torqvoice/.github/workflows/e2e.yml

name: End-to-end tests

on:
  pull_request:
    branches: [main]
  workflow_dispatch:

# A second push to the same branch makes the run in flight pointless.
concurrency:
  group: e2e-${{ github.ref }}
  cancel-in-progress: true

# The suite runs one test at a time against one seeded database, so more
# workers would only trip over each other. More databases do not: resetting
# and seeding one takes about fifteen seconds. So the specs are split across
# shards, each on its own runner with its own database, and the cloud specs
# run beside them. The last job merges what they found into one report and is
# the one check a pull request waits on.

jobs:
  e2e:
    name: Playwright (${{ matrix.name }})
    runs-on: ubuntu-latest
    timeout-minutes: 30

    strategy:
      # A failure in one shard says nothing about the others; let them finish.
      fail-fast: false
      matrix:
        include:
          - { id: shard-1, name: shard 1 of 8, shard: 1/8 }
          - { id: shard-2, name: shard 2 of 8, shard: 2/8 }
          - { id: shard-3, name: shard 3 of 8, shard: 3/8 }
          - { id: shard-4, name: shard 4 of 8, shard: 4/8 }
          - { id: shard-5, name: shard 5 of 8, shard: 5/8 }
          - { id: shard-6, name: shard 6 of 8, shard: 6/8 }
          - { id: shard-7, name: shard 7 of 8, shard: 7/8 }
          - { id: shard-8, name: shard 8 of 8, shard: 8/8 }
          # The same build started in cloud mode: plan limits, the sign-up
          # pitch and Google sign-in only exist there.
          - { id: cloud, name: cloud, mode: cloud }

    services:
      postgres:
        image: postgres:16-alpine
        env:
          POSTGRES_USER: torqvoice
          POSTGRES_PASSWORD: torqvoice
          # The name has to carry "e2e" or "test": the harness refuses to reset
          # a database whose name says nothing about being throwaway.
          POSTGRES_DB: torqvoice_e2e
        ports:
          - 5432:5432
        options: >-
          --health-cmd "pg_isready -U torqvoice -d torqvoice_e2e"
          --health-interval 10s
          --health-timeout 5s
          --health-retries 5

    env:
      # Playwright starts the app itself on this port and resets this database
      # before it does; both come from playwright.config.ts.
      E2E_DATABASE_URL: postgresql://torqvoice:torqvoice@127.0.0.1:5432/torqvoice_e2e
      E2E_MODE: ${{ matrix.mode }}
      # The cloud job mints a one-day cloud token from it; without the token the
      # app ignores TORQVOICE_MODE=cloud. Only the cloud job gets it.
      TORQVOICE_COM_LICENSE_SIGNING_PRIVATE_KEY: ${{ matrix.mode == 'cloud' && secrets.TORQVOICE_COM_LICENSE_SIGNING_PRIVATE_KEY || '' }}

    steps:
      # A pull request from a fork gets no secrets, so the cloud job has no key
      # to mint its cloud token with and the app would refuse cloud mode. That
      # says nothing about the contribution, so the job skips instead of failing.
      # The self-hosted shards need no key and always run.
      - name: Check for the cloud signing key
        id: gate
        run: |
          if [ "${{ matrix.mode }}" = "cloud" ] && [ -z "$TORQVOICE_COM_LICENSE_SIGNING_PRIVATE_KEY" ]; then
            echo "::notice title=Cloud specs skipped::No TORQVOICE_COM_LICENSE_SIGNING_PRIVATE_KEY here (a fork or Dependabot pull request), so the cloud specs cannot run."
            echo "run=false" >> "$GITHUB_OUTPUT"
          else
            echo "run=true" >> "$GITHUB_OUTPUT"
          fi

      - uses: actions/checkout@v4
        if: ${{ steps.gate.outputs.run == 'true' }}

      - uses: actions/setup-node@v4
        if: ${{ steps.gate.outputs.run == 'true' }}
        with:
          node-version: 22
          cache: npm

      # A fresh one per run: the sessions it signs live as long as the job, and
      # a short or guessable value makes better-auth warn on every request.
      - name: Make a session secret for this run
        if: ${{ steps.gate.outputs.run == 'true' }}
        run: echo "BETTER_AUTH_SECRET=$(openssl rand -base64 32)" >> "$GITHUB_ENV"

      - run: npm ci
        if: ${{ steps.gate.outputs.run == 'true' }}

      - run: npx prisma generate
        if: ${{ steps.gate.outputs.run == 'true' }}

      # Keyed on the pinned version, because the image tag and this download
      # have to be the same build.
      - name: Cache the browser
        if: ${{ steps.gate.outputs.run == 'true' }}
        uses: actions/cache@v4
        with:
          path: ~/.cache/ms-playwright
          key: playwright-${{ runner.os }}-${{ hashFiles('package-lock.json') }}

      - name: Install the browser
        if: ${{ steps.gate.outputs.run == 'true' }}
        run: npx playwright install --with-deps chromium

      # The seed copies bundled vehicle photos when they are in the tree and
      # downloads fifty of them from Unsplash when they are not, one after
      # another, which is minutes of a cold run and the flakiest thing in it.
      # Cached, a warm run copies them off disk instead.
      - name: Cache the seed's photos
        if: ${{ steps.gate.outputs.run == 'true' }}
        uses: actions/cache@v4
        with:
          # Where prepare-db.ts points the seed's DATA_ROOT, so a test run
          # cannot disturb a running instance's own uploads.
          path: e2e/.data
          key: seed-photos-${{ hashFiles('prisma/seed_dummy_data.ts') }}
          restore-keys: |
            seed-photos-

      # Next reuses its compiler cache across builds when it is given one.
      # Every job reads it; only the first shard writes it back, so five jobs
      # do not race to save the same entry.
      - name: Restore the build cache
        if: ${{ steps.gate.outputs.run == 'true' }}
        uses: actions/cache/restore@v4
        with:
          path: .next/cache
          key: next-${{ hashFiles('package-lock.json') }}-${{ github.sha }}
          restore-keys: |
            next-${{ hashFiles('package-lock.json') }}-
            next-

      # NEXT_PUBLIC_APP_URL is baked into the client bundle, and better-auth
      # refuses a sign-in from an origin the build was not made for, so it has
      # to match the base URL the suite uses.
      - name: Build
        if: ${{ steps.gate.outputs.run == 'true' }}
        run: npm run build
        env:
          NEXT_PUBLIC_APP_URL: http://127.0.0.1:3100

      - name: Save the build cache
        if: ${{ steps.gate.outputs.run == 'true' && (matrix.id == 'shard-1') }}
        uses: actions/cache/save@v4
        with:
          path: .next/cache
          key: next-${{ hashFiles('package-lock.json') }}-${{ github.sha }}

      - name: Run the specs
        if: ${{ steps.gate.outputs.run == 'true' }}
        run: npx playwright test ${{ matrix.shard && format('--shard={0}', matrix.shard) || '' }}

      # The blob carries the results with their traces, screenshots and videos;
      # the report job turns every job's blob into one HTML report.
      - name: Upload the blob report
        if: ${{ steps.gate.outputs.run == 'true' && (!cancelled()) }}
        uses: actions/upload-artifact@v4
        with:
          name: blob-report-${{ matrix.id }}
          path: blob-report/
          retention-days: 1
          if-no-files-found: ignore

  report:
    # Named as the single job used to be, so a required check keeps its name.
    name: Playwright
    needs: e2e
    if: ${{ !cancelled() }}
    runs-on: ubuntu-latest
    timeout-minutes: 10

    steps:
      - uses: actions/checkout@v4

      - uses: actions/setup-node@v4
        with:
          node-version: 22
          cache: npm

      - run: npm ci

      - name: Download the blob reports
        uses: actions/download-artifact@v4
        with:
          pattern: blob-report-*
          path: all-blob-reports

      # Each job's blob lands in a folder of its own. The merge wants them side
      # by side, and two jobs may give their file the same name.
      - name: Gather the blobs
        run: |
          mkdir -p blob-reports
          for dir in all-blob-reports/*/; do
            [ -d "$dir" ] || continue
            job=$(basename "$dir")
            for file in "$dir"*.zip; do
              [ -e "$file" ] || continue
              mv "$file" "blob-reports/${job}-$(basename "$file")"
            done
          done
          ls -la blob-reports

      - name: Merge the reports
        run: npx playwright merge-reports --reporter html ./blob-reports

      - name: Upload the report
        uses: actions/upload-artifact@v4
        with:
          name: playwright-report
          path: playwright-report/
          retention-days: 7
          if-no-files-found: ignore

      # Green only when every shard and the cloud specs are.
      - name: Every job passed
        if: ${{ always() }}
        run: |
          if [ "${{ needs.e2e.result }}" != "success" ]; then
            echo "The e2e jobs finished as: ${{ needs.e2e.result }}"
            exit 1
          fi