Improving Software Quality With Introspective Testing

我的研究计划的长期目标是帮助工业软件工程师提高他们的软件系统的质量和可靠性，为他们提供新的工具和技术来分析和加强他们的测试套件。软件工程师已经成功地构建了在现代社会中发挥关键作用的令人难以置信的规模和实用性的系统;然而，随着这些系统的复杂性不断增加，我们对这些系统质量的期望也在不断提高。这导致了对软件质量的更严格的审查和对系统故障的不容忍;故障也有直接的经济成本，估计每年超过600亿美元。开发人员主要使用软件测试来确认他们的系统按照预期运行。不幸的是，软件工程师用来验证测试套件“强度”的工具和技术对于现代软件系统来说是不够的。该计划中采用的研究方法将包括持续的工业参与，以确保我们的方法改善工业实践中的缺点，并扩展到真实的软件系统的复杂性。该研究计划将产生新的技术来表征软件测试的敏感性，一种通过将测试执行映射到真实的用户执行来评估测试关键性的方法，以及用于评估任何已识别故障的严重性的机制。最后，我们将开发新的方法来自动加强现有套件中代表弱点的测试。我们的方法将通过实验室实证研究和与我们的工业合作伙伴的验证（包括定量实验和定性案例研究）进行评估。*最终，这个研究项目的目标是帮助软件工程师评估和提高他们的测试套件的质量，这反过来又会提高我们每天使用的软件系统的整体质量。

The long-term objective of my research program is to help industrial software engineers improve the quality and reliability of their software systems by providing them novel tools and techniques for analyzing and strengthening their test suites.******Software engineers have managed to build systems of incredible size and utility that play a crucial role in modern society; however, as these systems have grown in complexity, so has our expectations for the quality of these systems. This has resulted in increased scrutiny of software quality and intolerance for system failure; failures also have a direct economic cost, estimated at in excess of $60B per year.******Developers predominantly use software testing to confirm that their systems behave as they expect. Unfortunately, the tools and techniques software engineers have at their disposal for validating the 'strength' of their test suites are insufficient for modern software systems.******The research methodology employed in this program will include continual industrial involvement to ensure our approaches improve shortcomings in industrial practice and scale to the complexity of real software systems. The research program will produce novel techniques for characterizing the sensitivity of software tests, a means to assess test criticality by mapping test executions to real user executions, and mechanisms for assessing the severity of any identified faults. Finally, we will develop new ways to automatically strengthen tests that represent weaknesses in existing suites. Our approaches will be evaluated in both empirical lab studies and through validations with our industrial partners (which will comprise of both quantitative experiments and qualitative case studies).******Ultimately, the goal of this research program is to help software engineers assess and improve the quality of their test suites which will in turn lead to improved overall quality for the software systems we use every day.**