A cover-based approach for configuration repair

Abstract

Feature models are often used to describe variability and commonality in Software Product Lines, specifying admissible configurations of valid products. However, invalid configurations may arise in some scenarios. These include feature model evolution that invalidates pre-existing products or collaborative configuration by multiple stakeholders with conflicting goals, among others. This problem has been acknowledged in the literature and some techniques for configuration repair have already been proposed. However, common optimization criteria such as proximity between original and repaired configurations can result in a significant number of alternative repair possibilities, easily attaining thousands of alternatives for models of practical dimension. Consequently, rather than just efficiently providing an exhaustive list of possibilities, an approach that specifically addresses this issue should be able to offer the user a manageable and comprehensible view of the configuration problems and potential repair options. We offer a novel approach for configuration repair, based on partitioning and cover analysis, with high performance and generating high quality solutions, which allows efficient identification and presentation of multiple competing repairs.