On Tue, May 10, 2016 at 04:25:23PM +0200, Pietro Abate wrote:
> On 10/05/16 15:09, David Kalnischkies wrote:
> > > 'APT::Solver::aspcud::Preferences=-removed,-changed,-new,-count(solution,APT-Release:=/experimental/)'
> > ^^^^^^
> >
> > The preferences string is missing as it isn't sent – and it isn't sent because it
> > doesn't apply to the specified solver: You requested 'dump' to resolve it, but
> > the preferences string is for 'aspcud'.
> >
> > So, seems to work as intended, doesn't it – or did I miss anything?
>
> I see. YEs, it respect the spec. But as it is the this option
> (APT::Solver::aspcud::Preferences) is actually useless because aspcud
> will never receive this information, correct ?
You answered that yourself in your next mail, I notice through that the
spec is disagreeing with itself: The configuration section mentions this
solver specific option, while later it only mentions the all-solver
option. It is all-solver for the strict-pinning only option.
That seems rather inconsistent and is also not really consistent with
how apt usually is configured ala value = config("apt::specific::foo",
config("apt::foo", default_value));
So, I will take that bugreport as a wish to implement & document that
consistently for both – lets just hope we never have a solver called
'preferences' or 'strict-pinning'. ;)
[apt would actually support that just fine, it would just be confusing
as hell for a human to look at]
> This make me think that maybe it would be worth adding the solver to
> be called to the edsp file as well. Something like :
>
> Solver: aspcud
> Preferences: ...
>
> This will make the edsp fully contained.
As you wish :)
Best regards
David Kalnischkies
Attachment:
signature.asc
Description: PGP signature