Repository navigation
test: observe all remote input resends after recovery - #29674
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Priority: ⬇️ Low Merge Risk: ⚪ Minimal · up to
|
| Check name | Status | Explanation | Resolution |
|---|---|---|---|
| Docstring Coverage |  |
Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 1 functions across 1 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. |
-
Check name Status Explanation Title check The title clearly identifies the main change: observing all remote input resends after recovery. Description check The description is detailed and covers the change, reason, testing, scope, proof boundary, and checklist. It marks visual proof as not applicable and explains the reported validation. The Linked Issue… Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Description check
-
Explanation
The description is detailed and covers the change, reason, testing, scope, proof boundary, and checklist. It marks visual proof as not applicable and explains the reported validation. The Linked Issue section does not include an actual issue reference, but the rest of the description is substantially complete.
-
- Fix all pre-merge checks with AI
✨ Finishing Touches: 💡 1
-
📝 Generate docstrings: 💡
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
ELI5
The remote terminal fault test could pass while input kept being resent forever. It counted distinct operation IDs, so repeated sends of the same input—or empty cancellation frames—were invisible.
What Changed
Count every outgoing input frame at the existing simulated network boundary and require that count to stop growing after recovery. Keep the ID set for the separate host-restart oracle. Production behavior is unchanged; all 2,000 schedules, 120,000 actions, provider faults, delivery ordering and accepted-write assertions remain.
Why
Observing actual sends fixes the assertion without adding another test, changing the production interface, or reducing the fault workload. A deliberate acknowledgement off-by-one error passes all eight original random blocks but fails all eight repaired blocks at the resend assertion. The separate refusal test already catches that error, so this is stronger composed quiescence proof rather than a claim that the full suite previously missed the entire defect.
Linked Issue
Part of the PR CI test audit.
Visual Proof
N/A — test assertions only; no UI or production change.
Testing
Linux, background launches enabled. Untouched baseline: 10 cases pass. After restoring the deliberate production mutation byte-for-byte, the model and journal, replay, host-ledger and chunking keepers pass all 68 cases across five suites, with no unhandled errors. Mutation runs are deliberately failing controls, excluded from performance comparisons.
Three alternating-order local pairs retain the same ten passing cases and full schedule statistics: median process wall 10,528→10,276 ms; module 8,955→8,710 ms. These observations do not establish a speedup; the purpose is to repair false confidence at the same workload. Formatting, targeted lint, changed-code quality and diff checks pass. The full PR workflow run 38905448586 and its aggregate passed at the exact merged head, as did all separate checks. All five shard inventories reconcile, with 12,362→12,368 files explained by seven inherited additions and one relocation, zero omissions or unhandled errors. Hosted module time was slower, 14.973→18.132s (test execution 6.468→9.083s), against baseline 38901382074 selected before candidate artifacts. Production dependency and checkout drift limit attribution; no hosted or whole-CI speedup is claimed.
Review
Independent preservation review approved. Known disconnected-grace oracle and aggregate-summary findings remain separate audit work; this PR makes no claim to repair them.
Agent skill upstream boundary
Notes
No production or protocol changes. The real renderer transport and host dispatcher/sequence ledger remain in the model. Actual OS PTY, SSH socket and mixed-version behavior remain outside this unit model's proof boundary.
Checklist
N/Awith reasonpnpm lint,pnpm typecheck,pnpm test, andpnpm buildpass (or CI will cover; local preferred)