{T-15} bin/main.dag: trampoline include! authority + B1 digest placeholders (TestClaim/Equals witness deferred: B1/T-22); Wave-6 not this lane

[briansrls]

Auto-opened by session-dashboard for session snappy-koi-363.
Pushing to session/snappy-koi-363 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]

Codex review (vacuous T-15 witness): addressed in 2f2b23b90.

• Removed the t_15_self_host_fixed_point TestClaim that used the same stub for input and expected, which could read as a false-green fixed-point check.
• Replaced with two structurally distinct digest placeholders (Conj vs Disj roots) so the eventual Equals/B1 witness has clear stage1 vs stage2 operand hooks.
• TestClaim is intentionally absent until real B1 content_hash(compile(…)) operands exist (TASKS.md T-15 + MOVE1/T-22 deferral); Status names that trigger.

No vacuous success path at the data layer; compile-only gate still clean (v2-compiler compile --source-root src/v4).

[briansrls]

Verified on current HEAD (2f2b23b90): src/v4/bin/main.dag still matches that summary — no t_15_self_host_fixed_point TestClaim, only the two digest placeholders and the deferred-witness Status line. No additional commit warranted for this item.

Merge readiness snapshot: GitHub mergeStateStatus is CLEAN and required checks on the latest run are green; dashboard-ops reviews still shows 1/2 distinct dashboard approvals on this SHA and openai-pro was in flight at last poll — re-check after that review posts.

— sent from snappy-koi-363

[briansrls]

Openai-pro (14574): No actionable findings in the posted artifact — APPROVE on 2f2b23b90 stands; nothing to change in-tree.

Claude exploratory (Conj vs Disj): The two stubs are not an arbitrary style choice. Once B1 content_hash is wired, empty Conj and empty Disj roots are different connective discriminants, so their canonical digests stay distinguishable while still being minimal type-node shapes. Collapsing both placeholders onto the same connective (or one shared data) would recreate the digest-collision / vacuous-equality hazard the codex thread flagged. I did not add another // body line here to stay inside Lane B’s terse header discipline (DECISIONS.md V-HDR); if we want that rationale in-file, the right place is a short clause on the existing Status line in a follow-up micro-edit (would need a new SHA + re-review cycle).

Merge readiness (dashboard + GitHub, current 2f2b23b90): ≥2 distinct api-review APPROVE (openai-pro, claude, cursor on this SHA), no open REQUEST_CHANGES on HEAD, mergeable clean, CI green → per operator policy I am stopping here and not running gh pr merge.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 2f2b23b9 · Trigger: schedule
• Thinking: 172s wall

---

BLOCKING (1)

Root Cause

• src/v4/bin/main.dag The no-premature-TestClaim cleanup conflated "wait for B1 content_hash operands before asserting fixed-point" with "move the trampoline contract out of bin/main.dag" → keep the digest-placeholder scaffold, but restore bin/main.dag as the T-15 trampoline authority.

⚠️ One blocking contract drift should be fixed before merge.

[briansrls]

Blocking review (trampoline authority / contract drift): fixed in 2bcf264e3.

• Restored bin/main.dag as the single trampoline authority: data main_rs_trampoline_authority: String = "include!(\"v4_main_generated.rs\");" — the one-liner lives here (STRUCTURE closed-system invariant 2 + TASKS T-15 / design-pure-bootstrap-zero:210), not “owned by” emit.
• Scope / Owns / Consumes / Status rewritten so 05_emit only materializes the artifact; it does not supersede this module’s contract (no more “trampoline is T-16/T-22” drift).
• Digest stubs unchanged; no premature TestClaim unchanged.

v2-compiler compile --source-root src/v4 → 0 diagnostics.

[briansrls]

Verified on current HEAD (2bcf264e3): src/v4/bin/main.dag already matches this note — main_rs_trampoline_authority plus the revised Scope/Owns/Consumes/Status are present; digest stubs unchanged; no premature TestClaim.

No additional commit from this pass (nothing left to apply for this item).

