Bug Localization Based on Code Change Histories and Bug Reports

摘要

A bug report is mainly used to find a fault location in software maintenance. It contains several fields such as summary, description, status and version. The description field includes detail scenario and stack traces if exceptional messages are presented. Recently researchers have proposed several approaches for automatic bug localization by using information retrieval and data mining. We propose BLIA, a statically integrated analysis approach of IR-based bug localization by utilizing texts and stack traces in bug reports, structured information of source files, and source code change histories. We performed experiments on three open source projects, namely AspectJ, SWT and ZXing. Compared with prior tools, our experiment results showed that BLIA outperforms the existing tools in terms of mean average precision. Our approach on average improved the metric of BugLocator, BLUiR, BRTracer and AmaLgam by 34%, 23%, 17% and 8%, respectively.