Repository navigation
fix(fixtures): refresh official plugin versions and provenance - #460
Conversation
Update the 19 pending official plugin pins to 2026.9.7–2026.9.9 and retain their matching published-source provenance. Keep the promoted host and moving-track policy unchanged. Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
5343005 to
6afc7b9
Compare
|
ClawSweeper reviews are opt-in for maintainers. If a review is required, finish the work and validation first, then comment |
|
@clawsweeper re-review Please review exact head Final controlled exact-version qualification: https://github.com/openclaw/crabpot/actions/runs/38532618877. All 19 isolated fixtures completed successfully on OpenClaw Acting and independent review verified all 76 report handoff hashes, 43 capture artifacts, and 43 synthetic artifacts: 48 passed probes, zero failures, 80 explicit prerequisite/unsupported-callable blocks. All 19 execution profiles have no non-audit failure. Discord and Matrix now have actual published-artifact capture and retained callback proof. This is mock-SDK qualification, not authenticated channel/provider or full supported-host-range proof. The carrier's static result is deliberately not called green: 775 tests passed and one existing Matrix source-pack assertion failed because the controlled carrier selects its npm artifact. No test was weakened. Exact branch cleanup succeeded; the ephemeral ref is absent. Five inherited audit package entries remain: Matrix music-metadata (moderate), Codex smol-toml (moderate), and the development/optional LanceDB → Hugging Face → sharp graph (three high, including propagated parents). Independent comparison found the same affected baseline versions/graph and no expanded image-parser entrypoint. Upstream dependency refreshes and correct dev/optional audit attribution remain follow-ups; this PR does not claim those vulnerabilities fixed. The production fixture-lock audit actually ran and passed. Codex evidence is bound to the actual package: the published 2026.9.8 tarball's SHA512 matched the registry; its declared SDK and compiled managed/native version checks are 0.158.0, while the attested source commit's checked-in pin is 0.160.0. No cause for that distinction is assumed. I personally inspected the exact upstream 0.158 initialize contract, initialized notification, and pre-initialization rejection, plus the published managed resolver. No native Codex/account E2E invocation is claimed. Source review and whole-branch P1 review are clear. Normal production CI is still completing the four static lanes; its 19 changed-fixture lanes and actual fixture-lock audit passed. The full 19 original PR links and Dependabot credit are retained in the PR body and commit history. No release, publication, or original PR closure is part of this request. |
|
🦞👀 Re-review progress:
|
|
Codex review: needs maintainer review before merge. What this changesThis PR refreshes nineteen official plugin fallback packages and their provenance, and adopts Inspector capture and import-scanning repairs. Example: Discord materialization falls back to its committed package pin.
Review scores
ProductKind: Maintenance · Worth it: Yes · Fix scope: Complete Merge readiness✅ Ready for maintainer review Keep open: this maintainer-authored PR contains useful fixture updates absent from main, and no actionable introduced defect was found. Priority: P3 Before mergeNone. FindingsNone. Agent review detailsHow this fits togetherCrabpot materializes configured plugin fixtures, passes their code to Inspector, and produces compatibility reports from static inspection and isolated capture probes. flowchart LR
A[Fixture manifest and shims] --> B[Package materializer]
B --> C[Plugin artifacts]
D[Pinned Inspector source] --> E[Inspection and capture]
C --> E
E --> F[Compatibility reports]
Technical reviewBest possible solution: Land the coherent fixture refresh through the existing required checks, preserving explicit qualification limits and unsupported-operation failures. Do we have a high-confidence way to reproduce the issue? This maintenance refresh has no single local reproduction requirement; source inspection confirms the older main inputs, and no target code was executed during review. Is this the best way to solve the issue? Consolidating the overlapping package updates and fixing capture behavior in Inspector respects the existing ownership boundary and avoids plugin-specific workarounds. AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning medium; reviewed against 73dfebff1f41. Provenance checked
TestingProof path: in-process harness. SecurityNone. EvidenceWhat I checked:
Likely related people:
Root-cause clusterRelationship: Members:
Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything. LabelsLabel changes: No label changes. Label justifications:
Rating scale6/6 🦀 challenger crab · 5/6 🦞 diamond lobster · 4/6 🐚 platinum hermit · 3/6 🦐 gold shrimp · 2/6 🦪 silver shellfish · 1/6 🧂 unranked krab. Overall follows the weaker of proof and patch quality; ✨ marks media proof (a screenshot, video, or linked artifact) that directly shows the changed behavior. WorkflowClawSweeper edits this one comment on every review. Comment HistoryReview history (1 earlier review cycle)
Reviewed October 14, 2026, 5:05 AM ET / 09:05 UTC (Revision 2). |
|
@clawsweeper re-review Please review exact head This source-pin update consumes merged Inspector #108 at Five focused source-pin/resolution tests passed; independent Astra/high P1 delta review was scoped-clean. Fresh normal CI and fresh exact19 qualification are running. The latter uses the same controlled exact package versions/npm selectors and frozen OpenClaw host The old-head Windows static suite hit its unchanged 600-second deadline; this is a new source head, not a blind retry or a claim that Windows has passed. Historical19 results and inherited-audit attribution remain historical until the fresh artifacts are inspected. Existing prerequisite blocks, the five inherited affected audit entries, and the published Codex 0.158.0 versus checked-in managed-pin distinction remain documented in the preceding proof. No native Codex/account/channel-provider E2E claim, release, or package publication is made. |
|
🦞👀 Re-review progress:
|
|
Scope correction after reading the current original PR files: #456, #452, #451, #450, #448, #447, #446 and #445 now select 2026.10.1 in both their package manifests and locks. This PR deliberately retains their earlier 2026.9.9 snapshots in the nineteen-input qualification. It does not incorporate or supersede those eight current updates; they remain open for separate review. The body now distinguishes those eight from the eleven original targets that still match this cohort. Current head |
What Problem This Solves
Nineteen frozen official plugin updates overlap in the shared fixture manifest and provenance fields. This PR preserves those selected 2026.9.7–2026.9.9 snapshots and corrects their published source references. Eight original PRs have since advanced to 2026.10.1; those newer updates are not incorporated here.
User Impact
Refresh the official fallback pins to 2026.9.7–2026.9.9 and report their actual published source provenance. The frozen promoted OpenClaw host and normal moving plugin tracks remain unchanged.
Why This Change Was Made
This consolidates the selected overlapping Dependabot snapshots after #401, preserving their package/lock updates and credit to @dependabot. It adopts merged Inspector source
b56782860cbef840e8892f18f58c0c3a38c02dd6for channel registration and bounded namedcreateRequirediscovery. The published CLI pin remains 0.3.26; no release or publication is part of this PR.2b988ee83444f08ccaf37aa5e98f370726c75f6eaa6008ad198ef99c43f9d89dbd016947087129742b4055f8dd703e9c09985940dce9a9c6fc3c29bfEvidence
39d1781e561cb394562651973468b0746a58acd2.10bb24270d8dad0f85781980248c5ba9a1d5cd71, separate from production configuration. Its static suite reported 775 passed tests, one expected Matrix npm-versus-source-pack assertion failure and 16 skips. No production test was changed. Cleanup succeeded and the exact ephemeral ref is absent.39d1781e561cb394562651973468b0746a58acd2: Ubuntu, macOS, Windows, container, all nineteen isolated fixtures and the actual fixture-lock audit passed. The separate advisory Windows HEAD canary failed with EOWNERSTART during git initialization before Inspector checkout/import, followed by EOWNERCLEANUP and preserved-lock waits. This remains the Windows command-supervisor follow-up; no retry, lock clearing or timeout change was used. It is not the promoted Default Track gate. The full current-head ClawSweeper review has no findings or before-merge moves; its reviewer could not download raw artifacts, so the separate executor and independent artifact checks above provide that receipt.This is mock-SDK contract qualification, not real-account end-to-end proof or a claim that every plugin's declared host range includes the frozen host. Existing explicit prerequisite blocks and actual failures remain visible. For Codex, the published 2026.9.8 artifact and its managed/native-version checks consistently select SDK 0.158.0; the attested source commit's checked-in version is 0.160.0. Those are distinct observations. The exact 0.158 protocol was personally inspected at upstream Rust commit
064c6b8c737f5b41d171fdda80bd9ef10ad06eb3; no native Codex/account end-to-end invocation is claimed.Five inherited affected package entries remain: Matrix's music-metadata (moderate), Codex's smol-toml (moderate), and the LanceDB development/optional Hugging Face → sharp graph (three high entries, including propagated parents). The canonical baseline already has the same affected versions/graph; no expanded image-parser entrypoint was observed. Upstream follow-ups are music-metadata and smol-toml refreshes, and a reviewed LanceDB/Hugging Face/sharp dependency refresh with correct audit attribution. This is not an audit-clean or vulnerability-fixed claim. The fresh per-fixture audit vulnerability objects exactly match the previous qualification. The actual fixture-lock security check and normal CI remain separate from these retained warnings.
Fixture Impact
Fixture Verification
npm testequivalent materialization and test sequence (fresh normal static CI)node scripts/sync-fixtures.mjs --check(fresh normal static CI)node scripts/run-contract-smoke.mjs --strict(fresh normal static CI)Notes
Original PRs whose selected target versions still match this cohort; closure requires a merged replacement and fresh exact equivalence verification:
Original PRs that have advanced to 2026.10.1. This PR includes their historical 2026.9.9 snapshots only; these current PRs are not superseded and must remain open:
External plugin upgrades, held input-safety reviews, Tencent publisher changes and DingTalk runtime repair are outside this cohort. No live-provider credentials, releases or deployments are requested.