fix(r3): gate #28 receipt includes Shape A emit_rust on shared Dag

[briansrls]

Summary

Resolves codex BLOCKING feedback on merged #2412: gate #28 omni_layers_share_one_node_tree must match r3-structure.md acceptance — every emitted layer (Shape A backend + Shape B artifacts) deriving from one compile_to_dag result.

Changes

• omni_layers_share_one_node_tree: after projections, call emit_rust(&dag) on the same Dag as routes/OpenAPI/Markdown; compile_count still 1.
• Fixture: minimal top-level let omni_emit_anchor: Int = 0 so emit_rust is defined for this module (emit_rust requires a top-level value bind).
• docs/r3-program-plan.md §1.8 row Fix Value serialization and implement is_empty() method #28 Notes + docs/thesis/r2-r3-thesis-mapping.md coherence row: spell out emit_rust alongside route extraction + Shape B projections.

Verification

cargo test -p v3-compiler m1_5_omni_shape_b_openapi_test

Made with Cursor

[briansrls]

Reply — codex review (dashboard stdout artifact)

Verification against HEAD:

• emit_rust: Re-export of emit_rust_text (src/v3/compiler/src/emit_rust.rs) — emit_rust(&dag) is &Dag -> Result<String, EmitError> on the same compiled graph as the Shape B helpers.
• Program-mode requirement: emit_rust_with_mode rejects empty top-level value binds in EmitRustMode::Program (rust_target.rs 2969–2973 — exact error string matches UnsupportedBehavior("emit_rust requires at least one top-level value Bind")).
• Fixture change: let omni_emit_anchor: Int = 0 satisfies that precondition without altering service-operation data rows used by route/OpenAPI/Markdown tests.

Verdict: APPROVE — codex assessment matches the implementation; no further code change from this thread.

Merge readiness: Dashboard merge policy asks for ≥2 distinct api-review APPROVE threads (Verdict: APPROVE grep). Codex satisfies one provider; second provider (e.g. openai-pro) still needed before squash-merge per operator policy. Confirm fmt / ci / v3 green + mergeStateStatus: CLEAN before gh pr merge.

— sent from wise-cat-637

[briansrls]

Reply — codex follow-up review (sha 454a41a44, dashboard artifact /api/reviews/8467/...)

Verification: Re-checked HEAD 454a41a44 — unchanged from the prior codex pass: omni_layers_share_one_node_tree still calls emit_rust(&dag) after routes/OpenAPI/Markdown on the same Dag with compile_count == 1; omni_emit_anchor remains the minimal top-level bind for program-mode emit_rust. docs/r3-program-plan.md §1.8 #28 Notes and docs/thesis/r2-r3-thesis-mapping.md coherence row still describe emit_rust + interim route extraction + Shape B projections.

Verdict: APPROVE — second codex pass is consistent with the first; no code/doc amend required from this artifact alone.

Merge readiness: This adds a second inline Verdict: APPROVE line from codex only — merge policy calls for ≥2 distinct providers (openai-pro + codex, etc.). v3 CI was still in flight on 25613778112 last check; mergeStateStatus BLOCKED until checks/reviews clear. No squash-merge from this session until openai-pro (or another distinct api-review) APPROVE + v3 green + CLEAN.

— sent from wise-cat-637

[briansrls]

Reply — cursor/composer-2 dashboard review (8489)

Verification on HEAD d0eae976: The diff still wires omni_layers_share_one_node_tree as emit_rust(&dag) plus extract_rest_routes / project_openapi_yaml / project_markdown_documentation on one compile_to_dag result (compile_count == 1). omni_emit_anchor remains the minimal top-level bind so program-mode emit_rust is defined on an endpoint-list fixture. Ledger/thesis lines still describe emit_rust alongside the interim route projection + Shape B artifacts.

INVARIANTS P5(b) note:** Composer flags that exactly-one-checkable-receipt-in-PR-body is not verifiable from repo diff alone — agreed; no contradiction inferred from changed lines.

Verdict: APPROVE — aligns with implementation and r3-structure.md gate #28 strengthening path.

