ci.yaml

name: CI

on:
  workflow_call:
  pull_request:
    types: [ opened, synchronize, reopened, ready_for_review ]
  workflow_dispatch:
  push:
    paths-ignore:
      - '**.md'
      - '**.kf'
    branches:
      - main

permissions:
  contents: read
  # Optional: allow read access to pull request. Use with `only-new-issues` option.
  # pull-requests: read

jobs:
  golangci:
    name: lint
    runs-on: ubuntu-latest
    if: ${{ !github.event.pull_request.draft }}
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
          go-version: '1.25.3'
          cache: false
      - name: golangci-lint
        uses: golangci/golangci-lint-action@v8
        with:
          # Require: The version of golangci-lint to use.
          version: v2.10.1
          args: --timeout=30m --issues-exit-code=0 --verbose

  acceptance-test:
    runs-on: ubuntu-latest
    if: ${{ !github.event.pull_request.draft }}
    steps:
      - name: Clear cache and show disk space
        if: ${{ !env.ACT }} # skip during local actions testing
        run: |
          echo "Initial disk space:"
          df -h /
          sudo rm -rf /usr/share/dotnet
          sudo rm -rf /opt/ghc
          sudo rm -rf "/usr/local/share/boost"
          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
          # Additional cleanup for Android, Azure, and other pre-installed tools
          sudo rm -rf /usr/local/lib/android || true
          sudo rm -rf /usr/local/.ghcup || true
          sudo rm -rf /usr/share/swift || true
          echo "Disk space after cleanup:"
          df -h /

      - uses: actions/checkout@v4

      - name: Setup Go
        uses: actions/setup-go@v5
        with:
          go-version: '1.25.3'

      - name: Install Taskfile
        uses: arduino/setup-task@v2
        with:
          repo-token: ${{ secrets.GITHUB_TOKEN }}

      - name: Run Taskfile
        run: task build

      - name: Clean caches to free disk space
        run: |
          # Clean Go build cache to free space before test compilation
          # NOTE: Keep .build/ directory - needed later for single-node stack
          go clean -cache -testcache -fuzzcache || true
          # Remove Docker build cache and images
          docker builder prune -af || true
          docker image prune -af || true
          # Show remaining disk space
          df -h /

      - name: Run Go Tests (with retry)
        run: |
          set -o errexit -o nounset -o pipefail

          attempt=1
          max_attempts=3
          backoff=10

          # Ensure Docker daemon is ready before first attempt (max 60s)
          wait_count=0
          max_wait=30
          until docker info >/dev/null 2>&1; do
            if [ $wait_count -ge $max_wait ]; then
              echo "❌ Docker daemon failed to become ready after ${max_wait} attempts"
              exit 1
            fi
            echo "Waiting for Docker daemon to be ready... (attempt $((wait_count + 1))/${max_wait})"
            sleep 2
            wait_count=$((wait_count + 1))
          done
          echo "✅ Docker daemon is ready"

          while [ $attempt -le $max_attempts ]; do
            echo "Test attempt $attempt of $max_attempts"

            # Always try to cleanup lingering Kwil DB resources before each attempt
            bash scripts/ci-cleanup.sh || true

            # Additional cleanup only on retries
            if [ $attempt -gt 1 ]; then
              docker compose -f compose.yaml down -v || true
              docker system prune -af --volumes || true
              sleep 5
            fi

            # Run unit tests only on PRs (skip integration tests and benchmarks to avoid disk space issues)
            # Integration tests create Docker containers which rapidly fill disk (~50+ containers)
            # Benchmarks are long-running performance tests (not needed for every PR)
            # Full integration suite runs on main branch via slow-integration-tests job
            if go test -failfast -p 1 -timeout=15m -count=1 -tags=kwiltest \
              $(go list -tags=kwiltest ./... | grep -v -E '^github\.com/trufnetwork/node/(tests/(streams|database_size|extensions/(erc20|database-size))|internal/benchmark)'); then
              echo "✅ Tests passed on attempt $attempt"
              break
            else
              echo "❌ Tests failed on attempt $attempt"
              # Surface Docker diagnostics
              echo "::group::Docker diagnostics"
              docker ps -a || true
              docker system df || true
              docker info || true
              echo "::endgroup::"

              if [ $attempt -eq $max_attempts ]; then
                echo "All test attempts failed"
                exit 1
              fi
              attempt=$((attempt + 1))
              echo "Waiting ${backoff}s before retry..."
              sleep "$backoff"
              backoff=$((backoff * 2))
            fi
          done

      - name: Cleanup Docker resources before acceptance test
        run: |
          docker compose -f compose.yaml down -v || true
          docker system prune -af --volumes || true

      - name: Start Single-Node Stack
        run: task single:start

      - name: Wait for node to warm up
        run: |
          echo "⏳ giving the node 10s to initialize…"
          sleep 10

