R3 Gap 7 T-WAD Slice 6 BinaryShim runtime + slow-test-exemptions dissolution — gates #98-#103 cascade

[briansrls]

Auto-opened by session-dashboard for session wise-gull-139.
Pushing to session/wise-gull-139 advances this PR.

SG-0 hand-path delta: +2

SG-0 pairing: (c) dispatch-tracked T-WAD Full R3-close program (operator/Director ratification) covers this census net-add for Slice 6 / gate #102 substrate; P5(b) receipt — deleted scaffold scripts/test-node-wall-clock-ratchet.jsonl (warn policy sole authority → dsl/gunbc/test_node_wall_clock_ratchet.dag).
gunbc#828

Worker attestation

• Title describes the change (not the session id or branch).
• PR body summarises what and why (below).
• Tests: cargo test -p v3-compiler ratchet_dag_warn_manifest, bash scripts/test-check-test-timeout.sh, cargo clippy -p v3-compiler --bin gunbc-ci -- -D warnings, gen_gunbc_ci_workflow_dag drift check — local PASS before push.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises.
• No secrets / credentials / large binaries staged.

Summary

Moves Phase-0 slow-test warn policy from checked-in scripts/test-node-wall-clock-ratchet.jsonl to substrate authority dsl/gunbc/test_node_wall_clock_ratchet.dag, projected fail-closed through new gunbc-ci wall-clock-warn-manifest for scripts/check-test-timeout.sh. Adds the gunbc-ci binary (BinaryShim entrypoint stub: --workflow ci --event …) and CI cargo build --bin gunbc-ci before the per-test ratchet. Regenerates dsl/gunbc/ci_github_actions_workflow.dag from .github/workflows/ci.yml.

Test plan

• cargo test -p v3-compiler ratchet_dag_warn_manifest
• bash scripts/test-check-test-timeout.sh
• cargo build -p v3-compiler --bin gunbc-ci + smoke wall-clock-warn-manifest / --workflow ci --event
• cargo test -p v3-compiler gunbc_ci_github_actions_workflow_dag_matches_yaml_generator_output
• CI: full ci workflow on PR branch after SG-0 body fix

[briansrls]

codex review (artifact / APPROVE) — verified

Re-checked the branch against the stated claims: warn rows are substrate-modeled in dsl/gunbc/test_node_wall_clock_ratchet.dag, projected via gunbc-ci wall-clock-warn-manifest into the same JSONL shape check-test-timeout.sh already consumed; checked-in scripts/test-node-wall-clock-ratchet.jsonl is removed; INVARIANTS / modeling / CODING / TESTING discipline on the touched surface still holds.

CI / process (no new commit): the failing ci job was SG-0 PR-body discipline (SG-0 hand-path delta: + SG-0 pairing: for the +2 census net-add). Updated the PR description accordingly so the gate can go green on the next run.

Not squash-merging yet: dashboard merge rule still needs a second distinct approving provider (currently only codex) and a green full ci / v3 run on the updated HEAD.

— sent from wise-gull-139

[briansrls]

claude-opus-4-7 review (dashboard artifact) — verified on current session/wise-gull-139

1. Cargo.toml / R1C-E comment pairing — Confirmed: the [[bin]] r1c_e_emit_gates block sits immediately under the R1C-E ExecuteCommand comment, and [[bin]] gunbc-ci follows after it (no longer sandwiched between comment and r1c_e_emit_gates).

2. --workflow / --event stub fail-open — Addressed: the dispatch path now exits 2 by default once inputs validate, with stderr explaining that gate-matrix dispatch is not wired. Explicit smoke is opt-in via GUNBC_CI_ALLOW_DISPATCH_STUB=1 (or true); documented in the gunbc-ci module header. Current YAML still only cargo builds the binary for the ratchet projector, so CI behavior is unchanged until a step actually invokes dispatch.

Exploratory notes (substrate projector shape, TEST_TIMEOUT_MANIFEST split, shared JSON helper) — acknowledged; no further code change in this pass.

— sent from wise-gull-139

[briansrls]

cursor/composer-2 review (dashboard artifact) — addressed

1. wall_clock_ratchet_manifest test — Valid: lines.len() > 50 was an implementation-detail backlog ratchet. Replaced with per-line JSON parse (test string + policy == "warn"), which stays correct as rows are payed down (including to zero lines). Test renamed to ratchet_dag_warn_manifest_lines_are_parseable_warn_policy_objects. Landed in 4093762c1.

2. P5(b) / sg0-pr-body-append.3032.txt — Valid: append now enumerates the deleted scaffold scripts/test-node-wall-clock-ratchet.jsonl as the checkable dissolution receipt alongside (c) / gunbc#828. Same sentence synced into the live PR body SG-0 block + attestation/test-plan filters updated for the renamed test.

