How can variant B imply variant A?

[Jeremy Lavergne]

On Aug 21, 2012, at 3:55 PM, Clemens Lang wrote:

>> This is just a bad idea: your package will install differently based
>> on what's available
> 
> Why is that a problem? In this case, the maintainer wants to
> specifically support both wxWidgets and wxWidgets-devel. The end-user
> either already made his choice on that (and chances are he'd select the
> same variants on this port), or he doesn't care, in which case the port
> can specify a sane default.
> 
>> and not even consistently (e.g. archives).
> 
> Yeah, there would only be one of both variants available as binary
> archive. I don't see that as a problem either, we're having users with
> non-standard variants build the software on their own anyway. You could
> consider having wxWidgets-devel "non-standard".
> 
> OTOH, not having variants for this case, but just supporting whatever
> the user has without specifying a variant _will_ break stuff when a user
> gets the buildbot archive, which might be built against wxWidgets but
> has wxWidgets-devel installed.

The variant itself is a good idea; but having default variants dynamically change has users unsuspectingly getting to build from source without expecting it.

As we just discussed on IRC, why not just prompt the user to pick a variant rather than deciding for them? This ensures people get what they expect, and avoids running into bad assumptions.

Let's copy the outcome of our chat here when we're done :-)