ci-unit-tests-os.yml

name: Multi-OS Unit Tests

# Ensures that only one workflow task will run at a time. Previous builds, if
# already in process, will get cancelled. Only the latest commit will be allowed
# to run, cancelling any workflows in between
concurrency:
  group: ${{ github.workflow }}-${{ github.head_ref || github.run_id }}
  cancel-in-progress: true

on:
  workflow_dispatch:
  # we build Rust and Zcash parameter caches on main,
  # so they can be shared by all branches:
  # https://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows#restrictions-for-accessing-a-cache

  pull_request:
    paths:
      # code and tests
      - '**/*.rs'
      # hard-coded checkpoints and proptest regressions
      - '**/*.txt'
      # test data snapshots
      - '**/*.snap'
      # dependencies
      - '**/Cargo.toml'
      - '**/Cargo.lock'
      - '**/deny.toml'
      # configuration files
      - '.cargo/config.toml'
      - '**/clippy.toml'
      # workflow definitions
      - '.github/workflows/ci-unit-tests-os.yml'

  push:
    branches:
      - main
    paths:
      # production code and test code
      - '**/*.rs'
      # hard-coded checkpoints
      # TODO: skip proptest regressions?
      - '**/*.txt'
      # test data snapshots
      - '**/*.snap'
      # dependencies
      - '**/Cargo.toml'
      - '**/Cargo.lock'
      - '**/deny.toml'
      # configuration files
      - '.cargo/config.toml'
      - '**/clippy.toml'
      # workflow definitions
      - '.github/workflows/ci-unit-tests-os.yml'
      - '.github/workflows/sub-build-docker-image.yml'

env:
  CARGO_INCREMENTAL: ${{ vars.CARGO_INCREMENTAL }}
  RUST_LOG: ${{ vars.RUST_LOG }}
  RUST_BACKTRACE: ${{ vars.RUST_BACKTRACE }}
  RUST_LIB_BACKTRACE: ${{ vars.RUST_LIB_BACKTRACE }}
  COLORBT_SHOW_HIDDEN: ${{ vars.COLORBT_SHOW_HIDDEN }}

jobs:
  ########################################
  ### Build and test Zebra on all OSes ###
  ########################################
  test:
    name: Test ${{ matrix.rust }} on ${{ matrix.os }}${{ matrix.features }}
    # The large timeout is to accommodate:
    # - macOS and Windows builds (90 minutes, typically 30-70 minutes), and
    # - parameter downloads (an extra 100 minutes, but only when the cache expires)
    timeout-minutes: 190
    runs-on: ${{ matrix.os }}
    strategy:
      fail-fast: false
      matrix:
        # TODO: Windows was removed for now, see https://github.com/ZcashFoundation/zebra/issues/3801
        # TODO: macOS tests were removed for now, see https://github.com/ZcashFoundation/zebra/issues/6824
        os: [ubuntu-latest]
        rust: [stable, beta]
        # TODO: When vars.EXPERIMENTAL_FEATURES has features in it, add it here.
        #       Or work out a way to trim the space from the variable: GitHub doesn't allow empty variables.
        #       Or use `default` for the empty feature set and EXPERIMENTAL_FEATURES, and update the branch protection rules.
        #features: ${{ fromJSON(format('["", "{0}"]', vars.EXPERIMENTAL_FEATURES)) }}
        features: [""]
        exclude:
        # We're excluding macOS beta for the following reasons:
        # - the concurrent macOS runner limit is much lower than the Linux limit
        # - macOS is slower than Linux, and shouldn't have a build or test difference with Linux
        # - macOS is a second-tier Zebra support platform
          - os: macos-latest
            rust: beta

    steps:
      - uses: actions/checkout@v4.1.1
        with:
          persist-credentials: false
      - uses: r7kamura/rust-problem-matchers@v1.4.0

      - name: Install last version of Protoc
        uses: arduino/setup-protoc@v2.1.0
        with:
          # TODO: increase to latest version after https://github.com/arduino/setup-protoc/issues/33 is fixed
          version: '23.x'
          repo-token: ${{ secrets.GITHUB_TOKEN }}

      # Setup Rust with ${{ matrix.rust }} toolchain and minimal profile
      - name: Setup Rust
        run: |
          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain=${{ matrix.rust }} --profile=minimal


      - uses: Swatinem/rust-cache@v2.7.0
        # TODO: change Rust cache target directory on Windows,
        #       or remove this workaround once the build is more efficient (#3005).
        #with:
        #  workspaces: ". -> C:\\zebra-target"
        with:
          # Split the experimental features cache from the regular cache, to avoid linker errors.
          # (These might be "disk full" errors, or they might be dependency resolution issues.)
          key: ${{ matrix.features }}

      - name: Change target output directory on Windows
        # Windows doesn't have enough space on the D: drive, so we redirect the build output to the
        # larger C: drive.
        # TODO: Remove this workaround once the build is more efficient (#3005).
        if: matrix.os == 'windows-latest'
        run: |
          mkdir "C:\\zebra-target"
          echo "CARGO_TARGET_DIR=C:\\zebra-target" | Out-File -FilePath "$env:GITHUB_ENV" -Encoding utf8 -Append

      - name: cargo fetch
        run: |
          cargo fetch

      - name: Install LLVM on Windows
        if: matrix.os == 'windows-latest'
        run: |
          choco install llvm -y
          echo "C:\Program Files\LLVM\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
          echo "LIBCLANG_PATH=C:\Program Files\LLVM\bin" | Out-File -FilePath $env:GITHUB_ENV -Encoding utf8 -Append

      - name: Skip network tests on Ubuntu and Windows
        # Ubuntu runners don't have reliable network or DNS during test steps.
        # Windows runners have an unreliable network.
        shell: bash
        if: matrix.os != 'macos-latest'
        run: echo "ZEBRA_SKIP_NETWORK_TESTS=1" >> $GITHUB_ENV

      - name: Minimise proptest cases on macOS and Windows
        # We set cases to 1, because some tests already run 1 case by default.
        # We keep maximum shrink iterations at the default value, because it only happens on failure.
        #
        # Windows compilation and tests are slower than other platforms.
        # macOS runners do extra network tests, so they take longer.
        shell: bash
        if: matrix.os != 'ubuntu-latest'
        run: |
          echo "PROPTEST_CASES=1" >> $GITHUB_ENV
          echo "PROPTEST_MAX_SHRINK_ITERS=1024" >> $GITHUB_ENV

      # Modified from:
      # https://github.com/zcash/librustzcash/blob/c48bb4def2e122289843ddb3cb2984c325c03ca0/.github/workflows/ci.yml#L20-L33
      #
      # TODO:
      # - split Fetch/Cache Zcash parameters into their own job,
      #   and use `concurrency:` to limit it to one job per OS
      # - split get-params-path and download-params examples into their own crate,
      #   to speed up compilation
      - name: Fetch path to Zcash parameters
        working-directory: ./zebra-consensus
        shell: bash
        run: echo "ZCASH_PARAMS=$(cargo run --release --example get-params-path)" >> $GITHUB_ENV
      - name: Cache Zcash parameters
        id: cache-params
        uses: actions/cache@v3
        with:
          path: ${{ env.ZCASH_PARAMS }}
          key: ${{ runner.os }}-sprout-and-sapling-params
      - name: Fetch Zcash parameters
        if: steps.cache-params.outputs.cache-hit != 'true'
        working-directory: ./zebra-consensus
        run: cargo run --release --example download-params

      # Run unit and basic acceptance tests, only showing command output if the test fails.
      #
      # If some tests hang, add "-- --nocapture" for just that test, or for all the tests.
      - name: Run tests${{ matrix.features }}
        run: |
          cargo test --features "${{ matrix.features }}" --release --verbose --workspace

      # Explicitly run any tests that are usually #[ignored]

      - name: Run zebrad large sync tests${{ matrix.features }}
        # Skip the entire step on Ubuntu and Windows, because the test would be skipped anyway due to ZEBRA_SKIP_NETWORK_TESTS
        if: matrix.os == 'macos-latest'
        run: |
          cargo test --features "${{ matrix.features }}" --release --verbose --package zebrad --test acceptance -- --nocapture --include-ignored sync_large_checkpoints_

  # Install Zebra with lockfile dependencies, with no caching and default features
  install-from-lockfile-no-cache:
    name: Install zebrad from lockfile without cache on ubuntu-latest
    timeout-minutes: 60
    runs-on: ubuntu-latest

    steps:
      - uses: actions/checkout@v4.1.1
        with:
          persist-credentials: false
      - uses: r7kamura/rust-problem-matchers@v1.4.0

      # Setup Rust with stable toolchain and minimal profile
      - name: Setup Rust
        run: |
          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain=stable --profile=minimal

      - name: Install zebrad
        run: |
          cargo install --locked --path ./zebrad/ zebrad

