The current method used in our company is to have a long-term release branch, which we all develop, with developers checking any incompatible test code both on the branch and on the external line, while only the verification code is checked on the branch. Then, as soon as we exit, the winning test cells merge into the trunk, while most of the ab test codes are thrown away. This has been working for us for a long time, but we are starting a new project, and I wonder if we can improve our workflow with this new project. Everyone who checks both the branch and the trunk has pain. Ideally, we would have a more standard approach to developing against the trunk and branching for stabilization, but that would mean polluting our trunk with the ab test code.
source
share