Coder powers secure, scalable development across key industries — automotive, finance, government, and technology — enabling faster builds, tighter compliance, and seamless AI adoption in enterprise-grade cloud environments.
The workflow used by code-server maintainers aims to be easy to understood by
the community and easy enough for new maintainers to jump in and start
contributing on day one.
Milestones
We operate mainly using
milestones. This was heavily
inspired by our friends over at vscode.
Here are the milestones we use and how we use them:
"Backlog" -> Work not yet planned for a specific release.
"On Deck" -> Work under consideration for upcoming milestones.
"Backlog Candidates" -> Work that is not yet accepted for the backlog. We wait
for the community to weigh in.
"<0.0.0>" -> Work to be done for a specific version.
With this flow, any un-assigned issues are essentially in triage state. Once
triaged, issues are either "Backlog" or "Backlog Candidates". They will
eventually move to "On Deck" (or be closed). Lastly, they will end up on a
version milestone where they will be worked on.
Triage
We use the following process for triaging GitHub issues:
Create an issue
Add appropriate labels to the issue (including "needs-investigation" if we
should look into it further)
Add the issue to a milestone
If it should be fixed soon, add to version milestone or "On Deck"
If not urgent, add to "Backlog"
Otherwise, add to "Backlog Candidate" for future consideration
Project boards
We use project boards for projects or goals that span multiple milestones.
Think of this as a place to put miscellaneous things (like testing, clean up
stuff, etc). As a maintainer, random tasks may come up here and there. The
project boards give you places to add temporary notes before opening a new
issue. Given that our release milestones function off of issues, we believe
tasks should have dedicated issues.
Project boards also give us a way to separate the issue triage from
bigger-picture, long-term work.
Versioning
<major.minor.patch>
The code-server project follows traditional semantic
versioning, with the objective of minimizing major changes
that break backward compatibility. We increment the patch level for all
releases, except when the upstream Visual Studio Code project increments its
minor version or we change the plugin API in a backward-compatible manner. In
those cases, we increment the minor version rather than the patch level.
Pull requests
Ideally, every PR should fix an issue. If it doesn't, make sure it's associated
with a version milestone.
If a PR does fix an issue, don't add it to the version milestone. Otherwise, the
version milestone will have duplicate information: the issue and the PR fixing
the issue.
Merge strategies
For most things, we recommend the squash and merge strategy. There
may be times where creating a merge commit makes sense as well. Use your
best judgment. If you're unsure, you can always discuss in the PR with the team.
Changelog
To save time when creating a new release for code-server, we keep a running
changelog at CHANGELOG.md.
If either the author or reviewer of a PR believes the change should be mentioned
in the changelog, then it should be added.
If there is not a Next Version when you modify CHANGELOG.md, please add it
using the template you see near the top of the changelog.
When writing your changelog item, ask yourself:
How do these changes affect code-server users?
What actions do they need to take (if any)?
If you need inspiration, we suggest looking at the Emacs
changelog.
Releases
With each release, we rotate the role of release manager to ensure every
maintainer goes through the process. This helps us keep documentation up-to-date
and encourages us to continually review and improve the flow.
If you're the current release manager, follow these steps:
After release is published, close release milestone
Publishing a release
Create a new branch called release/v0.0.0 (replace 0s with actual version aka v4.5.0)
If you don't do this, the npm-brew GitHub workflow will fail. It looks for the release artifacts under the branch pattern.
Run yarn release:prep and type in the new version (e.g., 3.8.1)
GitHub Actions will generate the npm-package, release-packages and
release-images artifacts. You do not have to wait for this step to complete
before proceeding.
Run yarn release:github-draft to create a GitHub draft release from the
template with the updated version. Make sure to update the CHANGELOG.md.
Bump chart version in Chart.yaml.
Summarize the major changes in the release notes and link to the relevant
issues.
Change the @ to target the version branch. Example: v3.9.0 @ Target: v3.9.0
Wait for the npm-package, release-packages and release-images artifacts
to build.
Run yarn release:github-assets to download the release-packages artifact.
They will upload them to the draft release.
Run some basic sanity tests on one of the released packages (pay special
attention to making sure the terminal works).
Publish the release and merge the PR. CI will automatically grab the
artifacts, publish the NPM package from npm-package, and publish the Docker
Hub image from release-images.
Update the AUR package. Instructions for updating the AUR package are at
coder/code-server-aur.
Wait for the npm package to be published.
AUR
We publish to AUR as a package here. This process is manual and can be done by following the steps in this repo.
Docker
We publish code-server as a Docker image here, tagging it both with the version and latest.
This is currently automated with the release process.
We hope to eventually hit 100% test coverage with our unit tests, and maybe one day our scripts (coverage not tracked currently).
If you're ever looking to add more tests, here are a few ways to get started:
run yarn test:unit and look at the coverage chart. You'll see all the uncovered lines. This is a good place to start.
look at test/scripts to see which scripts are tested. We can always use more tests there.
look at test/e2e. We can always use more end-to-end tests.
Otherwise, talk to a current maintainer and ask which part of the codebase is lacking most when it comes to tests.
Documentation
Troubleshooting
Our docs are hosted on Vercel. Vercel only shows logs in realtime, which means you need to have the logs open in one tab and reproduce your error in another tab. Since our logs are private to Coder the organization, you can only follow these steps if you're a Coder employee. Ask a maintainer for help if you need it.
Taking a real scenario, let's say you wanted to troubleshoot this docs change. Here is how you would do it:
