Mining Function Call Sequence Patterns Across Different Versions of the Project for Defect Detection

Large scale programs usually imply many programming rules, which are missing from specification documents. However, if programmers violate these rules in the process of programming, they may introduce software defects. Mining programming rules for detecting defect is an effective way to alleviate this problem. However, previous works suffer from a large number of candidate rules and suspicious defects which need manual validation. This issue affects the applicability and scalability of these previously proposed approaches. This paper proposes a novel approach to detect defects based on programming rules mined from different versions of a project. Firstly, it mines function call sequence patterns from the version under analysis and a previous stable version; secondly, it filters useful function call sequence patterns based on the patterns contained in the previous version; thirdly, the programs are automatically checked against filtered patterns for detecting suspicious defects. Experiments are carried out on three open source projects varies from 12k to 142k LOC to evaluate the effectiveness of our proposed approach. The experiment results show that the approach can improve the efficiency of defect detection by reducing 55% suspicious defects for the three projects without comprising the defect detection capability.