软件产品线测试中具有优先级的特征配置生成方法

Generation of Prioritized Feature Configurations for Software Product Line Testing

在软件产品线(SPL)中,尽早针对领域工程中核心资产及其组合进行测试是保证软件产品线成功的必要手段.采用特征驱动(feature-driven)方法,可以从业务视角对可变点进行组合及测试,能够尽早发现潜在的由特征交互引起的软件错误.组合交互测试(CIT)是避免组合数量爆炸问题的一种抽样技术.但生成的特征配置应该进一步地基于优先级排序来帮助合理分配有限的时间和经费.提出一种面向产品线集成测试,具有优先级的特征配置生成方法.其核心算法遵循CIT技术,且基于成对组合测试覆盖标准.此外,生成的特征配置还会基于其对SPL测试的重要度排序.该方法作用于一个标记了优先级的特征模型,包含特征模型扁平化、配置模型转化、约束推导和P-AETG算法执行活动.我们通过一个在线考试产品线验证了此方法的有效性;并通过SPLOT网站上提供的特征模型对此方法进行性能和有效性的评估.

Early validation of core assets as well as their combinations in domain engineering is an essential assurance for the success of software product line （ SPL ） engineering. By feature-driven approaches, variability configurations can be constructed and tested from the business perspective, thus supports to uncover defects caused by the unintended feature interactions. Combinatorial interaction testing （CIT） is adopted as a sampling technique to avoid the combination explosion problem. However, the generated feature configu- rations should be further prioritized to help allocate restricted time and budget in domain testing. In the paper, we propose a method to generate and priorifize feature configurations for SPL integration testing. The method follows the idea of CIT and is based on the pair- wise feature interaction coverage. In addition, the feature configurations are sorted as they are built according to their importance to SPL testing. The method applied on weighted feature models involves the activities including feature model flattening, configuration model transformation, constraint deduction and P-AETG algorithm execution. The method is applied and validated by an Online Exam Product Line. In addition,it is further applied to weighted feature models from SPLOT to evaluate its performance and effectiveness.