Test Regression Report

Table of Contents

1. The baseline
2. Where it shows up
3. Blame: which commit broke it
4. Relation to the quality gate

The regression report answers one question for a run: what changed versus the baseline? It diffs the run’s per-test outcomes against the latest run on the base branch and splits the result into three buckets:

New failures — tests failing now that were not failing in the baseline. These are the regressions this run introduced.
New passes (fixed) — tests that were failing in the baseline and now pass.
Still failing — tests failing in both, i.e. pre-existing failures carried over.

Where the quality gate gives a single pass/fail verdict, the regression report gives the itemised diff behind it.

1. The baseline

The baseline is the same one the quality gate uses: the most recent run on the configured base branch (unitrack.quality-gate.base-branch, default main) for the same project and the same coverage flag, created no later than the run being viewed. If there is no prior run to compare against, the report says so and lists the run’s current failures as "new".

Tests are matched by class name + test name, so renaming a test reads as one removed and one added.

2. Where it shows up

Run page — a "Regression vs baseline" panel above the failures list, summarising N new · M fixed · K still failing with a link to the baseline run, and listing the new failures (with their failure type) and the fixed tests.
REST — GET /api/v1/runs/{id}/regression:

{
  "baselineFound": true,
  "baselineRunId": 41,
  "baseBranch": "main",
  "newFailures": [
    { "className": "com.x.G", "name": "c", "failureType": "java.lang.AssertionError", "failureMessage": "boom" }
  ],
  "newPasses": [
    { "className": "com.x.G", "name": "a", "failureType": null, "failureMessage": null }
  ],
  "stillFailing": []
}

404 if the run does not exist; baselineFound is false (and baselineRunId omitted) when there is no prior run on the base branch.

3. Blame: which commit broke it

On the run page, each failing test carries a first failed: <sha> link pointing at the run where its current failing streak began on the same branch — the first failure since the test was last green. It answers "which commit broke this?" without leaving the dashboard. The same data is available at GET /api/v1/runs/{id}/blame:

[ { "className": "com.x.G", "name": "a", "firstFailingRunId": 41,
    "firstFailingCommit": "c2broke", "firstFailingShortSha": "c2broke" } ]

This uses only the stored run history (commit SHAs) — UniTrack does not read git diffs.

4. Relation to the quality gate

The quality gate’s new-failures rule excludes quarantined flaky tests, because its job is to decide whether to fail the build. The regression report is purely informational and shows the raw diff — including quarantined tests — so you see everything that actually changed.