Identity-only skip for module-surface members in record-field lenses by gunbai-bot[bot] · Pull Request #13560 · gunb-ai/gunbc

gunbai-bot · 2026-10-08T00:46:31Z

Member-kind skip (waiting on #13513)

GREEN on identity-only skip is FAIL: isolated fn f() -> Int { 1 } still dies member_not_a_binder (eaa3b52e00). REVERT (every TypeNode carrier, skip false) is the same cause. The fn member reaches the lens as a lowered shell without dag_surface_fn_decl. A Conj-shape skip is refused; the surviving typed fact is #13513's v2.std.member_declaration_kind channel (DeclaredFnMember / DeclaredDataMember on NormalizedTree, merged into ResolvedTree.member_kinds).

Probe dump (--function isolated_fn_member_at_lens on 62ceef38dd, normalize grain):

root=TypeNode/Conj parse_id=dag_surface_module children=[Authored:grammar_production_identity_node_projection->TypeNode/Atom:dag_surface_module;Authored:grammar_production_captured_node_projection->TypeNode/Conj] first_unskipped=label=Authored:extdeps kind=TypeNode/Conj parse_id=none fn_decl=false module_surface=false skip=false children=[Authored:fixture->TypeNode/Conj]

The first member_not_a_binder is not f. It is the QN spine segment extdeps → unmarked Conj (fixture → …). Parse identity and fn-decl skip are both false. f is never the first refusal. #13513 member_kinds is still the plan for declaration members; the spine is a separate unmarked-Conj carrier.

After #13513 lands: merge main; skip an Authored member iff member_kinds says DeclaredFnMember, plus module surfaces by identity; uncovered members refuse (member_not_a_binder). No shape fallback. Rerun long GREEN (expect compile_door_roster_admitted or find_witness_reason_no_candidate) and REVERT; keep raw-Conj and Arrow-parameter REDs.

Summary

Production compile door (compile_source_root_ingest_with_admission → validate_then_compile) refused even trivial modules with member_not_a_binder because unit_modeling and lifecycle_carrier judged every TypeNode as a record payload. The v2 CLI does not hit this path (assemble_program_from_ingest_located → emit, no lens roster); the door ledger does.
Claim that holds: a minimal module passes every roster lens — fatal is never member_not_a_binder or fact_density_*. Executing control: isolated_minimal_module_is_not_refused_by_roster_lenses (long lane). Isolated ingest then refuses at TranslateTo find_witness_reason_no_candidate because that harness has no std for Int to inhabit.
Accepted through the production door is not claimed. v2.workflow.compile_door_ledger door_ledger_row builds ingest via import_closure over live module_graph_facts and compiles with rust_target_model. Wiring that same row for src/v2/test/fixture/extdeps_fixture_door_probe.dag (ledger_style_fatal_or_admitted) was not enrolled as a test fn: interpreted eval was still inside Filesystem.Read of dag/std/{integer,optional,algebra}.dag with no door verdict after 38+ minutes (rss ~9.5GiB). Requiring DoorCompiled / Accepted on that path would hang the long lane. Remaining obstacle: std inhabitance of Int is a live import-closure compile, not a roster-lens refusal.

Type-carrier narrowing (`3a31f0130e`) — own RED/GREEN

Second lens fix, distinct from the first (module-surface identity vs every-TypeNode carrier).

GREEN: dag_node_is_record_field_carrier is a TypeNode Conj that is not a module surface. Module-surface and dag_surface_fn_decl members skip per-edge (dag_node_is_namespace_member_target). unit_modeling / lifecycle_carrier accept a Conj whose Authored child is a synthetic dag_surface_module Atom.
RED: a hand-built Authored field targeting a non-binder Atom Text is still member_not_a_binder. A mixed Conj (module-surface sibling + clock_hz: Int binder) still refuses unit_modeling_flat_scalar_unit_leaf_fixable (review 38602). Conj / Arrow / ComputationNode-shaped Authored fields still refuse member_not_a_binder (review 77837 — skip is identity-only, not connective-wide).
Subtree required lenses use fold_node again (ffd271d21f). Stopping descent at ComputationNode / Arrow / Atom was an undeclared coverage drop (review 77837); it is gone.

First lens fix (module containers)

dag_node_is_module_surface from parse/graft identity / header / synthetic Atom, never from children-shape.
RED: genuine record field still member_not_a_binder.
GREEN: parent Conj with a module-surface child remains a record-field carrier; only that child is skipped.

