Evolutionary Variant Selection to Reduce Redundancy in Behavioral-Aware SPL Testing

In software product lines, behavioral models are often represented as super-imposed or 150% models that encode the combined behavior of all product variants within a single artifact. Generating test cases from a superimposed behavioral model through a feature model has its own challenges. For instance, a feature model may not have a product/variant that represents the superimposed model. Therefore, multiple variants must be selected to test the behavioral superimposed model, which defines the problem as a combinatorial (selection) optimization problem. Selecting variants can play an important role in affecting the number of generated test cases. Bad selections of variants can result in multiple redundant test case generations. There are several factors that could lead to redundant test cases: (i) Shared features from selected variants, (ii) Shared covered components (events, event-pairs, states, state transitions, etc.) from the behavioral model mapped to features. This study proposes a variant selection approach that uses evolutionary algorithms to reduce test generation redundancy from behavioral models. Fur- thermore, we have used Event Sequence Graphs (ESGs) as our behavioral model. Our results show that, with 100% event-pair coverage on the behav- ioral model, we are able to significantly reduce the number of generated test cases.