[samtrion]

Problem

In the latest mutation run (run 40488214820) the score is 59.95% (Killed 793, Timeout 5, Survived 491, NoCoverage 42). 202 of the 491 survivors are mutants inside static field initializers (static=true in the report: 202 Survived / 3 Killed / 3 CompileError):

File	Static survivors
Helpers/CSharpKeywords.cs	127
Usage/RequireCancellationCheckCodeFixProvider.cs (CollectionExpressionTargetTypes)	46
Maintainability/AvoidRedundantCollectionClearAnalyzer.cs (AllowedCollectionMetadataNames)	11
Helpers/CancellationTokenCallSites.cs	5
DiagnosticDescriptors, UseLangwordSuppressor, RequireNumericLiteralSuffixSuppressor, TargetFrameworkMonikers, others	13

They survive although existing tests would fail if the value were really changed: "System.Collections.ArrayList" -> "" and "System.Collections.Generic.List1" -> ""survive whileAvoidRedundantCollectionClearAnalyzerTestshasArrayList_ImmediateClear_Reportsand about 20Listtests. The three killed static mutants are inDiagnosticIds.HelpLink`, an expression-bodied static method that is evaluated at test time, not a field initializer.

Root cause

Consistent with the Stryker 5.0.0 source and not verified by a dedicated run: the MTP runner (preview, log: "Microsoft Test Platform testrunner is currently in preview", "Starting per-test coverage capture for MTP runner") switches the active mutant by writing the mutant id into a warm test host and only restarts the host after a timeout. A static field initializer has therefore already run before the mutant is read, so the mutation never takes effect. Static mutants are run against the whole suite and still survive; this is a tool artefact, not missing tests. Local confirmation was not possible: the repository does not build on the maintainer's Windows machine (NetEvolve.Analyzer.Dogfood.dll copy error, MSB3030), so Stryker aborts with "Failed to analyze project builds". The branch run below is the confirmation.

Change

// Stryker disable all : ... / // Stryker restore all pragma pairs placed directly around the individual static table declarations only (7 files, comments only, no behaviour change):

• CSharpKeywords: all five tables
• RequireCancellationCheckCodeFixProvider.CollectionExpressionTargetTypes
• AvoidRedundantCollectionClearAnalyzer.AllowedCollectionMetadataNames
• CancellationTokenCallSites.SupportedAsyncReturnTypeMetadataNames
• UseLangwordSuppressor.Suppression, RequireNumericLiteralSuffixSuppressor.Suppression
• TargetFrameworkMonikers.UnifiedTfmPattern

Surrounding logic in those classes stays mutated. DiagnosticIds is untouched (its lazy members are killable). DiagnosticDescriptors fields are not covered (the per-field mutants there were not identified), so the expected effect is slightly below the full 202. This repository has no earlier "Stryker disable" usage, so this introduces a new convention.

Before / After

Metric	Before (run 40488214820)	After (run 40815422362)
Mutation score	59.95% (798/1331)	70.37% (798/1134)
Killed / Timeout	793 / 5	793 / 5 (unchanged)
Survived	491	294
NoCoverage	42	42
Ignored by the new pragma	0	198
Mutants created / compile errors	2110 / 474	2110 / 474
Duration	25 min	about 13 min

Run: https://github.com/dailydevops/analyzer/actions/runs/40815422362

198 of the 202 static survivors moved to Ignored (the rest are outside the scoped tables, e.g. DiagnosticDescriptors). The Killed count did not change, so no real test strength was lost or gained. Duration is not a controlled comparison (different runner load). The score gain is purely a denominator change.

Risk / what this does not change

• Trade-off: a mutation-test pragma on these tables hides any real regression in them from mutation testing, for example a wrongly edited keyword. The ordinary unit tests for the tables stay and still guard them.
• It is a workaround. Remove the pragmas once Stryker's MTP runner handles static initializer mutants. An upstream issue and a pipeline-level fix (the same problem affects other repos using the MTP runner, e.g. defaults, frameshift) are separate work, not part of this PR.
• Per-test-isolation coverage (coverage-analysis perTestInIsolation) was not tried; mutation runs would still use the warm host.
• No production or test code behaviour changes.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.72%. Comparing base (8c86c2b) to head (785b989).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #229   +/-   ##
=======================================
  Coverage   94.72%   94.72%           
=======================================
  Files          48       48           
  Lines        2557     2557           
  Branches      449      449           
=======================================
  Hits         2422     2422           
  Misses         58       58           
  Partials       77       77           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.