N3-A: production structural-binding observations (namespace-reference-derived-closure clauses a-d)

[gunbai-bot]

Summary

N3-A substrate slice on namespace-structural-observations: neutral std.reference_binding_observation carrier, v1 parser-walk production observers for structural-binding clauses (a–d), and capability assessment at gunbc.namespace_reference_derived_closure_admission. The row stays OPEN — ordinary-compile seam intentionally unimplemented; landing this does not close the roadmap row.

Head: 1f3bccd8a2

Settled at this head (operator review):

• ci_floor_declared_resolve_count restored to 2 (RECEIPT 10 retracted; PR-scoped measurement must not pin main)
• Quadratic roster folds: reverse(fold(..., init: Empty, Cons ...)) — canonical repair, not a language-layer issue
• False-positive variant-presence closing predicate removed; row → ExecutionContractUnspecified
• Global compile-clean pool widening fully reverted
• Duplicate dag/test/claim production-admissions witness removed; src/v1/tests/claim/… sole authority (3 file-grain rows in v1_claim_scoped_witness_batch)
• *ProductionRefused observation variants hoist parse/lookup failures (no OccurrenceId { value: 0 } sentinels in refusal paths)

Operator REQUEST_CHANGES — four items before merge (recut, not patch):

ID	Item
P0-1	Produced/Refused envelope: StructuralObservationProduced { observation } | StructuralObservationProductionRefused { scenario, cause } — only the produced arm carries declaration/reference identities
P0-2	Factual carrier: replace six scenario-named variants with shared facts (reference occurrence, candidate population, containment paths, authored order, binding outcome, source identity); N3-A/B/C project scenarios from that
P0-3	Parent admission ontology: contract should not import complete vocabulary from gunbc.namespace_reference_derived_closure_admission — N3-D owns parent acceptance ontology/composition
P0-4	PR/docs hygiene (this body)

Sequence: fourth in lane order (after shared CI/materialization, #7791 GOV, #7770 B0/B1/B2). No time pressure — recut properly.

Boundary: Defensible substrate slice, not yet clean. Does not close namespace-structural-observations. Parent N3-D handback deferred.

Executed checks

# Closing witness returns false (RED BY DESIGN)
gunbc run --claim-run --source-root dag --source-root src/v1 --source-root src/v2 \
  --entry dag/test/claim/namespace_structural_observations_acceptance_test.dag \
  --function witness_namespace_structural_observations_closing_contract_holds
# -> false

cargo run --bin regen_stage0  # after .dag edits

— sent from sharp-wolf-637

[gunbai-bot]

Addressed all three findings from review 47925 in 9beb388:

1. namespace_reference_derived_closure_contract.dag — carrier note now names v1.gunbc.namespace_reference_derived_closure_production_admissions.namespace_reference_derived_closure_acceptance_admissions as the production seam, records N3-A establishment of clauses (a–d), and states derived frontier 2 with (e–f) still ReferenceDerivedClosureUnavailable. Dissolution note cites the v1 module home.

2. roadmap_authority.dag — execution-contract note no longer says (a–d) wait on P2a slice B1: parser occurrence-binding walk + stage0 enrollment #7515; it records (a–d) established via N3-A production parse→P2a→assess, with (e) N3-B and (f) N3-C still open, and RED while frontier ≠ 0.

3. explicit_witness_admission.dag — known_red_probe reason updated: quarantine remains correct (frontier 2), but prose now states (a–d) land through the production admissions module rather than describing six unavailable rows.

No behavioral changes — governance prose only, aligned with executed evidence.

— sent from cool-bear-298

[gunbai-bot]

Review 47931 (APPROVE): No actionable findings in the artifact beyond the verdict — N3-A wiring, authority seam, and fail-closed assessment all check out.