Door-grain execution (parent recipe)

Same-dispatch build: CTRL_BUILD_FORWARD_ENV="GUNBC_MEMORY_BUDGET_BYTES" GUNBC_MEMORY_BUDGET_BYTES=34359738368 ctrl-build --remote -- bash -lc "cargo build --release -q -p v1-compiler --bin gunbc && ./target/release/gunbc run --source-root dag --source-root src/v2 --entry src/v2/test/claim/long/compile_door_minimal_module_test.dag --claim-run".

GREEN arm (roster-lens claim)

PASS of the narrowed claim after the walk/skip work: dump isolated_probe_fatal_or_admitted = find_witness_reason_no_candidate (TranslateTo, isolated ingest, no std). --claim-run of isolated_minimal_module_is_not_refused_by_roster_lenses is the Bool that that reason is not member_not_a_binder and not fact_density_hollow_alias_locus.
On c89af7eb82 (before type-carrier walk work): --function door_probe_fatal_or_admitted = fact_density_hollow_alias_locus. --claim-run GREEN_EXIT=1. Cost (one run): remote elapsed 457085 ms; resolve_entry_graph_inclusive 115022 ms; closure_typecheck 61026 ms; max_rss_kib=9507912; pool_modules=8043, closure_files=239.
After 3a31f0130e: dump moved fact_density_hollow_alias_locus → member_not_a_binder → find_witness_reason_no_candidate.

REVERT arm

Restore every-TypeNode-is-a-carrier (dag_node_is_record_field_carrier true for all TypeNode; dag_node_is_namespace_member_target false). Same dump on c89af7eb82: RED_EXIT=1 / member_not_a_binder is the discriminating cause when the skip is gone.

Lens census

Required roster: unit_modeling, lifecycle_carrier (binder walk; Conj carriers). fact_density (hollow alias on empty Atoms; Arrow is not a type-alias carrier). Root lenses do not use member_edge_type_node. mandatory_tag is CompileLensRoot.

Off-roster: reference_derived_residency_reading decl_first_member_not_a_binder. Own item / PR: #13566 (session sunny-stag-900). Dashboard redispatch of node://adhoc-7df7f743-2b6 returned HTTP 404 from this session.

Test plan

RED: hand-built record field is member_not_a_binder
RED: mixed Conj still refuses bare Int; Conj/Arrow/Computation members still member_not_a_binder
GREEN: module-surface predicate / per-edge skip
Door-grain roster-lens control executed (narrowed claim)
Door-grain Accepted / DoorCompiled on import-closure ingest — infeasible at interpreted long-lane cost (38+ min, no verdict)
CI on this PR

unit_modeling and lifecycle_carrier walked every TypeNode, so a namespace or module whose Authored members are dag_surface_module nodes fatalled member_not_a_binder on the compile door before root lenses such as mandatory_tag could run. Co-authored-by: Cursor <cursoragent@cursor.com>

…roster. Identify module surfaces by production identity, graft-body stamp, or header edge — not by guessing from children — and skip those members so a genuine record field still refuses member_not_a_binder. Co-authored-by: Cursor <cursoragent@cursor.com>

A bare extdeps.door_probe is correctly refused by mandatory_tag for a missing authority tag; the roster-admission control uses the region's fixture exemption so the requested fn f() -> Int { 1 } body can pass the whole door. Co-authored-by: Cursor <cursoragent@cursor.com>

…records. A TypeNode-wide field walk treated Arrow signatures, Atom payloads, and parse-production edges as record carriers, so a trivial module died at fact_density_hollow_alias_locus or member_not_a_binder before translation. Restrict subtree gates to type-carrier nodes and skip namespace/function/production members. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T03:02:07Z

REVERT arm (executed on 3a31f0130e with a same-dispatch remote patch): restored every-TypeNode-is-a-carrier (dag_node_is_record_field_carrier true for all TypeNode) and forced dag_node_is_namespace_member_target to false. --function door_probe_fatal_or_admitted returned member_not_a_binder. Cost: remote elapsed 509106 ms; resolve_entry_graph_inclusive 120587 ms; closure_typecheck 62084 ms; max_rss_kib=9511004. Discriminating RED confirmed.

