PhD Oral Exam - SayedHassan Khatoonabadi, Computer Science

When studying for a doctoral degree (PhD), candidates submit a thesis that provides a critical review of the current state of knowledge of the thesis subject as well as the student’s own contributions to the subject. The distinguishing criterion of doctoral graduate research is a significant and original contribution to knowledge.

Once accepted, the candidate presents the thesis orally. This oral exam is open to the public.

Abstract

Pull-based development is a common paradigm for contributing to and reviewing code changes in numerous open-source projects. However, a considerable amount of Pull Requests (PRs) with valid contributions are not finalized because their contributors have left the review process unfinished. Such abandoned PRs waste a considerable amount of time and effort from both their contributors and their maintainers. Furthermore, PRs that are neither progressed nor resolved, clutter the list of PRs, and eventually make it difficult for the maintainers to manage and prioritize unresolved PRs. Recognizing these challenges, this thesis aims to investigate the underlying dynamics of abandoned PRs, evaluate the helpfulness of common solutions to PR abandonment, and propose ways to mitigate PR abandonment in large open-source projects. We start by studying the characteristics of abandoned PRs, the reasons why contributors abandon their PRs, and the perspectives of project maintainers on dealing with PR abandonment. Our findings indicate that contributors and the review process play a more prominent role in PR abandonment than projects and PRs themselves. Our survey with project maintainers also indicates that Stale bot is commonly adopted by many open-source projects to deal with abandoned PRs. However, there are ongoing debates on whether using Stale bot alleviates or exacerbates PR abandonment. Therefore, in our next study, we investigate the reliance of projects on Stale bot to deal with their PR backlog, the impact of Stale bot on pull-based development, and the kind of PRs usually intervened by Stale bot. Our findings indicate that despite its benefits, Stale bot tends to further aggravate contributor abandonment. To help better mitigate PR abandonment, in our last study, we propose a machine learning approach to predict the first response latency of the maintainers and the contributor of a PR. We demonstrate the effectiveness of our approach in both project-specific and cross-project settings and also discuss the importance and impact of different features on the predicted waiting times. The awareness fostered by these predictions enables both the maintainers and the contributor to take proactive actions to mitigate potential challenges during the review process of the PR before it gets abandoned.