Lightweight fault localization combined with fault context to improve fault absolute rank  

作　　者：Yong WANG Zhiqiu HUANG Yong LI Bingwu FANG 

机构地区：[1]College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics [2]College of Computer and Information, Anhui Polytechnic University

出　　处：《Science China(Information Sciences)》2017年第9期174-189,共16页中国科学（信息科学）（英文版）

基　　金：supported by National High Technology Research and Development Program of China (863) (Grant No. 2015AA015303);National Natural Science Foundation of China (Grant Nos. 61272083, 61562087, 71371012, 61300170, 61572033);Key Support Program Projects for Outstanding Young Talents of Anhui Province (Grant No. gxyq ZD2016124);Advanced Research of National Natural Science Foundation (Grant No. 2016yyzr10);Anhui Natural Science Foundation (Grant Nos. KJ2016A252, 1608085MF147)

摘　　要：Lightweight fault localization （LFL）, which outputs a list of suspicious program entities in descend- ing order based on their likelihood to be a root fault, is a popular method used by programmers to assist them in debugging. However, owing to the nature of program structures, it is impossible for LFL to always rank the root faulty program entity at the top of a ranking list. Recently, Xia et al. noted in their study that programmers inspect the first top-K-ranked program entities outputted by an LFL tool in sequence. Therefore, it is of practical significance to further improve the absolute rank of those buggy programs by using LFL if the root fault is ranked higher. To solve this issue, we propose a new LFL combined with a fault context to improve the fault absolute rank. We conduct experiments in which we apply our proposed approach to seven Siemens benchmark programs. The results show that by combining the suspiciousness scores of program entities with their fault-context suspiciousness scores that are based on an LFL called Dstar, our approach can improve the fault absolute rank with an effectiveness rate of 35.7% for 129 faulty versions from the seven benchmark pro- grams. It should be noted that our approach can obtain an average improvement of 65.18% for those improved programs to which LFL can be effectively applied, and that there were improvements to seven top-ranked root faults of buggy programs.Lightweight fault localization （LFL）, which outputs a list of suspicious program entities in descend- ing order based on their likelihood to be a root fault, is a popular method used by programmers to assist them in debugging. However, owing to the nature of program structures, it is impossible for LFL to always rank the root faulty program entity at the top of a ranking list. Recently, Xia et al. noted in their study that programmers inspect the first top-K-ranked program entities outputted by an LFL tool in sequence. Therefore, it is of practical significance to further improve the absolute rank of those buggy programs by using LFL if the root fault is ranked higher. To solve this issue, we propose a new LFL combined with a fault context to improve the fault absolute rank. We conduct experiments in which we apply our proposed approach to seven Siemens benchmark programs. The results show that by combining the suspiciousness scores of program entities with their fault-context suspiciousness scores that are based on an LFL called Dstar, our approach can improve the fault absolute rank with an effectiveness rate of 35.7% for 129 faulty versions from the seven benchmark pro- grams. It should be noted that our approach can obtain an average improvement of 65.18% for those improved programs to which LFL can be effectively applied, and that there were improvements to seven top-ranked root faults of buggy programs.

关 键 词：fault context fault localization program spectrum absolute rank debugging 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]