Alpha Core 3

[Sergey Tikhonov]

On Sat, 2007-03-17 at 11:32 +0100, Oliver Falk wrote:
> Sergey Tikhonov schrieb:
> > On Fri, 2007-03-16 at 17:25 -0500, Jay Estabrook wrote:
> >   
> >> On Fri, Mar 16, 2007 at 09:27:31PM +0100, Oliver Falk wrote:
> >>     
> >>> Check out /etc/rpm/platform; We need to fix this for the different 
> >>> sub-archs I believe... And then check /usr/lib/rpm/redhat/*.
> >>>
> >>> Arg. Even uname -m reports the wrong thing on my DS10. :-( Havn't 
> >>> checked the AS1000 yet.
> >>>       
> >> If we want it to behave like i*86, which does somerthing like:
> >>
> >> 	uname -mpi
> >> 	i686 i686 i386	
> >>
> >> should Alpha report on (667MHz) DS10:
> >>
> >> 	uname -mpi
> >> 	alphaev67 alphaev67 alpha
> >>
> >>
> >> Here's from the manpage:
> >>
> >>        -m, --machine
> >>               print the machine hardware name
> >>
> >>        -p, --processor
> >>               print the processor type
> >>
> >>        -i, --hardware-platform
> >>               print the hardware platform
> >>
> >> I assume we'd keep the generic "alpha" name for hardware platform (-i),
> >> but should the sub-architecture show up in either or both of the others?
> >>
> >> BTW, I'm VERY interested in this area, as one can get a BIG benefit
> >> from running packages built for a specific Alpha sub-architecture on
> >> appropriate machines (eg alphaev56 for PWS/LX/SX, alphaev6 for DS10).
> >>
> >> Right now, "rpm" will make you say "--ignorearch" to get ANY sub-arch
> >> package installed; that's not right.
> >>
> >> It'd be nice, not only if "rpm" didn't force you to say that, but also
> >> if "rpm" OR "yum" would consider it a valid potential "update" to just
> >> install the same version package built for a (more appropriate)
> >> sub-arch on top of an installed generic one.
> >>
> >> I've rebuilt a lot of SRPMS targeted for alphaev67, and I'd like to
> >> make it easy to get a more optimal system after installing from a
> >> generic set.
> >>
> >> Anyone have any additional thoughts on this?
> >>     
> > I was able uname to output for uname -mpi:
> > alpha alphaev56 alpha
> >
> > uname -m show information returned by uname glibc call (which in turns
> > calls kernel.
> Kernel-Bug :-P
May be. Actually for Athlon kernel returns i686 as "machine" and than
uname does additional magic for -p option. I am not sure what should be
returned for "-m".

> 
> > I remember tracing that a long time ago).
> > The -p and -i is handled by "uname" and have some altering code for
> > other arches. I'll install modified version of uname and see how other
> > programs would like the new output. :)
> >   
> Let us know!
So far, no problems for my usual build procedures.

Regards,
Sergey.