基于混杂偏倚消除的谓词统计错误定位方法  

Predicate-Level Statistical Fault Localization Based on Confounding Bias Mitigation

作　　者：王兴亚[1] 姜淑娟[1,2] 鞠小林[1,3] 曹鹤玲[1,4] WANG Xing-Ya;JIANG Shu-Juan;JU Xiao-Lin;CAO He-Ling(School of Computer Science and Technology,China University of Mining and Technology,Xuzhou,Jiangsu221116;Guangxi Key Laboratory of Trusted Software,Guilin University of Electronic Technology,Guilin,Guangxi541004;School of Computer Science and Technology,Nantong University,Nantong,Jiangsu226019;College of Information Science and Technology,Henan University of Technology,Zhengzhou450001)

机构地区：[1]中国矿业大学计算机科学与技术学院,江苏徐州221116 [2]桂林电子科技大学广西可信软件重点实验室,广西桂林541004 [3]南通大学计算科学与技术学院,江苏南通226019 [4]河南工业大学信息科学与工程学院,郑州450001

出　　处：《计算机学报》2017年第12期2671-2687,共17页Chinese Journal of Computers

基　　金：国家自然科学基金(61673384;61502497);广西可信软件重点实验室研究课题(kx201530);南京大学计算机软件新技术国家重点实验室开放课题(KFKT2014B19);江苏省研究生培养创新工程(KYLX_1390);河南省高等学校重点科研项目(16A520005);南通市应用研究计划(BK2014055)资助~~

摘　　要：识别并消除统计错误定位过程中的混杂偏倚效应可以有效提高错误定位结果的精度.该文对谓词可疑度度量过程中的混杂偏倚现象进行了研究,提出一种基于混杂偏倚效应消除的谓词统计错误定位方法.首先,提出一种基于变量类型的错误候选谓词筛选方法来提高错误相关谓词的识别能力;其次,通过分析影响谓词取值和程序执行结果的条件,识别定位过程中的数据依赖和控制依赖混杂偏倚元素;再次,采用一种静态切片与动态约减相结合的谓词信息收集方法,提高谓词信息的收集效率;最后,针对收集的谓词信息、混杂偏倚元素信息和程序执行结果,使用线性回归分析度量谓词的可疑度,降序排列后提供给开发人员开展程序调试.实验表明该文方法可以有效识别错误相关谓词,提高错误定位的精度和效率.Identifying and reducing the confounding effect can improve the effectiveness of statistical fault localization.In this paper,we present a novel approach that accounts for the effects of program dependencies to mitigate the confounding effect during Predicate based Statistical Fault Localization(PBSFL).First,we present a variable type based predicate pre filtering technique to improve the ability of fault relevant predicate identification.Then we recognize both the control dependence confounding bias and the data dependence confounding bias by analyzing the mutual dependence between the program predicate and the executing result based on the causal inference.To improve the efficiency and accuracy of the PBSFL,we also design a combined predicate collection technique with static slicing and dynamic reducing.With the run time information and the executing result,for each predicate we estimate its causal effect by a linear regression analysis and treat the failure causing effect as the suspiciousness.Finally,they are provided to developers for program debugging.Experimental studies demonstrate that the fault relevant predicate can be identified effectively by the proposed variable type based predicate pre filtering technique,and the effectiveness as well as the efficiency of the predicate based statistical fault localization can be significantly improved after mitigating the dependence confounding effect.

关 键 词：谓词错误定位 混杂偏倚元素 因果推理 程序依赖关系 软件测试 程序调试 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]