dotnet-sdk-tests.yml

name: ".NET SDK Tests"

on:
  workflow_dispatch:
  workflow_call:

permissions:
  contents: read

jobs:
  validate:
    name: ".NET SDK Build and Format"
    if: github.event.repository.fork == false
    runs-on: ubuntu-latest
    timeout-minutes: 20
    defaults:
      run:
        shell: bash
        working-directory: ./dotnet
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
      - uses: actions/setup-dotnet@26b0ec14cb23fa6904739307f278c14f94c95bf1 # v5.4.0
        with:
          dotnet-version: "10.0.x"
      - uses: actions/setup-node@249970729cb0ef3589644e2896645e5dc5ba9c38 # v6.5.0
        with:
          node-version: "22"

      - name: Restore .NET dependencies
        run: dotnet restore

      - name: Run dotnet format check
        run: |
          if ! dotnet format --no-restore --verify-no-changes; then
            echo "❌ dotnet format produced changes. Please run 'dotnet format' in dotnet"
            exit 1
          fi
          echo "✅ dotnet format produced no changes"

      # Build the whole solution once to validate every SDK target framework.
      # Matrix cells below only need to build frameworks consumed by the tests.
      - name: Build SDK
        run: dotnet build --no-restore

  test:
    name: ".NET SDK Tests (${{ matrix.os }}, ${{ matrix.transport }}, ${{ matrix.backend }}, ${{ matrix.shard }})"
    if: github.event.repository.fork == false
    env:
      POWERSHELL_UPDATECHECK: Off
      COPILOT_SDK_E2E_BACKEND: ${{ matrix.backend }}
      DOTNET_TEST_FILTER: ${{ matrix.test-filter }}
    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-latest, macos-latest, windows-latest]
        transport: ["default", "inprocess"]
        backend: [capi]
        shard: [full]
        # TODO: Re-enable after fixing in-process sqlite file locking on shutdown on Windows.
        exclude:
          - os: windows-latest
            transport: "inprocess"
          - os: windows-latest
            transport: default
            shard: full
          # TODO(cli-1.0.81-2): CLI 1.0.81-5 still stops completing in-process
          # CAPI model turns, causing repeated per-test timeouts until the
          # 30-minute job limit. Stdio CAPI and in-process BYOK remain enabled.
          - os: ubuntu-latest
            transport: inprocess
          - os: macos-latest
            transport: inprocess
          # The macOS default/capi host runs the whole suite on the smallest
          # runner in the matrix (3 vCPU / 7 GB vs ubuntu's 4 / 16). Since the
          # 1.0.81-2 bump it stopped finishing: the job ran 50+ minutes until
          # the runner lost communication with the server, and a runner that
          # dies that way uploads no logs at all, so the failure could not be
          # diagnosed from CI output. Split it across isolated hosts so one
          # poisoned runtime cannot consume the entire job without diagnostics.
          - os: macos-latest
            transport: default
            shard: full
        include:
          # Keep xUnit serial within each process, but split the slow Windows
          # default-transport suite across two isolated test hosts. Keep both
          # target frameworks in each shard: separate framework jobs did not
          # shorten the critical path and doubled the Windows job count.
          - os: windows-latest
            transport: default
            backend: capi
            shard: "1"
          - os: windows-latest
            transport: default
            backend: capi
            shard: "2"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "1"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2a"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-pending"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-permission"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-auth"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-hooks"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-unit-p"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-provider"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-additional"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-agent"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-event-log"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-event-effects"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-mcp-skills"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-mcp-config"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-mcp-lifecycle"
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2b-rpc-q-z"
          # Standalone extension tests require the legacy JavaScript entrypoint.
          # Keep its subprocess tree isolated from the rest of the O/P/R shard.
          - os: macos-latest
            transport: default
            backend: capi
            shard: extensions
          - os: macos-latest
            transport: default
            backend: capi
            shard: "2c"
          - os: ubuntu-latest
            transport: inprocess
            backend: anthropic-messages
            shard: full
            test-filter: "FullyQualifiedName~GitHub.Copilot.Test.E2E&E2EBackend!=SelfConfiguredBackend&E2EBackend!=CapiOnly"
          - os: ubuntu-latest
            transport: inprocess
            backend: openai-responses
            shard: full
            test-filter: "FullyQualifiedName~GitHub.Copilot.Test.E2E&E2EBackend!=SelfConfiguredBackend&E2EBackend!=CapiOnly"
          - os: ubuntu-latest
            transport: inprocess
            backend: openai-completions
            shard: full
            test-filter: "FullyQualifiedName~GitHub.Copilot.Test.E2E&E2EBackend!=SelfConfiguredBackend&E2EBackend!=CapiOnly"
    runs-on: ${{ matrix.os }}
    # A hung test used to run until the runner died (~50 min) and the dying
    # runner never uploaded its logs, so the failures were undiagnosable.
    # Every healthy cell finishes well under 15 min.
    timeout-minutes: 20
    defaults:
      run:
        shell: bash
        working-directory: ./dotnet
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
      - uses: actions/setup-dotnet@26b0ec14cb23fa6904739307f278c14f94c95bf1 # v5.4.0
        with:
          dotnet-version: "10.0.x"
      - uses: actions/setup-node@249970729cb0ef3589644e2896645e5dc5ba9c38 # v6.5.0
        with:
          node-version: "22"
          cache: "npm"
          cache-dependency-path: |
            ./nodejs/package-lock.json
            ./test/harness/package-lock.json

