The Effect of Test Suite Type on Regression Test Selection

Regression test selection (RTS) techniques reduce the cost of regression testing by running only test cases related to code modifications. RTS techniques have been extensively researched, and the effects of several context factors on techniques have been empirically studied, but no prior work has explored the effects that might arise due to differences in types of test suites. We believe such differences may matter, and thus, we designed an empirical study to investigate them. Specifically, we consider two types of test suites obtained with automated test case generation techniques-feedback-directed random techniques and search-based techniques-along with manually written test suites. We assess the effects of these test suite types on two RTS techniques: a "fine-grained" technique that selects test cases based on dependencies tracked at the method level and a "coarse-grained" technique that selects test cases based on dependencies tracked at the file level. We performed our study on eight open-source projects across 800 commits. Our results show that on average, fine-grained RTS was more effective for test suites created by search-based test case generation techniques whereas coarse-grained RTS was more effective for test suites created by feedback-directed random techniques, and that commits affect RTS techniques differently for different types of test suites.