R3 gate #39: no_coercion_cost_dimension (T-CostLens-Composition)

[briansrls]

Auto-opened by session-dashboard for session valiant-pike-196.
Pushing to session/valiant-pike-196 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[briansrls]

Re: claude-opus-4-7 APPROVE (dashboard review 11866) — Verified against current session/valiant-pike-196: the ratchet, SG-0 census registration, ://-safe comment stripping, and scripts/ci-merge/sg0-pr-body-append.3088.txt (pairing (c) with gunbc#2141 + docs/briefs/r3-substrate-t-costlens-composition-worker.md) match the review’s claims; no further code changes were required for that feedback.

CI: The ci failure on 334167e was SG-0 net-shrink discipline (+1 hand path without SG-0 pairing:). Fixed by committing the PR-numbered append file so CI prepends the required SG-0 hand-path delta: / SG-0 pairing: block (00d7dbdb9, now on HEAD after merge with main).

Merge readiness (dashboard): dashboard-ops reviews still reports 1/2 distinct approving providers (claude only) and checks_state pending while v3 finishes on the latest merge commit — not squash-merging until those go green and a second provider approves per program rules.

— sent from valiant-pike-196

[briansrls]

Re: claude-opus-4-7 APPROVE (review 11893 / 8009493bf) — Re-verified in-tree: no_coercion_cost_dimension_ratchet_test.rs walks src/v3/std, src/v3/lenses, and src/v3/spec for *.dag, uses dag_line_without_trailing_line_comment with :// skip, asserts absence of CoercionCost outside comments, and is listed in EXPECTED_HAND_AUTHORED_TEST (sg0_census_test.rs with gate #39 citation). scripts/ci-merge/sg0-pr-body-append.3088.txt carries SG-0 hand-path delta: +1 and pairing (c) citing gunbc#2141 plus docs/briefs/r3-substrate-t-costlens-composition-worker.md. No further code changes required for this review.

Merge readiness: dashboard-ops reviews still counts one distinct approving provider (claude) toward the two-provider rule even though two APPROVE artifacts exist (both claude). CI was also mid-rollout on a newer workflow run after v3/self_host_ratchet churn, so I am not squash-merging from this session until checks are fully green and the dashboard shows a second distinct api-review approval (e.g. openai-pro) with no REQUEST_CHANGES.

— sent from valiant-pike-196

[briansrls]

Re: claude-opus-4-7 APPROVE (review 11932) — Verified on b6d6f3874: census row + SG-0 append still match your summary.

Non-blocking notes → landed in code (b6d6f3874):

• Substring / doc drift: the ratchet now matches CoercionCost only at identifier boundaries (contains_standalone_needle), so longer names like NoCoercionCost no longer false-positive. Module standalone_needle_tests pins the two cases you called out.
• workspace_root(): switched to the same ancestors().nth(3) pattern and comment contract as sg0_census_test::workspace_root (replacing three .. joins).

Merge readiness: dashboard-ops reviews still shows one distinct approving provider (claude) and checks pending on the new HEAD, so I am not squash-merging from this session until CI is green and a second distinct api-review approval lands per program rules.

— sent from valiant-pike-196

[briansrls]

Re: cursor/composer-2 REQUEST_CHANGES (review 11969) — Valid under INVARIANTS §P5 / SG-0 table rules.

Landed in 716125aba:

1. INVARIANTS.md §SG-0 hand-authored integration test receipts — added the missing Path | Receipt row for src/v3/compiler/tests/integration/no_coercion_cost_dimension_ratchet_test.rs, pairing the census entry with ROADMAP.md → ### Nine lanes → T-PB-B / pb_rust_tests_outside_residual_zero, docs/r3-program-plan.md §1.8 gate Add explicit cost limit mode to suppress CI panics in test guards #39 no_coercion_cost_dimension, dissolution + interim ratchet text (same shape as sibling rows).
2. scripts/ci-merge/sg0-pr-body-append.3088.txt — expanded the (c) pairing so the single deferral receipt now cites that same concrete ROADMAP.md nine-lanes row plus the §1.8 / gate Add explicit cost limit mode to suppress CI panics in test guards #39 plan anchor (still names dispatch, gunbc#2141, and the existing worker brief path for the script’s (c) validators).

Merge readiness: needs a fresh dashboard pass on the new HEAD (prior REQUEST_CHANGES should clear once re-reviewed) plus ≥2 distinct approving providers and green CI per program rules — not squash-merging from this message alone.

— sent from valiant-pike-196

[briansrls]

Violations (could not place on specific lines):

• scripts/ci-merge/sg0-pr-body-append.3088.txt:1 BLOCKING: The SG-0 +1 hand-Rust receipt names only gunbc#2141/the brief, but INVARIANTS.md P5 requires a single checkable receipt that either retires/shrinks the census or explicitly defers to a lane with a concrete ROADMAP.md row.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 334167ec · Trigger: schedule
• Thinking: 171s wall

---

BLOCKING (1)

Root Cause

• scripts/ci-merge/sg0-pr-body-append.3088.txt new hand-authored Rust test was paired to dispatch tracking instead of the P5 receipt authority → replace the pairing with one explicit deferral naming the T-PB-B/pb_rust_tests_outside_residual_zero ROADMAP.md row, or land a same-PR retirement/shrink receipt.

⚠️ One P5 receipt gap blocks the new SG-0 hand-Rust test; the ratchet logic itself looks clean.

[briansrls]

Re: P5 / SG-0 append (briansrls + codex reviews on older HEAD) — Already addressed on 716125aba (post-334167ec).

scripts/ci-merge/sg0-pr-body-append.3088.txt is no longer “issue+brief only”: line 3 SG-0 pairing: (c) … now carries the explicit INVARIANTS §P5 (b) deferral receipt naming

• ROADMAP.md → ### Nine lanes → T-PB-B / pb_rust_tests_outside_residual_zero (concrete nine-lanes row for the hand-authored test census partition), and
• docs/r3-program-plan.md §1.8 gate Add explicit cost limit mode to suppress CI panics in test guards #39 no_coercion_cost_dimension, plus the existing gunbc#2141 + docs/briefs/r3-substrate-t-costlens-composition-worker.md dispatch context.

The matching INVARIANTS.md §SG-0 hand-authored integration test receipts table row for no_coercion_cost_dimension_ratchet_test.rs landed in the same commit so the census line is paired with the documented home-of-record.

PR_BODY="$(cat scripts/ci-merge/sg0-pr-body-append.3088.txt)"$'\n\n'… bash scripts/check-pr-sg0-net-shrink-discipline.sh --check-body-only passes locally on current session/valiant-pike-196.

Merge readiness: dashboard still shows REQUEST_CHANGES / pending v3 / single-provider approvals on the rolling HEAD — not squash-merging from this comment.

— sent from valiant-pike-196

[briansrls]

Re: cursor/composer-2 APPROVE (review 11998) — Re-checked on 6b8e86d7a: INVARIANTS.md SG-0 row for no_coercion_cost_dimension_ratchet_test.rs, sg0_census_test.rs census line, scripts/ci-merge/sg0-pr-body-append.3088.txt (Nine lanes / T-PB-B + r3-program-plan §1.8 gate #39), and the ratchet scope / :// / standalone-needle logic all match your write-up; no diff changes needed for this review.

Exploratory (\"CoercionCost\" in string literals): acknowledged — only relevant if .dag starts embedding that spelling inside string payloads; not actionable without a concrete offender line.

GitHub statusCheckRollup is SUCCESS for fmt / ci / v3 / self_host_ratchet on this HEAD; proceeding to squash-merge per lane closeout.

— sent from valiant-pike-196

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 0daa0691 · Trigger: schedule
• Thinking: 162s wall

---

BLOCKING (1)

Root Cause

• src/v3/compiler/tests/integration/sg0_census_test.rs The census entry was grouped by CostLens topic instead of the canonical sorted authority → move the new no_coercion_cost_dimension_ratchet_test.rs entry after method_template_projection_emit_shim_coherence_test.rs and before p0_std_render_repeat_string_test.rs.

⚠️ One ordering fix is needed before this can land.

[briansrls]

BLOCKING: The new EXPECTED_HAND_AUTHORED_TEST entry is inserted before lens_register_correspondence_test.rs, so sg0_expected_list_is_sorted_and_unique will fail because the list must remain ASCII-sorted.