committer_ci.yml

name: Committer-CI

on:
  pull_request:
    types:
      - opened
      - reopened
      - synchronize
      - edited
    paths:
      - ".github/workflows/committer_ci.yml"
      - "Cargo.toml"
      - "Cargo.lock"
      - "crates/starknet_committer_and_os_cli/**"
      - "crates/starknet_api/**"
      - "crates/starknet_committer/**"
      - "crates/starknet_patricia/**"
      - "crates/starknet_patricia_storage/**"
  push:
    branches:
      - main
      - "main-v*"
      - "release-*"
    paths:
      - ".github/workflows/committer_ci.yml"
      - "Cargo.toml"
      - "Cargo.lock"
      - "crates/starknet_committer_and_os_cli/**"
      - "crates/starknet_api/**"
      - "crates/starknet_committer/**"
      - "crates/starknet_patricia/**"
      - "crates/starknet_patricia_storage/**"
  workflow_dispatch:
    inputs:
      branch:
        description: "Branch to benchmark (leave empty for current)"
        required: false
        type: string

env:
  RUSTFLAGS: "-D warnings"
  FULL_COMMITTER_FLOW_TIME_LIMIT_NS: "1000000000"
  TREE_COMPUTATION_FLOW_TIME_LIMIT_NS: "1800000000"
  BENCHMARK_REGRESSION_LIMIT_PERCENT: "20"

# On PR events, cancel existing CI runs on this same PR.
# On push events, group by branch to avoid gh-pages conflicts (no cancellation).
concurrency:
  group: ${{ github.workflow }}-${{ github.ref_name }}
  cancel-in-progress: ${{ github.event_name == 'pull_request' }}

jobs:
  run-regression-tests:
    runs-on: namespace-profile-medium-ubuntu-24-04-amd64
    if: ${{ github.event_name == 'pull_request' }}
    steps:
      - uses: actions/checkout@v4

      # Setup pypy and link to the location expected by .cargo/config.toml.
      # Python + requirements are needed to compile the OS.
      - uses: actions/setup-python@v5
        id: setup-pypy
        with:
          python-version: "pypy3.9"
          cache: "pip"
      - run: ln -s '${{ steps.setup-pypy.outputs.python-path }}' /usr/local/bin/pypy3.9
      - env:
          LD_LIBRARY_PATH: ${{ env.Python3_ROOT_DIR }}/bin
        run: echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}" >> $GITHUB_ENV
      - run: pip install -r scripts/requirements.txt

      - uses: ./.github/actions/bootstrap
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}

      - id: auth
        uses: "google-github-actions/auth@v2"
        with:
          credentials_json: ${{ secrets.COMMITER_PRODUCTS_EXT_WRITER_JSON }}
      - uses: "google-github-actions/setup-gcloud@v2"
      - run: echo "BENCH_INPUT_FILES_PREFIX=$(cat ./crates/starknet_committer_and_os_cli/src/committer_cli/tests/flow_test_files_prefix)" >> $GITHUB_ENV
      - run: gcloud storage cp -r gs://committer-testing-artifacts/$BENCH_INPUT_FILES_PREFIX/* ./crates/starknet_committer_and_os_cli/test_inputs
      - run: cargo test -p starknet_committer_and_os_cli --release -- --include-ignored test_regression

  # Benchmarking job - behavior depends on event type:
  #
  # ┌─────────────────────┬───────────────────────────────────────────────────────┐
  # │ Event               │ Behavior                                              │
  # ├─────────────────────┼───────────────────────────────────────────────────────┤
  # │ pull_request        │ Benchmark base branch, then head branch, compare.     │
  # │                     │ Enforces regression limits and absolute time limits.  │
  # ├─────────────────────┼───────────────────────────────────────────────────────┤
  # │ push (main branches)│ Benchmark current branch, publish results to gh-pages │
  # │ workflow_dispatch   │ for historical tracking and visualization.            │
  # └─────────────────────┴───────────────────────────────────────────────────────┘
  benchmarking:
    runs-on: namespace-profile-medium-ubuntu-24-04-amd64
    permissions:
      contents: write # Required for pushing to gh-pages on push events.
    steps:
      # PR-only: Checkout the base branch to benchmark the old code.
      - name: Checkout base branch (PR only)
        if: github.event_name == 'pull_request'
        uses: actions/checkout@v4
        with:
          ref: ${{ github.base_ref }}

      # Push/dispatch: Checkout the specified or current branch.
      - name: Checkout repository (push/dispatch)
        if: github.event_name != 'pull_request'
        uses: actions/checkout@v4
        with:
          ref: ${{ inputs.branch || github.ref }}

