Bug localisation through diverse sources of information

Abstract

Many approaches have been proposed to address the problem of bug localisation - taking a bug report and recommending to developers the possible locations of the bug in the project. However, these can often require significant up-front work from developers, and are not widely adopted. Furthermore, those techniques which do not require this up-front investment are often far from accurate, and do not take advantage of all of the information that they could. We propose a technique for combining information from multiple, novel sources of information about a project and a bug, and use this to recommend bug locations to developers. We also identify how this technique could be used to create a low-effort tool for bug localisation, with the aim of increasing developer adoption. We evaluate the technique on 1143 bugs in three open-source projects, and find that it can be used to increase the number of bugs where the first relevant method recommended to developers is the top result from 98 to 132 and in the top-10 from 271 to 322.