Structural testing techniques for the selective revalidation of software

Abstract

The research in this thesis addresses the subject of regression testing. Emphasis is placed on developing a technique for selective revalidation which can be used during software maintenance to analyse and retest only those parts of the program affected by changes. In response to proposed program modifications, the technique assists the maintenance programmer in assessing the extent of the program alterations, in selecting a representative set of test cases to rerun, and in identifying any test cases in the test suite which are no longer required because of the program changes. The proposed technique involves the application of code analysis techniques and operations research. Code analysis techniques are described which derive information about the structure of a program and are used to determine the impact of any modifications on the existing program code. Methods adopted from operations research are then used to select an optimal set of regression tests and to identify any redundant test cases. These methods enable software, which has been validated using a variety of structural testing techniques, to be retested. The development of a prototype tool suite, which can be used to realise the technique for selective revalidation, is described. In particular, the interface between the prototype and existing regression testing tools is discussed. Moreover, the effectiveness of the technique is demonstrated by means of a case study and the results are compared with traditional regression testing strategies and other selective revalidation techniques described in this thesis.