Port-vax archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: NetBSD/vax kernel seems stable now

Tom Ivar Helbekkmo <tih%Hamartun.Priv.NO@localhost> writes:

> I'm once again trying to complete the build without swap, now...

It's working great without swap -- within the limits of available real
memory, of course.  Thus, as in June 2007, my only observed problem is
that swapping makes the machine unstable.  Back then, it would crash
immediately upon swapping something out, whereas now, there seems to be
an unknown likelihood of a crash when it tries to swap something in.

I find that the build progresses nicely when I let it run until it runs
out of memory (like when building libc, for instance), and then enable
swap just long enough to manually perform the operation that needed more
memory than is physically available.

Just learned something new: I thought that the cleandir done as part of
a 'build.sh' run without '-u' would remove everything, letting me ignore
any "you have to remove this-and-that from /usr/obj" stuff in UPDATING,
but I just got bitten by this one:

        The posix_fadvise system call has been changed from an assembly
        stub, to a c file that calls an assembly stub. You need to
        'rm -f posix_fadvise.* .depend' in the libc build directory to
        avoid using the old assembly stub.

Turns out that even after the cleandir, posix_fadvise.S was left in the
objdir, tricking make into doing the wrong thing during the build...

That set the build back a few hours... *sigh*

Self documenting code isn't. User application constraints don't. --Ed Prochak

Home | Main Index | Thread Index | Old Index