Re: -= PROPOSAL =- Release sarge with amd64
Raul Miller <moth@debian.org> writes:
> On Sat, Jul 17, 2004 at 10:23:13AM -0400, I wrote:
>> Maybe you could explain how you want the debs to look like -- in
>> particular, do you expect that sid is going to have to replace /usr/bin
>> with /usr/i486-linux/bin and /usr/amd64-linux/bin/?
>
> Ok, after reading a few more multiarch docs, I see that that part of
> the proposal has another option, you'd instead have all programs that
> support multiarch have names of the form /usr/bin/amd64-linux-$basename
>
> That makes a bit more sense, but it still is going to require buyin from
> the maintainers of those packages.
>
> Also, which are those packages? At minimum, binutils, gcc and glibc need
I think that covers it already for the binaries.
gcc-defaults, build-essential, lsb, dpkg dpkg-dev, apt, dselect,
aptitude, deborphan, debfoster, python-apt are other candidates that
need to be patched.
> to support multiarch, and be built this way -- and that's mostly because
> gcc's cross compiling support is in such poor shape when contrasted
> with native compiling. [And it's probably worth noting that multiarch
> doesn't seem to address cross compiling at all.]
>
> Anyways, the big issue which I don't see being addressed (and the
> reason I'm still Cc-ing debian-devel) is: most packages in debian
> (everything which offers a shared lib) is going to have to be rebuilt
> to support multiarch. Just about every single developer is going to
> have to support the idea, and rebuild their packages to this new standard.
Eventually yes. We are talking about ~2000 source packages. If all
libraries have to be coinstallable or just move the libs remains to be
decided (also see below).
> Or have I missed something major?
No, you are right on point.
We also not only want to change every package in Debian but change
every package on every linux distribution. We want to change the LSB
to multiarch.
Changing Debian to multiarch will probably go in stages:
1. the build-essential tools are changed to support building multiarch
packages
2. build-essential packages are changed
3. base packages are changed
4. the rest is changed
If step 4 is taken, how far it is taken remains to be seen. But for
simplicity sake it is probably better to get all packages to use the
multiarch directories even if they don't support coinstallation for
multiple archs.
Getting multiarch into debian will be a huge undertaking and is likely
to take years to complete. That adds to the unwillingness to wait for
it before getting amd64 added.
MfG
Goswin
Reply to: