Global and Geographically Distributed Work Teams: Understanding the Bug Fixing Process and Potentially Bug-prone Activity Patterns

Abstract

This dissertation has studied the roots of the errors introduced involuntary into the source code through the analysis of the history of source code management systems (SCMs). The event of introducing an error has been named as bug seeding commit, and its place in the bug life cycle has been studied from several perspectives: a) the bug life cycle has been expanded and considers not only the bug tracking system (BTS), b) the time to fix a bug or c) the estimation of the finding rate, among others. In addition, we study the potential causes of the seeding. This thesis has focused on two potential human related aspects such as experience and time of the day. As main results, the time span from a bug being introduced to its discovery represents on average 60% to a 90% of its total life. In addition, the discovery rate of bugs follows in most of the cases a lineal distribution, a result that can help in estimating further effort in the project. Regarding to human related factors, no relationship was found between experience and the ratio of buggy activity of the developers. And finally, there are specific timeslots of the day where the changes to the source code are more prone to be buggy: specifically during the evening timeframe of activity.