      - name: Install Node.js dependencies
        working-directory: ./nodejs
        run: npm ci --ignore-scripts

      - name: Restore .NET dependencies
        run: dotnet restore test/GitHub.Copilot.SDK.Test.csproj

      - name: Install test harness dependencies
        working-directory: ./test/harness
        run: npm ci --ignore-scripts

      - name: Warm up PowerShell
        if: runner.os == 'Windows'
        run: pwsh.exe -Command "Write-Host 'PowerShell ready'"

      - name: Select inprocess transport
        if: matrix.transport == 'inprocess'
        run: echo "COPILOT_SDK_DEFAULT_CONNECTION=inprocess" >> "$GITHUB_ENV"

      - name: Run .NET SDK tests
        env:
          COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
          DOTNET_TEST_SHARD: ${{ matrix.shard }}
        run: |
          # --blame-hang names the offending test instead of letting it wedge
          # the runner. No single test legitimately runs for 10 minutes; the
          # whole suite normally finishes in about five.
          # The validation job performs the full analyzer-enabled SDK build.
          # Build only test-consumed frameworks here and do not repeat analyzers.
          args=(
            --no-restore
            -v n
            --blame-hang
            --blame-hang-timeout 10m
            --blame-hang-dump-type none
            --logger "trx;LogFilePrefix=test-results"
            --results-directory "$GITHUB_WORKSPACE/dotnet/TestResults"
            -p:RunAnalyzers=false
          )

          filter="$DOTNET_TEST_FILTER"
          if [[ "$DOTNET_TEST_SHARD" != "full" ]]; then
            case "$DOTNET_TEST_SHARD" in
              1)
                # M moved here after recent runs showed shard 2 was about two
                # minutes slower; this keeps Windows and macOS balanced.
                initials=(A C D H I J K L M N Q S U W Y)
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.ConnectionToken"
                ;;
              2)
                initials=(B E F G O P R T V X Z)
                shard_filter=""
                ;;
              2a)
                initials=(B E F G)
                shard_filter=""
                ;;
              2b-pending)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.PendingWorkResumeE2ETests"
                ;;
              2b-permission)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.PermissionE2ETests"
                ;;
              2b-auth)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.PerSessionAuthE2ETests"
                ;;
              2b-hooks)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.PreMcpToolCallHookE2ETests"
                ;;
              2b-unit-p)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.Unit.P"
                ;;
              2b-provider)
                initials=(O)
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.ProviderEndpointE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RewindE2ETests|FullyQualifiedName~GitHub.Copilot.Test.Unit.R"
                ;;
              2b-rpc-additional)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcAdditionalEdgeCasesE2ETests"
                ;;
              2b-rpc-agent)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcAgentE2ETests"
                ;;
              2b-rpc-event-log)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcEventLogE2ETests"
                ;;
              2b-rpc-event-effects)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcEventSideEffectsE2ETests"
                ;;
              2b-rpc-mcp-skills)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcMcpAndSkillsE2ETests"
                ;;
              2b-rpc-mcp-config)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcMcpConfigE2ETests"
                ;;
              2b-rpc-mcp-lifecycle)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcMcpLifecycleE2ETests"
                ;;
              2b-rpc-q-z)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcQueueE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcRemoteE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcScheduleE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcServerE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcServerMiscE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcServerPluginsE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcServerRemoteControlE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcSessionStateE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcSessionStateExtrasE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcShellAndFleetE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcShellEdgeCaseE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcShellUserRequestedE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcTasksAndHandlersE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcUiEphemeralQueryE2ETests|FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcWorkspaceCheckpointsE2ETests"
                ;;
              extensions)
                initials=()
                shard_filter="FullyQualifiedName~GitHub.Copilot.Test.E2E.RpcExtensionsLoadedE2ETests"
                ;;
              2c)
                initials=(T V X Z)
                shard_filter=""
                ;;
            esac

