my kernel headers are hosed!

[Kurt Wall]

In a 1.0K blaze of typing glory, Net Llama! wrote:
> On 12/21/03 13:44, Collins wrote:
> >>Seems that Linus issued this edict back in the late 2.2.x days.
> >
> >
> >Yeah, I read that edict.  Correct me if I am wrong, but I thought that he 
> >was merely stating the obvious (i.e. existing behavior) rather than 
> >introducing something new.  The same memo also discussed his preference 
> >for doing kernel compiles as normal user, switching to root only for the 
> >modules_install.
> 
> I honestly don't know if you're right or wrong.  I just found out about 
> this last week when i was trying to build 2.6.0 on an ancient RH-6.2 
> box, and it didn't go too well.

Two separate issues. The first is not to comingle user space headers 
with kernel headers, which is a perfectly sensible dictum. User space
headers should *never* depend on the kernel. Glibc is a special case
because it needs kernel headers to build, so the solution there is
to give glibc static copies, against which it will run fix-includes.sh
during the build to munge the static headers in a special way.

The second issue is not to build the kernel as root. Again, this is
perfectly sensible because the only part of the kernel "build" process
that needs root access is the installation. I've been building kernels
as a mortal user for quite some time (years) now and have had no problems.


Kurt
-- 
Hatred, n.:
	A sentiment appropriate to the occasion of another's
superiority.
		-- Ambrose Bierce, "The Devil's Dictionary"