View article

Testing should be one of the key activities of every software development process. However it requires up to half of the software development effort when it is properly done. One of the main problems is the generation of smart tests to probe the system, which is both difficult and timeconsuming. The research community has been proposing several ways to automate the generation of these tests; among them, the search-based techniques recently achieved significant results. This doctoral dissertation presents TestFul, our evolutionary testing approach for stateful systems; it is tailored to work on object-oriented systems. It uses a holistic approach to make the state of object evolve, to enable all the features the class provides, and to generate the shortest test with the utmost coverage for the class under test. We employ several complementary coverage criteria to drive the evolutionary search. We aim to generate tests with high fault detection effectiveness. To this end, we consider the system from complementary perspectives and we combine white-box analysis techniques with black-box ones. The evolutionary search is completed with a local one, and we establish a synergic cooperation between them. The evolutionary search concentrates on evolving the state of objects, while the local search detects the functionality not yet exercised, and directly targets them. All the proposal were subject to an extensive empirical validation. We devised a benchmark composed of independent benchmarks for tests, public libraries, and third party studies. As comparison, we consider both search-based, symbolic, and traditional (ie, manually generated by human …