Looking For Novelty in Search-Based Software Product Line Testing

Testing software product lines (SPLs) is difficult due to a huge number of possible products be tested. Recently, there has been growing interest in similarity-based testing SPLs, where similarity used as surrogate metric for the $t$ -wise coverage. In this context, one primary goals sample, by optimizing metrics using search-based algorithms, small subset test cases (i.e., products) dissimilar possible, thus potentially making more combinations covered. Prior work shown, means empirical studies, great potential current approaches. However, rationale technique deserves rigorous exploration. To end, we perform correlation analyses investigate how are correlated with We find that generally have significantly positive correlations This well explains why works, improvement on will increase Moreover, explore, first time, use novelty search (NS) algorithm SPL testing. The rewards “novel” individuals, i.e., those being different from individuals discovered previously, and matches goal score NS (much) stronger coverage than previous approaches relying genetic (GA) fitness function. Experimental results 31 validate superiority over GA, other state-of-the-art approaches, concerning both fault detection capacity. Finally, whether it useful combine two satisfiability solvers when generating new NS, performance affected its key parameters. summary, looking provides promising way sampling diverse SPLs.