Rapid Releases and Patch Backouts: A Software Analytics Approach

Abstract

Mozilla's decision to release a new version of its products every six weeks (instead of every year) profoundly affected developers and users and was accompanied by significant changes in the release process. Were such changes enough to allow Mozilla to move faster without breaking things? What lessons can be learned from Mozilla's adoption of rapid releases? To answer these questions, researchers analyzed tens of thousands of commits and bug reports from Firefox and talked to its developers. The results show that, because of integration repositories, build sheriffs, and better testing tools, broken patches were backed out (reverted) earlier, rendering the release process more stable.