A Semi-Automated Iterative Process for Detecting Feature Interactions

• Larissa Rocha UFBA
• Ivan Machado UFBA
• Eduardo Almeida UFBA
• Christian Kästner Carnegie Mellon University
• Sarah Nadi University of Alberta

Resumo

For configurable systems, features developed and tested separately may present a different behavior when combined in a system. Since software products might be composed of thousands of features, developers should guarantee that all valid combinations work properly. However, features can interact in undesired ways, resulting in failures. A feature interaction is an unpredictable behavior that cannot be easily deduced from the individual features involved. We proposed VarXplorer to inspect feature interactions as they are detected and incrementally classify them as benign or problematic. Our approach provides an iterative analysis of feature interactions allowing developers to focus on suspicious cases. In this paper, we present an experimental study to evaluate our iterative process of tests execution. We aim to understand how VarXplorer could be used for a faster and more objective feature interaction analysis. Our results show that VarXplorer may reduce up to 50% the amount of interactions a developer needs to check during the testing process.

Palavras-chave: Experimental Study, Runtime Analysis, Configurable Systems, Feature interaction