Black-Box Test Case Selection and Prioritization for Software Variants and Versions
Remo Lachmann · Digitale Bibliothek Braunschweig (Verbundzentrale Göttingen (VZG)) · 2017
Software testing is a fundamental task in software quality assurance. Especially when dealing with several product variants or software versions under test, testing everything for each variant and version is infeasible due to limited testing resources. To cope with increasing complexity both in time (i.e., versions) and space (i.e., variants), new techniques have to be developed to focus on the most important parts for testing. In the past, regression testing techniques such as test case selection and prioritization have emerged to tackle these issues for single-software systems. However, testing of variants and versions is still a challenging task, especially when no source code is available. Most existing regression testing techniques analyze source code to identify important changes to be retested, i.e., they are likely to reveal a failure. To this end, this thesis contributes different techniques for both, variants and versions, to allow more efficient and effective testing in difficult black-box scenarios by identifying important test cases to be re-executed. Four major contributions in software testing are made. (1) We propose a test case prioritization framework for software product lines based on delta-oriented test models to reduce the redundancy in testing between different product variants.(2) We introduce a risk-based testing technique for software product lines. Our semi-automatic test case prioritization approach is able to compute risk values for test model elements and scales with large numbers of product variants. (3) For black-box software versions, we provide a test case selection technique based on genetic algorithms. In particular, seven different black-box selection objectives are defined, thus, we perform a multi-objective test case selection finding Pareto optimal test sets to reduce the testing effort. (4) We propose a novel test case prioritization technique based on supervised machine learning. It is able to imitate decisions made by experts based on different features, such as natural language test case descriptions and black-box meta-data. All of these techniques have been evaluated using the Body Comfort System case study. For testing of software versions, we also assesses our testing techniques using an industrial system. Our evaluation results indicate that our black-box testing approaches for software variants and versions are able to successfully reduce testing effort compared to existing techniques.