[trissim]

P300: partial live parity, not acceptance

Draft. Implements comparisons for exactly rows 5, 6, 11, 12, 13, 27. The 23:05 passing deadline was missed. Nothing here is claimed installed.

Current exact head

Head 2abf5c4ff97be15df71b57df747b77a221115911, rebased onto ef4cc72da8d52548b2a594a78200749f0fc878bf. Main's other parity implementations are preserved.

• 5, 6, 12 pass in the actual configured six-row run.
• 11 fails: current/native call coverage differs. Every shared call's digest matched in the diagnosed run; current had one extra newest call on several problems. A source-selection/frontier mismatch is suspected, not proved. pi-cn-phase6 is tracing the watcher/history boundary read-only; I retain the verifier implementation.
• 13 and 27 stay explicitly unverified until their native implementations land. P309/pi-story-tool now owns the story tools and row13 comparison; P310/pi-prediction owns prediction/outcome and row27 comparison on their assigned shared base. I sent each the exact overlapping row site and will not edit these rows concurrently. Integration must retain their real comparisons over these missing-owner declarations. No placeholder passes.

Working repair

The real run exposed recorded coordination calls losing their original argument digest. NativeRequestTools now uses its existing native call factory with original arguments and call identity. CoordinationCall retains its existing presentation; UndeclaredTool owns external-name/raw-argument presentation. ToolStartTranscript.record_call delegates to those declarations. The optional-tool ToolDeclaration.call_digest selector is deleted.

Search on this head:

rg -n 'recorded_digest|call_digest' src tools

There are four capability implementations (ToolDeclaration, NativeRequestTools, CoordinationCall, UndeclaredTool) and one production delegation in ToolStartTranscript.record_call; no production call_digest remains. This is presentation only, not tool execution or original-history mutation.

Shared parity behavior lives in commander_parity/fleet.py. Board, branch and unjudged comparisons observe actual native returns consumed by the current path, avoiding independent second reads. Board's activity clock, historical handwritten branch artifacts and legacy terminal call states are not independently compared.

Exact-head verification

• Live six-row family plus parity/digest/watcher tests: 14 passed, 3 failed. Failures are row11 and two watcher refusal-text expectations. Those same two expectations fail with unchanged baseline watcher/codec sources in the other isolated checkout. The two new original-argument/undeclared-call regressions pass.
• Separate --root src and --root tools ratchets against pinned base above: both exit 0, no increases or introduced lines.
• Actual full --verify: failed before structured output, because existing row17 LoadComparison.read timed out its configured installed capacity child after 30 seconds. This is not a full verifier pass or merely an expected unverified-row exit. Ownership correction5539 supersedes5527: pi-cn-phase7 already published the timeout/shared-observer repair in P302: verify rules and contact parity through live owners #2098 at 1a0b3e337; pi-plan-claims retains original context and reviews/consumes that implementation, with no competing load edits. A load-only checkpoint including its observation owner is requested from pi-cn-phase7. No repair, integration, verifier retry or UNKNOWN replay here.
• AST after-census through existing audit.findings.Package / audit.repository.Repository: 397 production, 47 tools, 509 test modules; zero parse omissions. Before/after owner/consumer evidence is retained below. Dynamic receiver and TypeScript wire consumption are not statically resolved; the wire protocol is unchanged.

Evidence: /home/ts/.cache/agent-scratch/pi-cn-phase3/p300/: ownership-before.txt, digest-ownership-before.txt, ownership-after.txt, family-exact.log, watch-baseline-exact.log, ratchet-{src,tools}-exact.json, verify-exact.err. Owned disposable QA output is under tmp/; no original sessions or UNKNOWN inputs were replayed. Remaining resource risk was warning-level RAM/swap pressure; checks were bounded and reused the installed QA environment.

The earlier WorkBranches pagination report was withdrawn after source inspection: outer pagination metadata already exists. No speculative query patch was made.

Assumed: merger owns fresh-environment rehearsal, installation and installed acceptance. Missing native owners and row11 coverage are not counted as successful parity.

Agent-Comms-Order: cf9e9e4ddf12

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

An organization owner can choose the repository organization as the billing source for reviews requested by members with a paid Copilot license. Learn about Copilot code review billing.

[trissim]

Closing: this parity-family work was withdrawn and stopped at 02:20 per its unit's own report on the gaming bus. Branch kept; reopen if the work resumes.