"Brian J. Murrell" <[EMAIL PROTECTED]> writes:
> >
> > The problem is also that it's not explicited clearly somewhere: you
> > must install a new kernel with rpm -i and not rpm -U :(
>
> Can one actually do that? Without a --force I mean. I have never been
> able to rpm -i a package of the same name, different version/release on
> top of an older one.
>
Yes you can. All files are different from one kernel package to
another. This is why it works in this special case.
>
> I came across a couple of other things while upgrading. What is /boot
> supposed to look like? I like mine to look like this when all is done:
>
[description of /boot]
>
> assuming that I upgraded from 2.2.13-29mdkfb to 2.2.14-1mdkfb. That way
> I don't have to screw around with lilo.conf before I run lilo.
>
Yes but you also lose the possibility to bot on your old kernel.
> Also I really do think that the kernel install should run the mkinitrd.
> That way the modules will still be there when it is run. Any reason why
> to not do it? Also any reason to not make /boot follow the stucture
> above?
>
Two things:
* on boot, rc.sysinit makes the System.map symlink point to the one
corresponsing to the running kernel - a .old symlink wouldn't be of
any help. It looks for a System.map named System.map-$(uname -r);
* it won't solve the problems of multiple booting kernels anyway. One
(for now) must still do it by hand.
--
fg
# rm *;o
o: command not found
