Tests

This directory holds the theme's automated tests. It is the proof-of-concept
for the testing system tracked in the testing epic.

Layout

tests/
├── fixtures/site/      # a minimal Hugo site that uses the theme
├── e2e/                # Playwright specs (E2E, accessibility, no-JS)
├── support/            # build-the-fixture + static-server helpers
├── catalog.yaml        # feature -> test manifest for the coverage gate
└── coverage-gate.mjs   # fails CI when a param/shortcode has no test

The legacy build/HTML harness still lives at
scripts/test-hugo-quickstart.ts and is being migrated into this structure.

Running

# Browser end-to-end, accessibility, and progressive-enhancement (no-JS) tests.
# Builds the fixture site against the local theme working tree, serves it, and
# runs Playwright. Requires browsers: `npx playwright install chromium`.
npm run test:e2e

# Feature -> test coverage gate: fails if an [ananke] param or a shortcode has
# no entry in tests/catalog.yaml.
npm run test:coverage-gate

# Legacy quickstart build/assertion harness.
npm test

How the fixture is built

tests/support/prepare-site.mjs creates a temporary themes directory with a
symlink to the repository root and runs Hugo against tests/fixtures/site, so
the suite always exercises the current theme code (including uncommitted
changes) without committing an absolute-path symlink. The built site is served
by tests/support/dev-server.mjs, which Playwright launches via
webServer in playwright.config.ts.

Adding a test for a feature

When you add or change a feature:

1. If it has a config parameter or shortcode, add fixture content that
   exercises it under tests/fixtures/site/.
2. Add a spec under tests/e2e/ (interaction, output, or accessibility).
3. Register it in tests/catalog.yaml so the coverage gate is satisfied.

The coverage gate (npm run test:coverage-gate) fails if a new [ananke]
parameter or shortcode is not listed in the catalog with at least one test, so
tests stay in sync with features.

Accessibility note

The axe suite catches roughly 30-40% of WCAG issues. It is a floor, not a
guarantee — pair it with manual keyboard and screen-reader checks.