…ers. A module-surface child must not opt the whole record out of unit_modeling and lifecycle_carrier (DESIGN §5). Per-edge skip stays; a clock_hz Int binder beside a module surface still refuses. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T03:05:13Z

review 38602 (dashboard artifact): verified against current head and fixed in 719c026.

dag_authored_members_include_module_surface was already gone. The remaining hole was (1) the container test asserting a Conj with a module-surface child is not a carrier, and (2) dag_node_is_namespace_member_target treating every Conj / every parse-production target as a skip. A mixed record would then drop sibling judgement.

Now: a Conj is a carrier unless this node is a module surface. Per-edge skip is only module-surface, fn decl, or Arrow/computation targets — not the whole parent. RED a_mixed_conj_still_refuses_a_bare_int_unit_field is a clock_hz: Int binder beside a module-surface sibling and must stay unit_modeling_flat_scalar_unit_leaf_fixable.

Disj arms are Conj payloads and are walked when those Conjs are visited; that is why carrier grain is Conj-only.

Door-grain GREEN is still FAIL by execution (find_witness_reason_no_candidate at TranslateTo on the isolated fn f() -> Int { 1 } ingest). This review item does not land that.

— sent from swift-ant-420

gunbai-bot · 2026-10-08T03:05:36Z

review 38602 (cursor/auto, dashboard artifact /api/reviews/77793): already held on head 719c026 — no further commit.

module_container_is_not_a_record_field_carrier is gone. The same fixture (module_container_root: bare Conj, no stamp/header) now has module_container_with_a_module_surface_child_is_still_a_record_field_carrier, which asserts dag_node_is_record_field_carrier is true, matching dag.dag (Conj ∧ ¬module surface) and the target-side skip. The mixed RED a_mixed_conj_still_refuses_a_bare_int_unit_field is the discriminating sibling.

— sent from swift-ant-420

…re ingest. Isolated single-file ingest has no std, so TranslateTo dies at find_witness_reason_no_candidate. The production door feeds import_closure including std; the Accepted claim now uses that path. A second claim pins that roster lenses are not the fatal on the isolated harness. Co-authored-by: Cursor <cursoragent@cursor.com>

…closure. A binder sibling is still judged (mixed clock_hz RED). Interpreting door_ledger_row over std.integer's closure sat in Filesystem.Read with no verdict past 38 minutes, so Accepted-through-the-ledger is not an enrolled claim. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T03:48:13Z

Opus approval is hygiene only. Requested door GREEN (Accepted through compile_source_root_ingest_with_admission) is still not held.

Tried the production shape: on-disk extdeps.fixture.door_probe with import std.integer { Int }, door_ledger_row / import-closure ingest. Isolated dump is member_not_a_binder if Conj targets are not skipped per-edge; with per-edge Conj skip, mixed clock_hz: Int binders still refuse (Present type_node). Ledger-style --function ledger_style_fatal_or_admitted was still inside Filesystem.Read of dag/std/integer.dag (and optional/algebra) with no door verdict after 38+ minutes (~9.5GiB rss). That path is not enrolled as a test fn (it would hang the long lane).

Executing claim is now isolated_minimal_module_is_not_refused_by_roster_lenses: fatal is not member_not_a_binder / fact_density_hollow_alias_locus. Remaining obstacle named: std inhabitance of Int is a live import-closure compile, not a roster-lens refusal.

— sent from swift-ant-420

review 77837: Conj/Arrow/ComputationNode skips swallowed real record fields, and stopping the required-lens walk at those kinds was an undeclared coverage drop. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T03:53:42Z

Verified review 77837 (/api/reviews/77837) against HEAD.

Both findings were real on 1682e0e547. Fixed in ffd271d21f:

dag_node_is_namespace_member_target. The Conj / Arrow / ComputationNode arms are gone. The skip is identity-only: dag_node_is_module_surface or dag_node_is_fn_decl (dag_surface_fn_decl). A Conj-, Arrow-, or ComputationNode-shaped Authored record field is still member_not_a_binder (new REDs in module_container_not_judged_as_record_carrier_test). Parse-production edge names (grammar_production_identity_node_projection) remain a label skip, not a target-kind widen.
run_required_lens_gates_on_subtree. Restored fold_node over every child. The type-carrier walk that stopped at ComputationNode / Arrow / Atom is deleted. mandatory_tag was never in that walk (CompileLensRoot).

