I think the linux crash is due to some memory leak. I try to run it on Amazon Linux and top shows that RES memory is constant but VIRT continuously increases. I followed Fasan's directions too, but nothing changed..
I can confirm the virtual allocation keeps climbing on my machines as well (compiled from newest git source); perhaps my 16+ core boxes are increasing much faster and therefore running out of RAM sooner than the others, so my smaller core machines are just biding their time.
The error I'm getting:
terminate called after throwing an instance of 'boost::exception_detail::clone_impl >'
what(): boost::thread_resource_error: Resource temporarily unavailable
I've tried messing with ulimits to no avail, so a leak makes sense... threads are not closing properly perhaps?
just sat and watched one older 4-core box, and it's like clockwork: virtual memory increases by 32MB every 4 minutes 15 seconds (not sure the cpu speed on this one offhand, but I assume this coincides with completion of whatever the equivalent is of a getwork session is?), and that's PER CORE. No wonder my 32-core boxes are giving up after just one hour.