Reusing Test Cases on Graph Product Line Variants: Results from a State-of-the-Practice Test Data Generation Tool

• Willian D. F. Mendonça UFPR
• Wesley K. G. Assunção UTFPR
• Silvia R. Vergilio UFPR

Resumo

Software testing is an essential activity for quality assurance, but, it is an error-prone and effort consuming task when conducted manually. Because of this, the use of automated tools is fundamental, as well as, the evaluation of these tools in practice. However, there is not so much evidence on how such tools perform on highly-configurable systems. Highly-configurable systems are commonly observed in industry as an approach to develop families of products, where products have different configuration options to meet customer needs. To fulfill such a gap, this paper reports results on the use of the tool Randoop, which is widely used in industry, to test variants of the Graph Product Line (GPL) family of products. Our goal is to evaluate reusability of a test data set generated by Randoop for one product when reused for testing other GPL products. Besides, we also investigate the impact of using different values of runtime, the main Randoop parameter, on the number of reused test data. The results show that the used value for runtime in general does not contribute to increase the coverage of test data reused in different products. Furthermore, similarity among products does not ensure a greater reusability.

Palavras-chave: Test Data Generation, Test Coverage, Software Reuse, Highly-configurable systems, Family of Products