Merge readiness note: GitHub currently shows mergeStateStatus: BLOCKED with affected pending on the post-2bcf264e3 CI run; dashboard-ops reviews may still show a stale head_sha (2f2b23b90) until it refreshes — please re-poll after checks complete and expect fresh api-reviews on 2bcf264e3 if required.

— sent from snappy-koi-363

[briansrls]

Codex APPROVE_WITH_COMMENTS (kernel-ambient String): applied in 6e5739d92 — removed import v4.std.text { String } and dropped v4.std.text from Consumes per STRUCTURE.md §kernel-ambient (raw literal only). v2-compiler compile --source-root src/v4 still 0 diagnostics.

[briansrls]

Verified on HEAD (6e5739d92): this item is a relay of the prior resolution — main.dag already has Consumes: v4.std.node only, no import v4.std.text, and main_rs_trampoline_authority: String = ... unchanged. No further commit.

Merge readiness snapshot: GitHub shows mergeStateStatus: CLEAN and the latest gh pr checks run is green (fmt, affected, ci, v4 pass). dashboard-ops reviews still had checks_state: pending / ready: false on last poll (likely lag). Also: the dashboard’s approval_providers list may not all correspond to reviews pinned to 6e5739d92 (several posted APPROVE rows are on older SHAs); if policy is SHA-strict, schedule/confirm fresh api-reviews on 6e5739d92.

— sent from snappy-koi-363

[briansrls]

Verified: claude review 14669 matches current HEAD (6e5739d92) — src/v4/bin/main.dag is still the single-file trampoline + digest-placeholder scaffold described in the artifact; no findings ⇒ no code change from this pass.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 6e5739d9 · Trigger: schedule
• Thinking: 278s wall

---

Non-blocking — Strengths

• src/v4/bin/main.dag The B1 digest placeholders are documented as bounded stubs with a named content_hash/Equals witness trigger and no premature TestClaim.

✅ No blocking concerns.

[briansrls]

Verified on HEAD (6e5739d92, matches codex 6e5739d9): src/v4/bin/main.dag still matches the relayed strengths — named digest placeholders, Status names B1 content_hash / Equals witness deferral and no premature TestClaim, plus main_rs_trampoline_authority as the trampoline line authority. No blocking concerns ⇒ no commit from this pass.

Merge readiness: mergeStateStatus: CLEAN, required checks green, dashboard merge_criteria.ready: true with ≥2 api-review lane satisfied; PR remains OPEN — no gh pr merge per policy.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 2ba84e8b · Trigger: schedule
• Thinking: 101s wall

---

✅ No blocking concerns; this is a narrow .dag scaffold authority update and I did not see a new invariant violation in the changed lines.

[briansrls]

Verified on HEAD (2ba84e8b): git diff 6e5739d92859eea85a4f2b4ed1415573c08734a4..HEAD -- src/v4/bin/main.dag is empty (merge-of-main only); src/v4/bin/main.dag still matches the prior T-15 scaffold (trampoline String authority, Conj/Disj digest placeholders, no premature TestClaim, Consumes: v4.std.node only). Codex “no blocking concerns” on this SHA ⇒ no worker commit.

Merge readiness (mechanical checklist): GitHub mergeStateStatus: CLEAN, required checks green; dashboard reports merge_criteria.ready: true, checks_state: passing, has_request_changes: false, ≥2 distinct api-review providers in the summary (with vivid-ratified A carry-forward from 6e5739d9 for this merge shape). No gh pr merge per policy; lane still tracks operator resolution of the unresolved BLOCKING review thread as the binding human gate.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 2aa84e3c · Trigger: schedule
• Thinking: 177s wall

---

Non-blocking — Strengths

• src/v4/bin/main.dag The scaffold is narrow, keeps T-15 single authority for the trampoline, and names the B1/T-22 dissolution trigger for the digest placeholders.

✅ No blocking concerns.

[briansrls]

Verified on HEAD (2aa84e3c3f07ccb276eec22f2e4b54505475f3af, matches codex 2aa84e3c): another operator merge-of-main tip; git diff 2ba84e8bd8b793607bfd6f1337460c388f963323..HEAD -- src/v4/bin/main.dag is empty — src/v4/bin/main.dag unchanged vs prior held tip (still T-15 trampoline String authority, B1/T-22 + Equals deferral in Status, no premature TestClaim). Codex “no blocking concerns” ⇒ no worker commit (per hub hold: no churn on routine base staleness).

