Contributing to Eclipse Platform

Thank you for your interest in this project.

Developer Resources

The project issues and source code are maintained in multiple GitHub repositories:

• https://github.com/eclipse-platform/eclipse.platform
• https://github.com/eclipse-platform/eclipse.platform.swt
• https://github.com/eclipse-platform/eclipse.platform.ui
• https://github.com/eclipse-platform/eclipse.platform.releng.aggregator

Please search for existing issues before you create a new one.

Setting up Your Eclipse and GitHub Account

All contributors must have an Eclipse account with an associated Eclipse Contributor Agreement.

Details…

Create an Eclipse account if you don't already have one. See the "Eclipse Foundation Account" section in the Eclipse Committer Handbook.

All contributors who are not committers on this Eclipse project must electronically sign the Eclipse Contributor Agreement (ECA) via their Eclipse account. For more information, please see the Eclipse Committer Handbook.

To verify the Eclipse Contributor Agreement, GitHub contributions must use the same email address as your Eclipse account. If your GitHub account was registered with a different address, you can add your Eclipse email address to the account instead.

Also add your GitHub account name to your Eclipse account to enable automated management of access rights for Eclipse project members. You can do this by logging into your Eclipse account, choosing "Edit Profile", and add your GitHub account name in the social media links section.

If you are a committer in any of the Eclipse projects, you should receive an email containing an invite to join that project's GitHub organisation within 2 hours. You should accept the invite to maintain committer status in the GitHub organisation.

It is recommended to add SSH public keys to your GitHub account.

Creating an Eclipse Development Environment

You can set up a pre-configured IDE for development of the Eclipse SDK using the following link.

Recommended Workflow

Every change must be made by creating a pull request from a fork of the main project repository; see fork and pull model as well as working with forks in the GitHub documentation.

Creating and Managing a Fork

You will need a fork for each repository to which you wish to contribute.

Details…

Use the Fork button (top right) on the GitHub repository page and choose Create a new fork to create a forked repository under your GitHub account. Whenever changes are pushed to the main project repository, you can click Sync fork in the forked repository to get the latest source code from the main project repository.

Then you have two options:

1. Clone the main project repository and add the forked repository as an additional remote

   With this option, you can easily retrieve the most recent changes from the main project repository by simply pulling the master branch, as your default remote repository will be the main project.

   Clone the main repository to your Eclipse workspace or, if you use the Eclipse installer as explained in Creating an Eclipse Development Environment, the repository is already cloned that way. The repository URL can be found by clicking the <> Code button on the top right of the <> Code tab within the GitHub page for the main project repository. When you clone the repository, it will automatically be added as the "origin" remote to your local clone.

   You then have to add your fork as an additional remote to push your changes to a branch of that fork. To this end, add the forked repository as an additional remote called "fork" (or whatever you like):

   • Go to your fork on GitHub.
   • Click the <> Code button on the top right of the <> Code tab and copy the URL.
   • Add the URL as the "fork" remote to your local clone.

   Whenever the master branch of the main project repository changes, pulling the master branch of your local clone will give you the current state. In case you have created a branch and committed changes to it, you can rebase it onto the updated master branch.

2. Clone the forked repository and add the main repository as an additional remote

   With this option, you can easily create new branches for committed changes in your fork, as the default remote repository will be your forked repository.

   Clone the forked repository to your Eclipse workspace or, if you use the Eclipse installer as explained in Creating an Eclipse Development Environment, you have to change the URL for the repository during the setup. The repository URL can be found by clicking the <> Code button on the top right of the <> Code tab within the GitHub page for the forked repository. When you clone the forked repository, it will automatically be added as the "origin" remote to your local clone.

   You then have to add the main project repository as an additional remote. To this end, add the main project repository as an additional remote called "upstream":

   • Go to the main project repository on GitHub.
   • Click the <> Code button on the top right of the <> Code tab and copy the URL.
   • Add the URL as the "upstream" remote to your local clone.

