rawhide report: 20050205 changes

[Arjan van de Ven]

On Mon, Feb 07, 2005 at 07:27:35AM -0200, Alexandre Oliva wrote:
> On Feb  7, 2005, Arjan van de Ven <arjanv at redhat.com> wrote:
> 
> > On Mon, Feb 07, 2005 at 12:16:58AM +0000, Tim Waugh wrote:
> >> On Sun, Feb 06, 2005 at 10:21:08AM +0100, Arjan van de Ven wrote:
> >> 
> >> > this is already wrong. The "athlon" hack is *WRONG*.  Athlon is a
> >> > marketing name, it could just as well be a duron or a sempron.
> >> > It was a mistake to put the athlon hack in uname (but hindsight is
> >> > easy); it's imo very wrong to repeat that mistake.
> >> 
> >> Should I take it out?
> 
> > hard dilemma in that it risks breaking existing stuff....
> 
> The most common case in which I've seen scripts depend on athlon or
> ia32e is to build kernels and kernel modules.  Since the removal of
> athlon- and ia32e-specific kernels, having uname print athlon or ia32e
> no longer helps; it actually gets in the way.  So my vote would be to
> get rid of it.

those scripts are really broken anyway if they do that. uname is NOT the way
to get the rpm architecture of the kernel, you HAVE to use 
rpm -q --queryformat %{ARCH} kernel-`uname -r`

for that. Simlply because you can install an i686 kernel on an athlon....