dotnet-test-quality（点网测试质量）

dotnet-test-quality

Test quality analysis for.NET projects. Covers code coverage collection with coverlet, human-readable coverage reports with ReportGenerator, CRAP (Change Risk Anti-Patterns) score analysis to identify undertested complex code, mutation testing with Stryker.NET to evaluate test suite effectiveness, and strategies for detecting and managing flaky tests.

Version assumptions: Coverlet 6.x+, ReportGenerator 5.x+, Stryker.NET 4.x+ (.NET 8.0+ baseline). Coverlet supports both the MSBuild integration (coverlet.msbuild) and the coverlet.collector data collector; examples use coverlet.collector as the recommended approach.

Out of scope: Test project scaffolding (creating projects, package references, coverlet setup) is owned by [skill:dotnet-add-testing]. Testing strategy and test type decisions are covered by [skill:dotnet-testing-strategy]. CI test reporting and pipeline integration -- see [skill:dotnet-gha-build-test] and [skill:dotnet-ado-build-test].

Prerequisites: Test project already scaffolded via [skill:dotnet-add-testing] with coverlet packages referenced..NET 8.0+ baseline required.

Cross-references: [skill:dotnet-testing-strategy] for deciding what to test and coverage target guidance, [skill:dotnet-xunit] for xUnit test framework features and configuration.

Code Coverage with Coverlet

Coverlet is the standard open-source code coverage library for.NET. It instruments assemblies at build time or via a data collector and produces coverage reports in multiple formats.

Packages

<!-- Data collector approach (recommended) -->
<PackageReference Include="coverlet.collector" Version="8.0.0">
  <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
  <PrivateAssets>all</PrivateAssets>
</PackageReference>

Collecting Coverage

# Collect coverage with Cobertura output (default for ReportGenerator)
dotnet test --collect:"XPlat Code Coverage"

# Specify output format explicitly
dotnet test --collect:"XPlat Code Coverage" \
  -- DataCollectionRunSettings.DataCollectors.DataCollector.Configuration.Format=cobertura

# Multiple formats
dotnet test --collect:"XPlat Code Coverage" \
  -- DataCollectionRunSettings.DataCollectors.DataCollector.Configuration.Format=cobertura,opencover

Coverage results are written to TestResults/<guid>/coverage.cobertura.xml under each test project's output directory.

Filtering Coverage

Exclude generated code, test projects, or specific namespaces:

dotnet test --collect:"XPlat Code Coverage" \
  -- DataCollectionRunSettings.DataCollectors.DataCollector.Configuration.Exclude="[*.Tests]*,[*.IntegrationTests]*" \
  DataCollectionRunSettings.DataCollectors.DataCollector.Configuration.ExcludeByAttribute="GeneratedCodeAttribute,ObsoleteAttribute,ExcludeFromCodeCoverageAttribute"

Or configure via a runsettings file for repeatability:

<!-- coverlet.runsettings -->
<?xml version="1.0" encoding="utf-8"?>
<RunSettings>
  <DataCollectionRunSettings>
    <DataCollectors>
      <DataCollector friendlyName="XPlat Code Coverage">
        <Configuration>
          <Format>cobertura</Format>
          <Exclude>[*.Tests]*,[*.IntegrationTests]*</Exclude>
          <ExcludeByAttribute>
            GeneratedCodeAttribute,ObsoleteAttribute,ExcludeFromCodeCoverageAttribute
          </ExcludeByAttribute>
          <ExcludeByFile>**/Migrations/**</ExcludeByFile>
          <IncludeTestAssembly>false</IncludeTestAssembly>
        </Configuration>
      </DataCollector>
    </DataCollectors>
  </DataCollectionRunSettings>
</RunSettings>

dotnet test --settings coverlet.runsettings

Merge Coverage from Multiple Test Projects

When a solution has multiple test projects, merge their coverage into a single report:

# Run all tests, collecting coverage per project
dotnet test --collect:"XPlat Code Coverage"

# Find all coverage files and merge via ReportGenerator (see next section)

Coverage Reports with ReportGenerator

ReportGenerator converts raw coverage data (Cobertura, OpenCover) into human-readable HTML reports with line-level highlighting.