      # Setup pypy and link to the location expected by .cargo/config.toml.
      # Python + requirements are needed to compile the OS.
      - uses: actions/setup-python@v5
        id: setup-pypy
        with:
          python-version: "pypy3.9"
          cache: "pip"
      - run: ln -s '${{ steps.setup-pypy.outputs.python-path }}' /usr/local/bin/pypy3.9
      - env:
          LD_LIBRARY_PATH: ${{ env.Python3_ROOT_DIR }}/bin
        run: echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}" >> $GITHUB_ENV
      - run: pip install -r scripts/requirements.txt

      - uses: ./.github/actions/bootstrap
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}

      # Auth with GCS for bench_tools to download inputs.
      - id: auth
        uses: "google-github-actions/auth@v2"
        with:
          credentials_json: ${{ secrets.COMMITER_PRODUCTS_EXT_WRITER_JSON }}
      - uses: "google-github-actions/setup-gcloud@v2"

      # PR-only: Benchmark the base branch code.
      - name: Benchmark base branch (PR only)
        if: github.event_name == 'pull_request'
        run: cargo run -p bench_tools -- run --package starknet_committer_and_os_cli --out /tmp/base_results

      # PR-only: Save the input files from the base branch to use in the second benchmark.
      - name: Save test inputs (PR only)
        if: github.event_name == 'pull_request'
        run: cp -r crates/starknet_committer_and_os_cli/test_inputs /tmp/test_inputs

      # PR-only: Checkout the current branch to benchmark the new code.
      - name: Checkout current branch (PR only)
        if: github.event_name == 'pull_request'
        uses: actions/checkout@v4
        with:
          clean: false

      # PR-only: Re-install pip requirements in case they changed between branches.
      - name: Re-install pip requirements (PR only)
        if: github.event_name == 'pull_request'
        run: pip install -r scripts/requirements.txt

      # PR-only: Benchmark the current branch and compare to the previous run.
      # Use the saved inputs and enforce absolute time limits and regression limit.
      - name: Benchmark and compare (PR only)
        if: github.event_name == 'pull_request'
        run: |
          cargo run -p bench_tools -- run-and-compare \
            --package starknet_committer_and_os_cli \
            --out /tmp/new_results \
            --regression-limit ${BENCHMARK_REGRESSION_LIMIT_PERCENT} \
            --input-dir /tmp/test_inputs \
            --set-absolute-time-ns-limit full_committer_flow ${FULL_COMMITTER_FLOW_TIME_LIMIT_NS} \
            --set-absolute-time-ns-limit tree_computation_flow ${TREE_COMPUTATION_FLOW_TIME_LIMIT_NS}

      # Push/dispatch: Run benchmarks and publish to GitHub Pages.
      - name: Run benchmarks (push/dispatch)
        if: github.event_name != 'pull_request'
        run: |
          cargo run -p bench_tools -- run \
            --package starknet_committer_and_os_cli \
            --out /tmp/benchmark_results \
            --github-action-benchmark-output-file benchmark_output.json

      # Push/dispatch: Sanitize branch name and compute alert threshold.
      - name: Prepare publish variables (push/dispatch)
        if: github.event_name != 'pull_request'
        id: branch
        env:
          BRANCH_NAME: ${{ inputs.branch || github.ref_name }}
        run: |
          SANITIZED="${BRANCH_NAME//\//-}"
          echo "name=${SANITIZED}" >> $GITHUB_OUTPUT
          echo "alert_threshold=$((100 + BENCHMARK_REGRESSION_LIMIT_PERCENT))%" >> $GITHUB_OUTPUT
          echo "Branch: ${BRANCH_NAME} -> Directory: ${SANITIZED}"

      # Push/dispatch: Clean up downloaded test inputs to avoid conflicts when switching to gh-pages.
      - name: Clean up test inputs (push/dispatch)
        if: github.event_name != 'pull_request'
        run: git checkout -- crates/starknet_committer_and_os_cli/test_inputs/

      # Push/dispatch: Store benchmark results to GitHub Pages.
      - name: Store benchmark results (push/dispatch)
        if: github.event_name != 'pull_request'
        uses: starkware-libs/github-action-benchmark@v1
        with:
          tool: "customSmallerIsBetter"
          output-file-path: benchmark_output.json
          benchmark-data-dir-path: dev/bench/${{ steps.branch.outputs.name }}
          gh-pages-branch: gh-pages
          github-token: ${{ secrets.GITHUB_TOKEN }}
          auto-push: true
          # Alert if performance regresses by more than the threshold.
          alert-threshold: ${{ steps.branch.outputs.alert_threshold }}
          # Comment on commit if threshold exceeded.
          comment-on-alert: true
          fail-on-alert: false

      # Push/dispatch: Upload raw benchmark results as artifact for debugging.
      # always() ensures artifacts are uploaded even if benchmarks fail, to help diagnose issues.
      - name: Upload benchmark artifacts (push/dispatch)
        if: github.event_name != 'pull_request' && always()
        uses: actions/upload-artifact@v4
        with:
          name: benchmark-results-${{ steps.branch.outputs.name }}-${{ github.sha }}
          path: target/criterion/*/new/estimates.json
          retention-days: 30