基于对数几率回归的函数级软件缺陷定位

Logistic Regression-based Software Fault Localization in Function Level

现有的基于程序频谱的缺陷定位方法是通过利用语句覆盖信息计算可疑度从而确定其检查次序的,但在系统测试时,待定位对象代码量庞大,导致这类方法效果不佳。针对以上情况,提出一种基于对数几率回归的函数级别软件定位方法,其主要是分析失败测试用例的子系统和模块的执行信息,区分缺陷根源互异的失败测试用例,缩小定位范围;依次进行模块级别和函数级别的缺陷定位,计算每个模块和函数与失败测试用例的关联度,根据可疑值确定检查次序。实验表明,提出的方法能够有效缩小缺陷定位范围,提高缺陷函数的定位效率。

The method of fault localization based on spectrum determines the check order by calculating the suspicious of the coverage information of statements. However,during the system test,the size of the object to be located is large,which leads to the poor effect of those methods. This paper proposed a method to locate the functions faults during system test. In order to narrow the range of location,we analyzed the execution information of failed test cases subsystem and module,and distinguished the failed test cases with different faults. The correlation degree is calculated for each module and function,and the order of functions to be checked is determined by locating the faults at module level and function level in sequence. The experiments show that the proposed method can reduce the range of fault localization and improve the efficiency of fault localization.