Feature cardinalities overwrite mandatory group rule

[FAuersw]

Hello,
unfortunately i could not find a lot of literature on the exact semantics of feature cardinalities in UVL. As far as i understood, a single feature can be enabled multiple times. In this feature model example ..

features 
    Lunchbox 
        mandatory 
            Sandwich cardinality [0..3]

.. i would expect to have at least one sandwich in every configuration, as the feature is mandatory. But UVLS allows me to disable all sandwich at the same time in a valid configuration.
If i add a constraint Lunchbox => Sandwich it works as i would expect.
I am not sure if this is the intended behaviour but it seems unintuitive to me. What do you think about this?

[SundermannC]

Thanks again for your report @FAuersw! :)

In this preprint https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4764657, we have some discussion on the semantics of feature cardinality. However, we did not go into detail for the problem you described.
For UVLS, we discussed exactly this issue and we were not happy with either option.
Currently, the group of a feature with cardinality is not tied to the "cardinality selection". This means an optional feature with a cardinality of [1..3] can still be deselected, a mandatory with [0..3] can be deselected, in an alternative group you can still select the one feature multiple times, and so on.

[FAuersw]

Hi @SundermannC thanks for your explanation. I have read the preprint some time ago and am curious about the final version. Also I believe there is more work required on the semantics of feature cardinalities. The current implementation inflates the configuration space quite a bit ... but this may be a general issue of the transformation of feature cardinalities into lists of features at the boolean level.