      - name: Run CI Tests Script
        run: |
          scripts/ci-tests.sh

  check-slow-test-paths:
    runs-on: ubuntu-latest
    if: github.event_name == 'pull_request'
    outputs:
      slow_tests_changed: ${{ steps.filter.outputs.slow_tests }}
    steps:
      - uses: actions/checkout@v4

      - name: Check for affected paths
        uses: dorny/paths-filter@v3
        id: filter
        with:
          filters: |
            slow_tests:
              - 'tests/streams/**'
              - 'tests/database_size/**'
              - 'tests/extensions/erc20/**'
              - 'tests/extensions/database-size/**'

  slow-integration-tests:
    runs-on: ubuntu-latest
    needs: check-slow-test-paths
    # Runs on:
    # - main branch (always)
    # - workflow_dispatch (manual trigger)
    # - pull_request when:
    #   * Label "run-slow-tests" is added to the PR, OR
    #   * Files in order_book/, digest/, erc20/, or streams/ directories are modified
    #
    # To trigger on a PR:
    # 1. Add label "run-slow-tests" to the PR, OR
    # 2. Modify files in tests/streams/order_book/, tests/streams/digest/,
    #    tests/extensions/erc20/, or tests/streams/ directories
    if: |
      always() && (
        github.event_name == 'workflow_dispatch' ||
        github.ref == 'refs/heads/main' ||
        (
          github.event_name == 'pull_request' &&
          (
            contains(github.event.pull_request.labels.*.name, 'run-slow-tests') ||
            needs.check-slow-test-paths.outputs.slow_tests_changed == 'true'
          )
        )
      )
    steps:
      - uses: actions/checkout@v4

      - name: Clear cache and show disk space
        run: |
          echo "Initial disk space:"
          df -h /
          sudo rm -rf /usr/share/dotnet
          sudo rm -rf /opt/ghc
          sudo rm -rf "/usr/local/share/boost"
          sudo rm -rf "$AGENT_TOOLSDIRECTORY"
          sudo rm -rf /usr/local/lib/android || true
          sudo rm -rf /usr/local/.ghcup || true
          sudo rm -rf /usr/share/swift || true
          echo "Disk space after cleanup:"
          df -h /

      - name: Setup Go
        uses: actions/setup-go@v5
        with:
          go-version: '1.25.3'

      - name: Install Taskfile
        uses: arduino/setup-task@v2
        with:
          repo-token: ${{ secrets.GITHUB_TOKEN }}

      - name: Run Taskfile
        run: task build

      - name: Clean build artifacts to free disk space
        run: |
          sudo rm -rf .build/ || true
          go clean -cache -testcache -fuzzcache || true
          docker builder prune -af || true
          docker image prune -af || true
          df -h /

      - name: Run Slow Integration Tests (with retry)
        run: |
          set -o errexit -o nounset -o pipefail

          attempt=1
          max_attempts=3
          backoff=10

          # Ensure Docker daemon is ready before first attempt (max 60s)
          wait_count=0
          max_wait=30
          until docker info >/dev/null 2>&1; do
            if [ $wait_count -ge $max_wait ]; then
              echo "❌ Docker daemon failed to become ready after ${max_wait} attempts"
              exit 1
            fi
            echo "Waiting for Docker daemon to be ready... (attempt $((wait_count + 1))/${max_wait})"
            sleep 2
            wait_count=$((wait_count + 1))
          done
          echo "✅ Docker daemon is ready"

          while [ $attempt -le $max_attempts ]; do
            echo "Integration test attempt $attempt of $max_attempts"

            # Always cleanup lingering resources before each attempt
            bash scripts/ci-cleanup.sh || true

            # Additional cleanup only on retries
            if [ $attempt -gt 1 ]; then
              docker compose -f compose.yaml down -v || true
              docker system prune -af --volumes || true
              sleep 5
            fi

            # Run ALL integration tests (everything excluded from PR fast tests)
            # These create Docker containers and need significant disk space
            if go test -v -p 1 -timeout=120m -count=1 -tags=kwiltest \
              ./tests/streams/... \
              ./tests/database_size \
              ./tests/extensions/erc20 \
              ./tests/extensions/database-size; then
              echo "✅ Integration tests passed on attempt $attempt"
              break
            else
              echo "❌ Integration tests failed on attempt $attempt"
              # Surface Docker diagnostics
              echo "::group::Docker diagnostics"
              docker ps -a || true
              docker system df || true
              docker info || true
              echo "::endgroup::"

              if [ $attempt -eq $max_attempts ]; then
                echo "All integration test attempts failed"
                exit 1
              fi
              attempt=$((attempt + 1))
              echo "Waiting ${backoff}s before retry..."
              sleep "$backoff"
              backoff=$((backoff * 2))
            fi
          done