Comprehensive Static Analysis for Configurable Software via Combinatorial Instantiation

Equipped with customized parameters, configurable software is more flexible when facing various hardware platforms and scenario options. The configurability can tailor the source code to different instances. Consequently, it is difficult for developers to enumerate all possible configurations for finding bugs, especially for large-scale configurable software systems. In this paper, we propose a method to efficiently detect bugs of such systems with static analysis techniques. The method takes advantage of combinatorial testing techniques to generate sufficient configurations. It first extracts required parameters and the corresponding constraints from a configure file. The parameters together with constraints are employed to generate configurations with required coverage. Considering the features of configuration options, we further classify the parameters into clusters, according to the tightness of their relations. Inspired from the idea of divide-and-conquer, every cluster can be assigned with a local strength, such that the tightly coupled options can be covered, without incurring other unnecessary options. Such improvement can reduce the number of required configurations, thus improving the efficiency of static analysis. The experimental results over four real-world configurable systems demonstrate the efficiency, scalability and practicality of our method.