variants

[Daniel J. Luke]

On Apr 12, 2016, at 4:08 PM, Christopher Jones <jonesc at hep.phy.cam.ac.uk> wrote:
>> the current situation is basically the same as what upstream provides when building from source.
> 
> No, it is very different.
> 
> reinstalling with a new set of variants is a lot easier than figuring but by hand what options need to be specified to achieve that, and what other dependencies are required. MP, handles all that just fine.

ok, so MP is better than upstream?

> Reinstalling with new variants is, as far as I am concerned, not a major issue.

so, I'm not sure what the problem you're solving is? Is it just "new install requires a specific variant on a dependency"?

>> Am I missing somthing? Changing how default variants works only fixes one case:
>> new install, installing something that depends on a variant of some other port
>> 
>> It breaks existing behavior:
>> port install A (which installs B as a dependent) is currently the same as port install B && port install A
>> 
>> It doesn't fix things when the dependent is already installed.
>> 
>> This could be fixed by adding variants to the dependency engine OR by making use of the existing dependency engine (ie, breaking the port up into pieces so that things can depend on what they actually need) OR by just getting rid of variants (batteries-included install).
> 
> Variants aint going away any time soon, as far as I am concerned.

-- 
Daniel J. Luke