Bug Localization for Version Issues With Defect Patterns

Abstract

Version issues are becoming more and more prominent with the continuous development of software. Bug localization for version issues is time-consuming and labor-intensive. Although some bug localization techniques, such as those based on information retrieval (IR), have been proposed, they cannot handle these bugs very well as the version-related bugs have their own defect patterns. However, few existing works have focused on revealing these defect patterns and utilizing them for localization of bugs. To fill this gap, we propose a new approach by leveraging the version-related defect patterns to localize the version-related issues integrated with the IR technique. First, we extract version-related bugs from bug repositories and build a version-related bug repository. Given a new version-related bug report, we identify the defect patterns of corresponding similar historical bug reports from the version-related bug repository. Then, we combine these defect patterns with the IR technique to rank the candidate code snippets as suspicious code for developers to fix. The evaluation demonstrates that our approach is more effective to identify the faulty code related to version issues than the existing IR-based bug localization technique.