I don't know yet whether they'll be any help, but please keep them in a safe place in case we need to call for them. Just to be certain, can you be sure (from file timestamps or however) that this stack trace comes from the time when you were running v7.8.2 under, I think you said, Ubuntu 17.04?

I have been running Ubuntu 17.04 for several weeks, and BOINC 7.8.2 since at least 12 September, which I know from the CPDN results page; probably longer, though I can't tell from the file dates on stderrdae.txt and stdoutdae.txt since they were lost on copying. As for the time stamps, I don't really know, except that the first thing that looks like one is

======= Memory map: ========
564646ee0000-564646fc6000 r-xp 00000000 08:05 3145802  /usr/bin/boinc

and the last one is

7fd550ce0000-7fd550ce1000 rw-p 00026000 08:05 2752569                    /lib/x86_64-linux-gnu/ld-2.24.so

If that is referring to 08:05 UTC (04:05 EDT), then that is the right time, or for the reboot after I detected it if BOINC was still operational at that point. That would not be more than a couple of hours after it occurred. Beyond that, I will certainly save all the logs and you can PM me here or on BOINC and I will be glad to send them for your expert inspection.

In later Linux kernels vsyscall is disabled. I'm running Debian and can't go past the 4.9 kernel (without fiddling) due to it. Ubuntu 17.04 ships with the 4.10 kernel as default. My machines are Ryzen 1700 and running BOINC 7.8.2 from the Stretch-backports repo. See this thread at Einstein.

---

BOINC blog

I have (or had) BOINC 7.8.2 installed on three Ubuntu 17.04 machines and one Win7 64-bit machine with no problems. That is, until BOINC crashed (manager could not connect to client) on one of the Ubuntu machines. Even after a reboot it did not work, which I don't recall ever seeing before. So I uninstalled BOINC, and went back to 7.6.33, but it was still borked. The only other thing I can think of is that VirtualBox 5.1.28 was installed, but not attached to any projects, and removing it did not fix anything.

Just reporting back on this one for completeness.

All reported cases of "won't connect, won't run, won't even run with old version" have now been traced to a newly released batch (batch 658) of CPDN climate models - sprecifically, WAH2 for the PNW region. These tasks all fail after one simulation month under Linux and OS X (CPDN are trying to track down the reason for that - their problem). When the tasks crash, they leave behind a huge crash dump in stderr_txt, and 51 failed upload messages.

BOINC - all current versions - can't cope with that much error information, and fails with the symptoms described here. There are two known recovery routes:

a) Delete the file 'account_climateprediction.net.xml' from BOINC's data directory. This detaches you temporarily from the CPDN project, until the problems are resolved and you can re-attach.

b) Very carefully, edit client_state.xml to remove the <workunit> and <result> sections for any WAH2 PNW tasks you may have. Set 'no new tasks' for CPDN as soon as you get back control of BOINC.

BOINC v7.8.2 is NOT, it turns out, implicated in this problem. A fix has been written, and will be included in the next BOINC release - whenever that is.

That is a very nice summary, and I (and a lot of other people) are fortunate that Richard visited this forum at the right time. I would add only that the problem does not appear to affect the Windows version of BOINC on CPDN, though it is not clear why not.

I would add only that the problem does not appear to affect the Windows version of BOINC on CPDN, though it is not clear why not.

Because the Windows version of the CPDN application doesn't crash after the first month, and doesn't produce the huge crash dump.

All reported cases of "won't connect, won't run, won't even run with old version" have now been traced to a newly released batch (batch 658) of CPDN climate models - sprecifically, WAH2 for the PNW region. These tasks all fail after one simulation month under Linux and OS X (CPDN are trying to track down the reason for that - their problem). When the tasks crash, they leave behind a huge crash dump in stderr_txt, and 51 failed upload messages.

BOINC - all current versions - can't cope with that much error information, and fails with the symptoms described here. There are two known recovery routes:[...]

I am pretty sure, that this was the problem in my case, as I am running 14 WUs of climateprediction.net alongside of gpugrid.net.

It is not the first time, that climateprediction.net shut down one of my computers, because the model crashes.

But as I wanted to install Lubuntu 17.04 and overclock my RAM anyway, I was quick to install everything anew.

And now it works without any problems for three days. I will handpick the WUs of climateprediction.net at this moment.