Research Of Software-defect Localization Based On Function Calling Sequence Mining

With the rapid development of the computer software industry and the growth ofsoftware scale, software quality was paid more and more attention. Even the little defectsin the software can be used by malicious attackers, and cause huge loss to enterprise andordinary users. Therefore, localizing defects to help developers to modify the defects ofprograms by testing techniques is a compulsory task.The number of defects in the software is apparently related to software reliability.Localizing defects based on dynamic call sequences can be combined with automation test,and this has greatly practical significance. We derive program traces, includingmethod-invocation structure and the spot information, from program correct and failingexecutions by using the aspect-oriented programming language (such as AspectJ). Thencalculate the probability of the defects in functions by mining and analyzing the callingsequences obtained from the program traces. In this paper, a new method for localizingdefects based on function calling sequences is proposed, and this method is validated withan example. The calling sequence not only reflects the chronological order of the functioncalling structure, but also contains the spot information, including the parameters, implicitparameters and the return values. The rationale for defect localization is that the functionwhich has a greater rate of information gain in the decision tree algorithm has a strongerinfluence on the execution of the program, and is more likely to have defects.In this paper, software prototype for the Software Defect Location Algorithm basedon calling sequence was designed and developed according to the theory, and the functionmodules and the design of the system were introduced in great detail. At last, wedemonstrated the qualities of our approach with a study on defects introduced into theopen source software–Weka. And sampled datas were derived from UCImachine-learning repository. The experiments showed that, the method proposed in this paper was an effective method to locate software defects, and it maked the defects locationmore accurately.