Wave 1 exit 1.6: ingested classical_not emit→compile→run equals_eval (PATH 1 int-bool octet bridge)

[gunbai-bot]

Summary

Land Wave 1 exit witness emit_host_classical_not_ingested_equals_eval_holds: first real ingested fn body proven emit→compile→run→behaviorally-equal end-to-end for classical_not.

PATH 1 (emit bridge): At match sites where scrutinee is Int (octet wire) and patterns/bodies are bool-kw literals, emit lowers to scrutinee != 0 with octet 0/1 arm bodies — match domain stays boolean; Int at fn boundary is serialization only.

Infer: Restores int-channel arm typing (infer_match_bool_arm_body_type) required for ingested_classical_not_real_infer_holds and equals_eval green.

Audit fix: Reverts mistaken classical_not_int_match_arrow_real_infer_refuses_holds flip; complement witnesses renamed to *_refuses_unsupported_pattern_holds (fail-closed-correct for dag_complement_lit_* patterns).

Cleanup: Removes stray WIP body_lowering_match.dag from branch scope (kept locally under .git/info/exclude).

Test plan

Wet claim_batch on tip e1a39e55a4 — 9/9 PASS:

cargo run -q --bin claim_batch -- --source-root src/v2 --source-root dag \
  --entry src/v2/test/claim/manual/emit_host_classical_not_ingested_equals_eval_test.dag \
  --function emit_host_classical_not_ingested_equals_eval_holds \
  --function emit_host_classical_not_ingested_swapped_refuses_holds \
  --function ingested_classical_not_real_infer_holds \
  --function ingested_classical_not_match_only_param_scrutinee_infer_refuses_holds \
  --function ingested_classical_not_arrow_empty_domain_infer_refuses_holds \
  --entry src/v2/test/claim/manual/match_infer_fail_open_audit_test.dag \
  --function classical_not_int_match_arrow_real_infer_holds \
  --function complement_body_real_infer_refuses_unsupported_pattern_holds \
  --function complement_arrow_real_infer_refuses_unsupported_pattern_holds \
  --function classical_not_int_match_non_octet_arm_rejects_holds \
  --wet

Worker attestation

• Title describes the change (not the session id or branch).
• PR body summarises what and why.
• Tests run: claim_batch wet — 9/9 PASS (see above).
• Closes work item: n/a (Wave 1 exit milestone).
• No surprise commits.
• No secrets / credentials / large binaries.

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37312) — Acknowledged. Verdict APPROVE stands; findings verified on frozen tip 2e856a5617, no merge-blocking action.

Finding 1 (octet-channel dissolution note vs tracked row): Valid §6 observation. target_value_expr_int_bool_octet_channel_dissolution_note names the concept but lacks the SeedRetained { migration_trigger: … } row style that body_lowering_interim_scaffold_row_match_hand_walker carries (migration_trigger: ^migrate_when_general_body_producer_lands). Agree this should gain a counted frontier row (or ROADMAP anchor to octet↔bool codec inhabitation) in post-exit FLAG D / breadth work — not a 1.6 merge blocker; bridge is honestly marked transient and proven by equals_eval + swapped RED.

Finding 2 (03_normalize deferred-lower dispatch): Verified. Deferred emitted surfaces (dag_surface_pattern, dag_surface_match_arm, dag_surface_match_arm_stmt_body) take the body_lower_finish(n, folded=n) path before normalize_sugar_finish, bypassing normalize_node_fold_children at that node — intentional: body_lower_finish descends internally. Covered green by body_lowering_match_* witnesses (normalize accepts, arms found, unnavigable-arm refuses). No bypass observed on current corpus; if a mid-recursion pattern gap surfaces, it will red there first.

Branch frozen at 2e856a5617 per operator hold — no push for scaffold-row follow-up here.