   Whenever the master branch of the main project repository changes, you have to synchronize the master branch of your fork. Go to the GitHub page of your forked repository and click Sync fork. You can then pull the master branch of your local clone to retrieve the current state. In case you have created a branch and committed changes to it, you can rebase it onto the updated master branch.

Creating a Pull Request

You will need to create a pull request in order to contribute.

Details…

When you have set up a fork of a repository to which you wish contribute and prepared a local clone for it, you can make changes, commit them, and propose them via a pull request according to the standard "pull requests from a fork" workflow. In the following, the workflow is sketched with some specific information for the Eclipse platform projects.

1. Preparing the PR: Create a branch in your fork and apply the desired changes to it. Push the branch with the changes to your forked repository (not to the original upstream project repository).

   • Make multiple commits if necessary as history can be retained, but note that you will usually be asked to combine your changes to a single commit per pull request. So keep changes as small as possible and try to split them up.

2. Creating the PR: Open the page of the forked repository in your browser, switch to the branch where the fix is developed. Click the Contribute button to start the Open Pull Request workflow.

   • On the new page verify the target branch is the master branch of the main project repository.
   • Verify that the list of commits contains the changes you intend to merge.
   • If you are a committer, you can add reviewers if you want certain people to review a change. If you leave this empty, anyone from the project team will review.

3. Updating the PR: To update your pull request, you can push more commits to the same branch in your fork. They automatically get added to the pull request (and trigger a new round of builds and reviews). You can also amend your previous commits and force push them to your branch.

   • You will usually be asked to squash your commits to a single one before a PR will be merged.
   • Exception: when you bump versions in the Manifest/POM files, you could provide 2 separate commits in the same PR: one with the functional changes (i.e., in the code) and one with the version bump (i.e., in the Manifest/POM files). This is not a rule but, if necessary, it would help cherry-picking commits when merging your PR, and avoids problems if there is ever a need to revert the code changes.

4. Validating and verifying the PR:

   • Reviewers can review the pull request on the GitHub website or fetch the PR using the Fetch GitHub PR menu entry in the Git Repositories view inside the Eclipse IDE.
   • Every PR is automatically verified to check that contributors have a valid Eclipse Contributor Agreement (ECA).
   • What does a valid PR look like? (Checklist) These are the basic criteria that need to be fulfilled for a PR to be considered valid.
     • All checks need to pass.
       • If tests are broken, they should either be fixed or documented, i.e., create a new issue or reference an existing issue.
       • The code must compile.
       • The versions of the affected bundles should be properly bumped.
       • There should be no new API warnings. If a warning can’t be fixed, suppress it and write a comment explaining why.
       • If there are other (obscure) errors in the checks, please find them (read the logs) and actively ask what is happening and how to fix them.
     • The commit text should be well-written (see Commit Message Recommendations).
     • If tests were added, they should meet the following criteria:
       • They should run: check the logs and make sure they are being executed.
       • They should pass: do not add broken tests.
     • If the PR has been reviewed and the reviewer left some comments, address them and clearly document what you did. If you have simply done the obvious thing recommended by the reviewer, a "thumbs up" emoji is sufficient and helps avoid unnecessary notifications. Please resolve each review comment when it has been addressed by a follow up commit.

5. Merging the PR: Once the PR is approved, it can be merged by a committer.

   • One of two options can be selected, based on what fits best given these criteria:
     • Rebase and Merge: retains all commits separately; useful when developing a feature consisting of multiple independent commits.
     • Squash and Merge: all commits in the PR get squashed into a single commit; useful to avoid separate "fix comments" commits.
   • After the PR is merged, the source branch used for the PR in the fork can be deleted.
   • Your fork is now out-of-date with the main (upstream) project repository. In case the "origin" remote of your local clone is the forked repository, you should get it back in sync. You can either use the Sync Fork function on GitHub to update the fork and pull the master branch to your local clone, or pull the latest changes from "upstream" (the main project repository) into your local repository.

