Chmouel Boudjnah wrote:
>
> Jeff Garzik <[EMAIL PROTECTED]> writes:
>
> > One thing I was wondering -- RedHat has always had the problem where a
> > kernel upgrade via RPM was not complete, ie. I could not do
> > rpm -U kernel*.rpm
> > reboot
> > and get a new kernel.
> > Basically it did not re-run LILO, if I recall correctly, but there may
> > have been other problems too.
> > Does Mandrake have the same problem?
>
> This is what the rpm do :
>
> if [ -x /sbin/lilo -a -f /etc/lilo.conf ]; then
> /sbin/lilo > /dev/null 2>&1
> exit 0
> fi
>
> but it's assume you have a lilo entry pointing to /boot/vmlinuz.
>
> I look to make a script who add a entry.
That's what I forgot: lilo.conf specifies the kernel version. IMHO rpm
needs to update lilo.conf by doing something like
if [ -x /sbin/lilo -a -f /etc/lilo.conf ]; then
sed -e 's#/boot.*kernel.*mdk#/boot/kernel-2.2.13-20mdk#' \
< lilo.conf > lilo.conf.new
if cmp lilo.conf*; then
/sbin/lilo
fi
fi
That way lilo isn't run unless it sees an mdk kernel. (note - you don't
want to use above script verbatim, because rpm needs to be smart enough
to _only_ do s/old-version/new-version/)
