is.
I wouldn't want to recompile the code, unless somebody can tell me this with great certainty.
(Bump) I've just answered
)
(Bump)
I've just answered my own question. I've given the command
./configure --disable-server --enable-unicode --with-x
instead of my earlier
./configure --disbale-server --enable-unicode
and of course I did this after doing a "make clean", in the directory where I compile BOINC. Then I stopped the core client, gave "make install", and restarted the core client.
Alas, there was no effect on the behavior of the client program.* When I click on "Run based on preferences" from the "Activity" command-menu in the BOINC manager, full activity resumes, even though my computer has also been in full use.
Yet, when I tell the BOINC manager to throttle the CPU usage, for example, it simply does ! But I'd be throttling the CPU usage, even for times when my computer is idle. And I tend to use process priorities over throttling CPU usage.
So for now I'll just need to suspend the BOINC client when I want to use my computer in the forgeground, and then to let it run immediately again, when I'm done using it...
*) On which subject, my make output also suggests that no changed files were reinstalled. It would normally tell me that some reinstalled file is not in its exact previous state.
Edit: Now you see, the next possibility I might need to explore, depends on the fact that my BOINC client is running as user 'boinc'. As such, it might not be able to peer in to the X-server timers, to observe that user 'user' has input keyboard strokes and mouse-movements to my own X-server session. After all, the X-server maintains its own security protocols and segregates the sessions.
It's possible that while v5.4.11 was only able to detect whether keyboard usage had taken place, the developers of v5.10.28 might have been more ambitious, and might be trying to detect whether mouse input has taken place as well. And the developers for v5.10.28 might no longer have taken into consideration that the BOINC client is running as different user.
I think that I've read a lot
)
I think that I've read a lot about problems of BOINC detecting keyboard/mouse activity under linux, but that's the problem with using such old versions: you'll have to dig out 2-3-year-old threads to find the entries. Additionally, you'd probably find better response at the BOINC dev boards.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
Well it turns out, that while
)
Well it turns out, that while you were responding to my posting, I was doing some research myself. Of greatest interest, I found this Web article:
http://boinc.berkeley.edu/trac/ticket/463
From this ticket, I followed to patch #15049 by "davea".
I found that even though I'm using an equivalent older version of BOINC, I was able to apply this patch to the source code by hand, and that eventually I was able to get it to work.
There was just one problem with the patch as shown: It still seems to assume, that the interrupts indicated by the file /proc/interrupts , as 1 column corresponding to 1 CPU. I needed to change the code, to keep track of the fact that such interrupts are handled each by 1 out of 2 CPUs on my system.
But my problem is solved!
Dirk