What is pull request testing?
Pull request testing runs checks against the exact revision in a pull request and reports the results before anyone merges. Reviewers see what was tested, what passed or failed and what remains unchecked.
Start with the requirement and the revision under review.
Record the base and head of the proposed change, then state what must still be true after it merges. In the owned delivery fixture, the rule is free delivery from 5000p. Testing only 4999p and 5001p misses a > boundary defect. A test at exactly 5000p connects the requirement to a failure a reviewer can reproduce.
// Requirement: delivery is free from £50.
assert.equal(deliveryPence(5000), 0);Run checks and preserve their native evidence.
The downloadable example exercises this requirement with Node tests, fast-check, Cucumber and Stryker. Its verification command deliberately expects the negative cases to fail and the corrected cases to pass. The archive contains the source, package lockfile and recorded native reports. These executions are local teaching material, not a run on GitHub or Cue April’s hosted product. The separate product-engine evidence shows actual Cue April engines executing an owned invoice fixture locally, with revision identity and native outputs.
node --version
npm ci --ignore-scripts
npm run verifyKnown bad boundary: fast-check counterexample [5000]
Gherkin: 1 failed scenario before the fix, 0 after
Boundary mutation: survived weak tests, killed by boundary assertion
Corrected property: 1000 generated cases passedMake the review record easy to inspect.
A PR comment or check summary should identify what ran and point to evidence, so a reviewer does not have to reconstruct the run from a green icon. Do not copy private test source or credentials into a public discussion. Keep artifacts accessible only to the people who should see them.
- Name the tested commit and whether it is the PR head or a test merge commit.
- Link the requirement and selected files, methods and input scope.
- Report passing, failing, blocked and unfinished work separately.
- Attach native outputs, counterexamples and reproduction commands where available.
- Identify what changed after a fix and retain the previous failure.
Check that success still applies to the latest change.
A successful run on an earlier commit does not establish the state of a new one. GitHub explains that required checks must pass for the relevant latest revision; where a test merge commit has status, that merge result takes precedence over head-only status. Read GitHub’s status-check guidance. A skipped or neutral check can satisfy some GitHub rules, so reviewers should still inspect what actually executed.
Use this checklist at the merge decision.
The reviewer decides whether the evidence addresses the risk in the change. Repository rules govern what can merge, and a test report does not replace those rules.
- The evidence names the revision GitHub is currently evaluating.
- The agreed behaviour includes important boundaries and denied actions.
- The tests were actually executed in the intended environment.
- Failures have a reproduction path and incomplete checks remain visible.
- The regression assertion fails against the known defect and passes after the fix.
- Changes since the last run have been checked, including changes to test configuration.
- An authorised reviewer has considered the remaining risk.
Bring testing evidence into your Cue April workflow.
Cue April’s implemented GitHub integration can publish an App-owned evidence comment for an approved PR run, linking the tested revision to protected results. Setup confirms repository permissions, supported suites, qualified execution and agreed scope. Watched changes can reuse standing approved intent where configured; your repository rules and reviewers retain the merge decision. Read the workflow or compare evaluation options.
Primary documentation
Framework behaviour and configuration are documented by their maintainers.