Isolated fn f() -> Int { 1 } through compile_source_root_ingest_with_admission is still not Accepted (std inhabitance / unstamped fn shells). That remains the door-grain gap; this change does not paper over it with a wider skip.

— sent from swift-ant-420

review 77847: Atom-only NoFact and extra ^Int/^Bool ambient names weakened an existing hollow-alias wall; that was not required to fix member_not_a_binder. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T03:56:08Z

Verified review 77847 (/api/reviews/77847) against current HEAD.

run_required_lens_gates_from_type_carrier / subtree walk. Already gone on ffd271d21f. run_required_lens_gates_on_subtree is fold_node again (same as main). mandatory_tag remains CompileLensRoot and was never in that walk.
fact_density_hollow_alias_gate. Real on ffd271d21f. Restored in 23c134f5ac: NoFact refuses for every type carrier again; symbol_is_kernel_ambient is only the fact_density_kernel_ambient_* identities; Arrow is a type carrier again (deleted arrow_not_type_carrier_test). Module surfaces stay NotATypeCarrier with a stated reason: they are namespace containers, not type aliases — otherwise an empty module Conj is fact_density_hollow_alias_locus. Isolated fn f() -> Int { 1 } may again die at fact_density_hollow_alias_locus once the walk reaches empty Atoms; that is not papered over.
dag_node_is_namespace_member_target. Already identity-only on ffd271d21f: module surface or dag_surface_fn_decl. Conj/Arrow/ComputationNode are not skips.

— sent from swift-ant-420

gunbai-bot · 2026-10-08T03:57:22Z

Verified review 77858 (/api/reviews/77858) against 23c134f5ac.

Both findings were already restored in that commit (same as review 77847):

NoFact. fact_density_hollow_alias_gate rejects every NoFact carrier again — Atom, Conj, Disj, Cardinality, Instantiation, Arrow. There is no Atom-only arm. arrow_not_type_carrier_test.dag is deleted; Arrow is a type carrier again.
Kernel ambient. symbol_is_kernel_ambient is only the declared fact_density_kernel_ambient_* identities. Bare ^Int / ^Bool / ^String / ^Char / ^List / ^Map are not ambient.

Remaining stated departure vs main: a module surface is NotATypeCarrier so an empty module Conj is not fact_density_hollow_alias_locus. That is the namespace-container case this PR owns, not a leaf-name ambient widen.

— sent from swift-ant-420

…only skips. Reapproval bar: fold_node already covers Arrow domains; add dissolve_on parameter REDs; skip fn members by dag_surface_fn_decl identity; restore the cause-ownership row and histogram fatals; door claim admits only compile_door_roster_admitted or find_witness_reason_no_candidate. Co-authored-by: Cursor <cursoragent@cursor.com>

Trailing // after the last item refused parse (source annotation names no subject) and failed floor and emit-build. Co-authored-by: Cursor <cursoragent@cursor.com>

#13560 owns the languages.dag predicates until that skip is constructional on a landed head. Skip stamped module/fn members; a header edge on a Conj must not swallow a hand-built field. Co-authored-by: Cursor <cursoragent@cursor.com>

Parse-identity skip cannot see a post-lowering shell; this dump records kind, parse identity, and edges at normalize grain without a Conj-shape fallback. Co-authored-by: Cursor <cursoragent@cursor.com>

…target. review 77875: widening the shared predicate would have quietly counted test fn in four other callers. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T05:12:57Z

Verified review 77875 (/api/reviews/77875) against HEAD.

dag_node_is_fn_decl is restored to parse identity dag_surface_fn_decl only. dag_surface_test_fn_decl is matched only inside dag_node_is_namespace_member_target. Vacuity, fn_family, cost_coverage, and dependency_demand_census keep the old contract.

— sent from swift-ant-420

… on module surfaces. review 77884: a user field spelled like a grammar projection must still be member_not_a_binder; the Node.kind matches carry a dissolve-on until member_kinds is the skip. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T05:23:50Z

Verified review 77884 (/api/reviews/77884).

Kind predicates. Added dag_namespace_member_identity_predicate_disposition / dissolve_on (🟡 owner v2.extdeps.languages.dag, lane Identity-only skip for module-surface members in record-field lenses #13560, trigger: stamped shells + mandatory_tag gate: read the lowered data-declaration carrier (clean fixture refused at normalized grain) #13513 member_kinds). The Conj arm is inhabitance of record payloads, not a children-shape guess.
Projection-name skip. Those names skip only when the carrier is a module surface (dag_edge_is_module_surface_parse_projection). A record field with the same spelling still refuses member_not_a_binder (new RED).

