Re: Multi-Arch: allowed
-----BEGIN PGP SIGNED MESSAGE-----
Le 01/11/2016 à 15:57, David Kalnischkies a écrit :
> On Tue, Nov 01, 2016 at 02:43:21PM +0100, Thibaut Paumard wrote:
>> How do you actually use Multi-Arch: allowed? Should a dependent
>> package then specify either <package>:same or <package>:foreign?
> Neither is valid syntax. What you do with these is depending on a
> package with the literal architecture "same" (or "foreign"). Thats
> not going to work…
>> I was able to find documentation about what allowed is supposed
>> to do, but not on how to depend on such a package.
> The spec  linked from that page says how, but in summary:
Thanks, I was not able to parse it correctly apparently.
> If a package (lets say perl) is marked as Multi-Arch: allowed your
> package foo can depend on perl:any and a perl package from any
> (foreign) architecture will statisfy this dependency, while a
> 'simple' perl would have just accepted a perl from the architecture
> your package foo was built for (with arch:all mapped to
> If it helps: Instead of "perl having Multi-Arch: allowed" envision
> it to have a "perl provides perl:any" and you are depending on this
> virtual package – which also explains why such a missing provides
> causes perl:any to be unresolveable.
That makes things clearer, thanks.
> That said, the usecase for 'allowed' is small – mostly
> interpreters – and you are trying to use it on… a -dbg package? I
> haven't looked closely, but that smells wrong… What are you trying
> to express here?
The -dbg package is Multi-Arch same. It Depends on the packages for
which it provides debugging symbols, some of which are Multi-Arch:
allowed. Lintian complains when I don't specify an architecture for
W: gyoto source: dependency-is-not-multi-archified gyoto-dbg depends
on gyoto-bin (multi-arch: allowed)
N: The package is Multi-Arch "same", but it depends on a package
N: neither Multi-Arch "same" nor "foreign".
N: Refer to https://wiki.ubuntu.com/MultiarchSpec for details.
N: Severity: normal, Certainty: possible
N: Check: group-checks, Type: source
By the way, by the same logic that interpreters should (or can?) be
Multi-Arch: allowed, I expect that
- extensions for those interpreters also should (or can?);
- any tool that is able to process an input data file to produce a
arch-independent output also should (or can?) be either "foreign" or
Is that correct?
> (and have you heard that automatic debug packages are a thing
Yes. Last time I checked, it was not clear how to use them in
> Best regards
> David Kalnischkies
>  https://wiki.ubuntu.com/MultiarchSpec  There are ways around
> it. See the "If it helps" remark for a hint.
Kind regards, Thibaut.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
-----END PGP SIGNATURE-----