Dynamic Similarity Based Fault Localization Prioritization

With software becoming more popular and more complex in daily life, the cycle of software debugging is getting longer. Software fault localization, one of the most costly and time-consuming activity in software debugging process, his become more and more important. Therefore developing an automatic approach to help programmers locate bugs effectively is the main task in recent years.In software testing, fault localization prioritization is to reorder test cases that aims at improving efficiency of fault localization. It combines two key processes in software regression testing, fault detection and fault localization, in order to reduce the test cost.Among the existing fault localization prioritization techniques, the similarity sort approach calculates the coverage similarity between test cases and the first failed test, which only including the number of statements covered. This method does not consider differences among the statements, which should have different weight to test cases, thus the best execution order of test cases for fault localization may not able to be identified. When the fault is found, it can prioritize test.In this paper, we focus on the problem of similarity metric for fault localization prioritization. We propose a dynamic similarity based fault localization prioritization approach. In similarity calculation, the statement suspicious value is introduced as a weight of statements to improve the effectiveness of fault localization. We define feature vector of the failed test cases that is represented by the statement suspiciousness, and the metric of dynamic similarity to improve the effectiveness of fault Localization. In order to reduce the cost of fault localization, the corresponding optimization strategies are proposed to reduce the cost of calculating of similarity, and improve the accuracy of fault localization.The impact of different test case prioritization algorithms is empirically analyzed and validated. In the experiments, three widely used test case prioritization algorithms and two fault localization methods are used on six benchmark C programs. The results show that the proposed approach can improve the effectiveness and efficiency of fault localization.