Using Draft Pull Requests

GitHub allows you to mark a pull request as a draft. Please describe the purpose and current state of the contribution.

What to do…

A draft PR indicates that:

• The changes are not yet ready for submission/merging, e.g., work is still in progress, tests are incomplete, or functionality is not fully working.
• You are seeking early feedback from the community on your approach, design decisions, or implementation. Important: Clearly state in the PR description what kind of feedback you are seeking, as draft PRs may be overlooked until marked as ready for review.
• You may need help or guidance to resolve specific issues or to complete the work.
• You want to share your progress with others who might be interested or affected by the changes.

What not to do…

A draft PR is not:

• A place to dump incomplete or unclear changes without context. Every PR, including drafts, should have a clear purpose and description.
• A shield against comments or reviews. Marking a PR as draft does not mean "leave me alone" – it means you welcome early feedback and collaboration.
• A substitute for local development or working in a private branch. If you're not ready to share your work or discuss it with others, keep it in your local repository or in a branch of your fork without creating a PR.

Best practices…

Even when marked as a draft, your PR should meet these basic requirements:

• Clear intent and description: Explain what you are trying to achieve, why the changes are being made, and what the current state is.
• Understandable changes: Others should be able to understand what you have done so far and what remains to be done. Consider using a task list in the description to track progress.
• Specific questions or requests: If you need help or feedback on particular aspects, clearly state what you need, e.g., "I'm unsure about the approach in XYZ.java" or "Tests for feature X are still missing".
• Reasonable completeness: While not everything needs to be finished, the PR should represent a coherent chunk of work that others can review and comment on meaningfully.

When your draft PR is ready for final review, convert it to a regular PR using the "Ready for review" button. At that point it should meet all the criteria from the checklist above.

Commit Message Recommendations

Please use a commit message with a concise summary, a description, and a link to an issue, if there is an associated issue.

Details…

<issue title>

Short description/summary as to why this change is being made

Fixes https://github.com/<Full URL to issues/issue #>

Example:

Make memory view cell editor visible on Linux GTK3

Since SWT changed from GTK2 to GTK3 as its backend the cell
editing in the memory view has not worked properly
with the cell editor not visible. This change uses explicit
setVisible rather than adjusting Z-order of the controls.

Fixes https://github.com/eclipse-platform/eclipse.platform.debug/issues/132

The issue title must be a concise summary of the change. A more detailed description is provided in the body. Please do not forget to add the issue URL to the commit message. This is used to link with the GitHub issue.

Using Generative AI for Contributions

Please review the Eclipse Project Handbook for details: Generative Artificial Intelligence Usage Guidelines for Eclipse Committers.

Using GitHub Copilot

Using the GitHub Copilot Agent can boost your productivity and provide interesting insights as it allows to fire up automated agents.

Details…

Due to current limitations, using this directly against the base repository has some drawbacks:

• Copilot cannot currently use a fork, so the branch is created directly on the repository which is against the recommended fork-based workflow.
• Copilot currently cannot squash and force push a branch.
• GitHub does not support filtering of individual notifications, due to the high cadence and the incremental work mode in some cases, this can easily flood the inbox of subscribers of the repository.
• If the task is not properly written or the problem is too complex, it can result in creating something off the track that needs to be discarded and then iterate again which results in more churn.

Because of this we recommend the following workflow:

• Let Copilot work on your fork. For this you should sync the fork which can be done through the web UI.
• Instead of assigning Copilot an issue, simply reference the issue in the prompt, e.g., Analyze and propose a fix for <link to the original issue>.
• If you are happy with the result or would like to further refine it yourself, just fetch the branch from your fork, squash all commits, remove unnecessary files and force push the result to a new branch, e.g., usually Copilot creates copilot/fix-for-issue-123 so just push it as fix-for-issue-123 and then open a PR against the repo as usual. This has the advantage that you can even ask Copilot to refine some things later and then just merge the results into your local branch by a simple fetch.
• Usually you want to leave Copilot as the author and you as the committer, unless you have significantly rewritten the results, e.g., if it was mostly a research task and you did the actual implementation.