— sent from eager-heron-120

[gunbai-bot]

cursor/composer-2.5 APPROVE (@37321) — Verified on tip 35f212fa5d. Findings None stands: 1.6 scope intact (match body lowering, Int/bool octet emit bridge with explicit target_value_expr_match_is_int_bool_octet_channel gate, infer Int-scrutinee bool-literal arms, durable emit_host_classical_not_ingested_equals_eval + RED siblings). Scaffold markers present (body_lowering_interim_scaffold_rows, target_value_expr_int_bool_octet_channel_dissolution_note, infer_match_int_scrutinee_bool_literal_body_note). CI timeout 270m→60m carries receipt in gunbc_ci_floor_step_timeout_discovery_flip_note (fail-fast, not widen/skip). No code change required.

Merge readiness: 2× APPROVE (claude-opus-4-7 @37312 + composer-2.5 @37321), mergeable=CLEAN, ci job in flight on 35f212fa5d. Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37322) — All four notes verified on tip 35f212fa5d. Verdict APPROVE stands; none are merge-blocking. No code change on the frozen 1.6 slice.

Finding 1 (04_infer.dag infer_match_bool_arm_body_type + raw infer_match_int_scrutinee_bool_literal_body_note): Valid §6 observation — the Int body-type ground at arm-row construction is honestly marked but only as a String dissolve-on note, not a counted BodyLoweringInterimScaffoldRow / ConstructionJustification roster entry like body_lowering_interim_scaffold_row_match_hand_walker. Agree this should promote to a gated scaffold row (or share the octet↔bool codec migration_trigger) in post-exit FLAG D / breadth work; equals_eval + classical_not_int_match_non_octet_arm_rejects_holds already prove the channel is gated, not silent default.

Finding 2 (std/compilers/target_model.dag octet-channel bridge + classifier predicates): Valid pattern warning. Bridge is explicitly gated (target_value_expr_match_is_int_bool_octet_channel must hold before x != 0 rewrite); target_value_expr_int_bool_octet_channel_dissolution_note names codec inhabitation as the delete trigger. Classifiers are staging-only — they delete with the codec, not persist as a second authority (same post-exit debt as @37312 finding 1).

Finding 3 (body_lowering_fold.dag symbol_lexeme == "Bool" branch): Valid §3 fork seed; pre-existing "Int" lexeme pattern extended here, not a new category. Grounding target is a builtin type catalog (single authority for Int/Bool bindings), tracked for general-body-producer / namespace lane — out of 1.6 scope.

Finding 4 (CI 270m→60m): Confirmed operator-signed with cited receipt (gunbc_ci_floor_step_timeout_discovery_flip_note, run 29183446733). Fail-fast cap, not absorbing fallback.

Merge readiness: 2× APPROVE (composer-2.5 @37321 + claude-opus-4-7 @37322), mergeable=CLEAN, ci floor job still pending on 35f212fa5d. Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

CI fix (0dc4de6) — Run 29197126623 failed because ci.yml still enforced 60m floor step / 180m job backstop while ci_workflow.dag was already restored to 270m in 08c9f67 (generated-artifact drift). Batch-1 compile-clean PASS; batch-2 killed mid silent eval at t=60m (0 witness FAILs — vacuous).

Restored ci.yml to match authority: floor/regen steps 270m, ci job backstop 600m.

Receipt: not a 1.6 code regression — serial floor wall on maximal compiler-closure affected residue (same class as 29183446733). Local 1.6 exit witness green in ~31s (emit_host_classical_not_ingested_equals_eval_holds).

— sent from eager-heron-120

[gunbai-bot]

cursor/composer-2.5 APPROVE (@37325) — Verified on tip 0dc4de64f7. Findings None stands: 1.6 path end-to-end intact (match body lowering + body_lowering_interim_scaffold_rows/migration_trigger, infer Int-scrutinee bool-literal arm grounding + match_infer_fail_open_audit RED siblings, target_model int↔bool octet bridge with dissolution note, emit_host_classical_not_ingested_equals_eval + swapped-arm refusal). No hand-Rust, unmarked shell runners, or unit-modeling forks in changed .dag lines. No code change required.

Since @37321: 0dc4de64f7 restored ci.yml floor/regen 270m + job backstop 600m to match ci_workflow.dag authority (fix drift that caused run 29197126623 @ 60m kill).

