Using feature model knowledge to speed up the generation of covering arrays

Abstract

Combinatorial Interaction Testing has shown great potential for effectively testing Software Product Lines (SPLs). An important part of this type of testing is determining a subset of SPL products in which interaction errors are more likely to occur. Such sets of products are obtained by computing a so called t-wise Covering Array (tCA), whose computation is known to be NP-complete. Recently, the ICPL algorithm has been proposed to compute these covering arrays. In this research-in-progress paper, we propose a set of rules that exploit basic feature model knowledge to reduce the number of elements (i.e. t-sets) required by ICPL without weakening the strength of the generated arrays. We carried out a comparison of runtime performance that shows a significant reduction of the needed execution time for the majority of our SPL case studies. © 2013 ACM.