A Study Of Program Spectra-Based Software Fault Localization Method

Software is a collection of procedures, which is able to provide specified function and performance with complex logic. To ensure that the program can execute accurate specific requirements, repeated debugging is needed. fault localization of software is a rather difficult and time-consuming task in debugging process. fault localization aims to find the program statements that lead to abnormal behavior of the program. Software fault localization method, which bases on program spectrum, is an effective, scientific and automatic method. This method is on code level.Software fault localization method which bases on program spectrum uses dynamic information gained after program statements are executed to find fault program statements or statements which might contain faults. Software fault localization method based on program spectrum requires high coverage rate of test cases for program statements. To meet this requirement, large test case sets are often taken in actual process of fault localization. But large test case sets tend to cause the redundancy of test case, and fault localization effects present instability in different test case sets. The efficiency and accuracy of fault localization are influenced. In addition, it requires to collect a large number of program spectrum information in order to get the dynamic information of program execution. This is undoubtedly a huge overhead in the process of fault localization. Moreover, the accuracy of the program statement suspicion degree measurement is crucial in the fault localization process. How to chose contribution rate of each test case used in testing directly relates with the accuracy in calculating suspicion degree of each program statement. In order to solve these problems, this paper proposes a program spectrum-based software fault localization method by removing redundancy of test cases. We reduce the quantity of test cases by removing redundancy of coverage information table. This paper also proposes program spectrum-based software fault localization method with increment. By executing test cases group after group, we managed to reduce the overhead of collecting program spectrum as more as possible. Considering the marginal effect of the large number of test cases, we correct the suspicion degree of each program statement in a dynamic way. The innovation points of the thesis are listed as follows:This paper proposes a test case reduction method based on program spectrum. In order to guarantee code coverage rate of test cases and to alleviate the effect of redundant program spectrum information, from another point of view, to reduce the redundancy of test cases. This strategy ensures that the fault localization effect is not affected by the negative impact, and ti improves the efficiency of defect location in a certain extent.This paper proposes software fault localization using increment and executing test case group after group. We managed to save the overhead of executing test cases and collecting program spectrum by reducing the quantity of test cases for the positioning. Incremental execution of test cases is an iterative process, and the end condition of iteration is the stability of localization sequence.In software fault localization method of using increment, we considered the marginal effect on the condition of a large number of test cases, and adjust weighted factor of suspicion degree parameter in a dynamic way in real time. By doing this, we ensure the reasonable contribution rate of each test case for the statement suspicious degree calculation. It makes the suspicion degree calculation of program statements more accurate and the efficiency of fault localization higher.