Mechanical merge readiness: mergeStateStatus: CLEAN, CI green, dashboard merge_criteria.ready: true / checks_state: passing / has_request_changes: false. No gh pr merge per policy; binding operator thread remains the human gate when applicable.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: b2acb041 · Trigger: schedule
• Thinking: 92s wall

---

Non-blocking — Strengths

• src/v4/bin/main.dag The scaffold is documented, bounded to the T-15 trampoline plus B1 digest placeholders, and names a concrete dissolution trigger, so I found no blocking invariant concern.

✅ No blocking concerns.

[briansrls]

Verified on HEAD (b2acb04179e70094cc6d4879e413435f42cd15c0, matches codex b2acb041): latest commit is merge origin/main; git diff 2aa84e3c3f07ccb276eec22f2e4b54505475f3af..HEAD -- src/v4/bin/main.dag is empty — main.dag unchanged (T-15 trampoline String authority, B1 digest placeholders, Status dissolution triggers incl. no premature TestClaim). v2-compiler compile --source-root src/v4 → 0 diagnostics. Codex “no blocking concerns” ⇒ no worker commit (hub hold: routine merge-of-main, no churn).

Mechanical readiness: mergeStateStatus: CLEAN, CI green, dashboard merge_criteria.ready: true / checks_state: passing / has_request_changes: false. No gh pr merge per policy.

— sent from snappy-koi-363

[briansrls]

Verified on HEAD (c9a1eb6fb) — matches claude review 14941 (same SHA): main_rs_trampoline_authority is the single trampoline String authority (:12); digest carriers are stub_stage{1,2}_emitted_rust_digest_placeholder with empty children, Conj vs Disj to keep distinct B1 content_hash until real operands land; Status already binds dissolution to B1 content_hash operands + Equals witness (T-22) — i.e. replace with hash-bearing Nodes, not a cosmetic rename (agree with the exploratory; no new commit needed for APPROVE-only).

Merge readiness snapshot: GitHub mergeStateStatus CLEAN, mergeable MERGEABLE, fmt / affected / ci / v4 pass on latest run. dashboard-ops reviews still showed merge_criteria.ready: false / mergeable: UNKNOWN at poll time — likely stale vs GitHub; re-poll if needed.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 2034a713 · Trigger: schedule
• Thinking: 149s wall

---

Non-blocking — Strengths

• src/v4/bin/main.dag The T-15 scaffold is bounded to the trampoline authority plus B1 digest placeholders and names the B1 content_hash/Equals witness trigger.
• .github/workflows/ci.yml The affected-job timeout increase is confined to the existing manual CI bridge while T-24 owns the eventual ci.yml projection.

✅ No blocking concerns.

[briansrls]

Verified on HEAD (c9a1eb6fb):

1. cursor review 14951 — APPROVE on this SHA; src/v4/bin/main.dag matches the cited lines (:2–5, :15–23): single trampoline String authority, B1/Equals/T-22 deferral in Status, well-formed empty Conj/Disj placeholders. No commit (no findings).

2. codex relay (2034a713) — main.dag at HEAD is unchanged from 2034a713a for the scaffold hunks (merge c9a1eb6fb only advanced main + CI); codex “no blocking concerns” on the main.dag / ci.yml strengths still applies. No commit.

Merge readiness: dashboard-ops merge_criteria.ready: true, meets_two_approval_rule: true, has_request_changes: false, GitHub mergeStateStatus: CLEAN, mergeable: MERGEABLE, fmt / affected / ci / v4 green. No gh pr merge per policy — stopping worker-side churn unless new feedback.

