Re: smarter way to differ architectures needed?
On Sat, Mar 03, 2001 at 02:27:48PM +1100, Brian May wrote:
> Now I am starting to remember. Gordon's proposal was more then just a
> quick fix for "architecture: all" problems, it was a more general
> solution to the overall "architecture: any" hack, too (among other
> things).
Sorry, I wrote a reply to an earlier mail, but it got stuck.
You wrote:
> Do we have a good solid proposal how this could be done?
I just tried to start discussion at debian-dpkg, because Ben whispered me on
IRC that we likely won't get versioned provides at all, and having an
unversioned provide fulfill a versioned dependency is wrong (to which I
completely agree).
So, what we basically need is some finer way to specify the architecture,
and arch specific dependencies, like "makedev (>= 2.x) [linux]" (I'd say
like build depends, but note that build depdends can't say linux/hurd,
only hurd-i386 etc. Another bug! and a real headache when we get more than
one Hurd port)
We can do without arch specific dependencies, but for this to work we must
file bug reports against packages like gnupg to get rid of the dependency on
makedev in the Hurd version of the package, forcing it to create the control
file in debian/rules. A lot of packages will have to do this then, and I would
like to start to file bug reports if this is decided.
> So, for instance you would say:
>
> Arch-Depends: ${cpu}, ${system}, proc
>
> and the fields would automatically be filled in at compile time. Of
> course, I am still reading the details.
That's an entirely different proposal to fix the whole Architecture stuff as
a whole. It was originally suggested and drafted by me. It would be the
optimal solution, but require a complete reorganization of what a
distribution means. I don't see enough support for this vision in Debian.
(see http://master.debian.org/~brinkmd/arch-handling.txt for my original
proposal, which is still valid as far as I am concerned).
For a more concrete proposal for the all-* any-* hack, see
http://lists.debian.org/debian-dpkg-0102/msg00053.html
I actually have a longer reply for you, but it is not finished yet, sorry.
Just don't rush it. Mine and Gordons proposal would require some careful
thought to get working because of interesting design issues. But the all-*
any-* hack (which is all we can hope for currently) requires some careful
thought because we must have a smooth transition and remain backward
compatible, and many tools have to be changed in different locations.
That's because the Architecture notion in Debian was never designed, it
evolved in arbitrary directions out of the need to get something working
quickly. It's quite simplistic.
Thanks,
Marcus
--
`Rhubarb is no Egyptian god.' Debian http://www.debian.org brinkmd@debian.org
Marcus Brinkmann GNU http://www.gnu.org marcus@gnu.org
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de
Reply to: