I run BOINC for SETI@home on 3 different computers: one on Linux (Mandrake 10.1) and two on Windows XP Home Version.
While the Windows instances never error, the Linux version starts well each time I upgrade to a new version and in a matter of days/weeks, starts getting more an more "client error" instances. I noticed this constantly since I started with version 4.19, up to 4.25 and with all the development versions (right now I use 4.32). They all behave the same.

Any idea what can be the problem?

Thanks,
Jean

---

are you overclocking? are these the precompiled binaries or did you compile them yourself?

I am using the precompiled version.
I'm afraid I have no idea what overclocking means in this context.
Is there a configuration setting I can tune or something of that sort?

Thanks,
Jean

---

I am using the precompiled version.
I'm afraid I have no idea what overclocking means in this context.
Is there a configuration setting I can tune or something of that sort?

Thanks,
Jean

if the CPU is overclocked, you could be getting unpredictable results. what are the exact client error messages you are getting?

I am using the precompiled version.
I'm afraid I have no idea what overclocking means in this context.
Is there a configuration setting I can tune or something of that sort?

Thanks,
Jean

if the CPU is overclocked, you could be getting unpredictable results. what are the exact client error messages you are getting?

Oh, now I see. You mean the CPU being overclocked. As far as I know it is not.
The errors are something in this line:

2005-05-16 18:14:31 [SETI@home] Unrecoverable error for result 20dc04aa.28373.1808.584662.207_3 (process got signal 11)
2005-05-16 18:14:31 [SETI@home] Unrecoverable error for result 20dc04aa.28373.1808.584662.207_3 (process got signal 11)


Jean

---

2005-05-16 18:14:31 [SETI@home] Unrecoverable error for result 20dc04aa.28373.1808.584662.207_3 (process got signal 11)
2005-05-16 18:14:31 [SETI@home] Unrecoverable error for result 20dc04aa.28373.1808.584662.207_3 (process got signal 11)

sig 11 is similar to the infamous windows blue screen, it means the program crashed hard and the OS had to stop it from running.

The Linux OS on every system I've seen it running on is fairly stable in that the only time I have actually seen sig11 errors is with bad code (rare) or bad hardware (more common.) Bad code would be a problem with the BOINC client. Try and see if there is a newer compiled binary to download since if that is your issue it was probably identified and fixed and there will be a newer version to download than what you are running. If it is not the BOINC program causing the crash then it is likely a hardware error. The two most common hardware problems I've encountered that cause sig11's are overclocked CPU (you have a P4 2000 and you run it at 2400, etc) and bad RAM.

---

Hi,
Since I saw this constantly happening with at least 5 versions of BOINC (three proper releases and two development releases) it would mean that bad software is pretty much ruled out. Not only that, but it also seems that there are not that many other reported instances.
So I guess it's the overclocking (I just upgraded the memory a couple of weeks ago and while happening at a lower rate than before, it's still there nontheless), altough, as I said, I didn't recall tuning it up. I'll have to check again.

Thanks,
Jean

---