Characterizing and Predicting Blocking Bugs in Open Source Projects

TitleCharacterizing and Predicting Blocking Bugs in Open Source Projects
Publication TypeConference Paper
Year of Publication2014
AuthorsValdivia Garcia, H, Shihab, E
Secondary TitleProceedings of the 11th Working Conference on Mining Software Repositories
Pagination72–81
PublisherACM
Place PublishedNew York, NY, USA
ISBN Number978-1-4503-2863-0
KeywordsCode Metrics, Post-release Defects, Process Metrics
Abstract

As software becomes increasingly important, its quality becomes an increasingly important issue. Therefore, prior work focused on software quality and proposed many prediction models to identify the location of software bugs, to estimate their fixing-time, etc. However, one special type of severe bugs is blocking bugs. Blocking bugs are software bugs that prevent other bugs from being fixed. These blocking bugs may increase maintenance costs, reduce overall quality and delay the release of the software systems.

In this paper, we study blocking-bugs in six open source projects and propose a model to predict them. Our goal is to help developers identify these blocking bugs early on. We collect the bug reports from the bug tracking systems of the projects, then we obtain 14 different factors related to, for example, the textual description of the bug, the location the bug is found in and the people involved with the bug. Based on these factors we build decision trees for each project to predict whether a bug will be a blocking bug or not. Then, we analyze these decision trees in order to determine which factors best indicate these blocking bugs. Our results show that our prediction models achieve F-measures of 15-42%, which is a two- to four-fold improvement over the baseline random predictors. We also find that the most important factors in determining blocking bugs are the comment text, comment size, the number of developers in the CC list of the bug report and the reporter's experience. Our analysis shows that our models reduce the median time to identify a blocking bug by 3-18 days.

URLhttp://doi.acm.org/10.1145/2597073.2597099
DOI10.1145/2597073.2597099
Full Text
AttachmentSize
PDF icon garcia.pdf407.52 KB