Test Implementation: Code Coverage & Pipeline Tests

Implementing tests in pipelines

Test tasks in Azure Pipelines

Azure Pipelines provides built-in tasks for running tests across different tech stacks:

Key test tasks

Task	What It Runs	Coverage Built-in	Result Format
DotNetCoreCLI@2	dotnet test for .NET projects	Yes — add --collect:"XPlat Code Coverage"	VSTest (.trx) or JUnit
VSTest@2	Visual Studio Test runner for .NET Framework	Yes — enable code coverage in task settings	VSTest (.trx)
Npm@1	npm test for Node.js projects	No — configure Jest/Istanbul separately	Depends on test runner config
Maven@4	mvn test for Java projects	No — configure JaCoCo plugin in pom.xml	JUnit (surefire reports)
PythonScript@0	pytest or custom Python test scripts	No — configure pytest-cov separately	JUnit (with —junitxml flag)

PublishTestResults task

After tests run, the PublishTestResults@2 task uploads results to the pipeline’s Test tab:

- task: PublishTestResults@2
  inputs:
    testResultsFormat: 'JUnit'    # JUnit, NUnit, VSTest, XUnit, CTest
    testResultsFiles: '**/TEST-*.xml'
    mergeTestResults: true

Supported formats: JUnit, NUnit, VSTest (.trx), XUnit, CTest. JUnit is the most universal — most test frameworks can output JUnit XML.

💡 Exam tip: format matching

The exam may ask which format a specific test framework produces:

• .NET (dotnet test) — VSTest (.trx) by default, JUnit with --logger "junit"
• Java (Maven Surefire) — JUnit XML in target/surefire-reports/
• JavaScript (Jest) — JUnit with jest-junit reporter package
• Python (pytest) — JUnit with --junitxml=results.xml flag

If the question says “publish .NET test results,” the format is typically VSTest. If it says “a cross-platform format that works with any language,” the answer is JUnit.

Test steps in GitHub Actions

GitHub Actions doesn’t have built-in “test tasks” — you use run steps with your test framework’s CLI commands:

- name: Run tests
  run: dotnet test --logger "junit" --collect:"XPlat Code Coverage"

- name: Publish test results
  uses: dorny/test-reporter@v1
  if: always()
  with:
    name: 'Test Results'
    path: '**/*.xml'
    reporter: 'java-junit'

Key differences from Azure Pipelines:

• No built-in PublishTestResults equivalent — use community actions like dorny/test-reporter or EnricoMi/publish-unit-test-result-action
• Results appear as check run annotations on the PR, not a dedicated Test tab
• The if: always() condition is critical — without it, test publishing is skipped when tests fail (which is exactly when you need the results)

Azure Pipelines has richer built-in test integration; GitHub Actions relies on community actions and external services

Aspect	Azure Pipelines	GitHub Actions
Test execution	Built-in tasks (DotNetCoreCLI, VSTest, Maven, Npm)	run steps with CLI commands (dotnet test, npm test, pytest)
Result publishing	PublishTestResults@2 task — native integration with Test tab	Community actions (dorny/test-reporter) — results as PR check annotations
Coverage publishing	PublishCodeCoverageResults@2 task — native Code Coverage tab	Community actions or upload to external services (Codecov, Coveralls)
Supported formats	JUnit, NUnit, VSTest, XUnit, CTest (5 formats)	Depends on action — typically JUnit, plus framework-specific parsers
Test analytics	Built-in pipeline analytics — flaky test detection, pass rate trends	No built-in analytics — use third-party services or custom dashboards
Distributed testing	VSTest multi-agent parallelism with test slicing	Matrix strategy — run different test slices across parallel jobs

Code coverage analysis

Code coverage measures how much of your codebase is exercised by your test suite. It answers: “When tests run, which lines of code actually execute?”

Coverage metrics

Metric	What It Measures	Example
Line coverage	Percentage of executable lines hit by tests	450 of 500 lines executed = 90%
Branch coverage	Percentage of conditional branches (if/else, switch) taken	Both the if and else paths tested
Function coverage	Percentage of functions/methods called	38 of 40 functions invoked = 95%
Statement coverage	Percentage of statements executed (similar to line but counts multi-statement lines)	Each statement on a multi-statement line counted separately

Branch coverage is the most valuable metric — high line coverage with low branch coverage means your tests hit the happy path but miss edge cases and error handling.

Coverage tools by language

Language	Coverage Tool	Output Format	Pipeline Task
.NET	Coverlet (built-in with dotnet test)	Cobertura XML	PublishCodeCoverageResults@2
Java	JaCoCo	JaCoCo XML or Cobertura	PublishCodeCoverageResults@2
JavaScript/TypeScript	Istanbul (via nyc or Jest —coverage)	Cobertura, lcov, text	PublishCodeCoverageResults@2
Python	coverage.py (via pytest-cov)	Cobertura XML	PublishCodeCoverageResults@2

Cobertura is the universal format — every major coverage tool can output Cobertura XML, and both Azure Pipelines and most GitHub Actions coverage tools accept it.

