SHF: Small: Empirical Studies, Principles and Techniques for Software Systems with Complex Configuration Spaces

今天的软件系统越来越多地由可以以多种不同方式配置的组件的灵活组合构建。 例如，流行的Firefox Web浏览器有一千多个配置选项。 虽然高度的可配置性有很多好处，但它也使可配置软件的测试和推理成为一个主要挑战。 这项工作的目标是：1）开发新的技术，以有效地发现软件系统的配置空间的结构，和2）利用这些信息，以改善常见的软件工程任务，特别是测试和程序理解。 特别是，这项工作将允许软件开发人员快速准确地回答问题，如，什么是“正确的”配置下测试系统？ 我们如何避免在"不必要的"配置下测试系统？ 随着系统的发展，软件配置是如何变化的？我们预计，由于今天使用的软件是可配置的，在这个项目中所取得的进步，在理解和测试可配置的系统将有广泛的好处，在可靠性和可信度的关键software.The建议的技术方法将追求四个主要方向。首先，PI将开发一个“有效”配置空间的表示-实现特定目标所需的配置-既可由软件工程工具使用，又可为开发人员所理解。 第二，PI将开发计算系统有效配置空间的新技术。第三，PI将探索一些软件工程应用程序，这些应用程序将利用有效的配置空间信息，包括配置感知测试用例选择，配置感知回归测试，以及可配置系统的几个程序理解任务。 最后，PI将进行广泛的基础实证研究，以测试研究假设，并评估大规模学科系统的拟议方法。除其他成果外，这些研究将检查一系列系统的有效配置;调查这些配置如何随着时间的推移而演变;并确定故障如何与系统配置相关。其结果将影响实践和教育。

Today's software systems are increasingly built from flexible combinations of components that can be configured in a multitude of different ways. For example, the popular Firefox web browser has more than a thousand configuration options. While a high degree of configurability has many benefits, it also makes testing and reasoning about configurable software a major challenge. The goals of this proposed work are 1) to develop new techniques to efficiently discover the structure of software systems' configuration spaces, and 2) to exploit that information to improve common software engineering tasks, specifically testing and program understanding. In particular, this work will allow software developers to quickly and accurately answer questions such as, What are the "right" configurations to test a system under? How can we avoid testing a system under "unnecessary" configurations? How does software configuration change as systems evolve? We expect that, as much of the software used today is configurable, the advances made in this project in understanding and testing configurable systems will have widespread benefits in the reliability and trustworthiness of critical software.The proposed technical approach will pursue four main directions. First, the PIs will develop a representation of "effective" configuration spaces---the configurations needed to achieve a specific goal---that is both usable by software engineering tools and understandable to developers. Second, the PIs will develop novel techniques that compute the effective configuration space of a system. Third, the PIs will explore a number of software engineering applications that will make use of effective configuration space information, including configuration-aware test case selection, configuration-aware regression testing, and several program understanding tasks for configurable systems. Finally, the PIs will conduct a wide range of fundamental empirical studies that will test the research hypotheses and evaluate the proposed approach on large scale subject systems. Among others outcomes, these studies will examine the effective configurations of a range of systems; investigate how those configurations evolve over time; and determine how failures relate to system configurations. The results will impact both practice and education.