Investigação e avaliação experimental de técnicas de reteste seletivo para teste de regressão baseado em especificação.

Abstract

Specification-based selective regression testing (selective retesting) techniques can be used to increase the reliability of computer systems which functionalities and/or requirements have been modified. Several technique have been proposed, however, unlike the code-based techniques, there isn’t much knowledge about the benefits, limitations and characteristics from specification-based techniques, except for expectations. This work presents an experimental investigation of five specification-based selective retesting techniques, analyzed under five properties: inclusiveness, precision, efficiency, reduction potential and fault density. These properties are widely used when evaluating selective retesting techniques. The analyzed techniques use a Model-Based Testing approach, where test cases are automatically generated and selected. Besides the techniques presented in the literature, analyzed here, we propose a new technique (Weighted Similarity Approach for Regression Testing - WSA-RT), able to reduce costs for software testing and increase fault detection. From the results of the experiment we were able to identify the advantages and disadvantages of each technique, as well as describing aspects of applicability of these techniques by performing an analysis of generality. We followed all the steps of a process for an experimental study, therefore, the obtained conclusions concerning the performance of the analyzed techniques, are statistically significant.