Automated testing of installability of foreign architecture packages?

Scott Ritchie scott at open-vote.org
Fri Nov 9 22:05:46 UTC 2012


On 11/9/12 10:40 AM, Colin Watson wrote:
> On Fri, Nov 09, 2012 at 07:25:14AM -0800, Scott Ritchie wrote:
>> wine1.4:i386 is currently not installable on amd64, although
>> wine1.4:amd64 is.  I would, however, like it to be, for a person who
>> wanted 32 bit only Wine even on a 64 bit host OS.
>>
>>
>> It occurred to me that, while Britney is automatically testing
>> installability of packages, we may not be testing
>> cross-installability like this.
>>
>> Shoudl britney be extended in this way, or does down this path lie madness?
>
> At the moment I'm not really keen on complicating this further when
> we've only just introduced the system, particularly since I expect this
> would be a pretty complex feature to add.  Most of the
> multiarch-installability problems were due to builds being out of sync
> across architectures, and we should have got rid of that problem in one
> fell swoop with proposed-migration, so the situation should already be
> quite a lot better in raring than in quantal.
>

Perhaps we could just do this once semi-manually during the cycle then? 
  I'm sure there are other i386 packages that don't install on amd64 for 
one reason or another.

> FWIW, the current breakage seems local to wine1.4.  If I drill down far
> enough I get:
>
>    # apt-get install wine1.4:i386 wine1.4-i386:i386 libsm6:i386 wine1.4-common:i386 libice6:i386 x11-common:i386
>    [...]
>    The following packages have unmet dependencies:
>     wine1.4-common : Depends: wine1.4 (= 1.4.1-0ubuntu1) but it is not going to be installed
>
> I think that's lack of :any, probably.  However, given that
> wine1.4-common is marked as transitional, it seems to me that you could
> fix this simply by dropping the dependency on wine1.4-common from
> wine1.4-i386 (and probably also from wine1.4-amd64).  Do you agree?
>

Ahh yes, wine1.4-common was exactly to workaround the fact that we 
couldn't depend on :any in the precise cycle.  I'll clean it up then.





More information about the ubuntu-devel mailing list