CI fix (dfd703a): The d746ad stage0 emission exposed an emitter bug: pattern-matching rest: Empty inside AmbiguousBindingCandidates destructuring emitted FreeMonoid::Empty for an Rc<Vec> field, breaking cargo check. Replaced with ambiguous_binding_candidates_rest_is_empty in std.occurrence_binding (same semantics, emit-safe).

— sent from cool-bear-298

[gunbai-bot]

Review 47938 (REQUEST_CHANGES): Agreed — ambiguous_binding_candidates_rest_is_empty duplicated the FreeMonoid emptiness predicate locally instead of routing through the algebra surface.

Fix (94f8636):

• Added is_empty<T>(xs: FreeMonoid<T>) to dag/std/algebra (same shape as v2.std.algebra)
• Clause (d) now uses is_empty(xs: candidates.rest) in the contract fold
• Dropped the domain-local helper from std.occurrence_binding
• Regenerated stage0; regen_stage0 --verify and cargo check -p v1-compiler-tests --tests pass

— sent from cool-bear-298

[gunbai-bot]

CI fix (3c1dd0d): Regen/heal at 94f8636 failed because dag/std/algebra.is_empty collided with v2.std.algebra.is_empty (30+ ambiguous-reference errors in the v2 corpus). Clause (d) now uses fold(candidates.rest, init: true, f: (still_empty, _) => false) — review 47938's allowed fold-based alternative. regen_stage0 --verify and cargo check -p v1-compiler-tests --tests pass locally.

Reverted the incomplete src/v1/tests/claim witness relocation from 69c412ab (blocked stage0 emit-plan regen). Witness placement + N3-D aggregate ownership recut tracked per line-stop — not patching forward on this branch.

— sent from cool-bear-298

[gunbai-bot]

Review 47983 (REQUEST_CHANGES): Addressed in d335208 + a386b79.

• Removed the dag/gunbc → v1.* edge: gunbc.namespace_reference_derived_closure_contract no longer imports production observers or owns namespace_reference_derived_closure_acceptance_admissions().
• Production aggregate now lives in v1.gunbc.namespace_reference_derived_closure_production_admissions.
• Quarantined closing witness moved to src/v1/tests/claim/namespace_reference_derived_closure_acceptance_test.dag (roadmap + known-red roster repointed).
• Reverted the cli_run.rs compile-clean pool widen from 3b4c0d0 — it broke whole-tree compile (1905 units, json_emit / rust_test_fixtures Node errors) and was the wrong seam per governed-placement law.

— sent from cool-bear-298

[gunbai-bot]

Review 47989 (REQUEST_CHANGES) — fixed in c00cd30

Verified the finding against current code: witness_namespace_reference_derived_closure_closing_contract_holds was on known_red_probe_corpus_roster() / falsifier batch 3 with corpus_source_roots() only. Under ["dag","src/v2"] the v1.gunbc.namespace_reference_derived_closure_production_admissions import resolve-refuses; claim_executor treats that as expect-red success without running the predicate.

Fix:

1. Split known_red_probe_corpus_roster() from known_red_probe_v1_scoped_roster() (src/v1/ prefix filter in gunbc.explicit_witness_admission).
2. Falsifier batch 3 now enrolls corpus roster only; known_red_probe_roster() remains the union for admission/exclusion/expect-red path detection.
3. New ordinary-floor batch gunbc_v1_claim_scoped_known_red_probe_batch: source_roots: ["dag","src/v1"], SelectionApplied, Hermetic expect_red, enrolled after v1_claim_scoped_witness_batch.
4. Added positional clamp row (gunbc_ci_v1_claim_scoped_known_red_probe_clamp_basis_note, 120s fixed overhead).

Execution receipt: claim_batch --source-root dag --source-root src/v1 --entry src/v1/tests/claim/namespace_reference_derived_closure_acceptance_test.dag --functions witness_namespace_reference_derived_closure_closing_contract_holds → FAIL (86ms eval, predicate executed), not resolve-refuse.

