Bootstrap Dashboard.Tests project source (plan 3)

[erikdarlingdata]

Summary

Populates the previously scaffold-only Dashboard.Tests/ folder with the pure-algorithm subset of Lite.Tests — every test that exercises shared analysis code without a DuckDB or SQL Server dependency. Three ported files, 18 cases, ~1s wall time.

What changed

• New Dashboard.Tests/Dashboard.Tests.csproj — mirrors Lite.Tests/Lite.Tests.csproj byte-for-byte modulo the project reference (net10.0-windows, xunit.v3 3.2.2, Microsoft.NET.Test.Sdk 18.5.1, same NoWarn set, UseWPF=true).
• New Dashboard.Tests/FactScorerTests.cs — 12 cases: 10 ApplyThresholdFormula [InlineData] rows, Score_UnknownWaitType_GetsSeverityZero, Amplifier_SeverityCappedAt2.
• New Dashboard.Tests/InferenceEngineTests.cs — 3 cases: Graph_NoEdgesForUnknownFact, Graph_CxPacketEdgeFires_WhenSosIsHigh, Graph_CxPacketEdgeDoesNotFire_WhenSosIsLow.
• New Dashboard.Tests/BaselineBucketTests.cs — 3 cases on BaselineBucket.EffectiveStdDev (proportional floor, zero-activity, pass-through). Renamed from Lite's BaselineProviderTests because only the value-type cases port; the provider-level DuckDB cases are deferred. using DuckDB.NET.Data; stripped.
• Modified Dashboard/Dashboard.csproj — adds <InternalsVisibleTo Include="Dashboard.Tests" />. FactScorer.ApplyThresholdFormula is internal static, so the Theory port would not compile without this.
• Modified PerformanceMonitor.sln — adds Dashboard.Tests project block + per-config Debug/Release build mappings under a fresh GUID.

Test bodies, [InlineData] rows, and assertions are byte-identical to the Lite originals — the shared algorithms are byte-identical between Lite and Dashboard, so the ported tests pass without behavior change.

Out of scope (intentionally deferred)

• BlockingChainReconstructorTests — Dashboard/Analysis/BlockingChainReconstructor.cs is on dev, but Lite's matching test file lives only on the unmerged Lite Stage 3 worktree. The parity port rides in the same PR that brings Lite Stage 3's reconstructor tests to dev, not this bootstrap — keeping Lite and Dashboard symmetrical.
• All SQL-touching ports — SqlServerFactCollector, SqlServerFindingStore, SqlServerAnomalyDetector, the database paths of SqlServerBaselineProvider, full AnalysisService/Scenario pipelines. The plan documents these under §10 and recommends TestContainers as the fixture surface; that decision is its own PR.
• FinOpsHealthCalculator / HighImpactScorer parity — Dashboard FinOps has no pure scoring class to mirror Lite's; structurally different surface, gets bespoke tests later.
• UI / WPF tests — separate concern; UseWPF=true in the test project is for type loading only.

Test plan

• dotnet build Dashboard.Tests/Dashboard.Tests.csproj -c Debug — 0 warnings, 0 errors.
• dotnet test Dashboard.Tests/Dashboard.Tests.csproj — 18 passed, 0 failed, 0 skipped (~1s).
• dotnet build PerformanceMonitor.sln -c Debug — full-solution build clean, 0 warnings, 0 errors.
• dotnet test Lite.Tests/Lite.Tests.csproj — Lite.Tests still passes (260/260), confirming tests were COPIED + renamed, not moved.
• code-reviewer agent review — clean, no findings.

Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com