mcpField
D
dotnet-gha-build-test

dotnet-gha-build-test

github.com/wshaddixCloud & DevOps8676 installsUpdated 35w ago
Repository

About dotnet-gha-build-test

A skill for configuring .NET build and test pipelines in GitHub Actions. It covers SDK setup, NuGet caching and authentication, result and coverage reporting, multi-TFM testing, and test sharding.

Key features

  1. 01.NET SDK setup and NuGet authentication
  2. 02NuGet restore caching
  3. 03Test result publishing
  4. 04Coverage report uploads
  5. 05Multi-TFM matrix testing
  6. 06Test sharding strategies

Use cases

  1. 01Configuring a .NET GitHub Actions pipeline
  2. 02Caching restores across CI runs
  3. 03Publishing test and coverage results
  4. 04Splitting large test suites across jobs

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/wshaddix/dotnet-skills --skill dotnet-gha-build-test

After installing, type/dotnet-gha-build-testin Claude Code to confirm it loaded.

Details

From the project README

dotnet-gha-build-test

.NET build and test workflow patterns for GitHub Actions: actions/setup-dotnet@v4 configuration with multi-version installs and NuGet authentication, NuGet restore caching for fast CI, dotnet test with result publishing via dorny/test-reporter, code coverage upload to Codecov and Coveralls, multi-TFM matrix testing across net8.0 and net9.0, and test sharding strategies for large projects.

Version assumptions: actions/setup-dotnet@v4 for .NET 8/9/10 support. dorny/test-reporter@v1 for test result visualization. Codecov and Coveralls GitHub Apps for coverage reporting.

Scope boundary: This skill owns .NET build and test pipeline configuration for GitHub Actions. Starter CI templates (basic build/test/pack) are owned by [skill:dotnet-add-ci]. Composable workflow patterns (reusable workflows, matrix strategies, caching) are in [skill:dotnet-gha-patterns]. Testing strategy guidance (what to test, test architecture, quality gates) is owned by [skill:dotnet-testing-strategy]. Benchmark CI workflows are owned by [skill:dotnet-ci-benchmarking].

Out of scope: Starter CI templates -- see [skill:dotnet-add-ci]. Test architecture and strategy -- see [skill:dotnet-testing-strategy]. Benchmark regression detection in CI -- see [skill:dotnet-ci-benchmarking]. Publishing and deployment -- see [skill:dotnet-gha-publish] and [skill:dotnet-gha-deploy]. Azure DevOps build/test pipelines -- see [skill:dotnet-ado-build-test].

Cross-references: [skill:dotnet-add-ci] for starter build/test templates, [skill:dotnet-testing-strategy] for test architecture guidance, [skill:dotnet-ci-benchmarking] for benchmark CI integration, [skill:dotnet-artifacts-output] for artifact upload path adjustments when using centralized build output layout.


actions/setup-dotnet@v4 Configuration

Basic Setup
steps:
  - uses: actions/checkout@v4

  - name: Setup .NET
    uses: actions/setup-dotnet@v4
    with:
      dotnet-version: '8.0.x'
Multi-Version Install

Install multiple SDK versions for multi-TFM builds within a single job:

- name: Setup .NET SDKs
  uses: actions/setup-dotnet@v4
  with:
    dotnet-version: |
      8.0.x
      9.0.x

The first listed version becomes the default dotnet on PATH. All installed versions are available via --framework targeting.

NuGet Authentication for Private Feeds

Configure NuGet source authentication via actions/setup-dotnet@v4:

- name: Setup .NET with NuGet auth
  uses: actions/setup-dotnet@v4
  with:
    dotnet-version: '8.0.x'
    source-url: https://nuget.pkg.github.com/${{ github.repository_owner }}/index.json
  env:
    NUGET_AUTH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

For multiple private feeds, configure additional sources after setup:

- name: Setup .NET
  uses: actions/setup-dotnet@v4
  with:
    dotnet-version: '8.0.x'

- name: Add private NuGet feed
  run: |
    set -euo pipefail
    dotnet nuget add source https://pkgs.dev.azure.com/myorg/_packaging/myfeed/nuget/v3/index.json \
      --name AzureArtifacts \
      --username az \
      --password ${{ secrets.AZURE_ARTIFACTS_PAT }} \
      --store-password-in-clear-text

The --store-password-in-clear-text flag is required on Linux runners where DPAPI encryption is unavailable.

