Branching and merging codeΒΆ

Following the git flow structure, repositories contain two special branches: the main branch which contains released versions of the code, and the develop branch which will contain all developments to be included in future releases. The branching model also provides three categories for other branches: feature branches where most developments happen, release branches for preparing new code releases and hotfix branches for bug fixes in already released code. This is different from older common working practices where main and develop are in the same branch (trunk) which is also where preparation of releases happen and where bugs are fixed.

A typical development will start by creating a feature branch (named feature/great_new_stuff) where the development work will happen. In most cases, this branch is created from the develop branch. Once the feature has been developed and tested, the new code is merged back in the develop branch.

As work happens in parallel, other developments might have been merged into the develop branch in the meantime. It is in principle the responsibility of the developer to first merge the develop branch into the feature branch and resolve conflicts. If the conflicts are simple to resolve this will be the case. If the conflicts are more complex to resolve, this is done in collaboration with the developer who introduced the conflicting change, or as a last recourse, by gatekeepers of the overall code.

For early detection of conflicts and easier resolution, modern practice recommend frequent merges of feature branches into the develop branch and frequent merges of the develop branch into the feature branch as work progresses. It is much more efficient to fix several small conflicts one after the other than large conflicts coming from several independent developments all at once after long periods of disconnect work. Early detection encourages discussions between developers to plan merging of future code changes before merges become too difficult.

Modern software design will also help in reducing conflicts in the development and merging process (explained below).

At some point, it will become necessary to prepare releases of the code. This process happens in specific releasebranches. Once a release is ready and fully tested it is merged into both the main and develop branches and tagged. Typically, the tag will contain the release number (for example 2.0 for a major release, or 2.1 for a minor one). The main branch then contains the new official release and development work continues in the develop branch based on the release.

Despite all the care being taken in the testing, there will always be (hopefully rare) bugs needing fixing in any software. When such a bug is detected in a release, a hotfix branch is created from the main branch to implement and test the fix. Once a satisfactory fix is implemented, the hot fix branch is merged into both the main and develop branches and the release number is incremented (a new tag is created with a minor version number increased, for example from 2.1.0 to 2.1.1). For reproducibility, an existing tag should never be moved.

See the guide for getting started with git-flow for more details.