Charles M. Hannum writes:
 > 
 >    After successfully rebuilding the kernel, I decided to leave the
 >    machine compiling the userland overnight. This morning, the build
 >    window was stuck building getpwent.so (ie.. not too far into the
 >    build process).
 > 
 > Most likely, the combination of make(1) malloc()ing ad infinitum and
 > GCC eventually consumed all available swap space, and the pager
 > wedged.

It really shouldn't do that :)

It looks like that may well be what happened - it seems MFS will, over
time, grow to use up it's full size in swap space, even with activity
on just small files such that it really should never need to touch
more than a very few of it's pages. I'd consider this a bug.

Adam asked me to build a kernel with DDB, I'm doing that now, with a
smaller MFS.

--
Dan.