Validating Changes and AI Feedback

Please respect the time and attention of the humans involved in the review process, and reduce the load on the shared build infrastructure that verifies each commit. Use your fork for drafting a pull request and for initial copilot reviews. Create a pull request only after you have personally verified the quality.

Details…

Typically a pull request should not be opened while the content is effectively a draft. It should be fully reviewed by you as a human author, and you should have high confidence that the work is complete and ready for the review of another human being. Content that is still being refined, or is still being reviewed in source form by automated (non-human) reviewers, should not be pushed to the upstream repository.

The goal is to reduce notifications to the repository followers and to avoid thrashing the shared build infrastructure. Every push to an upstream PR triggers the verification builds and notifies all subscribers. The number of concurrent jobs on GitHub-hosted runners is limited, so intermediate states queue up builds for everyone. Actions on standard runners are free for public repositories, so the motivation is less noise and particularly less shared ci.eclipse.org load. Therefore it is generally preferable to get AI feedback before you open the PR against the upstream repository:

1. Locally (preferred): Run the build on the command line and use the local review features of your AI tool, e.g., the GitHub Copilot CLI offers the /review command to run a code review agent on your local changes. Other tools provide similar features. This might have less coverage than the full build matrix, but it is a quick feedback loop. If something does not work locally, please let us know.
2. On your fork: Push to your fork to run the GitHub Actions workflows there. This does not run the Jenkins build on ci.eclipse.org but gives good initial confidence. Workflows don't run in forked repositories by default, you must enable GitHub Actions in the Actions tab of your fork.
3. Pull request on your fork: If you want to use web-integrated tools like Copilot code review, you can open a PR against your own fork. When you are satisfied, close it and open a new one against the upstream repository.
4. Combine them: Iterate locally first, then push to your fork, and open the PR against the upstream repository only at the end.

Command Line Builds

It is sometimes beneficial to build on the command line as this is how we also execute the verification builds. Moreover some tests might not run at all in the IDE if they require a very specific setup. If you want to produce an update site from your changes or want to have local build artifacts of your changes, you will need to do this on the command line.

Required tools…

We are using Apache Maven for building the sources, usually it is fine to just use the latest release you can use whatever install method that is sufficient for your system.

Running the build…

Most project repositories are already set up so that you can run mvn clean verify directly. One exception is the aggregator build that builds all projects together and requires special attention.

Tips and tricks…

• If your changes do not seem to be applied this is usually a sign that you forgot to increment the version of the bundle, or haven't committed (you don't need to push) your changes; in which case the baseline replace might come into effect. You will see a warning in the log like this [WARNING] <project name>: baseline and build artifacts have same version but different contents.
• If you just want to build one specific module, this is usually possible without special consideration. Just keep in mind that if the tests are separated from the module you have changed you need to build both of them or the whole reactor. For example you can run the tests and the bundle under test with the command mvn clean verify -pl :<name of the bundle under test>,:<name of the test bundle>; one specific test can be run by adding -Dtest=<fully qualified class name of test>.
• You can perform a quick run to check everything compiles with -DskipTests in combination with -T1C — this can really speed things up; just keep in mind that the tests are not executed.
• If you want to make sure you didn't break the API you can specify -Papi-check.
• If you want to check the javadoc (JDT currently not find all issues in the IDE) you can specify -Pjavadoc additionally -DfailOnJavadocErrors=true will even fail the build, you can even try to fix them automatically by adding 'javadoc:fix' to your command

Contact

• For official announcements, use the project's "dev" mailing list.
• To ask for help, either use the current issue/PR or start a GitHub Discussion.