Selecting Tests and Identifying Test Coverage Requirements for Modified Software
G. Rothermel and M. J. Harrold
Proceedings of the ACM International Symposium on Software Testing and Analysis
August 1994, pages 169-184


Regression testing is performed on modified software to provide confidence that changed and affected portions of the code behave correctly. We present an approach to regression testing that handles two important tasks: selecting tests from the existing test suite that should be rerun, and identifying portions of the code that must be covered by tests. Both tasks are performed by traversing graphs for the program and its modified version. We first apply our technique to single procedures and then show how our technique is applied at the interprocedural level. Our approach has several advantages over previous work. First, our test selection technique is safe, selecting every test that may produce different output in the modified program. However, our selection technique chooses smaller test sets than other safe approaches. Second, our approach is the first safe approach to identify coverage requirements, and the first safe approach to do so interprocedurally. Third, our approach handles both structural and nonstructural program modifications, and processes multiple modifications with a single application of the algorithm. Finally, our approach can be automated, and is applicable to all programs for which program dependence graphs can be constructed.