            for namespace in E2E Unit; do
              for initial in "${initials[@]}"; do
                clause="FullyQualifiedName~GitHub.Copilot.Test.${namespace}.${initial}"
                shard_filter="${shard_filter:+${shard_filter}|}${clause}"
              done
            done
            filter="${filter:+(${filter})&}(${shard_filter})"
          fi

          if [[ -n "$filter" ]]; then
            args+=(--filter "$filter")
          fi
          dotnet test test/GitHub.Copilot.SDK.Test.csproj "${args[@]}"

      - name: Upload .NET test diagnostics
        if: failure()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: dotnet-test-diagnostics-${{ matrix.os }}-${{ matrix.transport }}-${{ matrix.backend }}-${{ matrix.shard }}-${{ github.run_attempt }}
          path: dotnet/TestResults/
          if-no-files-found: warn
          retention-days: 7

  # JavaScript actions use a glibc-linked Node runtime, so Alpine runs through Docker.
  alpine-arm64:
    name: ".NET SDK Tests (Alpine ARM64, ${{ matrix.transport }}, CAPI)"
    if: github.event.repository.fork == false
    runs-on: ubuntu-24.04-arm
    timeout-minutes: 20
    env:
      COPILOT_SDK_E2E_BACKEND: capi
    strategy:
      fail-fast: false
      matrix:
        transport: ["default", "inprocess"]
    steps:
      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2

      - name: Run ${{ matrix.transport }} tests on ARM64 Alpine
        uses: ./.github/actions/run-alpine-tests
        env:
          COPILOT_HMAC_KEY: ${{ secrets.COPILOT_DEVELOPER_CLI_INTEGRATION_HMAC_KEY }}
        with:
          image: mcr.microsoft.com/dotnet/sdk:10.0-alpine
          workdir: /workspace
          transport: ${{ matrix.transport }}
          command: |
            dotnet --info | grep -Eq "RID:[[:space:]]+linux-musl-arm64"

            apk add --no-cache nodejs npm
            (cd nodejs && npm ci --ignore-scripts)
            (cd test/harness && npm ci --ignore-scripts)

            export DOTNET_ROLL_FORWARD=Major
            dotnet restore dotnet/test/GitHub.Copilot.SDK.Test.csproj

            dotnet test dotnet/test/GitHub.Copilot.SDK.Test.csproj \
              --no-restore \
              --framework net8.0 \
              -v n \
              --blame-hang \
              --blame-hang-timeout 10m \
              --blame-hang-dump-type none \
              --logger "trx;LogFilePrefix=test-results" \
              --results-directory /workspace/dotnet/TestResults \
              -p:RunAnalyzers=false

      - name: Upload .NET test diagnostics
        if: failure()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: dotnet-test-diagnostics-alpine-arm64-${{ matrix.transport }}-capi-${{ github.run_attempt }}
          path: dotnet/TestResults/
          if-no-files-found: warn
          retention-days: 7