Repository navigation
fix: remove Windows command startup compilation - #502
Conversation
|
🦞👀 Pull request received. I will update this pull request when review starts. ClawSweeper review completeClawSweeper finished reviewing this revision. The review result is being finalized. |
|
Codex review: needs maintainer review before merge. What this changesThis PR replaces per-command C# compilation and JSON cmdlet loading with emitted native bindings, framework serialization, and explicit admission receipts. Example: Windows command preparation exceeds the existing 10000 ms startup budget before any request is issued.
Review scores
ProductKind: Bug fix · Worth it: Yes · Fix scope: Partial Merge readiness✅ Ready for maintainer review Keep this maintainer-authored draft open: it contains distinct work at the Windows command owner, and current main still uses the startup path it replaces; no definite introduced bug was found. Priority: P2 Before mergeNone. FindingsNone. Tests
Agent review detailsHow this fits togetherCrabpot's shared synchronous command supervisor runs checkout, inspection, and security commands; its Windows helper admits native processes into Jobs and returns output, status, and cleanup receipts that govern checkout-lock release. flowchart TD
A[Inspector checkout resolver] --> B[Owned command supervisor]
B --> C[Worker and control pipe]
C --> D[PowerShell native helper]
D --> E[Windows Job and command]
E --> D
D --> F[Admission and cleanup receipts]
F --> A
Technical reviewBest possible solution: Qualify the replacement through the actual cold Windows checkout flow while preserving bounded startup, atomic Job admission, and lock retention whenever cleanup remains uncertain. Do we have a high-confidence way to reproduce the issue? No deterministic reproduction of the natural cold-start failure is established; the linked investigation records real Windows failures, while this read-only review verified source paths without executing target code. Is this the best way to solve the issue? Removing compilation and cmdlet loading at the existing owner is a coherent narrow candidate, but neither source inspection nor separate successful controls establishes that these operations caused the reported stalls. AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning medium; reviewed against c82f74c2c423. Provenance checked
TestingProof path: unit tests only. SecurityNone. EvidenceWhat I checked:
Likely related people:
Review metrics
LabelsLabel changes:
Label justifications:
Rank-up movesOptional improvements that raise the rating; they are not merge blockers.
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 Reviewed October 15, 2026, 11:53 AM ET / 15:53 UTC. |
steipete
left a comment
There was a problem hiding this comment.
Independent P0–P2 review is clean. The required exact-head gate and all CodeQL analyses passed. Native ownership controls passed in the required Windows lane and with identical owner/test source in both Windows lanes of #503. Residual advisory cold-start and fixture-timing failures remain tracked in #305; this approval covers removal of compiler startup work and the verified admission/cleanup protocol, not a claim that every Windows timing failure is resolved.
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Post-merge qualification receipt for head A remaining Windows failure is recorded in the same-head advisory job https://github.com/openclaw/crabpot/actions/runs/38781124149/job/116582064943. Its first Follow-up: reproduce that cold original order with the existing startup/cleanup budgets and retain admission/closure truth for the failed surface. No timeout increase, weaker assertion or retry is proposed from these logs. The landed startup/custody repair and its required-suite proof remain distinct from this unresolved advisory behavior. No release operation was performed. |
Windows command startup could spend most of its 10-second budget loading PowerShell JSON cmdlets and compiling C#, then leave checkout custody uncertain. Replace per-command compilation with Reflection.Emit native bindings and direct framework JSON serialization. PREPARED gates the first request; possible admission still requires verified Job extinction, helper closure, and drained control output before custody is released. Production startup, execution, and cleanup budgets stay unchanged.
The compiler observer distinguishes command admission from its own instrumentation setup. Its unchanged C# source is a standalone fixture so CodeQL can analyze it after removal of the production C# bootstrap. Native ABI/error mapping, blocked preparation/parsing, cancellation, receipt validation, suspended real-compiler positive/negative controls, and interrupted checkout retry remain asserted.
Validation at 18f6d84:
Residual evidence remains explicit: this PR's Windows HEAD canary failed before pipe connection, during negative-control rescue, and while waiting for observer instrumentation. Those failures are not claimed resolved or attributed to a particular cause. #305 remains open for that cold-start/test-timing investigation. The earlier local Mac run also had three short-budget lifecycle failures, only one reproduced on unchanged main; hosted required Linux/macOS gates passed.
Related: #305. No timeout increase, prewarming, release, tag, version bump, or deployment.