Kernel Compiling Checklist, rev 3

[Robert P. J. Day]

On Wed, 14 Apr 2004, Matt Hansen wrote:

... snip ...
> Hi all,
> 
> Where would a "make rpm" target fit into these steps? I would think
> these kernel compilation steps should include this should it not?
> Wouldn't it be cleaner and more manageable to create an RPM out of this
> process?

i've built and installed a *lot* of kernels, and i've never found any
overwhelming need for creating an RPM.  regardless of whether it's a 
2.4 or 2.6 kernel, the recipe is pretty straightforward:

1) edit Makefile, change EXTRAVERSION variable to not clash with any
   existing builds
2) config and compile new kernel, copy to /boot under new name vmlinuz-foo
3) make modules and install them, verify new directory under /lib/modules
4) mkinitrd to build new initrd.img under /boot
5) edit grub.conf (or lilo.conf) to add new stanza for new kernel and
   initrd
6) reboot to test

  obviously, there are endless subtle variations on the above, but what i
think has been lost in this recent exchange is that the same recipe is not
going to be appropriate for everyone.

  what's missing, and i think a lot of linux documentation suffers from 
this, is that there is a dearth of documentation explaining *why* you
do certain things.  a lot of documentation has incredibly detailed recipes
for how to do something, with little rationale for those steps, why 
certain things are being done, and why you might want/need to do them
differently.

  to paraphrase, give a man a recipe for building a new kernel, you get
rid of him for a day.  *teach* him the intricacies of building that kernel
and why certain things are done the way they're done, and you get rid of
him for life.  :-)

rday