Research On Classified Regression Test Suite Reduction

With the continuously growing importance of the role that computer software is playing in our daily life, the quality is becoming a sensitive issue. Regression testing is a key but expensive method to guarantee the quality of software under evolution. How to alleviate the cost of testing and improve the confidence to the software has become a hot topic in regression testing. To tackle this issue, test suite reduction aims to identify and remove redundant test cases, according to a given test adequacy criterion, to produce a reduced test suite on behalf of the original one.Traditional test suite reduction can significantly decrease the size of test suite, but at the cost of seriously weakening the fault detection ability (FDA) of original test suite. To solve this problem, this thesis proposes a classified test suite reduction ap-proach, named CATESR. Different from traditional approaches, CATESR conjectures that test cases which are more likely to cover the changes after code modification, can gain a higher probability to reveal potential faults in the modified code. Based on this conjecture, CATESR originates test case classification utilizing the change impact to test requirements, and performs test suite reduction only on the most impacted test sub-suite. To assess the effectiveness of CATESR, we also conduct a set of empirical studies.The main contributions of this thesis can be summarized as follows:1. Topics on regression test suite reduction are reviewed. First, background and framework of regression test suite reduction are introduced. Then, some concepts and definitions for test suite reduction are presented. Finally, Related work about this field, including test suite reduction approaches and effectiveness analysis, is studied.2. A classified test suite reduction approach, CATESR, is proposed. First, the motivation of CATESR is introduced through a simple example. Then, the frame-work of CATESR is described. Finally, the implementation details of3core modules, including Test Requirement Labelling, Test Case Classification and Test Suite Reduction, are clarified.3. An empirical study on the effectiveness of CATESR is conducted. First, to assess the effectiveness of CATESR,3research questions with2metrics are de-signed. Then, CATESR is implemented and applied on8experimental subjects. Finally,3previous research questions are answered one by one, and an analysis about threats to the validity is performed. The results show that, CATESR can not only greatly decrease the size of reduced test suite, but also keep, sometimes even improve, the fault detection ability.