Understanding and predicting bugs fixed by API-migrations

Abstract

Bug tracking systems are standard repositories that preserve a large number of uncovered bugs. Once a bug is reported in these repositories, developers search for appropriate changes to fix the bug. However, discovering the changes that can fix the bugs has a negative influence on the schedule and cost of projects. Mainly, fixing bugs could be done by performing some other maintenance changes. In this work, we study and examine the role of adaptive maintenance in the context of API-migration during bug fixing activities through a case study on KOffice, Extragear/graphics, and Open Scene Graph projects. Our goal is to direct developers towards potential bugs early in development which are more likely to be fixed by performing adaptive changes as opposed to other maintenance tasks. We examined the reports of fixed bugs from the bug tracking systems of the studied projects, then we explored several factors related to variant dimensions of the reports and their relevant version history commits, in order to evaluate their effectiveness to decide whether a bug is likely to be fixed by adaptive changes. Our case study results show that bug residency time, textual contents of the report, the component that the bug was found in, and reporter/commenter experience show significant differences between the bugs that are fixed by adaptive changes and other fixed bugs.