Numair Mansur, M., Wüstholz, V., & Christaki, M. (2023). Dependency-Aware Metamorphic Testing of Datalog Engines. In ISSTA 2023: Proceedings of the 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis (pp. 236–247). Association for Computing Machinery. https://doi.org/10.1145/3597926.3598052
ISSTA 2023. 32nd ACM SIGSOFT International Symposium on Software Testing and Analysis
17-Jul-2023 - 21-Jul-2023
Seattle, United States of America (the)
Number of Pages:
Association for Computing Machinery, New York
Datalog; metamorphic testing; fuzzing
Datalog is a declarative query language with wide applicability, especially in program analysis. Queries are evaluated by Datalog engines, which are complex and thus prone to returning incorrect results. Such bugs, called query bugs, may compromise the soundness of upstream program analyzers, having potentially detrimental consequences in safety-critical settings.
To address this issue, we develop a metamorphic testing approach for detecting query bugs in Datalog engines. In comparison to existing work, our approach is based on rich precedence information capturing dependencies among relations in the program. This enables much more general and effective metamorphic transformations. We implement our approach in DLSmith, which detected 16 previously unknown query bugs in four Datalog engines.
Testing Program Analyzers Ad Absurdum: 101076510 (European Commission)