Testing a new kernel

[Joshua Lokken]

* James Long (list at museum.rain.com) wrote:
==> > After the buildworld/buildkernel/installkernel procedure, it is recommended
==> > to reboot the new kernel to ensure that it is proper.  What should one do
==> > to properly 'test' a kernel?  Is it enough to reboot it and login?  Thanks.
==> 
==> That's a good question, and I don't know for sure.  Mr. Chen says that
==> he thinks that's sufficient.  Personally, I like to manually configure
==> the network card with ifconfig (or dhclient if the system is a DHCP
==> client), then do some traceroutes to "distant" systems, ones that are
==> 10-20 hops away, which exercises the network stack, and also, the 
==> resolver (which maps names to and from IP addrs), since it has to do
==> a reverse lookup on each hop.

Has anything ever not worked for you after the kernel install?

==> 
==> One of the _first_ things I do after booting the new kernel into 
==> single-user mode:
==> 
==> adjkerntz -i     (load the timezone adjuster)
==> fsck -p		 (check all the filesystems)

I always do these things when going into single-user, but when I've 'tested'
my new kernel, I've been booting into multi-user, logging in, and then re-
booting back into single-user. 

==> 
==> I basically just try to "flex" the major muscles of the system:
==> file system, network stack, etc.

This makes sense.

=> 
==> I'm no expert, but just keep in mind that at this stage, it seems
==> possible that if any device nodes (/dev/ entries) are out of date
==> with the new kernel, you might have hardware issues with those
==> devices until you installworld and MAKEDEV all to regenerate the
==> /dev/ entries.  I hear that FreeBSD 5.0 will have a revised
==> device system, so things may change when you move to 5.0.
==> 
==> Jim
==> 

Yeah, five's a bit beyond for me right now, but 4 stable sure rocks!  Thanks.

--
Joshua