Maintenance of software product line using software testing optimization techniques

Abstract

The customer requirements and market competition lead to the development of highly configurable systems. Due to this demand, the development of configurable software emerges. Now software development moves to families of configurable software rather than the single implementation of a product. The testing of these families of software product lines (SPLs) is a challenging task because of the large number of combinations in a SPL. In addition, the testing process of SPLs becomes impractical with the existence of optional features. In SPLs, a large number of features are presented but these features become infeasible with respect to time and cost constraints environment. So, one solution to this problem is to test subsets of configured products. For this, there need different approaches to test the SPLs like combinatorial interaction testing (CIT) technique to minimize the testing exertion and generate better results. But in case of large size SPLs with excessive constraints, this approach generates unscalable results. Due to feature combinations, the CIT approach becomes expensive. Furthermore, some existing approaches discuss to optimize the multiple conflicting testing objectives like to reduce the cost and configurations number. This research proposes a search-based software engineering solution using multi-objective optimization algorithms (MOEAs). In particular, the research applied on different types of MOEA methods; Indicator-Based Evolutionary Algorithm (IBEA), Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D), Non-Dominated Sorting Genetic Algorithm II (NSGAII), NSGAIII and Strength Pareto Evolutionary Algorithm 2 (SPEA2). The ‘SPL features models’ were selected from the Software Product Line Online Tool (SPLOT) repository. The valid number of configurations for Feature Models (FMs) is generated with the help of SAT solver. This dissertation first optimized the three objectives and compared the results using four MOEAs (IBEA, MOEA/D, NSGAII, NSGAIII, and SPEA2) framework. Secondly, there was another proposed framework having five MOEAs (IBEA, MOEA/D, NSGAII, NSGAIII, and SPEA2) that optimized the four objectives to solve the SPL testing problems. For three Objective optimizations, results of four MOEAs were compared and concluded that MOEA/D generated better results while in the case of four objectives optimization approach, the performance of SPEA2 was better. Finally, research work was applied to an industrial forum on the State Bank of Pakistan (SBP) organization. The SBP transactions are considered as a SPL and proposed research approaches are applied to resolve the testing issues. The approaches demonstrated the following advantages: minimization of transactions set, transactions prioritization and transactions generations.