QATrax
View as Markdown

Test results format

JUnit XML, NUnit, TRX, TestNG, Robot Framework, CTest, Cucumber, Behave, Allure, CTRF, TAP or JSON — the common rule, what each format must carry, and templates to download.


QATrax does not run your tests: it receives what your continuous integration has run, and links it to your requirements. It reads the report as your tool writes it, recognised by its content — no converter to write, no tool to change. These are the formats also read by Xray, Allure TestOps, TestRail, Zephyr, ReportPortal and Azure DevOps: your report remains yours.

formatproduced byto upload
JUnit XMLpytest (--junitxml), Maven Surefire, Gradle, Jest (jest-junit), Go (gotestsum), GitLab, Jenkins, mjs-unit --junitthe .xml file — the most common
NUnit 2 and 3nunit3-console, NUnit3TestAdapterTestResult.xml
TRXdotnet test --logger trx (xUnit, MSTest, NUnit)the .trx file
TestNGTestNG (Maven, Gradle)testng-results.xml
Robot Frameworkrobot (6 and 7)output.xml
CTestctest -T Test (C, C++)Testing/<date>/Test.xml
Cucumber JSONcucumber-js, Cucumber-JVM, Cucumber Ruby (--format json)the .json file
Behave JSONbehave -f json -o behave.jsonthe .json file
Allure 2any language (allure-results/)one -result.json, or all of them combined: jq -s . allure-results/-result.json > allure.json
CTRFthe ctrf-io reporters, mjs-unit --ctrfthe .json file
TAPnode --test, prove, pytest-tapthe simplest to write by hand
JSONmjs-unit --json, or your own scriptwhen you build the report yourself

A status that is neither passed, nor failed, nor skipped — Inconclusive (NUnit, TRX), an undefined or pending step (Cucumber, Behave), unknown (Allure), notrun for a missing executable (CTest) — is counted as blocked: never as a success. A failed TestNG configuration method is shown, under its name. Allure: a test that was re-run (same historyId) counts only once, the last one.

The common rule: the test reference at the start of the name

Your plan gives each test a reference (T-MOD-01) and an autoRef: <test file>::<reference>. The name of each test, in the report, starts with this reference, followed by a separator (·, —, : or -):

T-MOD-01 · aucune vulnérabilité connue non traitée

Written with _ — as pytest and JUnit require, since they refuse the hyphen in a function name — it is recognised too: test_T_MOD_01_aucune_vulnerabilite.

In BDD (Cucumber, Behave), with Robot Framework or Allure, a tag equal to the reference is enough: @T-MOD-01 on the scenario. The name, if it carries a reference, takes precedence.

A test whose reference is not known to the plan is reported as outside the plan: it is shown, it proves nothing.

JUnit XML

Download the template: resultats-modele.xml.

With mjs-unit (0.8.0 and later): mjs-unit test-scripts/*.test.mjs --junit=junit.xml.

TAP

Download the template: resultats-modele.tap.

ok N - <name> passed, not ok N - <name> failed, # SKIP or # TODO not run. The file is not required: QATrax finds it in the plan.

JSON

Download the template: resultats-modele.json.

{
  "title": "Produit modèle — campagne de sécurité",
  "generatedAt": "2026-10-03T09:12:00.000Z",
  "suites": [
    { "file": "test/acces.test.mjs",
      "results": [
        { "name": "T-MOD-03 · un appel sans jeton est refusé (401)", "status": "pass", "durationMs": 8.3 },
        { "name": "T-MOD-04 · un jeton expiré est refusé (401)", "status": "fail", "error": "attendu 401, reçu 200" }
      ] }
  ]
}
fieldrequiredrule
suites[].fileyesthe test file, as in the autoRef; if it differs, the reference finds it when it exists in only one file
results[].nameyesthe reference first
results[].statusyespass, fail, skip (also passed, failed, error, skipped, todo); any other word becomes "blocked", never "passed"
error, durationMsnothe failure message, the duration in milliseconds
generatedAtrecommendedidentifies the run: a new date at each run, otherwise the upload is recognised as already received

Manual tests

Document review, inspection, bench test: many CRA requirements cannot be proven by the CI. Declare the case manual in the plan ("type": "manual", with its steps), then enter its result in the Manual tests tab: passed, failed or blocked, the date and place of the run, the observation (required), and attachments (report, photo; 5 MB at most each).

Cancel an erroneous upload

A report pushed by mistake (wrong branch, wrong project) is cancelled from the "Runs" tab, with a reason: it remains in the history, marked cancelled (by whom, when, why), but no longer counts for the verdict. It is restored the same way. An erroneous manual entry is cancelled in the same way. Deliverables already issued keep what they sealed; the next deliverable states the cancelled uploads.

The plan

Download the template: plan-modele.json — three CRA requirements, four tests. Upload it ("Requirements" tab), then one of the three results templates: the four tests are linked, and one of them fails — you see what becomes of a failing requirement, up to the verdict.

Upload

The response states the recognised format (dialecte), the tests linked, those outside the plan and those of the plan never run.