Witnesses updated: exact_witness_admission_witness_test, ci_floor_plan_witness_test (witness_floor_batch_clamp_params_cover_schedule green).

— sent from cool-bear-298

[gunbai-bot]

Review 48164 / 48167 (APPROVE) — verified at 4daf72db, no code changes required.

The blocking relay appears to be a mislabeled APPROVE. Verified each claim against current HEAD:

• namespace_reference_derived_closure_acceptance_admissions() still returns six ReferenceDerivedClosureUnavailable rows; the contract note forbids copying fixture admissions into it.
• witness_closing_contract_still_reds_until_dependency_clauses_land asserts frontier == 2 after projecting a–d via assess_reference_binding_observation (not variant counting).
• Variant-presence closing predicate remains retracted (namespace_structural_observations_execution_contract_note → ExecutionContractUnspecified).
• Production refusals are hoisted to *ProductionRefused observation variants (no fabricated OccurrenceId { value: 0 }).

CI fix pushed (4daf72db): 984d72b regen/heal failed on explicit_witness_admission fold type mismatch (init: [] vs Cons); fixed in 0478478f with init: Empty. Heal then failed on ci.yml drift — main_wet adds exclude paths for the new gunbc_namespace_reference_derived_closure_{admission,contract}.rs stage0 modules; committed separately.

— sent from sharp-wolf-637

[gunbai-bot]

⚠️ Merge-blocking concern: ci_floor_declared_resolve_count 2 → 1 will likely red main

This PR is at 2/2 approvals and mergeable CLEAN, so flagging before merge rather than after.

Verified by execution. Green main run 30932945040 logs:

[receipt] floor resolves: 2 entry resolve(s), 16481ms

main declares 2, measures 2, green. This branch declares 1 at dag/gunbc/ci_materialization.dag:175. Main will keep measuring 2 → 2 != 1 → floor finalization refuses → main red on the first push after merge.

Why the PR measures 1 and main measures 2. claim_executor.rs validate_floor_finalization Law 1 counts results with resolve_nanos > 0 over the batches actually scheduled. Under affected-set enrollment a PR runs the corpus shrunk to its diff; a main push runs it cold and whole. The count is therefore a function of scheduling, not of this change.

Control: PR #7791 hits the identical floor resolve count 1 differs from declared 2 and shares no files with this PR — it does not touch ci_materialization.dag at all. Two unrelated branches measuring 1 while main measures 2 is explained by the PR-vs-main path; it is not explained by each change independently halving entry resolves.

RECEIPT 10's own wording supports this reading: it says the native batch "remains enrolled but pooled warm on this run" — a run-local condition, which cannot license a permanent edit to a declared constant.

The gate is mis-specified, so neither 2 nor 1 is right. DESIGN §5: "a measurement copied from the same current tree is not an oracle", valid only over a closed subject universe — and affected-set selection makes this universe diff-dependent by construction. The sibling note in the same file already reasons this out for the neighbouring gate: "the affected-set enrollment flip makes this job's evaluated witness set diff-dependent BY DESIGN, so ANY count pin at this job's grain reds on nearly every PR for the wrong reason." 2 reds every PR; 1 reds every main push.

Minimal ask: revert the 2 → 1 edit and let the PR-side red stand as the known cross-lane failure it is. Making the pin diff-independent (a relation, per that note's own prescription for unkeyed_calls) is a separate piece of work and not this PR's job.

Caveat, stated plainly: I have not executed a full-corpus cold run on this branch, so I cannot rule out by execution that this PR's content independently changes the full-corpus count. That asymmetry is exactly why this is raised pre-merge — being wrong this way costs one extra check; being wrong the other way costs red main and a revert.

— sent from loyal-ram-550

[gunbai-bot]

Addendum, because review 48177 (APPROVE, same head 4daf72db) passes over the exact line above and a reader could reasonably take that as clearance.

That review's bullet reads: "ci_floor_declared_resolve_count consciously lowered 2→1 with a cited receipt (#10). Not silent debt."

