build.yml

name: Build

# GitHub Pages builds and deploys this site itself; this workflow only verifies
# that it still builds and that the generated output holds together. A Liquid
# error in assets/search_plus_index.json, or markup that breaks the version
# toggle, otherwise reaches production unnoticed.

on:
  push:
    branches: [master]
  pull_request:
  workflow_dispatch:

permissions:
  contents: read

jobs:
  build:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - uses: ruby/setup-ruby@v1
        with:
          ruby-version: "3.3"
          bundler-cache: true

      - name: Install PDF text tools
        run: sudo apt-get update && sudo apt-get install -y poppler-utils

      - name: Check authoritative text fixtures and variants
        run: |
          python3 script/build-westminster-fixtures.py --check
          python3 script/check-variant-markup.py

      # _scripture/ is generated from the proof callouts. Editing a proof without
      # regenerating would leave the index quietly citing the old passage.
      - name: Check the scripture index matches the proofs
        run: python3 script/build-scripture-index.py --check

      - name: Build
        run: bundle exec jekyll build --strict_front_matter --trace
        env:
          JEKYLL_ENV: production

      - name: Check generated output
        run: |
          set -euo pipefail

          # The search index is generated by a dense Liquid template; a silent
          # failure there produces valid HTML but a broken index.
          python3 -c "
          import json, sys
          with open('_site/assets/search_plus_index.json') as f:
              idx = json.load(f)
          print(f'search index: {len(idx)} entries')
          if len(idx) < 800:
              sys.exit(f'search index too small ({len(idx)}); section splitting likely broke')
          garbled = [k for k, v in idx.items()
                     if any(bad in v['body'].lower() for bad in
                            ('unexcusable;without', 'unto to', 'hath has',
                             'ofour', 'ourown', 'third epistle to timothy'))]
          if garbled:
              sys.exit(f'search index contains crawler-text corruption: {garbled[:3]}')
          missing_fields = [k for k, v in idx.items()
                            if not {'document', 'collection', 'proofs'} <= set(v)]
          if missing_fields:
              sys.exit(f'search index entries missing filter/proof fields: {missing_fields[:3]}')
          "

          # Every document, plus the pages that are easy to break.
          for path in index.html 404.html sitemap.xml robots.txt \
                      search assets/search.html assets/search_plus_index.json \
                      pages/apostles-creed pages/nicene-creed pages/athanasian-creed \
                      pages/wcf pages/wsc pages/wlc \
                      pages/belgic pages/heidelberg pages/canons-of-dort \
                      pages/fg pages/bd pages/dpw \
                      scripture scripture/romans scripture/psalms scripture/jude; do
            target="_site/$path"
            [ -d "$target" ] && target="$target/index.html"
            [ -s "$target" ] || { echo "missing or empty: $target"; exit 1; }
          done
          echo "all expected pages present"

          # Internal links must stay on the canonical domain.
          if grep -rn 'href="https://normanormata\.github\.io' _site/ ; then
            echo "found links to the github.io mirror instead of the canonical domain"
            exit 1
          fi
          echo "no links to the github.io mirror"

      - name: Check semantics and accessible names
        run: python3 script/check-generated-html.py _site

      - name: Check internal links and fragments
        run: python3 script/check-links.py _site