Publishing coverage in Azure Pipelines

- task: PublishCodeCoverageResults@2
  inputs:
    summaryFileLocation: '$(System.DefaultWorkingDirectory)/**/coverage.cobertura.xml'

This publishes coverage results to the pipeline’s Code Coverage tab, showing line-by-line highlights of covered and uncovered code.

💡 Scenario: Jordan's test configuration at Cloudstream Media

☁️ Jordan Rivera sets up the test pipeline for Cloudstream Media’s Node.js streaming API:

1. Unit tests — Jest with --coverage flag, outputs Cobertura XML
2. Integration tests — Supertest against a Docker Compose environment (API + Redis + PostgreSQL)
3. Coverage threshold — 80% branch coverage minimum, enforced in Jest config (coverageThreshold in package.json)
4. Pipeline publishing — PublishTestResults for JUnit XML, PublishCodeCoverageResults for Cobertura

Avery (dev lead) asks: “What happens if I add a new endpoint but don’t write tests?” Jordan’s answer: “Branch coverage drops below 80%, Jest fails with a threshold violation, the pipeline fails, and your PR can’t merge.”

Chen (SRE) adds load tests separately — they run on a schedule (nightly) rather than on every PR, because they take 15 minutes and hit shared infrastructure.

Distributed testing

Large test suites can take too long on a single agent. Both platforms support distributing tests across multiple machines.

Azure Pipelines — VSTest multi-agent

The VSTest@2 task supports test slicing across multiple agents:

• By test count — each agent runs an equal number of tests
• By previous run duration — agents get balanced workloads based on historical timing
• By test assemblies — each agent runs a different DLL

Configure a job with strategy: parallel and the desired agent count. The VSTest task automatically partitions tests.

GitHub Actions — matrix strategy

Use a matrix to split tests across parallel jobs:

strategy:
  matrix:
    shard: [1, 2, 3, 4]
steps:
  - run: npm test -- --shard=${{ matrix.shard }}/4

Each job runs a different shard. Combine results with a downstream job that aggregates coverage reports.

Question

What is the PublishTestResults task in Azure Pipelines, and what formats does it support?

Click or press Enter to reveal answer

Answer

PublishTestResults@2 uploads test results to the pipeline's Test tab for visualisation and trending. Supported formats: JUnit, NUnit, VSTest (.trx), XUnit, and CTest. JUnit is the most universal — most test frameworks can output JUnit XML regardless of language.

Click to flip back

Question

What is the difference between line coverage and branch coverage?

Click or press Enter to reveal answer

Answer

Line coverage measures the percentage of executable lines hit by tests. Branch coverage measures the percentage of conditional paths (if/else branches) taken. Branch coverage is more valuable — you can have 90% line coverage but miss error-handling branches entirely. High line + low branch coverage means tests only cover the happy path.

Click to flip back

Question

Which code coverage format is universal across languages?

Click or press Enter to reveal answer

Answer

Cobertura XML. Every major coverage tool (.NET Coverlet, Java JaCoCo, JavaScript Istanbul, Python coverage.py) can output Cobertura format. Azure Pipelines PublishCodeCoverageResults@2 task accepts Cobertura natively. It is the safest choice for cross-language pipelines.

Click to flip back

Question

Why is 'if: always()' important when publishing test results in GitHub Actions?

Click or press Enter to reveal answer

Answer

Without 'if: always()', the publish step only runs when the previous step succeeds. But when tests fail, the test-run step has a non-zero exit code — so subsequent steps are skipped by default. This means you lose test results exactly when you need them most. 'if: always()' ensures the publish step runs regardless of test outcome.

Click to flip back

Knowledge check

Knowledge Check

Jordan's .NET microservice pipeline runs 'dotnet test' and produces coverage data. He wants to see line-by-line coverage highlights in the Azure Pipelines UI. Which task and format should he use?

APublishTestResults@2 with JUnit formatBPublishCodeCoverageResults@2 with Cobertura formatCPublishTestResults@2 with VSTest formatDPublishCodeCoverageResults@2 with JaCoCo format

Check Answer

Knowledge Check

Nadia's team has a Python test suite using pytest. The tests produce JUnit XML and Cobertura coverage. In Azure Pipelines, which two tasks does she need to add after the test step?

AVSTest@2 and PublishCodeCoverageResults@2BPublishTestResults@2 (JUnit format) and PublishCodeCoverageResults@2 (Cobertura format)CDotNetCoreCLI@2 and PublishTestResults@2DPythonScript@0 and PublishCodeCoverageResults@2

Check Answer

Knowledge Check

Kai's team at Launchpad Labs has 2,000 Jest tests that take 20 minutes on a single GitHub Actions runner. What is the most effective way to reduce the test time?

ASwitch to a self-hosted runner with more CPU coresBUse matrix strategy to shard tests across 4 parallel jobsCMove all tests to a nightly scheduled pipelineDConvert Jest tests to use the VSTest runner

Check Answer

🎬 Video coming soon

Next up: Azure Pipelines: YAML from Scratch — master pipeline structure, triggers, and multi-stage CI pipelines in Azure DevOps.