Installation

# Install as a global tool
dotnet tool install -g dotnet-reportgenerator-globaltool

# Or as a local tool
dotnet tool install dotnet-reportgenerator-globaltool

Generating Reports

# Single coverage file
reportgenerator \
  -reports:"tests/MyApp.Tests/TestResults/*/coverage.cobertura.xml" \
  -targetdir:"coverage-report" \
  -reporttypes:"Html;TextSummary"

# Multiple test projects (glob pattern merges automatically)
reportgenerator \
  -reports:"**/TestResults/*/coverage.cobertura.xml" \
  -targetdir:"coverage-report" \
  -reporttypes:"Html;Cobertura;TextSummary"

Report Types

Example: Full Coverage Pipeline

#!/bin/bash
# clean previous results
rm -rf coverage-report TestResults

# run tests with coverage
dotnet test --collect:"XPlat Code Coverage" --results-directory TestResults

# generate merged HTML report
reportgenerator \
  -reports:"**/TestResults/*/coverage.cobertura.xml" \
  -targetdir:"coverage-report" \
  -reporttypes:"Html;TextSummary;Badges"

# display summary
cat coverage-report/Summary.txt

Setting Coverage Thresholds

Enforce minimum coverage in CI by parsing the text summary or using a threshold parameter:

# ReportGenerator does not enforce thresholds directly.
# Parse the summary or use dotnet-coverage (Microsoft) for threshold enforcement.

# Alternative: use coverlet's built-in threshold via MSBuild
dotnet test /p:CollectCoverage=true \
  /p:Threshold=80 \
  /p:ThresholdType=line \
  /p:ThresholdStat=total

Note: The /p:Threshold parameter requires the coverlet.msbuild package (not coverlet.collector). For coverlet.collector workflows, enforce thresholds by parsing the ReportGenerator text summary in your CI script.

CRAP Analysis

CRAP (Change Risk Anti-Patterns) scores identify methods that are both complex and poorly tested. A high CRAP score means the method has high cyclomatic complexity and low code coverage -- a risky combination.

Formula

CRAP(m) = complexity(m)^2 * (1 - coverage(m)/100)^3 + complexity(m)

Where:

• complexity(m) = cyclomatic complexity of method m
• coverage(m) = code coverage percentage of method m (0-100)

Interpreting CRAP Scores

Generating CRAP Reports

ReportGenerator includes CRAP analysis when using OpenCover format as input:

# Step 1: Collect coverage in OpenCover format
dotnet test --collect:"XPlat Code Coverage" \
  -- DataCollectionRunSettings.DataCollectors.DataCollector.Configuration.Format=opencover

# Step 2: Generate report with risk hotspot analysis
reportgenerator \
  -reports:"**/TestResults/*/coverage.opencover.xml" \
  -targetdir:"coverage-report" \
  -reporttypes:"Html;RiskHotspots"

The Risk Hotspots report highlights methods sorted by CRAP score, showing:

• Method name and containing class
• Cyclomatic complexity
• Code coverage percentage
• Computed CRAP score

Using CRAP Scores Effectively

// Example: a method with high complexity and low coverage
// Cyclomatic complexity: 12, Coverage: 20%
// CRAP = 12^2 * (1 - 0.20)^3 + 12 = 144 * 0.512 + 12 = 85.7 (Critical)
public decimal CalculateShipping(Order order)
{
    if (order.Items.Count == 0) return 0;

    decimal baseRate = order.DestinationCountry switch
    {
        "US" => 5.99m,
        "CA" => 9.99m,
        "UK" => 12.99m,
        _ => 19.99m
    };

    if (order.Total > 100) baseRate *= 0.5m;
    if (order.IsPriority) baseRate *= 2.0m;
    if (order.Items.Any(i => i.IsFragile)) baseRate += 4.99m;
    if (order.Items.Any(i => i.IsOversized)) baseRate += 14.99m;
    if (order.HasInsurance) baseRate += order.Total * 0.02m;
    if (order.IsExpedited && order.DestinationCountry != "US") baseRate *= 1.5m;

    return Math.Round(baseRate, 2);
}

Address high CRAP scores by:

1. Adding targeted tests for uncovered branches to reduce the score via higher coverage
2. Reducing complexity by extracting methods (e.g., separate CalculateBaseRate and ApplySurcharges methods)
3. Both -- the most effective approach combines better coverage with simpler methods

Mutation Testing with Stryker.NET

Mutation testing evaluates test suite quality by introducing small changes (mutations) to production code and checking whether tests detect them. If a mutation survives (tests still pass), the test suite has a gap.

