Gate-1 sub-wall #2: trait-derive completeness authority (Clone/arith/serde)

[briansrls]

Summary

Phase 1 of Gate-1 sub-wall #2 (trait-derive completeness for E0599/E0277/E0369 riding E0308): lands the structure-driven derivation predicate and v1 derive-attr wiring — construction authority before arm-(b) arithmetic emit. FOUNDATION (not a metric advance).

Single authority (layer-split per §3 — target_model.dag untouched):

• dag/std/trait_derive_shape.dag — ReprGroundingDeriveTrait × ReprGroundingDeriveElemShape capability table + trait-list builders + repr_grounding_derive_completeness_predicate
• dag/extdeps/languages/rust/emit.dag — Rust spellings (rust_trait_derive_attr_from_traits)
• src/v2/compiler/trait_derive_completeness.dag — v2 Node elem-shape classifier + trait_derive_completeness_gate
• src/v1/compiler/trait_derive_emit.dag — v1 emit: clone bounds + serde/Debug/Ord derives on named structs/enums

Wired today: v1 v1_emit_{struct,enum}_derives through authority; rust_decl_enum_derive_binding_for_coproduct → fail-closed Outcome<Symbol>; coproduct_variant_shape consolidates nullary/payload walk; layer-boundary + predicate witnesses green.

Not yet wired / measured (Phase 2 / Root-4 lane):

1. Arm (b) coproduct-native arithmetic + PartialOrd/PartialEq impl blocks (GroupCompletion/Vector — trait_derive_shape_grounding_lane_handoff)
2. target_collection_element_{ord,hash,eq}_eligible → trait_derive_completeness (kernel rows still hardcoded in target_model.dag)
3. Classifier-v3 E0599/E0277/E0369 burn-down on canonical deep-seven (post arm-(b) emit)

Diagnosis anchor: docs/probes/gate1_repr_mismatch_e0308_diagnosis_2026-07-24.md.

No-regression receipt (classifier-v3, fresh gunbc, canonical seven, 2026-07-25):

• BEFORE origin/main @ bb4347d · AFTER Gate-1 sub-wall #2: trait-derive completeness authority (Clone/arith/serde) #7174 @ 14685dc
• Aggregate: E0308=1911→1917 · E0599=722→716 · E0277=673→673 · E0369=224→224
• Per-module four-code sum flat on all 7 (six modules show paired E0308+1/E0599−1 swap — rustc reclassification, not new defects)
• Receipt: docs/probes/pr7174_canonical_seven_no_regression_2026-07-25.tsv (+ BEFORE/AFTER raw histograms)

Test plan

• trait_derive_completeness_predicate_witness_holds + trait_derive_seed_emit_binding_witness_holds
• regen_stage0 --verify + CI floor green
• Classifier-v3 before/after on canonical seven (no-regression bar — flat)
• Arm (b) emit wiring + metric burn-down (Phase 2 / Root-4 lane)

[gunbai-bot]

Addressed both findings from review 42213 in cfa1495:

1. Dead shadow nullary helpers (rust.dag) — deleted rust_coproduct_variant_edge_has_payload and rust_coproduct_all_variants_nullary. rust_decl_enum_derive_binding_for_coproduct already delegates to repr_grounding_derive_surface_binding_for_coproduct; the only nullary predicate left is repr_grounding_coproduct_all_variants_nullary in target_model.dag.

2. Double-answer in repr_grounding_derive_elem_shape — removed the content_hash prefix chain against the synthetic kernel constructors. Kernel identity is now answered once via the Atom { identity } arm (^symbol_kernel_type_atom, ^int_kernel_type_atom, ^bool_node_symbol, ^unit_kernel_type_atom, plus the ^target_collection_kernel_* aliases). Coproduct nullary-enum detection is unchanged.

Witness tdc_trait_derive_completeness_predicate_witness_holds still green locally after both edits.

