Improving Bug Localization by Mining Crash Reports: An Industrial Study

摘要

The information available in crash reports has been used to understand the root cause of bugs and improve the overall quality of systems. Nonetheless, crash reports often lead to a huge amount of information, being necessary to consolidate the crash report data into groups, according to a set of well-defined criteria. Recent research work have proposed different criteria and techniques to group crash report data, making more effective the process of finding the root causes of a bug and showing the performance of the approaches in the context of open source applications (such as IDEs and web browsers). In spite of that, it is still not clear how these approaches perform in other application domains, such as enterprise systems. In this paper, we present an industrial study in this field. We tailor existing approaches to find and group correlated crash reports, and identify buggy files in the domain of web-based systems. We then evaluate the performance of the resulting criteria and technique in industrial settings – identifying and ranking the classes that are more likely to contribute to a crash and thus might need a fix. We also check if the methods changed by the developers to fix a bug are present in the stack traces of the crash report groups used to identify the buggy classes. Our study provides new pieces of evidence of the potential use of crash report groups to indicate buggy classes and methods using stack traces information. For instance, we successfully identify buggy classes with recall varying from 61.4% to 77.3%, considering the top 1, top 3, top 5, and top 10 suspicious buggy files identified and ranked by our approach. We also found that 80% of changed methods from the closed bug fix issues appeared in related stack traces of the crash report groups. Finally, the approach also received positive response from the project leaders of the evaluated projects to help their bug resolution processes.