Installation

# Install as a global tool
dotnet tool install -g dotnet-stryker

# Or as a local tool (recommended for team consistency)
dotnet tool install dotnet-stryker

Running Stryker.NET

# From the test project directory
cd tests/MyApp.Tests
dotnet stryker

# Specify the source project explicitly
dotnet stryker --project MyApp.csproj

# Target specific files
dotnet stryker --mutate "src/Services/**/*.cs"

Configuration File

Create stryker-config.json in the test project directory:

{
  "$schema": "https://raw.githubusercontent.com/stryker-mutator/stryker-net/master/src/Stryker.Core/Stryker.Core/stryker-config.schema.json",
  "stryker-config": {
    "project": "MyApp.csproj",
    "reporters": ["html", "progress", "cleartext"],
    "mutation-level": "Standard",
    "thresholds": {
      "high": 80,
      "low": 60,
      "break": 50
    },
    "mutate": [
      "src/Services/**/*.cs",
      "!src/Services/Migrations/**/*.cs"
    ],
    "ignore-mutations": [
      "string",
      "linq"
    ]
  }
}

Understanding Mutation Results

Stryker reports mutations in four categories:

Mutation Score

Mutation Score = Killed / (Killed + Survived + NoCoverage) * 100

A mutation score of 80%+ indicates a strong test suite. Below 60% suggests significant gaps.

Example: Identifying Test Gaps

Given this production code:

public class PricingService
{
    public decimal CalculateDiscount(decimal price, CustomerTier tier) =>
        tier switch
        {
            CustomerTier.Bronze => price * 0.05m,
            CustomerTier.Silver => price * 0.10m,
            CustomerTier.Gold => price * 0.15m,
            CustomerTier.Platinum => price * 0.20m,
            _ => 0m
        };
}

If tests only verify Gold tier, Stryker generates mutations like:

• Replace 0.05m with 0.06m (survived -- no Bronze test)
• Replace 0.10m with 0.11m (survived -- no Silver test)
• Replace 0.15m with 0.16m (killed -- Gold test catches this)
• Replace 0.20m with 0.21m (survived -- no Platinum test)
• Replace 0m with 1m (survived -- no default test)

The HTML report highlights each surviving mutation with the exact code change, guiding where to add tests.

Stryker Thresholds

{
  "thresholds": {
    "high": 80,   // Green: mutation score >= 80%
    "low": 60,    // Yellow: 60% <= mutation score < 80%
    "break": 50   // Red: mutation score < 50% -> exit code 1
  }
}

The break threshold causes Stryker to return a non-zero exit code, useful for CI gates.

Flaky Test Detection

Flaky tests pass and fail intermittently without code changes. They erode trust in the test suite and slow development.

Common Causes

Detecting Flaky Tests

Repeated Runs

# Run tests multiple times to surface flakiness
for i in $(seq 1 10); do
  dotnet test --logger "trx;LogFileName=run-$i.trx" || echo "Run $i failed"
done

xUnit Conditional Skip

xUnit v3 has built-in conditional skip via Skip on [Fact]:

// xUnit v3 — built-in conditional skip
[Fact(Skip = "Requires external service")]
public async Task ExternalApi_ReturnsData()
{
    var result = await _client.GetDataAsync();
    Assert.NotEmpty(result);
}

// xUnit v3 — runtime skip via Assert.Skip
[Fact]
public async Task ExternalApi_ReturnsData()
{
    if (!await IsServiceAvailable())
        Assert.Skip("External service unavailable");

    var result = await _client.GetDataAsync();
    Assert.NotEmpty(result);
}

Time-Dependent Tests

Replace DateTime.Now/DateTime.UtcNow with.NET 8's TimeProvider:

