Guard CI fixer against no-op test rooting changes

[Copilot]

Runtime-missing failures can prompt ineffective rooting changes when the harness or project already roots the test assembly, regardless of runtime mode.

• Rooting guard
  • Requires the fixer to inspect existing roots before proposing DynamicDependency, descriptor, TrimmerRootAssembly, or app-bundle changes.
• Fallback
  • Requires a concrete product-side cause or routes the failure to a loop-in comment when the proposed root would be a no-op.

Note

This PR description was generated with GitHub Copilot.

[azure-pipelines]

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

[kotlarmilos]

/ci-eval

[kotlarmilos]

/ci-eval

Note

This retry was posted by GitHub Copilot at the user's request after the PAT pool selected an invalid token.

[dotnet-policy-service]

Tagging subscribers to this area: @agocke, @dotnet/ilc-contrib
See info in area-owners.md if you want to be subscribed.

[kotlarmilos]

/ci-feedback eval

Note

This command was posted by GitHub Copilot at the user's request to complete the feedback eval cancelled by a concurrent policy-service comment.

[Copilot]

Pull request overview

Updates the CI outer-loop fixer workflow guidance to prevent proposing ineffective “rooting” changes for NativeAOT/runtime-missing failures when test assemblies are already effectively rooted by the harness.

Changes:

• Refines the Mobile (ios/tvos/maccatalyst/android/wasm/wasi) policy row to refer readers to the Step 5.2 “already-rooted” guidance before proposing roots.
• Adds a new Step 5.2 caution that instructs the fixer to check for existing roots and avoid no-op rooting proposals for already-rooted test assemblies, otherwise routing to a concrete product-side cause or a loop-in comment.

Show a summary per file

File	Description
.github/workflows/ci-failure-fix.md	Adds/centralizes guidance to guard the fixer against proposing redundant NativeAOT trimming/rooting changes for already-rooted test assemblies.

Copilot's findings

• Files reviewed: 1/1 changed files
• Comments generated: 1

[Copilot]

Copilot's findings

• Files reviewed: 1/1 changed files
• Comments generated: 1

[azure-pipelines]

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

[kotlarmilos]

/ci-eval

Note

This eval retry was triggered by GitHub Copilot at the user request and uses the workflow from the latest main.

[github-actions]

Note

AI/Copilot-generated eval report (Vally), triggered by a maintainer eval command.

ci-failure-scan

✅ passed

details

Eval Results

Timestamp: 2026-07-15T08:05:27.637Z

Dynamic gate for the CI failure scanner's safe-output, run with the maintainer /ci-scan eval command. It runs the scanner against a live outer-loop failure it discovers itself, then checks the emitted Known Build Error against the static KBE format and that it actually grounded the KBE in a real failing build.

Stimulus	Environment	Graders	Pass Rate	Duration	Tokens	Turns	Tool Calls	Verdict
live-kbe-safe-output	2 files Files: ../ci-failure-scan.md → .github/workflows/ci-failure-scan.md, ../shared/create-kbe.instructions.md → .github/workflows/shared/create-kbe.instructions.md	✅ file-matches 1/1 ✅ file-not-matches 1/1 ✅ fetched-real-failing-build 1/1 ✅ searched-existing-kbe 1/1 ✅ kbe-format-conformant 1/1	1/1	2m 29s	757,565	17	17 calls bash: 10, view: 5, grep: 1, create: 1	✅

Model: claude-opus-4.8 | Judge: claude-opus-4.8 | Executor: copilot-sdk

ci-failure-fix

❌ failed

details

Eval Results

Timestamp: 2026-07-15T08:08:54.303Z

Dynamic behavior gate for the CI failure fixer's safe-output, run with the maintainer /ci-fix eval command. It points the fixer at a real open Known Build Error it discovers itself and checks that it either produced a fix PR or engaged owners, and never muted the failing test.

Stimulus	Environment	Graders	Pass Rate	Duration	Tokens	Turns	Tool Calls	Verdict
live-fix-or-engage-safe-output	2 files Files: ../ci-failure-fix.md → .github/workflows/ci-failure-fix.md, ../shared/create-kbe.instructions.md → .github/workflows/shared/create-kbe.instructions.md	✅ file-matches 1/1 ❌ file-not-matches 0/1 ✅ acted-on-real-kbe 1/1 ✅ fix-or-engages-owners 1/1	0/1	3m 09s	665,017	15	16 calls bash: 12, view: 3, create: 1	❌ [1]

[1] Failed grader(s): file-not-matches

Model: claude-opus-4.8 | Judge: claude-opus-4.8 | Executor: copilot-sdk

ci-failure-scan-feedback

❌ failed

details

Eval Results

Timestamp: 2026-07-15T08:13:58.499Z

Dynamic constructiveness gate for the CI feedback meta workflow's safe-output, run with the maintainer /ci-feedback eval command. The agent scans real recent scanner issues and fixer PRs itself, then must turn a real signal into constructive, grounded feedback.

Stimulus	Environment	Graders	Pass Rate	Duration	Tokens	Turns	Tool Calls	Verdict
live-constructive-feedback-safe-output	4 files Files: ../ci-failure-scan-feedback.md → .github/workflows/ci-failure-scan-feedback.md, ../ci-failure-scan.md → .github/workflows/ci-failure-scan.md, ../ci-failure-fix.md → .github/workflows/ci-failure-fix.md, ../shared/create-kbe.instructions.md → .github/workflows/shared/create-kbe.instructions.md	✅ file-matches 1/1 ✅ scanned-scan-issues 1/1 ❌ scanned-fix-prs 0/1 ✅ is-constructive 1/1 ✅ edit-is-grounded 1/1	0/1	4m 07s	1,138,927	24	23 calls bash: 18, view: 4, create: 1	❌ [1]

[1] Failed grader(s): scanned-fix-prs

Model: claude-opus-4.8 | Judge: claude-opus-4.8 | Executor: copilot-sdk