Apply Benchmark patch only where its needed

[JohnReedV]

Description

This PR modifies the benchmark flow so that only extrinsics that have drifted passed the threshold or have incorrect read/writes are updated rather than replacing the entire file and creating conflicts everywhere.

Also fixes a bug where it would fail extrinsics where it shouldn't.

[github-actions]

🛡️ AI Review — Skeptic (security review)

VERDICT: SAFE

Baseline scrutiny: established OpenTensor-associated contributor with repo write permission; no trusted-list Gittensor match found; branch apply-benchmark-to-only-bad-benches -> devnet-ready.

Reviewed the cached PR metadata, prior Skeptic comment, contributor signals, PR comments, and full diff. The PR does not modify .github/ai-review/*, .github/copilot-instructions.md, dependencies, lockfiles, or Cargo build scripts.

The runtime-facing changes are generated weight updates plus a spec_version bump. The tooling changes are limited to benchmark comparison and selective patch generation; I did not find credential access, hidden network access, shell injection, supply-chain changes, origin/permission changes, unsafe runtime code, or a credible malicious patch path.

Findings

No findings.

Conclusion

No malicious intent or security vulnerability was found in the reviewed diff. Residual risk is limited to benchmark-tooling correctness and generated-weight accuracy, which do not cross the Skeptic security threshold here.

---

🔍 AI Review — Auditor (domain review)

VERDICT: 👍

Gittensor UNKNOWN; author has write permission and substantial prior repo activity, so review focused on benchmark tooling correctness rather than intent.

PR body is substantive enough to leave unchanged.

Overlap check: several open PRs touch common generated weight/runtime files, but their titles and bases indicate unrelated runtime features or broad feature propagation rather than duplicate benchmark patch-generation work. I do not recommend closing this PR in favor of another.

Spec version: bumped from 423 to 424 for the runtime/weights changes.

Validation: bash -n scripts/benchmark_action.sh and git diff --check passed. I did not run a targeted cargo build/test because this sandbox cannot initialize the pinned Rust toolchain under read-only /home/runner/.rustup, and no finding required runtime confirmation. No auto-fixes were applied.

Findings

No findings.

Conclusion

The PR scopes benchmark patch application to comparator-reported drifted functions and updates the comparator to evaluate benchmark-level ref-time drift over component ranges while keeping storage I/O exact. I did not find a domain correctness issue that should block merge.

[github-actions]

🔄 AI review updated — Skeptic: SAFE Auditor: 👍

[github-actions]

🔄 AI review updated — Skeptic: SAFE Auditor: 👍

[github-actions]

🔄 AI review updated — Skeptic: SAFE Auditor: 👍