Reopen缺陷的分析、预测与自动化修复技术研究

Bug fixing is one of the most important activities in the process of software development and maintenance. Due to the complexity of software systems, some bugs which were fixed in the current version will re-appear in the future, and these bugs will be reopened to perform the supplementary fixes. Reopen bugs will take longer time to resolve, increase the difficulty of bug fixing, decrease the software quality, increase the cost of software development and maintenance, and even cause the failure of the project. To reduce the negative impacts of reopen bugs and increase the developers’ productivity, in this project, we analyze reopen bugs and propose a number of automated algorithms to predict and repair reopen bugs: (1) perform a large-scale cross-company cross-project empirical study on the impact, categories and root causes of reopen bugs; (2) propose a composite reopen bug prediction algorithm which considers features based on bug report, submitter and the initial bug fix; (3) propose a reopen bug change recommendation model built on the multiple change relationship graphs; (4) propose an advanced fault localization algorithm which combines multiple information sources, and leverage it to improve the performance of the automated reopen bug repair model.

缺陷修复是软件开发与维护过程中一项必不可少的活动。由于软件系统的复杂性，部分在当前版本“修复”的缺陷在未来某个时间再次出现，这些缺陷需要重新打开（Reopen）进行再次修复。Reopen缺陷会造成缺陷修复时间的延长、软件质量的下降、软件成本的增加甚至项目失败。为了降低Reopen缺陷的危害和提高开发人员生产效率，本项目拟对Reopen缺陷进行系统化地研究与分析，并提出一系列自动化算法以实现从Reopen缺陷预测到修复的过程：（1）针对Reopen缺陷的危害、类型以及根源等进行跨公司、跨项目的大规模实证分析；（2）提出基于缺陷报告特征、提交者特征和初次修复补丁特征的复合Reopen缺陷预测模型；（3）提出基于缺陷初次修复构建的多种修改关系图的Reopen缺陷修改推荐模型；（4）提出基于多种Reopen缺陷信息源的缺陷定位算法，并整合该算法以增强Reopen缺陷自动化修复的效率。

缺陷修复是软件开发与维护过程中一项必不可少的活动。由于软件系统的复杂性，部分在当前版本“修复”的缺陷在未来某个时间再次出现，这些缺陷需要重新打开（Reopen）进行再次修复。Reopen缺陷会造成缺陷修复时间的延长、软件质量的下降、软件成本的增加甚至项目失败。为了降低Reopen缺陷的危害和提高开发人员生产效率，本项目对Reopen缺陷进行系统化地研究与分析，并提出一系列自动化算法以实现从Reopen缺陷预测到修复的过程：（1）针对Reopen缺陷的危害、类型以及根源等进行跨公司、跨项目的大规模实证分析；（2）提出基于缺陷报告特征、提交者特征和初次修复补丁特征的复合Reopen缺陷预测模型；（3）提出基于缺陷初次修复构建的多种修改关系图的Reopen缺陷修改推荐模型；（4）提出基于多种Reopen缺陷信息源的缺陷定位算法，并整合该算法以增强Reopen缺陷自动化修复的效率。本项目共发表论文13篇，其中包括CCF A类期刊ACM Transactions on Software Engineering and Methodology和IEEE Transactions on Software Engineering各一篇， CCF B类期刊9篇，CCF B类会议1篇。