Re: port-arm/48193: On NetBSD/evbearmv6hf-el for Raspberry Pi, MACHINE_ARCH returns earm, but *hf* is expected in /usr/share/mk/bsd.own.mk .

[Matt Thomas <matt%3am-software.com@localhost>]

The following reply was made to PR port-arm/48193; it has been noted by GNATS.

From: Matt Thomas <matt%3am-software.com@localhost>
To: Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>
Cc: gnats-bugs%NetBSD.org@localhost
Subject: Re: port-arm/48193: On NetBSD/evbearmv6hf-el for Raspberry Pi, 
MACHINE_ARCH returns earm, but *hf* is expected in /usr/share/mk/bsd.own.mk .
Date: Mon, 16 Sep 2013 02:34:33 -0700

 On Sep 16, 2013, at 2:25 AM, Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost> 
=
 wrote:
 
 >> Uh, no.  It solves the PR.
 >=20
 > No. It just worked around the problem, and then causes another =
 problems.
 >=20
 > You should first define what MACHINE_ARCH and sysctl.hw.machine_arch =
 mean,
 > before touching sources. Kernel architectures? Userland binary ABI?
 
 latter otherwise we could collapse a lot of machine_arches.
 MACHINE_ARCH is also used by the pkgtools to delineate ABI as well.
 
 >> It even solves the problem of untarring a different
 >> implementation in a chroot and then getting the right MACHINE_ARCH =
 for that tree.
 >=20
 > If hf and sf can be switched dynamically, it will cause confusion
 > to have static MACHINE_ARCH in <arm/param.h> and dynamic one
 > returned by sysctl.
 
 MACHINE and MACHINE_ARCH in param.h should die.  If you want the value,
 use sysctl.  I've been arguing that on ICB currently.
 
 >> It can be backed out if and when someone comes up with a better =
 solution.
 >=20
 > Probably you should read the commit guidelines again. You didn't
 > have a prior discussion, and you didn't even test your changes.
 > If core members don't have to have reviews even for exported API =
 changes,
 > I have no further comments though.
 
 Well, I did test my changes.  I determined that sysctl was doing what
 I expected.  It never occurred to me that make didn't use sysctl.