Do API-Migration Changes Introduce New Bugs?

Abstract

Software quality is broadly dependent on the use of dependent platforms, compilers, and APIs. This research reports a case study exploring the risk of API-migration activities in the regard of bug-introducing changes and software maintenance quality. The study involves screening tens of thousands of commits for six large C++ open source systems to identify bug-introducing commits caused by undertaking adaptive maintenance tasks through using traditional heuristic approaches. The obtained results show that 14.5% to 22.2% of examined adaptive commits enclose buggy code changes and so developers have to consider the potential risk of introducing new bugs after undertaking API-migration practices. Moreover, from investigating the bug fixing activities made by API-migration tasks, we provide a demonstration that typically these fixing activities do not cause further bugs and hence are safe undertakings. We feel that this work has developed a data set that will be used for constructing approaches to identify, characterize, and minimize potential adaptive maintenance practices that introduce bugs into a software system.