9311967 Offutt This project is investigating the relative merits of unit-level software testing techniques, specifically various data-flow criteria and mutation testing. It is developing and implementing algorithms for automatically generating test data to satisfy data- flow criteria. Heuristic procedures for determining program sub paths that cannot be executed are also being investigated. The techniques studied will be evaluated on an experimental basis, using combination of fault detection ability and relative abilities of the studied criteria to cover each other. Recent results with constraint-based test data generation strongly indicate that test data can be automatically generated to test software program units according to the data-flow criteria. This project would develop algorithms that can be used to improve the current state of practice and research in testing software. It would also provide evidence concerning the efficacy of testing criteria with the objective of improving the quality of software. ***
