The Impact of Adopting Continuous Integration on the Delivery Time of Pull Requests: A Partial Replication and Extension
Examensarbete för masterexamen
The practice of Continuous Integration (CI) is indispensable in the software development process nowadays. In a distributed and collaborative context, CI is a paradigm for integrating all changes from different developers quickly and safely. Thus the adoption of CI is considered to fasten delivering new functionalities. A study by Bernardo et al. doubted this claim, conducted an empirical study and found that only half of the investigated projects deliver Pull Requests (PR) faster after CI. In this study, considering the threats to the validity in their study, as well as the generalizability of their findings, a partial replication study, a comparison to projects that never apply CI, and an extension study based on other projects with CI are carried out. Through analysis on 132,006 merged PRs of 82 GitHub projects, I find that the results of Bernardo et al.’s work still stand, that CI could not always fasten delivery activities after CI. In addition, by comparing the statistical performance of projects that apply CI with those that never apply CI, it is proved that similar changes in development activities can only be observed in projects with CI adopted.
continuous integration , pull-based development , replication study , delivery delay , Travis-CI