软件故障定位报告质量评估方法  

Quality assessment methods for software fault localizating reports

作　　者：文勇[1,2] 蔡铭[1] 戴建华[3] 陈刚[1] 

机构地区：[1]浙江大学计算机科学与技术学院,浙江杭州310027 [2]广西民族大学物理与电子工程学院,广西南宁530006 [3]浙江行政学院信息管理部,浙江杭州311121

出　　处：《浙江大学学报（工学版）》2011年第6期984-990,共7页Journal of Zhejiang University：Engineering Science

基　　金：航天科技创新基金重点资助项目(CASC04);航天支撑技术基金资助项目(GFJG-E10803)

摘　　要：通过对基于静态程序依赖图的软件故障定位报告质量评估方法的分析,指出由于其评分缺乏对程序运行时动态特征的描述,而影响评分质量,提出一种新的改进方法.该方法利用描述程序运行的特征谓词信息和基于失败运行测试用例的程序运行轨迹,将动态特征融入到静态程序依赖图中,构造动态程序依赖图,在此基础上运用基于广度优先搜索算法,得到能够更加真实反映在查错过程中需要排查的程序语句集合.结合软件故障定位模型的实证分析表明,该方法能够有效提高评估质量,促进定位模型的改进.The quality assessment methods for software fault localization reports based on static program dependence graph fail to take into account the dynamic characteristics,which results in far below expectation.A novel quality assessment method was presented to improve the performance.This method makes use of the predicates describing the characteristics of program running and the running track based on the test cases in which a run of the program failed,constructs a dynamic program dependence graph,utilizes the breadth-first search-based algorithm,and obtains a collection of statements which is a more realistic reflection of the process in finding the bug.This method was implemented and evaluated on a fault localization model.The results show that this novel method can more effectively give out the quality of evaluations and promote the improvement of the localization model.

关 键 词：软件故障定位 软件质量评估 程序依赖图 程序运行动态特征 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]