Trunk-based pull-request GitHub workflow

Carbon repositories follow a few basic principles:

• Development directly on the trunk branch and revert to green.
• Always use pull requests, rather than pushing directly.
• Changes should be small, incremental, and review-optimized.
• Preserve linear history by rebasing or squashing pull requests rather than using unsquashed merge commits.

These principles try to optimize for several different uses or activities with version control:

• Continuous integration and bisection to identify failures and revert to green.
• Code review both at the time of commit and follow-up review after commit.
• Understanding how things evolve over time, which can manifest in different ways:
  • When were things introduced?
  • How does the main branch and project evolve over time?
  • How was a bug or surprising thing introduced?

Note that this document focuses on the mechanical workflow and branch management. Details of the code review process are in their own document.

Trunk based development

We work in a simple trunk-based development model. This means all development activity takes place on a single common trunk branch in the repository (our default branch). We focus on small, incremental changes rather than feature branches or the "scaled" variations of this workflow.

Green tests

The trunk branch should always stay "green". That means that if tests fail or if we discover bugs or errors, we revert to a "green" state by default, where the failure or bug is no longer present. Fixing forward is fine if that will be comparably fast and efficient. The goal isn't to dogmatically avoid fixing forward, but to prioritize getting back to green quickly. We hope to eventually tool this through automatic continuous-integration powered submit queues, but even those can fail and the principle remains.

Always use pull requests (with review) rather than pushing directly

We want to ensure that changes to Carbon are always reviewed, and the simplest way to do this is to consistently follow a pull request workflow. Even if the change seems trivial, still go through a pull request -- it'll likely be trivial to review. Always wait for someone else to review your pull request rather than just merging it, even if you have permission to do so.

Our GitHub repositories are configured to require pull requests and review before they are merged, so this rule is enforced automatically.

Small, incremental changes

Developing in small, incremental changes improves code review time, continuous integration, and bisection. This means we typically squash pull requests into a single commit when landing. We use two fundamental guides for deciding how to split up pull requests:

1. Ensure that each pull request builds and passes any tests cleanly when you request review and when it lands. This will ensure bisection and continuous integration can effectively process them.

2. Without violating the first point, try to get each pull request to be "just right": not too big, not too small. You don't want to separate a pattern of tightly related changes into separate requests when they're easier to review as a set or batch, and you don't want to bundle unrelated changes together. Typically you should try to keep the pull request as small as you can without breaking apart tightly coupled changes. However, listen to your code reviewer if they ask to split things up or combine them.

While the default is to squash pull requests into a single commit, during the review you typically want to leave the development history undisturbed until the end so that comments on any particular increment aren't lost. We typically use the GitHub squash-and-merge functionality to land things.

Stacking dependent pull requests

Carbon uses pull requests in the common, distributed GitHub model where you first fork the repository, typically into your own private GitHub fork, and then develop on feature branches in that fork. When a branch is ready for review, it is turned into a pull request against the official repository. This flow should always be where you start when contributing to Carbon, and it scales well even with many independent changes in flight.

However, a common limitation to hit is when you want to create a stack of dependent, small, and incremental changes and allow them to be reviewed in parallel. Each of these should be its own pull request to facilitate our desire for small and incremental changes and review. Unfortunately, GitHub has very poor support for managing the review of these stacked pull requests. Specifically, one pull request cannot serve as the base for another pull request, so each pull request will include all of the commits and diffs of the preceding pull requests in the stack.

We suggest a specific workflow to address this (note, commit access is required):

1. Create your initial pull request from a branch of your fork, nothing special is needed at this step. Let's say you have a branch feature-basic in your clone of your fork, and that the origin remote is your fork.

   Push the branch to your fork:

   git checkout feature-basic
   git push origin

   And create a pull request for it using the gh tool:

   gh pr create

   Let's imagine this creates a pull request N in the upstream repository.