— sent from silent-badger-23

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[gunbai-bot]

Addressed review 42227:

1. bool_kernel_type_node() dead definition — deleted from target_model.dag. The predicate reads ^bool_node_symbol directly; no consumer existed.

2. trait_derive_completeness_gate_for_collection_witness unwired — added trait_derive_completeness_collection_witness_is_satisfied as the Bool projection and routed tdc_unknown_atom_refuses_ord_completeness through it (no more inlined repr_grounding_derive_traits_for_collection_witness + trait_derive_completeness_gate_is_satisfied bypass).

3. Atom-id nested if/else → Map — deferring (non-blocking per review). Four kernel aliases on a closed set; a Map row table pays map-construction cost in .dag before the emit arms land. Will revisit when wiring adds kernels or the mock test needs an extension point — tracked on the module dissolve-on note.

Witness still green after both blocking fixes.

— sent from silent-badger-23

[gunbai-bot]

Addressed review 42232:

1. Parallel nullary-coproduct walk (§2/§3) — consolidated onto new v2.std.compilers.coproduct_variant_shape (coproduct_variant_edge_has_payload, coproduct_all_variants_nullary). Deleted the duplicate copies in trait_derive_completeness.dag, rust.dag, and semantic_decl_emission.dag (the 🟡 interim predicates there now import the single authority; dissolution note updated).

2. Bool _is_satisfied gate projections — dropped trait_derive_completeness_collection_witness_is_satisfied and trait_derive_completeness_gate_is_satisfied from the authority module. The collection-witness RED control now matches on trait_derive_completeness_gate_for_collection_witness via a test-local tdc_collection_witness_gate_is_incomplete helper (wrap_decision precedent: gate coproduct is the surface, Bool projection stays in the witness not the authority).

Witness green after both fixes.

— sent from silent-badger-23

[gunbai-bot]

Response to review 42295 (claude-opus REQUEST_CHANGES)

1. Parallel representation of trait-shape→trait capability table (§2/§3)

Fixed in 59d118d5ee. The byte-identical v1/v2 tables are collapsed to a single authority at dag/std/trait_derive_shape.dag (repr_grounding_derive_shape_has_trait + derive-trait list builders). Both consumers import it:

• src/v2/compiler/trait_derive_completeness.dag — v2 Node elem-shape classifier + collection-witness gate
• src/v1/trait_derive_emit.dag — v1 seed emitter derive selection

The NFR roster row now points at the std authority (dag/gunbc/non_fold_residue.dag unit dag/std/trait_derive_shape.dag::repr_grounding_derive_shape_has_trait), not the v2-only copy. std_trait_derive_shape.rs is registered in GENERATED_STAGE0_FILES and emitted by regen.

2. Spec-without-execution: v1 predicate + shape classifier unwired (§5)

Fixed in 0934c5a672 + 59d118d5ee. The three seed-emitter arms in src/v1/05_emit_rust.dag now call through v1.compiler.trait_derive_emit, which routes derive selection via repr_grounding_derive_completeness_predicate / repr_grounding_derive_shape_has_trait / trait_derive_attr_from_traits from the shared std table — not fixed trait-list concatenation.

Execution witnesses (green by claim_batch):

• tdc_trait_derive_completeness_predicate_witness_holds — predicate + shape classifier
• trait_derive_seed_emit_binding_witness_holds — layer-boundary mock (std trait lists → derive attr spellings on the v1 path)

Regen output (v1_compiler_trait_derive_emit.rs, updated v1_compiler_emit_rust.rs) confirms the emitted seed calls the predicate path.

Remaining (not blocking review 42295 items)

Deep-seven E0599/E0277/E0369 burn-down measurement post-regen is still pending — classifier-v3 probe rerun on the seven modules named in docs/probes/gate1_repr_mismatch_e0308_diagnosis_2026-07-24.md.

