Kernel SRPM questions

Andre Costa acosta at ar.microlink.com.br
Wed Jan 5 16:27:56 UTC 2005


Hi Paul,

On Wed, 05 Jan 2005 15:40:19 +0000
Paul Howarth <paul at city-fan.org> wrote:

> Andre Costa wrote:
> > I am used to configuring my own kernels, because I drop a lot of
> > stuff I don't really use. Lately, kernel-source RPMs have been
> > deprecated in favor of SRPMs. I do can build new versions from them,
> > but there are a couple of issues:
> > 
> > 1. I generate kernel sources with
> > 
> > 	rpmbuild --target i686 -bp kernel-2.6.spec
> > 
> > and then I move the resulting
> > .../redhat/BUILD/kernel-2.6.9/linux-2.6.9 to /usr/src/linux , and
> > then do 'make bzImage modules modules_install install' (usually
> > after a 'make oldconfig' using my own .config file).
> 
> An alternative approach is described at 
> http://crab-lab.zool.ohiou.edu/kevin/kernel-compilation-tutorial-en/
> 
> This method results in a new kernel RPM that you can install/uninstall
> just like any other kernel RPM.

Nice doc, it definitely looks like it is what I need. Thks.

> > 2. there's no support for 'athlon' architecture
> > 
> > I have an athlon-XP CPU, and if I plainly run 'rpmbuild -bp
> > kernel-2.6.spec' it barfs that athlon is not supported and bails
> > out. Browsing through spec file I couldn't indeed find specific
> > support for it. Any special reason for that?
> 
> I know that RedHat say that an i686 build should be used for athlons, 
> and you can get that using --target i686 as you described earlier. It 
> would seem to make sense to me though for the spec file to
> automatically  select the right architecture to use if none were
> specified.

Yeah, I thought so, too. Anyway, as others have pointed, it seems
optimizations for athlon arch are done runtime (... I still have to
learn how this could be more efficient than compile-time optimizations,
but one thing I've learned is not to doubt Linux kernel developers ;))

Thks again for the help.

Best,

Andre

-- 
Andre Oliveira da Costa




More information about the fedora-list mailing list