An empirical study on a regression testing technique for OO software. (c2006)

Abstract

Regression testing is an essential activity aiming to prove that the unmodified sections of a software application are unaffected by changes. This paper presents an empirical study on a regression test selection technique for object-oriented sojiware. The technique is based on the Unified Modeling Language (UML) program design; il10re 5pecifically it makes use of the class diagrams, interaction overview diagram and the sequence diagrams. Nine versions of three case sludies test cases are used: the UML design is specified, test cases are drawn and then they are studied in terms of the number of selected test cases, inclusiveness and precision. We prove the feasibility of the technique and demonstrate the ability of the regression testing technique to select a small number of test cases for retest from the original set of test cases; We also provide evidence revealing high inclusiveness and precision rates; an amendment was also introduced to cope wilh a problem in detecting changes inside the fragments of the sequence diagrams.