// Production code
public class SubscriptionService(TimeProvider timeProvider)
{
    public bool IsExpired(Subscription sub)
    {
        var now = timeProvider.GetUtcNow();
        return sub.ExpiresAt < now;
    }
}

// Test code
[Fact]
public void IsExpired_PastExpiry_ReturnsTrue()
{
    var fakeTime = new FakeTimeProvider(
        new DateTimeOffset(2025, 6, 15, 0, 0, 0, TimeSpan.Zero));

    var service = new SubscriptionService(fakeTime);
    var sub = new Subscription
    {
        ExpiresAt = new DateTimeOffset(2025, 6, 14, 0, 0, 0, TimeSpan.Zero)
    };

    Assert.True(service.IsExpired(sub));
}

[Fact]
public void IsExpired_FutureExpiry_ReturnsFalse()
{
    var fakeTime = new FakeTimeProvider(
        new DateTimeOffset(2025, 6, 15, 0, 0, 0, TimeSpan.Zero));

    var service = new SubscriptionService(fakeTime);
    var sub = new Subscription
    {
        ExpiresAt = new DateTimeOffset(2025, 6, 16, 0, 0, 0, TimeSpan.Zero)
    };

    Assert.False(service.IsExpired(sub));
}

Note: FakeTimeProvider is available in Microsoft.Extensions.TimeProvider.Testing (NuGet).

Quarantine Strategy

When a flaky test cannot be fixed immediately:

// Mark as skipped with a tracking issue
[Fact(Skip = "Flaky: tracking in #1234 -- race condition in event handler")]
public async Task EventHandler_ConcurrentEvents_ProcessesAll()
{
    // ...
}

Do not delete flaky tests. Skip them with an issue reference and fix them systematically.

Key Principles

• Coverage is a lagging indicator, not a target. High coverage does not guarantee good tests. A test suite with 90% coverage can still miss critical bugs if the assertions are weak.
• Use CRAP scores to prioritize. Focus testing effort on methods with high complexity and low coverage rather than chasing overall coverage percentage.
• Run mutation testing on critical paths. Mutation testing is computationally expensive. Focus on business-critical code (pricing, authentication, data validation) rather than running it on the entire codebase.
• Fix flaky tests immediately or quarantine them. A flaky test that remains in the suite trains developers to ignore failures, undermining the entire test suite's value.
• Measure trends, not snapshots. Track coverage and mutation scores over time. A declining trend indicates test quality erosion even if absolute numbers look acceptable.
• Exclude generated code from coverage. Migrations, generated clients, and scaffolded code inflate or deflate coverage numbers without reflecting actual test quality.

Agent Gotchas

1. Do not confuse coverlet.collector with coverlet.msbuild. The coverlet.collector package uses the --collect:"XPlat Code Coverage" CLI flag. The coverlet.msbuild package uses /p:CollectCoverage=true MSBuild properties. Do not mix flags across packages -- they are independent integration points.
2. Do not hardcode coverage result paths. The GUID in TestResults/<guid>/coverage.cobertura.xml changes every run. Always use glob patterns (**/TestResults/*/coverage.cobertura.xml) when referencing coverage output files.
3. Do not set coverage thresholds too high initially. Starting with 90%+ thresholds on an existing project blocks all PRs. Begin with the current baseline and increase incrementally (e.g., 5% per quarter).
4. Do not run Stryker.NET on the entire solution for CI. Mutation testing is CPU-intensive. In CI, limit mutations to changed files (--since:main) or critical paths. Reserve full runs for nightly builds.
5. Do not ignore survived mutations in trivial code. While some survived mutations are in code that does not warrant testing (logging, ToString()), review each one. Configure ignore-mutations in stryker-config.json for categories you have consciously decided not to test.
6. Do not use [ExcludeFromCodeCoverage] as a blanket fix for low coverage. This attribute hides the problem rather than solving it. Use it only for genuinely untestable code (platform interop, generated code) and ensure the reason is documented.

References

• Coverlet GitHub repository
• ReportGenerator GitHub repository
• ReportGenerator usage guide
• Stryker.NET documentation
• Stryker.NET configuration
• TimeProvider in.NET 8
• Microsoft.Extensions.TimeProvider.Testing
• CRAP metric explanation