Safe Composition of Configuration Knowledge-Based Software Product Lines

Feature models and configuration knowledge drive product generation in a Software Product Line (SPL). Mistakes when specifying these models or in the implementation might result in ill-formed products-- the safe composition problem. This work proposes an automated approach for verifying safe composition for SPLs with explicit configuration knowledge models. We translate feature models and configuration knowledge into propositional logic and use SAT Solvers to perform the verification. We evaluate our approach using seven releases of the MobileMedia SPL, which generate up to 272 products in the 7th release. We report safe composition problems related to non-conformity with the feature model, bad specification of the configuration knowledge, and implementation not envisioning the full SPL scope, that affect over 40% of the products in the 7th release.