MEDIC: A static analysis framework for equivalent mutant identification

• Published in: Information and software technology Vol. 68; pp. 1 - 17
• Main Authors: Kintis, Marinos, Malevris, Nicos
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.12.2015; Elsevier Science Ltd
• Subjects: Automation; Data analysis; Data flow analysis; Equivalent mutant detection; Genetic testing; Mutation; Mutation testing; Programming languages; Static analysis; Stubborn mutant detection; Studies; Equivalent mutant detection; Stubborn mutant detection; Data flow analysis; Mutation testing; Static analysis
• DOI: 10.1016/j.infsof.2015.07.009
• ISSN: 0950-5849

Context: The equivalent mutant problem is a well-known impediment to the adoption of mutation testing in practice. In consequence of its undecidable nature, a complete automated solution is unattainable. To worsen the situation, the manual analysis of the generated mutants of a program under test is prohibitive due to their vast number and the complexity of determining their equivalence. Objective: This paper focuses on the automated identification of equivalent and partially equivalent mutants, i.e. mutants that are equivalent to the original program for a specific subset of paths. To this end, the utilisation of a series of previously proposed data flow patterns is investigated. This study also examines the cross-language nature of these patterns and the killability of the detected partially equivalent mutants. Method: A tool, named MEDIC (Mutants’ Equivalence DIsCovery), incorporating the aforementioned patterns was developed. Its efficiency and effectiveness were evaluated based on a set of manually analysed mutants from real-world programs, written in the Java programming language. Furthermore, MEDIC was employed to test subjects written in the JavaScript programming language. Results: MEDIC managed to detect 56% of the examined equivalent mutants in 125 s, providing strong evidence regarding both its effectiveness and efficiency. Additionally, MEDIC was able to identify equivalent mutants in the JavaScript test subjects, lending colour to the cross-language nature of the implemented patterns. Finally, the identified partially equivalent mutant set consisted largely of killable mutants, 16% of which were stubborn ones. Conclusion: It can be concluded that pattern-based equivalent mutant identification forms a viable approach for combating the equivalent mutant problem. MEDIC automatically detected a considerable number of the manually identified equivalent mutants and was successfully applied to test subjects in all examined programming languages.