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 180b21e7 · Trigger: schedule
• Thinking: 226s wall

---

BLOCKING (1)

Root Cause

• dsl/gunbc/test_node_wall_clock_ratchet.dag JSONL was dissolved into a .dag-projected warn-token bridge, but the gate-closure wording was promoted before policy actually flows from TestNodeCostDimension/TimingBudget facts → keep the bridge explicitly marked interim or wire the policy to those substrate facts now.

ROADMAP — Incomplete

• slow_test_exemptions_dissolved: The old and new warn lists match at 84 tokens, but this is still a warn-policy bridge rather than #102’s modeled timing-fact closure.

⚠️ One wording/scope fix is needed so the bridge does not get recorded as gate #102 closure.

[briansrls]

Re: CI @ cf43346 (dashboard relay) — already fixed upstream of that SHA.

• Failure: ci job regen_bootstrap --verify — src/v3/std/verification.dag comment edits shifted embedded spans; committed bootstrap_generated*.rs drifted from fresh compile.
• Fix: 88590f4b6 (fix(ci): refresh bootstrap snapshots after verification.dag comment span drift) — current PR HEAD 88590f4b61bf…. cf43346 is an ancestor; please re-key alerts to HEAD, not that commit.
• Now: ci / fmt / changes green on run 25828098179; v3 may still be running (long job).

— sent from wise-gull-139

[briansrls]

cursor/composer-2 review (artifact 11397) — Findings: none actionable against rubric (agree with APPROVE).

Exploratory (header in scripts/test-check-test-timeout.sh): addressed in 0e91d0ed1 — merged the duplicate draft lines into one short block (fixture TEST_TIMEOUT_MANIFEST vs production .dag + gunbc-ci, interim #102 bridge called out).

[briansrls]

Re: relay of briansrls comment (composer-2 artifact 11397) — no further code change.

• Rubric findings: none (composer APPROVE); nothing to implement.
• Exploratory header: already landed at 0e91d0ed1; current scripts/test-check-test-timeout.sh opens with a single 4-line block (fixture TEST_TIMEOUT_MANIFEST vs production .dag + gunbc-ci, interim BT-E1: fix cross-module transport node deduplication #102 bridge).

Merge readiness (this moment): ci / fmt / changes green on run 25828355292; v3 still in_progress → check suite not complete, mergeStateStatus remains BLOCKED. Cannot squash-merge until v3 finishes green. Dashboard ≥2 distinct api-review approvals are outside gh pr view here — please confirm in dashboard; GitHub formal reviews on this PR are not a reliable proxy for that rule.

— sent from wise-gull-139

[briansrls]

claude-opus-4-7 review (artifact 11417) — verified on current branch; no code change (your verdict is APPROVE with no rubric findings).

• BT-E1: fix cross-module transport node deduplication #102 vs interim bridge: docs + both verification.dag + test_node_wall_clock_ratchet.dag + wall_clock_ratchet_manifest rustdoc still distinguish pass target (TestNodeCostDimension timing facts) from the warn-token projection — no regression spotted.
• P5(b) / SG-0: sg0-pr-body-append.3032.txt still names deleted scripts/test-node-wall-clock-ratchet.jsonl; census includes bin/gunbc_ci.rs and wall_clock_ratchet_manifest.rs (sg0_census_test.rs paths cited).
• gunbc-ci dispatch stub: --workflow / --event still exits 2 unless GUNBC_CI_ALLOW_DISPATCH_STUB (gunbc_ci.rs ~99–112).
• emit_warn_policy_jsonl_lines: still fail-closed on missing declaration / shape drift (ValueBody::List of string test field).
• check-test-timeout.sh traps: first EXIT trap clears temp log + projected manifest when set; after warn_list is created the trap is replaced so one handler removes log, manifest, and warn_list (three paths) — matches your read.
• Modeled CI workflow .dag: still regenerated from ci.yml (generator drift gate unchanged).

Exploratory (one binary, two roles): acknowledged — splitting wall-clock-warn-manifest vs future dispatch into separate bins would be a later hygiene move if the surface grows; not required for this slice.

Merge readiness: ci / fmt / changes green on latest run; v3 still in flight → cannot claim full CI green or squash-merge from here yet. Please re-poll when v3 completes; branch protection + dashboard ≥2 api-review rule still applies outside this comment.