— sent from snappy-koi-363

[briansrls]

Verified codex review (dashboard artifact → 8bc808c45) against current HEAD — same commit; findings match the tree, no fix commit:

• src/v4/bin/main.dag (:2–5, :12–23, Status) — single main_rs_trampoline_authority trampoline literal; digest carriers are explicitly stub placeholders (Conj / Disj, empty children) with B1 / Equals / T-22 deferral spelled in Status (no premature TestClaim).
• v4_bin_main_dag_smoke_test.rs — parse-only receipt + module docs state compile_to_dag hermeticity deferred (import v4.std.node / bootstrap collision); grounded-indicator HOLD noted per neat-hawk-87 relay.
• P5 / SG-0 — sg0_census_test.rs census line + INVARIANTS.md §P5(b) row for v4_bin_main_dag_smoke_test.rs align; scripts/ci-merge/sg0-pr-body-append.3356.txt pairs the +1 hand-path delta for CI.

Merge readiness (snapshot): GitHub mergeable: MERGEABLE; mergeStateStatus: BLOCKED (protection / pending checks — ci / v4 / v3 were still running at poll time). dashboard-ops: checks_state: failing + ready: false on stale head_sha (75c47c7) vs HEAD (8bc808c45) — re-poll after current run green; codex 14986 shows verdict: unknown in dashboard JSON even though stdout is APPROVE (parser gap — treat artifact as authoritative for this thread). No gh pr merge per policy.

— sent from snappy-koi-363

[briansrls]

Verified cursor/composer-2 review 14993 (Verdict: APPROVE) against HEAD (8bc808c45) — no fix commit (no blocking findings).

Cross-check vs review bullets

• src/v4/bin/main.dag: Single trampoline authority main_rs_trampoline_authority + two digest placeholder Nodes; Status defers B1 content_hash / Equals / T-22** and disclaims premature **TestClaim` — matches the review.
• P5 (b) single surface: v4_bin_main_dag_smoke_test.rs module docs, EXPECTED_HAND_AUTHORED_TEST entry (sg0_census_test.rs), and INVARIANTS.md L369 describe the same tokenize/parse interim ratchet + dissolution path — aligned with neighboring smoke rows.
• Exploratory ci.yml note: affected is timeout-minutes: 15 at :44–47 (ubicloud-standard-2); operational only — no change requested.

Merge readiness (poll snapshot): dashboard-ops reports checks_state: passing, active_review_count: 0, approval_count: 3 on head_sha = 8bc808c45 (claude, cursor, openai-pro — satisfies ≥2 with openai-pro present). merge_criteria.ready: false here tracks mergeable: UNKNOWN from GitHub while mergeability is recomputing — re-check mergeStateStatus/mergeable in the UI/API once it settles to CLEAN/MERGEABLE. No gh pr merge per policy.

— sent from snappy-koi-363

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 75c47c7a · Trigger: schedule
• Thinking: 269s wall

---

BLOCKING (1)

Root Cause

• .github/workflows/ci.yml The v3 integration harness remains HOT-FIX zero-filtered while new SG-0 smoke receipts are added under tests/integration.rs → add a CI-executed exact gate for this receipt or move the assertion into a live v4/generated-TestClaim path.

ROADMAP — Incomplete

• T-PB-B / pb_rust_tests_outside_residual_zero: The census and INVARIANTS receipt are present, but the added Rust smoke is not a live CI ratchet while integration execution is skipped.

⚠️ The T-15 authority looks aligned, but the new receipt needs to execute somewhere live before it can count as the PR’s P5/T-PB-B ratchet.

[briansrls]

BLOCKING: The new P5/T-PB-B receipt is a Rust integration test, but CI still zero-filters the integration harness, so its tokenize/parse/source-anchor assertions are not actually run.

[briansrls]

Blocking review (inline @ v4_bin_main_dag_smoke_test.rs) — verified

The finding was correct: the receipt lives in the v3-compiler integration harness, while the v3 job still runs that harness under the gunbc#846 __HOT_FIX_NONEXISTENT_FILTER__ zero-filter, and the v4 job does not run this test—so v4-only PRs could ship without ever executing the tokenize/parse/source-anchor assertions.

Fix: #3392 adds an exact-filtered cargo test step to the always-on ci job (when v3/v4/workflow_policy is hot or on main push) and regenerates dsl/gunbc/ci_github_actions_workflow.dag.

— sent from snappy-koi-363