  # Check that Cargo.lock includes any Cargo.toml changes.
  # This check makes sure the `cargo-deny` crate dependency checks are accurate.
  check-cargo-lock:
    name: Check Cargo.lock is up to date
    timeout-minutes: 60
    runs-on: ubuntu-latest

    steps:
      - uses: actions/checkout@v4.1.1
        with:
          persist-credentials: false
      - uses: r7kamura/rust-problem-matchers@v1.4.0

      - name: Install last version of Protoc
        uses: arduino/setup-protoc@v2.1.0
        with:
          version: '23.x'
          repo-token: ${{ secrets.GITHUB_TOKEN }}

      # Setup Rust with stable toolchain and minimal profile
      - name: Setup Rust
        run: |
          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain=stable --profile=minimal

      - uses: Swatinem/rust-cache@v2.7.0
        with:
          shared-key: "clippy-cargo-lock"

      - name: Check Cargo.lock is up to date
        run: |
          cargo check --locked --all-features --all-targets

  cargo-deny:
    name: Check deny.toml ${{ matrix.checks }} ${{ matrix.features }}
    runs-on: ubuntu-latest
    strategy:
      matrix:
        checks:
          - bans
          - sources
        # We don't need to check `--no-default-features` here, because (except in very rare cases):
        # - disabling features isn't going to add duplicate dependencies
        # - disabling features isn't going to add more crate sources
        features: ['', '--features default-release-binaries', '--all-features']
      # Always run the --all-features job, to get accurate "skip tree root was not found" warnings
      fail-fast: false

    # Prevent sudden announcement of a new advisory from failing ci:
    continue-on-error: ${{ matrix.checks == 'advisories' }}

    steps:
      - uses: actions/checkout@v4.1.1
        with:
          persist-credentials: false
      - uses: r7kamura/rust-problem-matchers@v1.4.0

      - name: Check ${{ matrix.checks }} with features ${{ matrix.features }}
        uses: EmbarkStudios/cargo-deny-action@v1
        with:
          # --all-features spuriously activates openssl, but we want to ban that dependency in
          # all of zebrad's production features for security reasons. But the --all-features job is
          # the only job that gives accurate "skip tree root was not found" warnings.
          # In other jobs, we expect some of these warnings, due to disabled features.
          command: check ${{ matrix.checks }} ${{ matrix.features == '--all-features' && '--allow banned' || '--allow unmatched-skip-root' }}
          arguments: --workspace ${{ matrix.features }}

  unused-deps:
    name: Check for unused dependencies
    runs-on: ubuntu-latest

    steps:
      - name: Checkout git repository
        uses: actions/checkout@v4.1.1
        with:
          persist-credentials: false
      - uses: r7kamura/rust-problem-matchers@v1.4.0

      # Setup Rust with stable toolchain and minimal profile
      - name: Setup Rust
        run: |
          curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain=stable --profile=minimal

      - name: Install cargo-machete
        uses: baptiste0928/cargo-install@v2.2.0
        with:
          crate: cargo-machete

      - name: Check unused dependencies
        # Exclude macro and transitive dependencies by filtering them out of the output,
        # then if there are any more unused dependencies, fail the job.
        run: |
          echo "-- full cargo machete output, including ignored dependencies --"
          cargo machete --skip-target-dir || true
          echo "-- unused dependencies are below this line, full output is above --"
          if cargo machete --skip-target-dir 2>/dev/null | \
          grep --extended-regexp -e '^\\t' | \
          grep -v -e gumdrop -e humantime-serde -e tinyvec -e zebra-utils; then
              echo "New unused dependencies were found, please remove them!"
              exit 1
          else
              echo "No unused dependencies found."
          fi

  failure-issue:
    name: Open or update issues for OS integration failures
    # When a new job is added to this workflow, add it to this list.
    needs: [ test,  install-from-lockfile-no-cache, check-cargo-lock, cargo-deny, unused-deps ]
    # Only open tickets for failed or cancelled jobs that are not coming from PRs.
    # (PR statuses are already reported in the PR jobs list, and checked by Mergify.)
    if: (failure() && github.event.pull_request == null) || (cancelled() && github.event.pull_request == null)
    runs-on: ubuntu-latest
    steps:
      - uses: jayqi/failed-build-issue-action@v1
        with:
          title-template: "{{refname}} branch CI failed: {{eventName}} in {{workflow}}"
          # New failures open an issue with this label.
          label-name: S-ci-fail-os-integration-auto-issue
          # If there is already an open issue with this label, any failures become comments on that issue.
          always-create-new-issue: false
          github-token: ${{ secrets.GITHUB_TOKEN }}