— sent from wise-gull-139

[briansrls]

cursor/composer-2 review (artifact 11435) — Findings: none; Verdict: APPROVE — verified against current session/wise-gull-139 HEAD:

• Interim-bridge / not-BT-E1: fix cross-module transport node deduplication #102-closure language still present across TESTING.md, docs/r3-program-plan.md §1.8 row 102, docs/r3-structure.md, dsl/std/verification.dag, src/v3/std/verification.dag, wall_clock_ratchet_manifest.rs, gunbc_ci.rs.
• docs/r3-remaining-work-dependency-graph.md still has BT-E1: fix cross-module transport node deduplication #102 as DECLARED (not spurious CLOSED).
• P5(b): scripts/ci-merge/sg0-pr-body-append.3032.txt still pairs census +2 with deleted scripts/test-node-wall-clock-ratchet.jsonl + interim posture.
• check-test-timeout.sh still fail-closed on missing gunbc-ci / projection failure; TEST_TIMEOUT_MANIFEST path preserved for test-check-test-timeout.sh.

No commit — rubric-aligned as written.

Merge readiness: ci / fmt / changes green; v3 still pending on latest workflow → not full CI green / mergeStateStatus still BLOCKED here. Dashboard ≥2 api-review rule not asserted via gh alone. Squash-merge not executed this round.

— sent from wise-gull-139

[briansrls]

codex review (artifact 11449) — Verdict: APPROVE — spot-checked current HEAD against your rubric claims:

• Interim bridge / not BT-E1: fix cross-module transport node deduplication #102 closure: still explicit in docs + module rustdoc; no “BT-E1: fix cross-module transport node deduplication #102 green” misrepresentation found in the touched surfaces.
• JSONL scaffold removed: scripts/test-node-wall-clock-ratchet.jsonl absent from tree (substrate + gunbc-ci projection is the live path).
• gunbc-ci dispatch: unimplemented --workflow/--event path still exits 2 unless GUNBC_CI_ALLOW_DISPATCH_STUB — no silent success.

No commit — no concrete findings to implement.

Merge readiness: ci / fmt / changes green; v3 still pending → full CI not green, mergeStateStatus BLOCKED; squash-merge not run. Dashboard ≥2 api-review / no REQUEST_CHANGES / mergeable=CLEAN must still be confirmed there.

— sent from wise-gull-139

[briansrls]

cursor/composer-2 review (artifact 11464) — Findings: none; Verdict: APPROVE — re-verified on current origin/main...HEAD: warn tokens live in dsl/gunbc/test_node_wall_clock_ratchet.dag with gunbc-ci wall-clock-warn-manifest + wall_clock_ratchet_manifest.rs; checked-in JSONL scaffold removed; CI builds gunbc-ci before the ratchet; interim-vs-#102-closure posture still explicit across TESTING.md, program-plan / r3-structure, both verification.dag files, projector + bin rustdoc; docs/r3-remaining-work-dependency-graph.md still shows #102 DECLARED (not spurious CLOSED). P5(b) append + census pairing unchanged in intent.

No commit.

Merge readiness: ci / fmt / changes green; v3 pending → not full CI green / merge still BLOCKED by policy. Squash-merge not executed. Dashboard ≥2 api-review + no REQUEST_CHANGES + mergeable=CLEAN remains out-of-band for gh.

— sent from wise-gull-139

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 8d2fc16b · Trigger: schedule
• Thinking: 347s wall

---

Non-blocking — Strengths

• scripts/check-test-timeout.sh The timeout consumer still fails closed on unknown over-budget tests while moving the default warn-policy source to the modeled .dag bridge.
• src/v3/compiler/src/bin/gunbc_ci.rs The BinaryShim dispatch stub exits 2 unless explicitly smoke-enabled, so the unimplemented runtime path does not falsely report success.

ROADMAP — Verified

• slow_test_exemptions_dissolved: The roadmap/status text now preserves #102 as DECLARED/not GREEN while documenting the .dag-to-JSONL warn-policy bridge as interim.

✅ No blocking concerns; the PR preserves the #102 pass target and provides a checkable P5 receipt for the added hand-Rust surface.

[briansrls]

Re: codex scheduled api-review (sha 8d2fc16b in relay body) — No blocking concerns matches current branch (HEAD may have advanced; same slice).

• scripts/check-test-timeout.sh: default path still projects warn policy from the modeled .dag via gunbc-ci; unknown over-budget tests still fail closed; TEST_TIMEOUT_MANIFEST override preserved for hermetic/self-test.
• gunbc-ci dispatch stub: still exits 2 unless GUNBC_CI_ALLOW_DISPATCH_STUB is set — no silent green on unimplemented --workflow/--event.
• ROADMAP / program-plan / r3-structure: BT-E1: fix cross-module transport node deduplication #102 slow_test_exemptions_dissolved remains DECLARED / not-GREEN framing with the interim .dag→JSONL bridge called out (no false closure).

No commit — strengths/non-blocking notes only.

Merge readiness: ci / fmt / changes green; v3 still in progress on workflow 25830976660 → full CI not green, merge BLOCKED by required checks / policy. Squash-merge not run. Dashboard ≥2 api-review + mergeable=CLEAN still needs your dashboard snapshot.

— sent from wise-gull-139