2. If you end up needing to create a subsequent pull request based on the first one, we need to create a branch in the upstream repository that tracks the first pull request and serves as the base for the subsequent pull request. Assuming your fork $USER/carbon-lang is remote origin and carbon-language/carbon-lang is remote upstream in your repository:

   git checkout feature-basic
   git push upstream HEAD:pull-N-feature-basic

   Everyone marked as a contributor to Carbon is allowed to push branches if the name matches pull-*, skipping pull request review processes. They can be force pushed as necessary and deleted. These branch names should only be used for this ephemeral purpose. All other branch names are protected.

3. Create your stacked branch on your fork:

   git checkout -b next-feature-extension
   git commit -a -m 'Some initial work on the next feature.'
   git push origin

4. Create the pull request using the upstream branch tracking your prior pull request as the base:

   gh pr create --base pull-N-feature-basic

   This creates a baseline for the new, stacked pull request that you have manually synced to your prior pull request.

5. Each time you update the original pull request by pushing more commits to the feature-basic branch on your origin, you'll want to re-push to the upstream tracking branch as well:

   git checkout feature-basic
   git commit -a -m 'Address some code review feedback...'
   git push
   git push upstream HEAD:pull-N-feature-basic

   Then merge those changes into your subsequent pull request:

   git checkout next-feature-extension
   git merge feature-basic
   git push

   The merge will prevent disrupting the history of next-feature-extension where you may have code review comments on specific commits, while still allowing the pull request diff view to show the new delta after incorporating the new baseline.

6. Follow a similar process as in 5 above for merging updates from the main branch of upstream:

   git checkout trunk
   git pull --rebase upstream
   
   # Update your fork (optional).
   git push
   
   # Merge changes from upstream into your branch without disrupting history.
   git checkout feature-basic
   git merge trunk
   # Push to the first PR on your fork.
   git push
   # Synchronize the upstream tracking branch for the first PR.
   git push upstream HEAD:pull-N-feature-basic
   
   # Merge changes from the first PR (now including changes from trunk)
   # without disrupting history.
   git checkout next-feature-extension
   git merge feature-basic
   # And push to the second PR on your fork.
   git push

7. When the first pull request lands in the main upstream branch, merge those changes from upstream trunk into the stacked branch:

   # Pick up the first PR's changes from upstream trunk.
   git checkout trunk
   git pull --rebase upstream
   
   # Merge those changes into the stacked PR branch.
   git checkout next-feature-extension
   git merge trunk
   git push

   Then update the stacked PR's base branch to be carbon-language:trunk rather than the upstream tracking branch. To do this, go to the page for the PR on GitHub, click the "Edit" button to the right of the PR title, and then select trunk from the "base" drop-down box below the PR title.

   Once that's done, delete the upstream tracking branch:

   git push upstream --delete pull-N-feature-basic

8. When landing the second, stacked pull request, it will require actively rebasing or squashing due to the complex merge history used while updating.

Additional notes:

• If you need to create a third or more stacked pull requests, simply repeat the steps starting from #2 above for each pull request in the stack, but starting from the prior pull request's branch.

• If you want to split the two pull requests so they become independent, you can explicitly edit the base branch of a pull request in the GitHub UI. The result will be two pull requests with an overlapping initial sequence of commits. You can then restructure each one to make sense independently.

Managing pull requests with multiple commits

Sometimes, it will make sense to land a series of separate commits for a single pull request through rebasing. This can happen when there is important overarching context that should feed into the review, but the changes can be usefully decomposed when landing them. When following this model, each commit you intend to end up on the trunk branch needs to follow the same fundamental rules as the pull request above: they should each build and pass tests when landed in order, and they should have well written, cohesive commit messages.

Prior to landing the pull request, you are expected to rebase it to produce this final commit sequence, either interactively or not. This kind of rebase rewrites the history in Git, which can make it hard to track the resolution of code review comments. Typically, only do this as a cleanup step when the review has finished, or when it won't otherwise disrupt code review. It is healthy and expected to add "addressing review comments" commits during the review and then squashing them away before the pull request is merged.

Linear history

We want the history of the trunk branch of each repository to be as simple and easy to understand as possible. While Git has strong support for managing complex history and merge patterns, we find understanding and reasoning about the history -- especially for humans -- to be at least somewhat simplified by sticking to a linear progression. As a consequence, we either squash pull requests or rebase them when merging them.