Skip to content

Review symbol mappings in bounded independent batches - #1224

Merged
George Adams (gdams) merged 1 commit into
mainfrom
fix/symbolmap-bounded-reviews
Sep 29, 2026
Merged

George Adams (gdams) merged 1 commit into
mainfrom
fix/symbolmap-bounded-reviews

Conversation

@qmuntal

Copy link
Copy Markdown
Member

Summary

Follow up on #1223 and run 36579819302. The prepared API report included the history-persistence option from #1201, but the agent did not complete its assessment after a literal-name lookup missed the different .NET spelling. It also claimed complete coverage despite unfinished review and throttled source searches.

  • Delegate API changes, behavioral diffs, and the separate full-catalog sweep to fresh symbolmap-reviewer batches, with at most two concurrent workers and no recursive delegation.
  • Require an evidence-backed outcome for every assigned item. Missing items, uninspected members, and unresolved source reads remain pending rather than counting toward completion.
  • Broaden literal-name misses to declaring-type inspection, prefer direct pinned .NET file reads, and verify existing counterpart mappings and the actual implementing diff before editing.
  • Keep duplicate checks, catalog edits, and publication in the main agent after independent review. Preserve the full audit scope, behavioral review when the API report is empty, and catalog-only output restrictions.

No SDK code, catalog assessments, inventory, preparation steps, runtime configuration, permissions, budgets, or shared pins change. No custom review ledger or publication gate is added. The generated workflow change is limited to prompt metadata.

Validation

  • gh aw compile symbolmap-maintenance-weekly --strict --no-check-update passed twice on current main with zero warnings and reproducible output.
  • Exercised the pinned runtime's actual inline-agent extractor in a temporary directory; it separates the reviewer from the parent prompt and writes the correct native Copilot agent file.
  • A focused two-item local review trial identified the missing history-persistence counterpart and its implementing commit, while recognizing that function middleware already has a catalog mapping. This is a prompt smoke test, not hosted end-to-end validation or proof of full-catalog coverage.
  • Verified the frontmatter and preparation are unchanged, all generated lines except metadata match the base, and whitespace/editor checks pass.

The hosted workflow has not been rerun with this change. SDK, provider, E2E, replay, and benchmark tests were not run for this workflow-only update.

Delegate API, behavioral-diff, and original-catalog review to a fresh
symbolmap-reviewer for each bounded batch. Limit concurrency to two
workers and leave duplicate checks, edits, and publication to the
main agent.

Require per-item pinned evidence, broader declaration lookup after
literal-name misses, and direct file reads when code search is
throttled. Keep unresolved items pending and verify existing mappings
and implementing diffs before editing.

Preserve audit scope, preparation, runtime, and permissions. Update
the guide and regenerate workflow metadata with gh-aw v0.89.21.
@qmuntal
Quim Muntal (qmuntal) requested a review from a team as a code owner September 29, 2026 15:56
Copilot AI balanced review requested due to automatic review settings September 29, 2026 15:56
@github-actions github-actions Bot added area:github Changes files in the github area size:medium At most 100 changed lines across at most 5 files kind:ci Changes CI, build, or repository automation kind:docs Changes documentation or comments labels Sep 29, 2026

Copilot AI left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copilot review overview

🔵 Needs a closer look

The new multi-agent orchestration has only local smoke validation and has not yet run in the hosted workflow environment.

Review effort: Balanced
Findings: None

What changed in this PR

Refactors weekly symbol-map auditing into bounded, evidence-backed subagent batches.

Changes:

  • Adds independent API, behavioral, and catalog review phases.
  • Defines the inline symbolmap-reviewer agent and completion safeguards.
  • Documents the workflow and updates generated metadata.
File Description
.github/​workflows/​symbolmap-maintenance-weekly.md Adds batched review orchestration and reviewer instructions.
.github/​workflows/​symbolmap-maintenance-weekly.lock.yml Updates the generated prompt hash.
docs/​dotnet-go-sdk-feature-comparison.md Documents the delegated review process.

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@gdams
George Adams (gdams) added this pull request to the merge queue Sep 29, 2026
Merged via the queue into main with commit 2b10117 Sep 29, 2026
22 checks passed
@gdams
George Adams (gdams) deleted the fix/symbolmap-bounded-reviews branch September 29, 2026 16:24
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:github Changes files in the github area kind:ci Changes CI, build, or repository automation kind:docs Changes documentation or comments size:medium At most 100 changed lines across at most 5 files

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants