ci.yml

name: CI

on:
  push:
    branches: [ main ]
  pull_request:
    branches: [ main ]

permissions:
  contents: read

jobs:
  build-and-test:
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-latest, windows-latest]
        build_type: [Debug, Release]

    runs-on: ${{ matrix.os }}

    steps:
    - uses: actions/checkout@v4

    - name: Clone vcpkg
      shell: bash
      run: git clone --depth 1 https://github.com/microsoft/vcpkg.git "${{ runner.temp }}/vcpkg"

    # The SlickQuant dependency libraries (slick-net, hyperliquid-cpp, ...) build
    # their Release config with -march=native, so the binaries vcpkg produces are
    # tuned to whatever CPU the runner that built them happened to have. Restoring
    # such a cache onto a runner with a narrower instruction set makes the tests
    # die with SIGILL on an AVX-512 opcode inside the dependency:
    #
    #   Program received signal SIGILL
    #   => vmovdqu8 %xmm0,0x20(%rbx)
    #      in slick::net::Websocket<...>::Websocket(...)
    #      from hyperliquid::WebsocketManager::init(...)
    #
    # That is why the Linux Release job failed intermittently and passed on
    # re-run: it depended on which CPU the cache came from. Key the cache on the
    # runner's CPU features so a cache is only ever reused on compatible hardware.
    - name: CPU signature for the vcpkg cache key
      id: cpusig
      shell: bash
      run: |
        if [ "$RUNNER_OS" = "Linux" ]; then
          raw=$(grep -m1 '^flags' /proc/cpuinfo | cut -d: -f2 | tr ' ' '\n' \
                | grep -E '^(sse4_[12]|avx|avx2|avx512[a-z0-9_]*|bmi[12]|fma)$' | sort -u | tr '\n' ',')
        else
          raw=$(wmic cpu get Name /value 2>/dev/null | tr -d '\r' | grep '=' || echo "windows-unknown")
        fi
        echo "CPU signature source: $raw"
        echo "sig=$(printf '%s' "$raw" | sha256sum | cut -c1-16)" >> "$GITHUB_OUTPUT"

    # No vcpkg.json manifest is committed in this repo by design (see CLAUDE.md), so
    # dependencies are installed explicitly in classic mode below.
    - name: Cache vcpkg
      uses: actions/cache@v4
      with:
        path: |
          ${{ runner.temp }}/vcpkg/installed
          ${{ runner.temp }}/vcpkg/downloads
          ${{ runner.temp }}/vcpkg/packages
        key: vcpkg-${{ runner.os }}-cpu${{ steps.cpusig.outputs.sig }}-${{ hashFiles('.github/workflows/ci.yml') }}

    - name: Bootstrap vcpkg (Windows)
      if: runner.os == 'Windows'
      shell: pwsh
      run: |
        & "${{ runner.temp }}\vcpkg\bootstrap-vcpkg.bat" -disableMetrics

    - name: Bootstrap vcpkg (Unix)
      if: runner.os != 'Windows'
      shell: bash
      run: |
        "${{ runner.temp }}/vcpkg/bootstrap-vcpkg.sh" -disableMetrics

    - name: Install dependencies via vcpkg
      shell: bash
      run: |
        "${{ runner.temp }}/vcpkg/vcpkg" install \
          quickfix \
          nlohmann-json \
          jwt-cpp \
          foonathan-memory \
          boost-uuid \
          gtest \
          slick-logger \
          slick-socket \
          slick-object-pool \
          coinbase-advanced-cpp \
          slickquant-hyperliquid-cpp \
          "uwebsockets[core,ssl,zlib]"

    - name: Configure CMake
      shell: bash
      run: >
        cmake -B build
        -DCMAKE_BUILD_TYPE=${{ matrix.build_type }}
        -DCMAKE_TOOLCHAIN_FILE="${{ runner.temp }}/vcpkg/scripts/buildsystems/vcpkg.cmake"

    - name: Build
      run: cmake --build build --config ${{ matrix.build_type }} --parallel

    - name: Run tests
      working-directory: build
      run: ctest -C ${{ matrix.build_type }} --output-on-failure

    # The Linux Release job intermittently dies with SIGILL inside the tests that
    # construct an Exchange or a gateway; ctest only reports "Exception: Illegal"
    # with no location. Re-run the failures under gdb to capture where and on
    # which thread, since the crash cannot be reproduced on Windows or in Debug.
    - name: Capture backtrace on test failure (Linux)
      if: failure() && runner.os == 'Linux'
      working-directory: build
      run: |
        sudo apt-get update && sudo apt-get install -y gdb
        ulimit -c unlimited
        # Ask ctest which tests failed, then run each under gdb.
        failed=$(ctest -C ${{ matrix.build_type }} -N --rerun-failed 2>/dev/null \
                 | sed -n 's/^ *Test *#[0-9]*: //p' | head -5)
        if [ -z "$failed" ]; then
          echo "ctest reported no rerun-failed list; falling back to the first exchange test"
          failed="HyperliquidRestTest.PlaceOrder_BuyLimit_Gtc_ReturnsResting"
        fi
        for t in $failed; do
          echo "::group::gdb backtrace for $t"
          gdb -batch -return-child-result \
              -ex "run" \
              -ex "echo \n===== signal / faulting frame =====\n" \
              -ex "bt" \
              -ex "echo \n===== all threads =====\n" \
              -ex "thread apply all bt" \
              -ex "info registers rip" \
              -ex "x/8i \$rip" \
              --args ./tests/slick_sim_tests --gtest_filter="$t" || true
          echo "::endgroup::"
        done

    # Release-config artifacts are uploaded here so release.yml can reuse this
    # exact, already-tested build instead of recompiling from scratch on tag push.
    - name: Stage release artifact (Windows)
      if: matrix.build_type == 'Release' && runner.os == 'Windows'
      shell: pwsh
      run: |
        New-Item -ItemType Directory -Force -Path release-stage | Out-Null
        $exe = Get-ChildItem -Path build -Recurse -Filter slick-sim.exe | Select-Object -First 1
        Copy-Item $exe.FullName -Destination release-stage
        Copy-Item "${{ runner.temp }}\vcpkg\installed\x64-windows\bin\*.dll" -Destination release-stage -ErrorAction SilentlyContinue

    - name: Stage release artifact (Linux)
      if: matrix.build_type == 'Release' && runner.os == 'Linux'
      shell: bash
      run: |
        mkdir -p release-stage
        find build -type f -name slick-sim -exec cp {} release-stage/ \;
        cp "${{ runner.temp }}/vcpkg/installed/x64-linux/lib/"*.so* release-stage/ 2>/dev/null || true

    - name: Upload release build artifact
      if: matrix.build_type == 'Release'
      uses: actions/upload-artifact@v4
      with:
        name: slick-sim-${{ runner.os }}-release
        path: release-stage/*