Re: Architectures (Operating Systems and CPU Architectures)
Jonathan Yu <jonathan.i.yu@gmail.com> writes:
> Does that mean we should be able to just pick something from both
> lists, and turn that into a valid string to put in the Architecture
> field?
>
> solaris-armel, for example.
I think you have to distinguish between syntax and semantics here.
Syntactically, such as from the Policy perspective, any alphanumeric
word or two words separated by a hyphen is a valid architecture.
Semantically, dpkg knows about a specific set of elements and permits
various combinations of them, but in practice a far smaller set are
viable platforms.
> My question is, does anyone know of cases where a given operating
> system and architecture does not constitute a valid platform (ie,
> Architecture in the d/control file sense).
armel and lpia are special cases and don't combine with other kernels
from dpkg's perspective, which explains your count difference. I forget
off-hand why this is.
Syntactically, they're all valid, at least from my perspective.
Semantically, it's useful to warn about typos. In practice, checking
that the components are known to dpkg is probably sufficient to catch
typos, so I'm not sure there's any point in warning about the ones that
are known but highly improbable.
> Another question might be, why are the given operating systems
> unsupported on those architectures? Is that likely to change in the
> future?
Well, there are combinations that are "obviously" unlikely. For
example, it seems rather unlikely that anyone is going to bother to port
the Hurd to hppa.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: