Effectiveness of Code Contribution: From Patch-Based to Pull-Request-Based Tools ∗ Jiaxin Zhu, Minghui Zhou Audris Mockus Institute of Software, EECS, Peking University University of Tennesee Key Laboratory of High Confidence Software 1520 Middle Drive Knoxville, TN 37996-2250, Technologies, Ministry of Education USA Beijing 100871, China
[email protected] {zhujiaxin, zhmh}@pku.edu.cn ABSTRACT 1. INTRODUCTION Code contributions in Free/Libre and Open Source Software In Free/Libre and Open Source Software (FLOSS) projects, projects are controlled to maintain high-quality of software. non-core contributors have to go through the steps of code Alternatives to patch-based code contribution tools such as contribution process [29] from code creation to review and mailing lists and issue trackers have been developed with acceptance. These steps are essential to ensure the quali- the pull request systems being the most visible and wide- ty of contributions because of the diverse nature and skill- ly available on GitHub. Is the code contribution process s of FLOSS participants. Mailing lists and issue trackers, more effective with pull request systems? To answer that, where code contributions are submitted by patches, have we quantify the effectiveness via the rates contributions are been widely employed to support code contributions. The accepted and ignored, via the time until the first response participation in and efficiency of the code contribution pro- and final resolution and via the numbers of contribution- cess from the perspective of code reviews have been exten- s. To control for the latent variables, our study includes a sively studied (see, e.g., [3, 27, 25, 5, 21]).