Subject: Re: Processor types and uname data on AIX (proposed change)
To: Peter Schmiedeskamp <>
From: grant beattie <>
List: tech-pkg
Date: 05/07/2004 20:15:41
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Thu, May 06, 2004 at 11:00:54AM -0700, Peter Schmiedeskamp wrote:

> The long string is what gets associated with the
> package.  This is what happens when I take a package
> built on foo and attempt to install it on bar:
> root@bar:/home/root $ pkg_add libiconv-1.9.1nb2.tgz=20
> pkg_add: Package `/home/root/./libiconv-1.9.1nb2.tgz'
> OS mismatch:
> pkg_add: AIX/000c5c7a4c00 1 (pkg) vs. AIX/000c8d4f4c00
> 1 (this host)
> pkg_add: aborting.
> pkg_add: 1 package addition failed

is your pkg_add locally hacked? I would have expected to see the
AIX version number in this error, too.

> I suggest that the output of 'uname -prvs' is more
> useful when the target is AIX > v.5.  The output is as
> follows:
> root@bar:/usr/pkgsrc/packages/All $ uname -prvs
> AIX 1 5 powerpc

interesting output order given the argument order... I presume it's
-p that produces "powerpc" and not -s :-)

> Unfortunately, there is a catch.  Under AIX 4.3, the
> -p flag is not recognized:
> root@snafu:/home/root $ uname -prvs
> uname: Not a recognized flag: p
> Usage: uname [-snlrvmaxuMS:T:]
> Ugh!  Under AIX 4.3, there appears to be no good way
> of determining the processor type.  Maybe during
> bootstrap, we could try to set the uname flags to
> -prvs, and failing that, fall back to building binary
> packages on a per-machine basis?

check what bootstrap/bmake/ reports on AIX. that should be a
good starting point.

whatever you can find can be easily added into the bootstrap process.
thanks for looking into this.


Content-Type: application/pgp-signature
Content-Disposition: inline

Version: GnuPG v1.2.4 (NetBSD)