Not yet done: need to review (and probably) fix the modello-maven-plugin deps, as they are off now done

I would still prefer 90352d3 because there are tools out there that instantiate the Modello class. And also lookup of ModeloCore.ROLE is done all around the place in tests, so this would be consistent with that.

@fridrich ported, pls review

The lookup does not matter, the String ROLE is legacy, latest Plexus (not the sisu plexus shim) introduced lookup(Class) and the only difference is that you don't need casting.

Another thing. the slf4j-simple does not need to be runtime dependency. one can have there slf4j-nop if one does not need logging

Without slf4j backend, CLI will emit warning like you reported in the issue... So you say that no logging backend is needed at all? (noop?)

The slf4j-nop is an artifact that is noop logger

org.slf4j:slf4j-nop is the artifact

So, one can leave the choice for the user.

testing, will report in 30 minutes

Faster then that: it works. Thanks a lot.

nagyom szepem keszenem

ni za kaj (molim, i drugi put)

ni za kaj (molim, i drugi put)

In my original language it would be: "nemas za co (dakujem a aj druhy raz)"

Just there is a downgrade of plexus-utils 4.0.1 -> 3.5.1 Intended?

Just there is a downgrade of plexus-utils 4.0.1 -> 3.5.1 Intended?

Yes, keep this Maven3 compat for now (p-u 3.x is for Maven3, p-u 4.x is for Maven4)

Let's see what @slawekjaranowski has to say as well...

@slawekjaranowski ping

Reason for magic: we make modello-core "complete" as it provides now (and was providing) the CLI that needs to be runnable. And then later, in modello-maven-plugin we make sure that bits provided to maven plugins are properly provided/excluded. Before, it was "provided" in modello-core, so nothing was needed in modello-maven-plugin, but this breaks CLI. Now, modello-core is "complete" and modello-maven-plugin POM makes sure all is okay in plugin as well.

Would a working CLI justify a 2.4.1? @slawekjaranowski @cstamas