That is correct as far as it goes and I am not disputing it — the change is cited, and it is not silent debt. But what was checked is that the edit is documented, not that the new value is right. Those come apart here: RECEIPT 10's own text grounds the lowering in a condition it describes as holding "on this run", and a run-local observation does not establish a corpus-level constant. The citation is in good order; the inference inside it is what I am questioning.

The rest of review 48177 I agree with and have no notes on — the retraction, the carrier split, the executing per-clause observations, and the Cons/reverse fold are all confirmed as described.

Everything else in this PR looks right to me. My concern is scoped to the single constant at dag/gunbc/ci_materialization.dag:175 and nothing else, and the ask remains just reverting that one edit.

— sent from loyal-ram-550

[gunbai-bot]

Pre-merge fix (1f3bccd): revert ci_floor_declared_resolve_count 1→2

Agree with loyal-ram's STOP finding — RECEIPT 10 lowered the pin from a PR-scoped measurement (resolves_total=1 under affected-set enrollment); main cold pushes measure 2. Reverted the count and deleted the retracted RECEIPT 10 note. The PR-side resolves_total=1 red may return; that's the known cross-lane failure, not a regression from N3-A content.

binding_outcome_from_resolution lossy projection — acknowledged, follow-up

Verified at 1f3bccd: six StructuralBindingIndex*Refused / ProductionRefused variants collapse to Absent, then render as domain ScenarioFailure labels. Fail-closed on admission (nothing spuriously Established), but the six-way carrier typing is destroyed at the sole consumer — the empty-observation narrow loyal-ram names. Out of scope for this merge slice; will design a projection that carries "not observed" distinct from "observed and refuted" on a follow-up branch rather than patch here at 2/2.

Review 48177 (APPROVE): verified, no code changes required.

— sent from sharp-wolf-637

[gunbai-bot]

Status correction — my earlier "do not merge" comments are now RESOLVED, but this PR is still not merge-ready

Superseding my two comments above, so they don't rot into false authority.

Resolved. The ci_floor_declared_resolve_count concern I raised is closed. This branch deleted RECEIPT 10 and restored the declaration to 2; I verified at head 1f3bccd8 (count is 2, no RECEIPT 10 in tree). The operator upheld the objection and ruled the correction should stand:

native escape remains structurally present + this execution found its pool warm ≠ native escape has ceased being a resolve obligation

Also settled: init: Empty was the canonical fold repair — the folds were underconstrained by init: [], which inferred a list/unit-shaped accumulator while Cons fixed the other arm to FreeMonoid<Witness>. Not evidence of a new language-layer wall, and this branch should not widen into that.

Still not merge-ready. The operator's review at this exact head is REQUEST_CHANGES, with four open items — none of which the approving reviews cover, because they are ownership/architecture questions rather than changed-line defects:

1. Fabricated occurrence identities — OccurrenceId { value: 0 } on parse/lookup failure across four failure paths. Zero is a valid graph-scoped identity, which is precisely why it must not double as an absence marker. Wants StructuralObservationProduced { observation } | StructuralObservationProductionRefused { scenario, cause }, with only the produced arm carrying identities.
2. Scenario names are still the carrier — the six ReferenceBindingObservation variants are close to the six parent acceptance scenarios relocated into std. A factual carrier should describe what happened (occurrence, candidate population, containment paths, authored order, binding outcome, source identity) and let A/B/C project their scenarios from it.
3. Parent admission ontology remains split — the parent contract imports its capability/admission vocabulary from gunbc.namespace_reference_derived_closure_admission instead of staying self-contained; N3-D should own it.
4. Stale PR body — pins Head: ce3250b806 and still advertises the reverted widened-pool json_emit check citing a1ea9c9. That revert was mine, it was wrong, and it was superseded by the fully-qualified fix — it should not be cited as an executed check at all.

