Repairing Over-Constrained Models for Combinatorial Robustness Testing

Testing negative scenarios is important to evaluate robustness of software systems. Error-handling can terminate the system before all values are evaluated and faults can remain undetected. Therefore, extensions for combinatorial testing separate generation of positive and negative scenarios. Unfortunately, it is easy to create over-constrained models. Certain values or value combinations are prevented from appearing in the test suite and remain untested. In this paper, we define over-constrained models and present a technique to identify and repair them.