支撑统计故障定位的测试技术研究

软件调试中的故障定位是软件开发过程中一项困难而耗时的工作。统计故障定位虽然能够帮助开发人员提高调试效率，但定位的准确度仍然不够理想。而统计故障定位的有效性不仅仅依赖于故障定位算法，也依赖于软件测试所产生的测试用例集的执行信息。我们调研发现，已有的测试和调试的集成研究工作仅能针对测试用例集的某一个特性来加强对统计故障定位的支持，仍然缺乏全面和实用的支撑统计故障定位的测试完备性准则和相应的测试生成技术。因此，本项目拟通过数据挖掘技术找出有效支撑统计故障定位的测试用例集的关键特征集，并提出相应的有效支撑统计故障定位的测试完备性准则。在此基础上，通过对基于符号执行的测试生成算法的制导和优化，构建支撑测试完备性准则的测试生成技术及工具集。本项目将对测试技术和统计故障定位技术的有效集成提供理论指导，并能通过测试完备性准则和高效的测试生成算法切实提高统计故障定位的精确度、实用性和可扩展性。

Software debugging involves fault localization, bug fixing and retesting the fixed program. Fault localization is one of the most tedious, difficult and time-consuming activity during the software development process. Although statistical fault localization techniques can help developers improve their debugging efficiency, the precision of current fault localization techniques is still unsatisfactory for large application. Software testing and debugging are two interconnected and interrelated activities, the execution information of the test suite generated by testing techniques is the basis for statistical fault localization. Our preliminary study found that existing work on the integration of testing and debugging can improve only one aspect of the test suite for statistical fault localization. Comprehensive testing adequacy criteria supporting effective statistical fault localization and its supporting test case generation techniques are still lacking. Hence, in this work, we aim at finding the set of key features for a test suite supporting effective statistical fault localization through data mining techniques. Based on that, we propose a set of testing adequacy criteria supporting effective statistical fault localization. Then, we further propose directed test case generation techniques based on symbolic execution guided by the proposed adequacy criteria. The successful completion of the project will provide theoretical guideline for the effective integration of testing and statistical fault localization. Furthermore, the proposed testing adequacy criteria and test case generation techniques will significantly improve the precision, practicability and scalability of existing statistical fault localization techniques.