Change impact analysis for regression testing of java software

Abstract

Modified versions of software are regression tested to assure that the modified codes behave as intended and that no faults are newly introduced to the unchanged parts of codes. Change impact analysis must be done in the first step of regression testing to confirm that all parts impacted by changes during modi??cation behave correctly. In object-oriented programs, change impact analysis is done at different levels, and among these approaches, method-level analysis achieves reasonable precision and efficiency. This thesis presents an efficient change impact analysis for regression testing of Java programs at the method level. In order to gain efficiency, we observes the relationships of possible Java changes and identi??es only eight changes which have to be tested among them. Second, for the identified changes, we consider each change is independent of each other and omit redundant analysis. We have implemented a prototype tool to automate our approach, and conducted case studies with it to evaluate its precision and efficiency. Comparison with another method- level analysis by previous approach shows that it preserves the same precision with previous one with less analysis time. Moreover, the comparison with a class-level technique shows that our approach is more practical than the class-level analysis.