SYSML21 — Conjugated port definition specialization problem

Consider the following model:

part def P1;
part def P2 :> P1:

interface def I1 {
   end source : P1;
   end target : ~P1;
}
interface def I2 :> I1 {
   end :>> source : P2;
   end :>> target : ~P2; // Validation error!
}

The indicated end feature fails the KerML validation constraint validateAssociationEndTypes, which requires the association ends (connection ends in SysML) have exactly one type. In the above model, P2 specializes P1, so I2::source has just the one narrowed type P2. However, P2::'~P2' does not specialize P1::'~P1', so I2::target ends up having two types: ~P1 from its redefinition of I1::target and ~P2 by explicit declaration. And this violates the constraint.