[github-actions]

Backport of #56414 to release/11.0.2xx

Related to #56412.

No servicing-approved label is recorded on this PR yet.

Customer Impact

• Customer reported
• Found internally

In a project-scale reproduction, CA1508 accounted for 162.765 analyzer-seconds out of 187.173 aggregate core-project analyzer-seconds. Local-function calls repeatedly cloned and reanalyzed unrelated caller state, while nested functions could also be scanned as part of the enclosing method. This backport reduces that avoidable analyzer work for safe direct local-function calls and bounded static-local-function graphs.

Regression

• Yes
• No

This is an analyzer performance change intended to preserve diagnostics. State is trimmed only when safe; uncertain indirect effects retain the full-state analysis. The backport adds deterministic coverage for local-function state, aliases, cached scans, and nested-function handling.

Testing

The source PR's full CI passed. The release-branch pipeline for this backport is still running; release-branch test results are pending.

Risk

The production change is limited to analyzer flow-state processing and CA1508 candidate scanning. It does not change public APIs or diagnostic rules. Conservative fallbacks preserve full analysis where trimming safety cannot be established.

Note

This PR description was drafted with GitHub Copilot.

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
2 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[Copilot]

🟢 Approval recommended

It matches the reviewed source change, includes focused regression coverage, and release-branch CI passed.

0 open findings

What changed in this PR

Backports the approved CA1508 performance optimization to .NET 11 servicing.

Changes:

• Trims safe local-function flow state while preserving captures and aliases.
• Bounds and caches nested/static local-function scans.
• Adds diagnostic-correctness and deterministic performance regression tests.

File	Description
DisposeObjectsBeforeLosingScopeTests.cs	Tests disposal through local-function callbacks.
AvoidDeadConditionalCode_ValueContentAnalysis.cs	Adds value-flow, aliasing, and scan-caching tests.
AvoidDeadConditionalCode_NullAnalysis.cs	Tests unrelated reference aliases.
DataFlowOperationVisitor.cs	Propagates trimmed-state metadata.
AnalysisEntityDataFlowOperationVisitor.cs	Implements safe trimming and cached scans.
PointsToAnalysis.PointsToDataFlowOperationVisitor.cs	Adapts points-to analysis to trimming metadata.
GlobalFlowStateDataFlowOperationVisitor.cs	Preserves full global-flow state behavior.
DisposeAnalysis.DisposeDataFlowOperationVisitor.cs	Propagates trimming metadata during disposal analysis.
CopyAnalysis.CopyDataFlowOperationVisitor.cs	Adapts copy analysis to trimming metadata.
AvoidDeadConditionalCode.cs	Excludes nested functions from enclosing candidate scans.

🧠 Review effort: Balanced

[jeffhandley]

Looks good and I am supportive of the 11.0 servicing backport (post-GA). This addresses a substantial perf issue. /cc @artl93

[artl93]

Approved, but best to preflight with the full context of analyzers since the list we reviewed recently was mostly backwards looking.

[tannergooding]

@artl93, @jeffhandley for backports to 11.0, we only plan on things that are effectively "bug fix" level things.

That is, it should be things like this where we are significantly improving the perf/UX or where we are fixing false-positives. It should not include anything that introduces new warnings/errors, especially not on by default.

We could separately talk about introducing new off by default analyzers, but they would need to have their own opt-in so that existing latest-all doesn't pick it up on upgrade. I'm not aware of any in this bucket currently.