Skip to content

Latest commit

 

History

History
116 lines (73 loc) · 4.08 KB

README.md

File metadata and controls

116 lines (73 loc) · 4.08 KB

Automerger

Automatically merge changes from trunk (usually master) into feature branches with PRs on GitHub.

Using

If you don't want to set your server up, here's the companion app hosted by me.

If you want to run your own version, see the running section below.

Why

At @verloop, we use feature branches. Whenever a feature would get merged into master, we would reach one of two states.

  1. All our feature branches would lag behind.

  2. Someone would have to hand merge changes from master into each feature branch.

If feature branch lagged behind, QA would raise bugs with the title "feature X stopped working on dev/staging".

Needless bugs/issues raised, wasted time, you get the idea.

Hand merging could be automated, so I did that.

For more, scroll to FAQ

How it works?

Whenever a push is made to the trunk, automerger makes pull requests to all other branches and tries to merge the PRs.

If it fails to do so, it assigns to the pusher of the commit (configurable) for human intervention.

Running

You can use the executable from releases, or use the ofpiyush/automerger docker image.

You can use path / or /automerger for hosting the app. At the moment the http server makes no assumptions about the path it runs on, but it might add paths like /stats or /metrics and /ping in later versions.

Security comes first, so automerger refuses to run without a webhook secret and verifies signatures of all incoming events.

--secret is the only required flag. It must match the secret you provided while setting up the webhook

Automerger can be configured to run in many ways with flags.

Run with your github app

Use --integration-id flag to specify your app's id.

Run as a user

You can use a token and add a webhook to your repo. --token=XXXX

Set the webhook with the push event.

Track a different branch

Use the --branch option to configure the trunk.

All options

Usage of automerger:
  -address string
        Address to listen on. (default "0.0.0.0:3000")
  -api-url string
        URL of github installation. (default "https://api.github.com")
  -assignee value
        Assignees if PR merge fails
        Defaults to the pusher.
  -branch string
        Branch to start PRs from. (default "master")
  -integration-id string
        Github Integration's id. (default "9441")
  -key-file string
        Full path to the Github Integration's private key.
        Either this or token should be present. (default "private_key.pem")
  -merge-method string
        How to merge changes on default branch to current branch. (default "merge")
  -secret string
        Github Integration's secret.
  -token string
        Usable token with access to repo.
        Either this or the key file should be present.

Contributing

Contributions are very welcome and super encouraged.

Two requests:

  • If the change is big, open a proposal issue to discuss first.

  • Be nice to other people.

Interesting potential next steps and opportunities to contribute

  • Abstract the implementation enough to add support for Bitbucket, GitLab etc.

  • Add support for AWS lambda so that hosting this can be cheap/free.

  • Add structured logging

  • Add metrics endpoint

FAQ

Why do you need to keep feature branches updated?

Because we show our feature branches to the QA.

Why don't you use a develop branch and merge features into it first?

Using a develop branch inevitably leads to a cycle merge to develop -> revert to avoid blocking releases -> make mistake -> land buggy feature in production

If all changes from master were always merged into feature branches, every branch is master + feature. It is exactly like the develop branch, but without the scope for landing a bad feature in production.

License

Released under Unlicense. Check the LICENSE for more details.

No warranties and public domain for the source code, companion app and hosted version.