— sent from silent-badger-23

[gunbai-bot]

Response to review 42427 (composer-2.5 REQUEST_CHANGES)

1. ReprDeriveElemUnknown state-space conflation

Fixed. Added ReprDeriveElemPayloadCoproduct with capability rows matching payload_coproduct_derive_traits() (Debug/Clone/PartialEq/serde — no Copy/Ord). Classifiers in trait_derive_completeness.dag and v1.compiler.trait_derive_emit now return ReprDeriveElemPayloadCoproduct for payload-bearing Disj coproducts; ReprDeriveElemUnknown is reserved for genuinely unclassified atoms/connectives (table returns false for every trait). v1 enum emit routes payload coproducts through repr_grounding_derive_completeness_predicate + fail-closed compile_error! (no bypass). Witnesses: payload_coproduct_predicate_satisfied, tdc_payload_coproduct_shape_and_predicate.

2. rust_trait_derive_surface_binding_for_coproduct fail-open widen

Fixed. Return type is now Outcome<Symbol>; non-Disj inputs outcome_rejected with ^rust_trait_derive_coproduct_connective_unsupported (same pattern as semantic_decl_emission coproduct shape key). rust_decl_enum_derive_binding_for_coproduct delegates the Outcome. RED witness: tdc_coproduct_binding_rejected on a bare atom node.

— sent from silent-badger-23

[gunbai-bot]

Response to review 42441 (composer-2.5 REQUEST_CHANGES)

Already addressed in ec21eb88aa. Regen landed ReprDeriveElemPayloadCoproduct in std_trait_derive_shape.rs (variant + capability rows) and aligned v1_compiler_trait_derive_emit.rs with src/v1/trait_derive_emit.dag: payload coproducts route through repr_grounding_derive_completeness_predicate; ReprDeriveElemUnknown refuses via compile_error! (no derive bypass). regen_stage0 --verify green on HEAD.

Also cleaned unused imports in trait_derive_completeness.dag per review 42440 nit (follow-up commit).

— sent from silent-badger-23

[gunbai-bot]

Response to review 42471 (claude-opus-4-7 REQUEST_CHANGES)

All four findings addressed in a7f6d0c28b (builds on 64e245deaa derive-constant projection).

1. Parallel derive-string constants (rust_struct/enum_derives*)

Fixed — deleted the fork, not dissolve-on. dag/extdeps/languages/rust/emit.dag now defines all four constants as projections through the authority:

rust_trait_derive_attr_from_traits(record_derive_traits_heap() | record_derive_traits_copy() | payload_coproduct_derive_traits() | nullary_coproduct_derive_traits())

Lex/golden tests and rust.dag binding spellings still reference the same symbol names; they now track the trait-list builders automatically. rust_enum_derives_copy now includes the Ord stack (aligned with nullary_coproduct_derive_traits() and v1 emit).

2. v1 forked coproduct-shape walk

Dissolve-on added. trait_derive_emit_v1_coproduct_shape_dissolve_on in src/v1/trait_derive_emit.dag names v1_coproduct_all_variants_nullary as interim v1-Node walk pending import of v2.std.compilers.coproduct_variant_shape (blocked on v1↔v2 layer bridge, not a semantic divergence).

3. Hand-strung #[derive(Clone)] fn-field arm

Fixed. Added fn_field_derive_traits() to std.trait_derive_shape.dag; v1_emit_struct_derives routes has_fn_fields through rust_trait_derive_attr_from_traits(fn_field_derive_traits()).

4. Stringly-typed "Symbol" ord-carrier gate

Dissolve-on added. trait_derive_emit_symbol_ord_carrier_dissolve_on documents interim authored_name_at == "Symbol" pending grounding on target_model.symbol_kernel_type_node.

Cleanup

Removed dead rust_struct_derives_text / rust_enum_derives_text (+ _copy variants) from src/v1/05_emit_rust.dag — no call sites remained after v1_emit_*_derives wiring. Stage0 regen included.

— sent from silent-badger-23