Beyond testing configurable systems: applying variational execution to automatic program repair and higher order mutation testing

Abstract

Generate-and-validate automatic program repair and higher order mutation testing often use search-based techniques to find optimal or good enough solutions in huge search spaces. As search spaces continue to grow, finding solutions that require interactions of multiple changes can become challenging. To tackle the huge search space, we propose to use variational execution. Variational execution has been shown to be effective in exhaustively exploring variations and identifying interactions in a huge but often finite configuration space. The key idea is to encode alternatives in the search space as variations and use variational execution as a black-box technique to generate useful insights so that existing search heuristics can be informed. We show that this idea is promising and identify criteria for problems in which variational execution is a promising tool, which may be useful to identify further applications.