许多学者研究了运用测试集对程序错误语句定位的问题,并提出了许多行之有效的方法,这些方法统称为TBFL(testing based fault localization)方法.后来人们发现,测试集里如果出现冗余,则这些冗余测试用例会伤害这些定位方法的功效.为了解决这个问题,Hao等人提出了SAFL(similarity aware fault localization)方法.实际上完全避免冗余是不可能的,因此从另一个角度构造了一个新的TBFL方法,称为随机TBFL方法.该方法的基本思想是:测试前对程序的语句错误概率进行先验分布,并把测试集看成随机变量,用测试用例反映的程序语句有关信息对程序语句的概率作一些调整,调整后的概率称为后验校正概率,最后根据这个后验概率对错误语句进行定位.将传统的TB-FL方法如Dicing方法、TARANTULA方法、SAFL方法纳入随机信息分析并通过几个实例进行分析和比较,结果表明,随机TBFL方法不仅能够正确定位错误语句,而且冗余对该方法的功效伤害不大.%Fixing faults in software are an essential task in software development,and many approaches have been presented to automate fault localization. Among them, testing-based approaches are most promising. These approaches use the information of test cases to localize the faults, and they are called collectively as TBFL approach. But these TBFL approaches have ignored the similarity of the test cases, which may harm the effectiveness of these approaches. In fact it is impossible to completely avoid redundancy. Therefore this paper presented a new TBFL approach named random TBFL approach from a new view. The basic idea is that: the program is viewed as a random variable,and before testing,a prior distribution about the error probability of statements of the program is given, then some adjustments to the error probability of statements are made based on the execution information of the test suite,and the re-adjusted probability is called posterior probability,finally this posterior probability is used to localize the faults. This paper integrated the traditional TBFL approaches into the random framework,and compared and analyzed them on several instances. The analysis demonstrates that the random TBFL approach can correctly locate the faults,and redundancy has little influence on the effectiveness of the random TBFL approach.
展开▼