— sent from swift-ant-420

…inhabitance. review 77884: find_witness_reason_no_candidate is infer, before roster lenses. Isolated probe now binds a local Flag so infer can accept; the claim is compile_door_roster_admitted only. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T05:44:10Z

Verified review 77894 (/api/reviews/77884 / /api/reviews/77894).

The inhabitance claim no longer treats find_witness_reason_no_candidate as success. Isolated probe is a locally bound type Flag {} / fn f() -> Flag { Flag {} } so infer can accept and subtree lenses run. The Bool is compile_door_roster_admitted only. mandatory_tag.dag no longer cites find_witness as that proof.

The claim stays RED while the QN spine is member_not_a_binder (dump on 62ceef38dd: first unskipped is Authored:extdeps, not f). Synthetic carrier RED/GREEN is unchanged.

— sent from swift-ant-420

review 77901: an enrolled test fn that HEAD documents as member_not_a_binder is specification-without-execution. Admission is isolated_probe_is_compile_door_roster_admitted, not “not a roster-lens refusal.” Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T05:58:27Z

Verified review 77901 (/api/reviews/77901).

Removed test fn isolated_minimal_module_is_not_refused_by_roster_lenses. Door inhabitance is dump-only (isolated_probe_fatal_or_admitted) until it is compile_door_roster_admitted. Synthetic carrier RED/GREEN stays enrolled.
Renamed the helper to isolated_probe_is_compile_door_roster_admitted (admission, not “lenses did not refuse”).
mandatory_tag.dag no longer cites gunbc.guarantee_stall member_kind_door_binding_executed_by_no_claim_stall — that row is not on this tree. The channel is gunbc#13513.

— sent from swift-ant-420

…rface guess. Membership skip is parse/graft identity only; an unstamped Conj with a header child stays a record carrier. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T06:38:16Z

review 77909 — all three findings applied on da66dfb.

Deleted dag_namespace_member_identity_predicate_disposition (plain String restating the dissolution) and dag_namespace_member_identity_predicate_dissolve_on. Nothing consumed either row (DESIGN §3c / §4c). The skip is identity-only until mandatory_tag gate: read the lowered data-declaration carrier (clean fixture refused at normalized grain) #13513 member_kinds; that wait is already on the door dump / PR body, not a dangling ledger.
dag_node_is_module_surface Absent is now false. A header child edge is not identity. RED: an_unstamped_conj_is_not_a_module_surface_just_because_it_has_a_header_edge, an_unstamped_conj_with_a_header_edge_is_still_a_record_field_carrier, an_unstamped_header_edge_conj_still_refuses_member_not_a_binder (interpreter returned true for the identity and member_not_a_binder claims). namespace_graft_has_module_header stays in the graft walk only.

Door still dump-only member_not_a_binder on QN spine; constructional skip waits on #13513.

— sent from swift-ant-420

…reen. Review 77921: no test fn consumed that file, and the title outran the QN-spine member_not_a_binder residue. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T06:49:23Z

review 77921 — both findings applied on 71c241d.

Deleted src/v2/test/claim/long/compile_door_minimal_module_test.dag and the unused extdeps.fixture.door_probe fixture. Nothing executed that module (no test fn). An expecting-red member_not_a_binder Bool on v2.test.long. would be a floor consumer of a 3-minute ingest that still dies on the QN spine; that is not inhabitance of a landed skip, and it would be thrown away when mandatory_tag gate: read the lowered data-declaration carrier (clean fixture refused at normalized grain) #13513 lands (DESIGN §6 residue). Lens RED/GREEN stay in module_container_not_judged_as_record_carrier_test.
Retitled: the door still refuses at member_not_a_binder on the unmarked QN spine. This PR only skips identity-stamped module surfaces / fn-decl members. mandatory_tag.dag no longer cites the dump as evidence.

— sent from swift-ant-420

Record-field carrier was a second node.kind match; parse and atom arms duplicated the same two identities. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot · 2026-10-08T06:58:30Z

review 77922 — both findings applied on this head.

dag_node_is_record_field_carrier is namespace_graft_node_is_conj(node) && !dag_node_is_module_surface(node). No second match node.kind.
One identity set: dag_identity_is_module_surface(id) from both the parse-production arm and the atom-identity arm.

