基于频谱增强的软件多故障定位

Software Multiple Fault Localization based on Spectrum Enhancement

摘要:

检测故障是繁琐而耗时的，为了提高传统软件故障定位方法的效率以及精确度，提出了一种基于频谱的故障定位新方法；充分利用了失败的测试用例与故障之间的关系，通过使用频谱增强的方式，采取逻辑与运算的关键技术和方法，对失败的测试用例进行精简，优化失败测试用例中的频谱信息，从而得到频谱增强后的测试用例，克服了冗余测试用例对定位效果的消极影响，再根据新的频谱信息计算可疑度值，最后生成优化后的可疑度排序列表；首次将频谱增强的方法同时运用到单故障与多故障程序场景中，在包含植入故障的西门子程序和真实故障的Defects4j程序中，经实验检测证明本研究方法能够显著减少代码检查的范围，尤其是在高性能范围内(EXAM5%)，并且仅通过检查Top-1至少能有效地定位超过原有约20%的故障，结果表明基于频谱增强的故障定位方法有效提升了检测率，可以更好地帮助程序员精准定位故障位置。

Abstract:

Detecting faults is tedious and time-consuming，in order to improve the efficiency and accuracy of traditional software fault location methods, a new method of fault location based on spectrum is proposed; the relationship between failed test cases and faults is fully utilized, and the key techniques and methods of logic conjunction operations are adopted to streamline the failed test cases and optimize the spectrum information in the failed test cases by using spectrum enhancement. The method of spectrum enhancement is applied to both single-fault and multi-fault program scenarios for the first time. In Siemens programs with embedded faults and Defects4j programs with real faults, experimental tests have demonstrated that this research method can significantly reduce the scope of code checking, especially in the high performance range (EXAM 5%), and can effectively locate more than about 20% of the original faults by checking Top-1 only, showing that the spectrum enhancement-based fault location method effectively improves the checking rate and can better help programmers pinpoint the location of faults.