Improving Counterexample Quality from Failed Program Verification

A successful automated program proof is, in software verification, the ultimate triumph. In practice, however, the road to such success is paved with many failed proof attempts. If a proof fails, the prover usually gives little useful feedback, making it very hard to know how to proceed. The work reported here attempts to help in such cases by providing immediately understandable counterexamples. To this end, it introduces an approach called Counterexample Extraction and Minimization (CEAM): if the proof fails, it transforms the counterexample model generated by the prover to a user-friendly format; counterexamples can be further simplified by reducing the containing integer values to their minimal possible. We have implemented the CEAM approach as an extension in the AutoProof verifier and demonstrate its application on a collection of examples.