Re: Problem with dpkg-architecture
On Wed, May 05, 1999 at 07:34:02AM -0600, Gordon Matzigkeit wrote:
> >>>>> Guy Maor writes:
>
> GM> Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de> writes:
> >> 1) Make "-e" the default.
>
> GM> No way! -e is a horrible argument that makes makefiles break for
> GM> mysterious reasons.
(Note that dpkg-cross does exactly this, and it seemed to work for years
now, but I agree, it does conflict with concept of least surprise. I don't
want to be responsible for those bugs that crop up).
> GM> You should just arrange for the overrides to be set on the
> GM> command line.
>
> I agree with Guy here.
Ah, that's another solution. Thanks for pointing out.
So we will do in dpkg-buildpackage (from top of my head):
debian/rules `dpkg-architecture -a$ARCH`
and people who don't use dpkg-buildpackage (oh my dear!) can do one of the
following:
debian/rules `dpkg-buildpackage -ahurd-i386` build
or
eval `dpkg-architecture -ahurd-i386 -s`
export MAKEFLAGS=-e # at their responsibility
(I can probably support a flag to dpkg-architecture that sets MAKEFLAGS
variable optionally)
or
dpkg-architecture -a hurd-i386 -m -c "debian/rules build"
where "-m" is a new flag that makes the variables not to be set in an
environment, but passed as make variables.
Note that this solution only requires modification of dpkg-buildpackage and
the way how people can do it manually. I like it. It seems to get easier to
use it directly, and nothing else changes.
Ok, I will make the change to dpkg-buildpackage.
BTW: It is important that people don't use "dpkg --print-arhcitecture" in
there scripts anymore, but "--print-installation-architecture". The first
one has no sane definition for other systems beside Linux. That's the bug in
e2fsprogs that I mentioned. The consequence is that compilation will default
to native building when dpkg-architecture is circumvented (either
directly or when dpkg-buildpackage is not used)
Thanks,
Marcus
--
"The purpose of Free Software is Free Software.
The End and the Means are the same." -- Craig Sanders
Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de>
Reply to: