ci.yml

name: CI

on:
  pull_request:
  push:
    branches: [main]

permissions:
  contents: read

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v6
      - run: bash scripts/check-public-repository.sh
      - uses: subosito/flutter-action@v2
        with:
          # PINNED, not `channel: stable`. An unpinned stable meant the Flutter
          # SDK could move between releases, and a newer one rewrote the
          # checked-in analysis_options.yaml during `pub get` -- which
          # `dart pub publish` then refused as a dirty tree, failing a release
          # whose own diff was four clean files. Pin so a release reproduces
          # what was validated locally; bump deliberately.
          flutter-version: 3.44.0
          channel: stable
          cache: true
      - run: flutter pub get
      - run: flutter analyze
      - run: flutter test
      - run: dart pub publish --dry-run

  # Goldens are recorded on macOS and can only be compared there.
  #
  # Fonts are pinned for every host in test/flutter_test_config.dart, so the
  # glyphs and their positions are now identical everywhere. What is not
  # identical is how Skia rasterises them: CoreText on macOS and FreeType on
  # Linux disagree on the anti-aliased edge of every letter, which left 2-5% of
  # each image differing one pixel deep around glyph outlines. A comparator
  # loose enough to accept that would also accept a real regression, so the
  # images are enforced on the platform that records them. The Linux job above
  # still runs every other test; the golden group skips itself there with a
  # reason.
  goldens:
    runs-on: macos-latest
    steps:
      - uses: actions/checkout@v6
      - uses: subosito/flutter-action@v2
        with:
          flutter-version: 3.44.0
          channel: stable
          cache: true
      - run: flutter pub get
      - run: flutter test --tags golden