On Sat, 11 May 2002, Galen Sampson wrote:
> > > Don't mergemaster until after you've installworld'd.
> > Yeah, I bumped into this on the TrustedBSD branch lately also. It goes
> > side by side with the "don't installworld until you mergemaster" rule.
> > And the very basic rule which is "Don't run -CURRENT unless you not only
> > want to shoot your feet, but have hot spares so you can shoot them with a
> > lower recovery time".
> All is well now. It turns out I ran the following sequence: buildworld
> (failed), buildworld -DNO_WERROR, shutdown now, buildkernel -DNO_WERROR,
> installkernel, installworld, shutdown -r now (su error after reboot),
> mergemaster (still error), and after a restart all was well.
> Don't mind shooting myself in the feet...just interested in the
> development process of large projects. I'm a senior CS major graduating
> in June. This is just a hobby.
Yeah, part of the gist of my message was that occasionally, there's no
"right" upgrade path, you just have to slog through until your system
appears to run normally again :-). Occasionally there are cycles of
dependencies for upgrades -- you need a new users to install the files,
but you don't get the new user until you run mergemaster. Or, you need a
new binary to install a file via mergemaster, but you can't install the
binary until mergemaster has created the config file for it.
Last night I shot myself in the feet on my notebook by upgrading it to a
recent -CURRENT with the trustedbsd_mac branch. The binary format for MAC
labels changed, so init wouldn't exec, so the system would panic. If I
booted the old kernel, because world was upgraded, all binaries setting
credentials seg-faulted. Good stuff, experimental code. :-).
Fortunately, it usually doesn't work out this way.
Robert N M Watson FreeBSD Core Team, TrustedBSD Project
[EMAIL PROTECTED] NAI Labs, Safeport Network Services
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message