The JavaDoc of getConstraintsForConstructor() wrongly says "Constructor of super types are considered". This shouldn't be the case and it never has been implemented like that by the RI nor by Apache BVal. Hence it's no migration concern to remove this wrong sentence.