> In article <19960515123912.3853.qmail@Mail.UTexas.EDU>,
> lilo <TaRDiS@mail.utexas.edu> wrote:
> >On Tue, 14 May 1996, Hartmut Niemann wrote:
> >
> >> No, a _file_ /etc/kernel.conf is IMO a bad idea. Those who really need to know how a kernel
> >> was built usually have more than one kernel to boot, so this information must
> >> be stored _in_ the kernel image.
> >
> >Actually, that's a very sweeping statement. For example, I rarely compile
> >kernels for more than one machine on my system, and I suspect there are a
> >lot of us in the same boat (there being more Linux users than developers ;).
> >When I get a new kernel, I want it compiled with exactly the same
> >configuration as the old one, except of course for new options.
>
> Whereas I do compile many different kernels as I maintain a linux network
> of 8 machines + server. What I would like is for make config to recognize
> an old .config file and tell me about any new options, and hide all the
> questions about previously existing options. Then I could simply have a
> .config file for each significantly different computer, and answer a *much*
> reduced 'make quickconfig' to recompile the kernels. That way I know I can't
> make a slip whilst answering make config and miss out some important driver
> from the kernel for one computer and/or put an extra driver into the kernel
> for another.
Definitely a different problem. Right now the simple way to do that is to
copy the old source's .config file to the new source's arch/<blah>/defconfig
and do a `make mrproper'. There might be some neater way to set this up as
well, though. Not sure.
lilo