Global.json SDK Version Pinning

When global.json exists in the repository root, actions/setup-dotnet@v4 can read it automatically:

- name: Setup .NET from global.json
  uses: actions/setup-dotnet@v4
  with:
    global-json-file: global.json

This ensures CI uses the same SDK version as local development.


NuGet Restore Caching

Standard Cache Configuration
- name: Cache NuGet packages
  uses: actions/cache@v4
  with:
    path: ~/.nuget/packages
    key: nuget-${{ runner.os }}-${{ hashFiles('**/*.csproj', '**/Directory.Packages.props') }}
    restore-keys: |
      nuget-${{ runner.os }}-

- name: Restore dependencies
  run: dotnet restore MySolution.sln
Built-in Cache with setup-dotnet

actions/setup-dotnet@v4 has built-in caching support using packages.lock.json:

- name: Setup .NET with caching
  uses: actions/setup-dotnet@v4
  with:
    dotnet-version: '8.0.x'
    cache: true
    cache-dependency-path: '**/packages.lock.json'

Generate lock files locally first: dotnet restore --use-lock-file. Commit packages.lock.json files for deterministic restore.

Cache Key Strategy
Key ComponentPurpose
runner.osPrevent cross-OS cache collisions
hashFiles('**/*.csproj')Invalidate when package references change
hashFiles('**/Directory.Packages.props')Invalidate when centrally managed versions change
restore-keys prefixPartial match for incremental cache reuse

Test Result Publishing

dorny/test-reporter

Publish dotnet test results as GitHub Actions check annotations with inline failure details:

- name: Test
  run: |
    set -euo pipefail
    dotnet test MySolution.sln \
      --configuration Release \
      --logger "trx;LogFileName=test-results.trx" \
      --results-directory ./test-results
  continue-on-error: true
  id: test

- name: Publish test results
  uses: dorny/test-reporter@v1
  if: always()
  with:
    name: '.NET Test Results'
    path: 'test-results/**/*.trx'
    reporter: dotnet-trx
    fail-on-error: true

Key decisions:

  • continue-on-error: true on the test step ensures the reporter step always runs, even on failures
  • if: always() on the reporter step publishes results regardless of test outcome
  • fail-on-error: true on the reporter marks the check as failed when tests fail
Alternative: EnricoMi/publish-unit-test-result-action

For richer PR comment integration with test counts:

- name: Publish test results
  uses: EnricoMi/publish-unit-test-result-action@v2
  if: always()
  with:
    files: 'test-results/**/*.trx'
    check_name: 'Test Results'

Code Coverage Upload

Codecov
- name: Test with coverage
  run: |
    set -euo pipefail
    dotnet test MySolution.sln \
      --configuration Release \
      --collect:"XPlat Code Coverage" \
      --results-directory ./coverage

- name: Upload coverage to Codecov
  uses: codecov/codecov-action@v4
  with:
    directory: ./coverage
    fail_ci_if_error: false
    token: ${{ secrets.CODECOV_TOKEN }}
Coveralls
- name: Test with coverage
  run: |
    set -euo pipefail
    dotnet test MySolution.sln \
      --configuration Release \
      --collect:"XPlat Code Coverage" \
      --results-directory ./coverage

- name: Upload coverage to Coveralls
  uses: coverallsapp/github-action@v2
  with:
    file: coverage/**/coverage.cobertura.xml
    format: cobertura
    github-token: ${{ secrets.GITHUB_TOKEN }}
Coverage Report Generation with ReportGenerator

Generate human-readable HTML coverage reports alongside CI upload:

- name: Generate coverage report
  run: |
    set -euo pipefail
    dotnet tool install -g dotnet-reportgenerator-globaltool
    reportgenerator \
      -reports:coverage/**/coverage.cobertura.xml \
      -targetdir:coverage-report \
      -reporttypes:HtmlInline_AzurePipelines\;Cobertura

- name: Upload coverage report
  uses: actions/upload-artifact@v4
  with:
    name: coverage-report
    path: coverage-report/
    retention-days: 30

Multi-TFM Matrix Testing

Matrix Strategy for TFMs
jobs:
  test:
    strategy:
      fail-fast: false
      matrix:
        tfm: [net8.0, net9.0]
        os: [ubuntu-latest, windows-latest]
    runs-on: ${{ matrix.os }}
    steps:
      - uses: actions/checkout@v4

      - name: Setup .NET
        uses: actions/setup-dotnet@v4
        with:
          dotnet-version: |
            8.0.x
            9.0.x

