Research On Test Cases Selection Method For Fault Localization

With the increasing requirements of the software in the people’s daily life, the scaleand complexity of the software system is also constantly rising. The requirements of thesoftware testing and maintenance on complex large-scale systems are beyond the abilityof the current testing and maintenance methods. Therefore, there have been manyaccidents which result in software failures, and most of the software failures are causedby the fault statements. When software failure occurs, fault localization approach isused to automatically find the location of the fault which results in the failure or to findthe reasons of this failure. But the fault localization algorithms have a high complexity,and the redundant test cases may affect the accuracy of the fault localization. In order toreduce the complexity of the fault localization and improve the accuracy of faultlocation, it is necessary to study how to choose the useful test cases for faultlocalization.Currently, the existing test cases selection methods mostly consider therequirements of the software testing, they consider the adequacy of software testing.However, the fault-localization-oriented test case selection methods focus on theeffectiveness rather than the sufficiency of testing. Maybe only partial test cases usedfor testing are beneficial to fault localization, and the total test cases used for testing willreduce the accuracy of the fault localization. Traditionally, fault localization is a manualprocess which is tedious and error-prone. In recent years, there has been considerableresearch on fault localization techniques. However the existing test cases selectionmethods used for the fault localization do not consider the path information of the testcases, so they can not select the appropriate test cases for fault localization, and can notimprove the efficiency and accuracy of fault localization effectively. In order to solvethese problems, this paper proposes a new test case selection method for faultlocalization. This method considers the statement coverage information and pathinformation. It fully considered the impact of path information to fault localizationusing the idea of clustering. Firstly, it analyzes the statement coverage information, anddeletes the passed test cases with little relationship to the failed test cases; and thenanalyzes the path information and deletes the passed test cases which are redundancy onthe path information. Finally, it deletes the passed test cases based on the branchclustering.This paper uses the faulty versions and the test cases provided by the SiemensSuite as the experimental data. We design and implement experiments to verify theeffect of our approach. The experimental results show that our method can effectively reduce the test cases, and improve the effectiveness of fault localization.