A Model for Predicting Bug Fixes in Open Source Operating Systems: an Empirical Study

This paper proposes an adaptation to the open source environment (Linux Kernel and OpenSolaris) of a model for predicting which bugs get fixed in the Microsoft Windows operating system. We have analyzed the entire bug repositories containing 16,136 bug reports reported in about 8 years of activity of the project (from 2002 to 2010) for the Linux Kernel and 16,301 bug reports reported in about 3 years of activity of the project (from 2007 to 2010) for OpenSolaris. According to the data analyzed and the descriptive models produced, we have found that (a) bugs reported by people with better reputation and bugs in which more people are involved are more likely to get fixed, (b) reassigning or re-opening bugs are not affecting the fix likelihood, and (c) managing bugs in the same location increases the fix likelihood. The predictive model defined has a precision of 61% and a recall of 39% for the Linux Kernel and a precision of 76% and a recall of 73% for OpenSolaris. These results are comparable with the ones for Microsoft Windows. document. Keywords-bugs; predictive models; open source