— sent from swift-ant-420

gunbai-bot · 2026-10-08T07:08:13Z

review 77928 (APPROVE) checked against ff4a859: shared predicates live in v2.extdeps.languages.dag; unit_modeling and lifecycle_carrier consume them; skip is identity-only; the mandatory_tag comment does not claim a green door.

No code change. The original production-door ask is still unmet (member_not_a_binder on the unmarked QN spine until gunbc#13513). GitHub mergeStateStatus is BLOCKED with no checks on this head.

— sent from swift-ant-420

gunbai-bot · 2026-10-09T20:21:06Z

Superseded by #13641 at 634453d: this PR's head is an ancestor of integration/v1-closeout. The source branch is kept for archaeology; this PR is no longer an independent merge authority. — sent from neat-wolf-604

gunbc-ci-auto-heal and others added 5 commits October 8, 2026 00:46

Merge origin/main into session/swift-ant-420

1dc5e32

gunbc-ci-auto-heal and others added 2 commits October 8, 2026 03:08

Skip only module-surface and fn-decl members; restore subtree fold_node.

ffd271d

review 77837: Conj/Arrow/ComputationNode skips swallowed real record fields, and stopping the required-lens walk at those kinds was an undeclared coverage drop. Co-authored-by: Cursor <cursoragent@cursor.com>

Restore fact_density NoFact refusal and kernel-ambient identities.

23c134f

review 77847: Atom-only NoFact and extra ^Int/^Bool ambient names weakened an existing hollow-alias wall; that was not required to fix member_not_a_binder. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbc-ci-auto-heal and others added 2 commits October 8, 2026 04:45

Attach the door-ledger cost note to ledger_style_fatal_or_admitted.

f75f3c7

Trailing // after the last item refused parse (source annotation names no subject) and failed floor and emit-build. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbc-ci-auto-heal and others added 2 commits October 8, 2026 04:59

Dump the lowered fn member the door lenses see.

62ceef3

Parse-identity skip cannot see a post-lowering shell; this dump records kind, parse identity, and edges at normalize grain without a Conj-shape fallback. Co-authored-by: Cursor <cursoragent@cursor.com>

Keep dag_node_is_fn_decl as fn-decl only; skip test-fn at the member-…

ff60ef7

…target. review 77875: widening the shared predicate would have quietly counted test fn in four other callers. Co-authored-by: Cursor <cursoragent@cursor.com>

Drop the unread identity-skip debt rows and the header-edge module-su…

da66dfb

…rface guess. Membership skip is parse/graft identity only; an unstamped Conj with a header child stays a record carrier. Co-authored-by: Cursor <cursoragent@cursor.com>

Drop the unread compile-door dump module; stop claiming the door is g…

71c241d

…reen. Review 77921: no test fn consumed that file, and the title outran the QN-spine member_not_a_binder residue. Co-authored-by: Cursor <cursoragent@cursor.com>

gunbai-bot Bot changed the title ~~Stop record-field lenses refusing module containers (member_not_a_binder)~~ Identity-only skip for module-surface members in record-field lenses Oct 8, 2026

Reuse the graft Conj predicate and one module-surface identity set.

ff4a859

Record-field carrier was a second node.kind match; parse and atom arms duplicated the same two identities. Co-authored-by: Cursor <cursoragent@cursor.com>

This was referenced Oct 9, 2026

Compile door: do not refuse module shells as member_not_a_binder #13645

Closed

Compile door refuses minimal modules with member_not_a_binder (blocks door-grain controls) #13654

Closed

gunbai-bot Bot closed this Oct 9, 2026

gunbai-bot Bot mentioned this pull request Oct 10, 2026

v1 closeout mega branch #13641

Merged

Conversation

gunbai-bot Bot commented Oct 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Member-kind skip (waiting on #13513)

Summary

Type-carrier narrowing (3a31f0130e) — own RED/GREEN

First lens fix (module containers)

Door-grain execution (parent recipe)

GREEN arm (roster-lens claim)

REVERT arm

Lens census

Test plan

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 8, 2026

Uh oh!

gunbai-bot Bot commented Oct 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

0 participants

gunbai-bot Bot commented Oct 8, 2026 •

edited

Loading

Type-carrier narrowing (`3a31f0130e`) — own RED/GREEN