      - name: Cache NuGet
        uses: actions/cache@v4
        with:
          path: ~/.nuget/packages
          key: nuget-${{ runner.os }}-${{ hashFiles('**/*.csproj', '**/Directory.Packages.props') }}
          restore-keys: |
            nuget-${{ runner.os }}-

      - name: Test ${{ matrix.tfm }}
        run: |
          set -euo pipefail
          dotnet test MySolution.sln \
            --framework ${{ matrix.tfm }} \
            --configuration Release \
            --logger "trx;LogFileName=${{ matrix.tfm }}-results.trx" \
            --results-directory ./test-results

      - name: Publish test results
        uses: dorny/test-reporter@v1
        if: always()
        with:
          name: 'Tests (${{ matrix.os }} / ${{ matrix.tfm }})'
          path: 'test-results/**/*.trx'
          reporter: dotnet-trx
Install All Required SDKs

When running multi-TFM tests in a single job instead of a matrix, install all required SDKs upfront:

- name: Setup .NET SDKs
  uses: actions/setup-dotnet@v4
  with:
    dotnet-version: |
      8.0.x
      9.0.x

- name: Test all TFMs
  run: dotnet test MySolution.sln --configuration Release

Without the matching SDK installed, dotnet test cannot build for that TFM and fails with NETSDK1045.


Test Sharding for Large Projects

Splitting Tests Across Parallel Jobs

For large test suites, split test projects across parallel runners to reduce total CI time:

jobs:
  discover:
    runs-on: ubuntu-latest
    outputs:
      projects: ${{ steps.find.outputs.projects }}
    steps:
      - uses: actions/checkout@v4
      - id: find
        shell: bash
        run: |
          set -euo pipefail
          PROJECTS=$(find tests -name '*.csproj' | jq -R . | jq -sc .)
          echo "projects=$PROJECTS" >> "$GITHUB_OUTPUT"

  test:
    needs: discover
    strategy:
      fail-fast: false
      matrix:
        project: ${{ fromJson(needs.discover.outputs.projects) }}
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Setup .NET
        uses: actions/setup-dotnet@v4
        with:
          dotnet-version: '8.0.x'

      - name: Test ${{ matrix.project }}
        run: |
          set -euo pipefail
          dotnet test ${{ matrix.project }} \
            --configuration Release \
            --logger "trx;LogFileName=results.trx" \
            --results-directory ./test-results

      - name: Publish test results
        uses: dorny/test-reporter@v1
        if: always()
        with:
          name: 'Tests - ${{ matrix.project }}'
          path: 'test-results/**/*.trx'
          reporter: dotnet-trx
Sharding by Test Class Within a Project

For a single large test project, use dotnet test --filter to split by namespace:

jobs:
  test:
    strategy:
      fail-fast: false
      matrix:
        shard: ['Unit', 'Integration', 'EndToEnd']
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Setup .NET
        uses: actions/setup-dotnet@v4
        with:
          dotnet-version: '8.0.x'

      - name: Test ${{ matrix.shard }}
        run: |
          set -euo pipefail
          dotnet test tests/MyApp.Tests.csproj \
            --configuration Release \
            --filter "FullyQualifiedName~${{ matrix.shard }}" \
            --logger "trx;LogFileName=${{ matrix.shard }}-results.trx" \
            --results-directory ./test-results

Agent Gotchas

  1. Always set set -euo pipefail in multi-line bash run blocks -- without pipefail, piped commands that fail do not propagate the error, producing false-green CI.
  2. Use continue-on-error: true on the test step, not on the reporter -- the test step must not fail the job prematurely so the reporter can publish results, but the reporter should fail the check when tests fail.
  3. Include runner.os in NuGet cache keys -- NuGet packages have OS-specific native assets; cross-OS cache hits cause restore failures.
  4. Install all required SDK versions for multi-TFM -- dotnet test without the matching SDK produces NETSDK1045; list every required version in dotnet-version.
  5. Do not hardcode TFM strings in workflow files -- use matrix variables to keep workflow files in sync with project configuration; hardcoded net8.0 in CI breaks when the project moves to net9.0.
  6. Coverage collection requires --collect:"XPlat Code Coverage" -- the default dotnet test does not produce coverage files; the XPlat Code Coverage collector is built into the .NET SDK.
  7. TRX logger path must match reporter glob -- if the logger writes to test-results/results.trx, the reporter path must include that directory in its glob pattern.
  8. Never commit NuGet credentials to workflow files -- use ${{ secrets.* }} references for all authentication tokens; the NUGET_AUTH_TOKEN environment variable is the standard pattern.

Comments

Markdown supported