AnyUnit.Report 1.2.2

AnyUnit.Report

Standalone .NET tool (anyunit-report) that converts an AnyUnit JSON results file (the -o/-output <file> output of anyunit-runner run, or any other AnyUnit runner) into a report format your CI system, IDE, or dashboard already understands: JUnit XML, TRX, NUnit3 XML, xUnit2 XML, or CTRF JSON - or into a self-contained HTML report built around AnyUnit's multi-platform results (see HTML), or a GitHub Actions job summary (see Markdown).

Install

dotnet tool install --global AnyUnit.Report

(Prerelease builds of unreleased work are published as artifacts of the repo's own pack workflow run, or build from source: dotnet pack Report/AnyUnit.Report.csproj.)

Usage

anyunit-report convert -f|-format <junit|trx|nunit|xunit|ctrf|html|markdown> -o|-output <file> <results.json> [<results2.json> ...]

One format per run - to produce more than one, run it more than once against the same results.json. Example:

anyunit-runner run -o results.json MyTests.dll
anyunit-report convert -f junit -o results.junit.xml results.json
anyunit-report convert -f trx -o results.trx results.json

A single results.json already represents multiple platforms fine on its own (the same test run under more than one platform becomes one <testcase>/<test> per platform, disambiguated as "TestName [platform]") - that's the normal case, no extra step needed. Passing more than one results.json (e.g. one per runner, the way .github/scripts/run-tests.sh in this repo produces them) merges them into that same shape first, so a separate net10/net48/browser-wasm run of the same assemblies converts into one report:

anyunit-report convert -f junit -o results.junit.xml results-net10.json results-net48.json

HTML report

-f html is the one format here that isn't an existing external schema, so it's the only one free to be shaped around what actually makes AnyUnit's results different: the same test run under several platforms at once.

The other five formats all assume one result per test, so a multi-platform test has to be flattened into N separate "TestName [platform]" test cases - the honest mapping into those schemas, but it scatters one test's story across N unrelated-looking rows. The HTML report keeps platform as an axis instead:

• A test × platform matrix - one row per test, one column per platform - so "passes everywhere except net48-win-x86" is a glance rather than a diff of two reports.
• Gaps are visible. A test × platform combination with no result at all (a platform that never ran that assembly) renders as · rather than being silently omitted, which is the other thing the flattened formats can't express.
• Per-platform environment cards - runtime, OS and architecture as actually captured per result, not one global guess.
• Filtering by name, failures-only, and per-platform column toggles; fixtures with failures start expanded, green ones start collapsed.

It's a single self-contained file - inline CSS/JS, no external references, no network - so it works straight out of a CI artifact zip over file://. It's responsive down to phone widths, follows the reader's light/dark preference, and marks outcomes with a glyph as well as a colour so the matrix stays readable with a colour vision deficiency. The page renders completely with JavaScript disabled (the script only adds filtering and collapsing).

anyunit-report convert -f html -o report.html results-*.json

Markdown summary

-f markdown emits GitHub Flavored Markdown sized for a GitHub Actions job summary

• the run's headline numbers, a per-platform table, and the failing tests with an output excerpt each.

It's the companion to -f html, not a replacement: HTML is exhaustive and lives in an artifact you download, Markdown is the bounded "what happened, do I need to look?" view that renders straight into the Actions UI with nothing to click through.

In a workflow

- name: Run tests
  run: dotnet anyunit-runner.dll run -o results.json MyTests.dll

- name: Post test summary
  if: always()
  run: |
    anyunit-report convert -f markdown -o summary.md results.json
    cat summary.md >> "$GITHUB_STEP_SUMMARY"

if: always() matters: a failing test run is exactly when you want the summary, and without it the step is skipped on failure. Pass several files (results-*.json) to summarise a whole platform matrix in one post - the per-platform table is the point of doing so.

To publish the full report alongside it, add the HTML as an artifact:

- name: Upload full report
  if: always()
  uses: actions/upload-artifact@v4
  with:
    name: test-report
    path: report.html

Staying under the limit

A job summary is capped at 1 MiB per step, and exceeding it fails the upload with an error annotation while the job still reports success - so an oversized summary silently doesn't appear at all. This writer is therefore bounded by construction rather than by hoping:

• Failures are grouped per test, not per (test, platform) - a test failing on 18 platforms is one entry, not 18.
• A test that fails on every platform says so; the platform list is enumerated only when it's a subset, which is the case where which platforms is the actual diagnosis.
• At most 50 failing tests are listed, with a 20-line / 2000-character output excerpt each, taken from the first failing platform.
• A whole-document byte budget backstops the rest, rewinding to the last balanced point so truncation can't leave a dangling <details>.

Whenever anything is left out, the summary says so and points at the HTML report. For scale: this repo's own 18-platform CI run (374 tests, 6718 results, 2830 of them failing by design) produces a ~23 KB summary, about 2% of the limit.