Testing with invalid test inputs is important to evaluate the robustness of a system. Combinatorial robustness testing is an approach to generate valid and invalid test inputs separately. Unfortunately, it is easy to create over-constrained test models. As a result, not all specified invalid values or invalid value combinations appear in the test suite. Previous work proposed to repair the test model manually or semi-automatically based on conflict detection and diagnosis techniques. In this paper, we extend that work and present a fully-automatic approach that allows to generate invalid test inputs from over-constrained test models based on alternative constraint handling strategies.