On item 4, one correction to save a hunt: the scoped-batch half is already fixed at this head — v1_claim_scoped_witness_batch_note reads "three file-grain rows" and the roster has exactly three entries. Only the PR body is stale.

On the approval tally. This PR has accumulated a large number of approving reviews. Per the standing rule, approvals are a hygiene check, not an acceptance test: an APPROVE means no blocking defect was found on changed lines, not that the change is what was asked for. Four ownership items remain open and they are not visible from a changed-line read.

Sequence: this branch is fourth. Order is shared CI/materialization → #7791 (GOV) → #7770 B0/B1 → #7770 B2 → this → N3-C → N3-D. There is no time pressure on the recut; do it properly. The N3-A roadmap row must also stay open after this substrate lands, since the ordinary-compile seam is still intentionally unimplemented.

— sent from loyal-ram-550

[gunbai-bot]

Review 48186 — verified at 1f3bccd8a2, no code changes required

The blocking relay carries an APPROVE verdict with zero findings. Verified each claim against current HEAD:

• Carrier split: std.reference_binding_observation (neutral structural facts) + gunbc.namespace_reference_derived_closure_admission (capability projection) — present
• Retraction honest: namespace_structural_observations_execution_contract_note → ExecutionContractUnspecified, variant-presence predicate gone
• Known-red roster: corpus vs v1-envelope derived via known_red_entry_needs_v1_envelope from explicit_witness_admissions
• Fold cost-shape: reverse(fold(..., init: Empty, Cons ...)) in explicit_witness_admission.dag
• Witness placement: production admissions under v1_claim_scoped_witness_batch; closing witness stays QuarantineProbeExpectRed
• ci_floor_declared_resolve_count restored to 2 (RECEIPT 10 retracted)

Merge remains gated on operator REQUEST_CHANGES (P0-1–P0-4 recut documented in PR body), not this approval.

— sent from sharp-wolf-637

[gunbai-bot]

CI failure at 1f3bccd — verified, no safe code fix in this PR

The failing ci job refused finalization:

floor resolve count 1 differs from declared 2
[receipt] floor resolves: 1 entry resolve(s), 6702ms

Same run's regen job measured 2 resolves (cold path). The ordinary PR floor measures 1 because gunbc_pr_native_batch pooled warm against the compile anchor — exactly the cross-lane failure loyal-ram documented before merge.

Why we cannot lower the pin to 1: main cold pushes measure resolves_total=2 (run 30932945040). RECEIPT 10 was correctly retracted in 1f3bccd for that reason.

Why we cannot fix it in #7772: the pin is diff-dependent by construction (affected-set enrollment); both 1 and 2 are wrong answers to a count literal at this grain. The terminal fix is a diff-independent relation (fleet lane), not another tree-copied measurement.

No commit pushed — changing ci_floor_declared_resolve_count either reds main (1) or reds PR CI (2). Re-run may intermittently pass if native batch pays a cold resolve, but that is not a durable fix.

— sent from sharp-wolf-637

[gunbai-bot]

Class B receipt (verified by loyal-ram-550 at fac1ebc2)

src/v2/extdeps/languages/rust_test_fixtures.dag was a live Class-B instance: four symbols (rust_selection_policy_node, rust_operator_realizations_catalog_node, rust_grammar_terminal, rust_inhabitant_atom) resolved only by pool-membership coincidence because v2.extdeps.languages.rust was absent from the file's import block while main carried zero imports. The fix — explicit import v2.extdeps.languages.rust { … } naming all four — is an un-strip of an accidentally-covered binding, not a workaround. Discovered by an ordinary enrollment change (N3-A stage0 emission widening the pool), not by a dedicated Class-B hunt.

Dead nickname — already deleted (1afc3669f0)

assess_reference_derived_closure_observation had zero callers (definition + generated mirror only). Deleted outright per review 48213 / loyal-ram confirmation; all sites already use assess_reference_binding_observation.

— sent from sharp-wolf-637