Skip to main content
When a feature changes the UI, its test changes live on a test-repository branch with the same name as the application branch (see Branching). If the application pull request merges first, runs on your default branch use tests that don’t know about the feature yet, and they fail. The linked test changes reporter (github_linked_changes) tracks that test branch for the application pull request a run was triggered for, and keeps a check named Empirical linked test changes on the pull request up to date. The check fails while the test branch has commits that are not merged into the test repository’s default branch, and passes once there is nothing left to merge.

​
Results

The check links to the latest test run for the pull request.

​
When it updates

The check is written on the pull request’s current head commit:
  • when a test run is triggered for the pull request, so every new commit that triggers a run gets a check, including pull requests without any test changes
  • when the test branch changes: a push, a test PR opened, closed or merged, or the branch deleted

​
Set up

  1. Install the GitHub Reporter on your application repository, with access to checks.
  2. Trigger runs for pull requests with trigger.event: "pull_request" and a build with the repository and commit. The dispatch-action does both on pull_request workflow events; for API calls see Pull request runs.
  3. Keep github_linked_changes in the environment’s reporters. Installing the GitHub Reporter turns it on for your environments, with the other GitHub reporters. Turn it off under Settings → Environments, or leave it out of the reporters of a trigger request.
A run with github_linked_changes asks Empirical to track its pull request’s test branch. The pull request’s latest run decides: if that run does not have the reporter, the check stops updating. Runs not triggered by a pull request have no test branch to track, and the reporter does nothing for them.

​
Making it a required check

Every pull request that triggers a test run gets the check, and it passes when there are no test changes, so you can require it in your branch protection rules or rulesets.
A pull request that never triggers a test run, for example because your workflow skips it with a paths filter, gets no check. A required check then waits forever. Require the check only if every pull request to the protected branch triggers a run.

​
Merging test PRs

Merge test PRs with a merge commit, as the Empirical dashboard and agents do. After a squash or rebase merge in GitHub, the test branch still looks unmerged and the check keeps failing. Delete the test branch after such a merge to pass the check.