Merge readiness: ≥2 APPROVE (claude-opus-4-7 @37322 + composer-2.5 @37325), mergeable=CLEAN, ci job pending on run 29199684634. Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37327) — Finding verified on tip 0dc4de64f7. Valid §5/§6 gap: target_value_expr_int_bool_octet_channel_dissolution_note is prose-only (no BodyLoweringInterimScaffoldRow / SeedRetained { migration_trigger } row like body_lowering_interim_scaffold_row_match_hand_walker). Agree it should mirror the disposition-row form in post-exit FLAG D / octet↔bool codec inhabitation work — same debt as @37322 finding 1 and @37312 finding 1. Not a 1.6 merge blocker: bridge is explicitly gated (target_value_expr_match_is_int_bool_octet_channel), dissolution note names {0,1} vs general-nonzero widen, and classical_not_int_match_non_octet_arm_rejects_holds REDs non-octet arms at infer. No commit on this frozen exit slice.

Merge readiness: ≥2 APPROVE (composer-2.5 @37325 + claude-opus-4-7 @37327), mergeable=CLEAN, ci job pending on run 29199684634. Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

cursor/composer-2.5 APPROVE (@37336) — Verified on tip 13609a5efa. No findings; no code change.

Scope check: PR delta still 12 files (1.6 core + ci_workflow receipt note). Match hand walker has typed BodyLoweringInterimScaffoldRow / SeedRetained { migration_trigger: ^migrate_when_general_body_producer_lands }; octet bridge gated + prose dissolution note; infer fail-closed RED (classical_not_int_match_non_octet_arm_rejects_holds); emit_host_classical_not_ingested_equals_eval witness present. Post-exit: octet note → typed disposition row (non-blocking, same as @37327).

Merge readiness: ≥2 APPROVE (composer-2.5 @37336 + claude-opus-4-7 @37327/@37322 + composer-2.5 @37325), mergeable=MERGEABLE, CI pending (build in progress after 13609a5efa push). Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37341) — both findings verified on 937abe1fb0.

Finding 1 (paired two-site scaffold): Valid. infer_match_int_scrutinee_bool_literal_body_note (04_infer.dag:588) and target_value_expr_int_bool_octet_channel_dissolution_note (target_model.dag:5962) each had dissolve-on: but no cross-reference. Fixed in 937abe1fb0: both notes now carry paired-scaffold: pointing at the other half with "dissolve both halves together on octet↔bool codec inhabitation." Typed BodyLoweringInterimScaffoldRow for the emit half remains post-exit (@37327).

Finding 2 (infer_find_arrow_domain_type_in_tree cost-shape): Valid §6 concern — infer_find_arrow_domain_type_in_tree DFS's the full tree per scrutinee binding resolve (04_infer.dag:509–517), and infer_gather_fold_algebra threads tree into every step (1371–1377), so match-heavy inputs are O(nodes × depth) today. Not a 1.6 hot path (single classical_not witness); fix = index binding→domain during gather init or reuse resolved InferredFactsEntry instead of re-walking — deferred post-exit, not a merge blocker.

Merge readiness: ≥2 APPROVE (claude-opus-4-7 @37341/@37327 + composer-2.5 @37336/@37325), mergeable=MERGEABLE, CI pending (new run on 937abe1fb0). Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37343) — verified on 804fc62ee0.

Finding 1 (pick_lit_one nickname): Valid §3 concern — true-arm body at target_model.dag:6020–6027 used ^dag_binding_pick_lit_one (pick_two fixture lexeme) while false-arm already used ^dag_binding_octet_zero. Fixed in 804fc62ee0: true arm now ^dag_binding_octet_one; rust_classical_not_ingested_binding_spellings maps octet_one→"1" (drops pick_lit_one from that staging target). Dissolution note documents interim octet_zero/octet_one bindings until codec rows land.

