Research Of Fault Propagation-Based Software Fault Localization Method

Program debugging is a very tedious work in software development process and it mainly includes fault detection, fault localization, fault correction. Fault localization is one of the most time consuming and strenuous activity in the debugging process and it’s through the examination of the source program semantics and structure, combined with the analysis of the execution and the result of the program, helping developers find software fault location. Efficiently locating software faults can relieve the programmer’s workload of manual checking program statement and improve debugging speed and efficiency. In recent years, a lot of diagnosis technologies have appeared in the field of artificial intelligence and software engineering automation, among testing based fault localization(TBFL) has become a research hotspot. It only works on test cases of success and failure information, but can still detect program faults without knowledge of internal structure and logic of program or even have no knowledge of program. But the TBFL still have some limitations for different types of faults of the program.This paper summarizes the development of TBFL technology, presents a improved method of software fault localization based on fault propagation. It is mainly through getting the execution path and dependency information of program, modify the existing fault state of program, and ultimately obtained statement suspicious score. It through program instrumentation and execute test cases get the dynamic execution path coverage information and dependency information, calculate the initial suspicious score of each node and suspicious score of dependence pair, and establish the fault propagation model, use the method in this paper to calculate the suspicious score of each statement, get a program statement suspicious score rank. Finally achieve the purpose of helping programmer fault localization.This paper use the Siemens suite as the experimental program, with the widely used criteria to evaluate the accuracy of the algorithm, and compare with several existing program fault localization techniques. The experimental results show that, the method proposed in this paper compared with other fault localization methods has better performance, fault propagation factor existing in the program has been considered well, the improved algorithm in this paper is effective and can help programmers debug efficiently.