Guiding Random Test Generation for Intra-class Dataflow Coverage

Automatic generation of a good test suite is difficult, especially for object-oriented software. Feedback-directed random test generation is an approach that can achieve good branch coverage and has been used as a basis to systematically construct suites for testing realistic Java programs. We augment this random test generation method to create tests suites that satisfy an intra-class data-flow coverage criterion which is highly relevant for object orientation, although little addressed or achieved by tools in practice. We show that our approach can be used on real object-oriented software and that the technique for guiding test generation produces an increase in coverage.