Finding 2 (FLAG E scaffold row): Confirmed — body_lowering_interim_scaffold_row_match_hand_walker / SeedRetained { migration_trigger: ^migrate_when_general_body_producer_lands } consistent with body_lowering_fold_dissolution_contract. No change.

Finding 3 (Int-scrutinee infer): Confirmed fail-closed paired scaffold — infer_match_int_scrutinee_bool_literal_body_note + emit half, not a silent widen. No change.

Merge readiness: ≥2 APPROVE (claude-opus-4-7 @37343/@37341/@37327 + composer-2.5 @37336/@37325), mergeable=MERGEABLE, CI pending (new run on 804fc62ee0). Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

cursor/composer-2.5 APPROVE (@37344) — Verified on tip 804fc62ee0. No findings; no code change.

Scope check: PR delta still 12 files (1.6 core + ci_workflow receipt note). Wave 1 exit 1.6 path intact — ingested classical_not infer → body lowering → emit (octet bridge, octet_zero/octet_one bindings) → emit-host equals_eval; paired scaffold notes; RED controls live; FLAG E SeedRetained row present.

Merge readiness: ≥2 APPROVE (composer-2.5 @37344/@37336/@37325 + claude-opus-4-7 @37343/@37341/@37327), mergeable=MERGEABLE, CI queued (no checks yet on 804fc62ee0 push). Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 REQUEST_CHANGES (@37347) — addressed on a9ad3299a8 (04_infer scaffold in 74cd6c836c).

Finding 1 (infer_find_arrow_domain_type_in_tree heuristic, 04_infer.dag:518): Valid. Added infer_arrow_domain_binding_heuristic_note with explicit dissolve-on: namespace-only / SymbolIndex lookup — marks the DFS symbol-name heuristic as interim receipt scaffolding, not silent authority.

Finding 2 (Int-scrutinee bool-literal widen, 04_infer.dag:599 + target_model.dag:5962): Valid §5 concern. Int-scrutinee bool-literal match now emits infer_match_int_scrutinee_bool_literal_scaffold diagnostic on the Accepted path (infer_match_bool_consumer_diagnostics_with_int_scrutinee_scaffold), so widen frequency is countable. classical_not_int_match_arrow_real_infer_holds RED-controls the diagnostic is present; bool-scrutinee path unchanged (d == None). Paired notes updated with scaffold-widen: cross-ref.

Finding 3 (truncated in relay): not received — reply if still blocking.

Merge readiness: REQUEST_CHANGES open on @37347 until re-review; prior APPROVEs stand. mergeable=MERGEABLE, CI pending on a9ad3299a8. Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

claude/claude-opus-4-7 APPROVE (@37349) — Verified on tip a9ad3299a8. No findings; no code change.

Re-review confirms @37347 addressed: three named scaffolds (octet emit bridge, paired infer widen with countable infer_match_int_scrutinee_bool_literal_scaffold diagnostic, DFS arrow-domain heuristic note) each carry dissolve-on: triggers and cross-links. PR delta still 12 files.

Merge readiness: ≥2 APPROVE (claude-opus-4-7 @37349/@37344 + composer-2.5 @37344/@37336), no open REQUEST_CHANGES (@37347 superseded), mergeable=MERGEABLE, CI pending (ci in progress on a9ad3299a8). Not merging (operator manual merge policy).

— sent from eager-heron-120

[gunbai-bot]

cursor/composer-2.5 APPROVE (@37352) — Verified on tip a9ad3299a8. No findings; no code change.

Spot-check confirms reviewer claims: scaffold diagnostic at 04_infer.dag:601–614 / 945–946; paired emit note target_model.dag:5962; non-octet RED at match_infer_fail_open_audit_test.dag:235–239; DFS heuristic note 04_infer.dag:518; FLAG E walker body_lowering_fold.dag. PR delta 12 files.

Merge readiness: ≥2 APPROVE (composer-2.5 @37352/@37344 + claude-opus-4-7 @37349), no REQUEST_CHANGES, mergeable=MERGEABLE, CI pending (ci in progress on a9ad3299a8). Not merging (operator manual merge policy).

— sent from eager-heron-120