UC Irvine

Many newly-submitted bug reports are not actionable: they do not have sufficient and clear information for developers to start the process of understanding and fixing them. The presence of such non-actionable bug reports leads to: (1) a waste of developers’ time as they have to come up with the right questions to ask bug reporters in order to understand the essence of these bug reports, (2) slower overall progress, and (3) negative effects on the overall quality of software. Therefore, effective bug reporting is crucial, especially in projects where a large number of bug reports are submitted on a regular basis and a significant portion of the overall software development lifecycle is spent addressing these reports.To address this issue, the dissertation first looks at understanding the overall quality of bug reports, contributing the implementation of a model that classifies bug reports as actionable or non-actionable. Second, through an empirical study of 2,814,599 bug reports across five systems, the dissertation answers the question of whether the inclusion of videos in bug reports may lead to tangible potential incentives for the reporters (i.e., reduced time to resolution, leading to an actual fix, or reduced number of back-and-forth). Third, the dissertation examines whether certain characteristics (e.g., the presence of voice over, clear highlighting with the mouse, a video contains steps to reproduce) might differentiate videos that have a positive impact on the bug resolution process from those that do not, by studying how developers react to videos with different characteristics and whether these different characteristics have potentially observable benefits. The main contributions of this dissertation include: (1) a machine learning model that with a 92% accuracy (F-measure of 0.91) separates actionable from non-actionable bug reports, which is significantly higher than the best results to date, (2) new findings concluding that the inclusion of videos in bug reports does not always translate to tangible benefits for reporters: bug resolution time is barely impacted, the percentage of bug reports being successfully resolved with a patch is lower for bug reports with videos, and back-and-forth is higher for bug reports with videos, and (3) new findings about videos with different characteristics: bug reports with videos that are less than 30 seconds or contain actual results appear to have observable benefits for bug reporters. The findings can serve as a basis for future research and developing tools that assist reporters in improving their bug reports before final submission, either by helping reporters to turn their non-actionable bug reports into actionable ones or by supporting them in attaching helpful videos exhibiting the characteristics found to be beneficial to both developers and bug reporters.