Iterative Feedback-Based Fault Localization Approach

Abstract

Most current fault-localization approaches merely use the information from testing to locate faults, assuming that test cases satisfying a certain test adequacy criterion can provide adequate information for fault localization. Unfortunately however, due to different goals between fault localization and testing, the information from testing may be inadequate for fault localization, which can greatly influence the performance. In this paper, we present an iterative feedback-based fault localization approach to address the information inadequacy problem. Based on both testing information and current fault-localization results, this approach iteratively uses test data generation techniques to feed back new test cases to initial test suites to increase more useful information for fault localization. Experimental results on the Siemens suite demonstrate a significant improvement of our approach over three best types of spectrum-based fault localization technique.