Examining the effects of developer familiarity on bug fixing

Abstract

Background: In modern software systems’ maintenance and evolution, how to fix software bugs efficiently and effectively becomes increasingly more essential. A deep understanding of developers’/assignees’ familiarity with bugs could help project managers make a proper allotment of maintenance resources. However, to our knowledge, the effects of developer familiarity on bug fixing have not been studied.Aims: Inspired by the understanding of developers’/assignees’ familiarity with bugs, we aim to investigate the effects of familiarity on efficiency and effectiveness of bug fixing.Method: Based on evolution history of buggy code lines, we propose three metrics to evaluate the developers’/assignees’ familiarity with bugs. Additionally, we conduct an empirical study on 6 well-known Apache Software Foundation projects with more than 9000 confirmed bugs.Results: We observe that (a) familiarity is one of the common factors in cases of bug fixing: the developers are more likely to be assigned to fix the bugs introduced by themselves; (b) familiarity has complex effects on bug fixing: although the developers fix the bugs introduced by themselves more quickly (with high efficiency), they are more likely to introduce future bugs when fixing the current bugs (with worse effectiveness).Conclusion: We put forward the following suggestions: (a) managers should assign some “outsiders” to participate in bug fixing. (b) when developers deal with his own code, managers should assign more maintenance resource (e.g., more inspection) to developers.