optimization? (was: RE: RedHat kernel RPMs?)

[Gregory Leblanc <GLeblanc cu-portland edu>]

> -----Original Message-----
> From: Jeff Johnson [mailto:jbj@JBJ.ORG]
> 
> On Mon, Sep 25, 2000 at 01:28:19PM -0700, Gregory Leblanc wrote:
> > How does RedHat manage to produce kernel RPMs for 
> i586/i686?  I just ran rpm
> > --rebuild kernel-2.2.16-3.src.rpm, but it only came back 
> with i386 binaries,
> > and I'm wondering how badly mangled things will be if I just use my
> > home-grown optimizations on the kernel.  Anybody tried this 
> and care to
> > offer advice?  Thanks,
> 
> However, what's biting you is that nested %if handling (which 
> is present in
> the kernel-2.2.16-3 spec file) has broken in rpm since the 
> kernel-2.2.16-3
> package was released.

D'oh!  That's what I get for upgrading before I finish re-compiling all of
the updates.  Thanks for the pointers, I just temporarily downgraded to
3.0.4.

I was just reading the "new features" page for RedHat 7 (since everybody and
their mother is trying to download it right now, I can't get a copy), and I
noticed that more than just the kernel RPMs are supposed to be optimized for
different architectures.  Thus 3 questions:
1) What flags are being used on RedHat's build machine (porky?) to optimize
their regular and kernel packages?
2) What's a good resource for more information on the black art of choosing
those flags?  For example, why does adding -s break important packages like
glibc?
3) Is there a way to make 2 targets use different flags, or to have 2
different targets use the same filename/architecture?  Say I wanted 1 target
to have -s, and a second target -not- to have -s, is there a way to do that?
Thanks so much,
	Greg