Baital: Sampling configurable systems with high t-wise coverage

Abstract

Testing of highly configurable systems is challenging due to an immense number of configurations and is usually performed on a small sample set. Its quality is often measured with t-wise coverage. We propose a tool $\mathrm{Baital}$ for sampling highly configurable systems with thousands of features capable of achieving high t-wise coverage. The effectiveness and scalability of the tool are based on two novel techniques: the adaptive weighted sampling and the approximation techniques for the t-wise coverage computation. $\mathrm{Baital}$ can easily handle sampling and computation for 6-wise coverage. The latest version supports multi-valued features.