— sent from wise-cat-637

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-pro
• Commit: d0eae976 · Trigger: manual
• Comparison: main @ 390d7adb ... fix/r3-gate-28-shape-a-receipt @ d0eae976
• Conversation: View conversation

---

1. Story of the diff

This PR tightens gate #28’s receipt from “Shape B projections share one lowered Dag” to “Shape A Rust emission and Shape B projections share the same lowered Dag.” The docs now describe omni_layers_share_one_node_tree as covering emit_rust, route extraction, OpenAPI YAML, and Markdown projections from one compile result (docs/r3-program-plan.md:252, docs/thesis/r2-r3-thesis-mapping.md:129). The test fixture is made Rust-emittable by importing Int and adding a tiny top-level let omni_emit_anchor: Int = 0, then the existing omni_layers_share_one_node_tree test calls emit_rust(&dag) alongside the Shape B projection calls (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:21, :25, :352). The load-bearing mechanism is not a new emitter path; it is a receipt that every layer is consuming the same already-compiled Dag.

2. Invariant categories

1. LAYER MODEL (substrate vs implementation).

N/A — the diff is docs plus an integration-test receipt; it does not change Dag substrate types, cross-pass substrate fields, or dag.rs. The added Shape A consumer is a read of the existing Dag via emit_rust(&dag), not a new representation (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:352).

1. INVARIANTS.md + modeling-discipline.md.

Compliant — Boundary Discipline / single authority is handled correctly: the test extends the receipt by adding another consumer of the same dag value, not a second compilation or parallel fact source (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:352, :356-358). This is “shared read of one resolved fact,” not duplicate authority.

1. CODING.md.

Compliant — the new dependency is explicit and functional: emit_rust is imported as a free function and called with &dag, matching the data + free-functions style rather than hiding state or introducing a method/object wrapper (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:13, :352).

1. TESTING.md.

Compliant — this is an appropriate integration-level regression receipt because the claim is cross-layer: one compiled fixture feeds Shape A emission and multiple Shape B projections. The assertion message now names the full behavioral contract being protected: one compile result for routes, OpenAPI, Markdown, and emit_rust (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:356-358). I do not see a missing lower-level unit test obligation here; the point is the shared pipeline artifact.

1. LOCKED DESIGN DECISIONS.

Compliant — the diff preserves the Shape A / Shape B split instead of collapsing it. The changed docs and test commentary explicitly call Rust emission “Shape A” and OpenAPI/Markdown “Shape B” (docs/r3-program-plan.md:252, src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:4-8).

1. TRACKED vs UNTRACKED DEBT.

Compliant — no new scaffold, TODO, temporary API, or bridge is introduced. The only “interim” language remains the pre-existing cross-target predicate caveat, and this patch does not expand it beyond documenting the added emit_rust consumer (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:7-8).

2.5. Top-down PM intent review

Compliant. The highest-level omni-emission intent is that multiple layers derive coherently from the same modeled structure, with Shape A compiler targets and Shape B user-artifact projections remaining distinct. This patch strengthens the receipt in exactly that direction: the plan row now says emit_rust, route extraction, OpenAPI YAML, and Markdown all share the same Dag (docs/r3-program-plan.md:252), and the test actually calls emit_rust(&dag) after the Shape B projections (src/v3/compiler/tests/integration/m1_5_omni_shape_b_openapi_test.rs:352). I do not see semantic dilution: no must-have target becomes optional, no dissolution goal becomes permanent scaffolding, and no new hand-written implementation path is added.

3. Verdict

APPROVE. The PR is a narrow receipt-alignment change: it updates the docs and test so gate #28 covers Shape A emit_rust plus Shape B projections over one shared Dag. I found no substrate issue, no duplicate-authority problem, and no untracked debt introduced by the diff.

[briansrls]

Reply — openai-pro (sha d0eae976) + dashboard stdout artifact

Substantive [api-review] thread: Re-verified on merged main — §1.8 gate #28 Notes still list Shape A emit_rust + routes + Shape B projections on one Dag; omni_layers_share_one_node_tree still runs emit_rust(&dag) with compile_count == 1 and omni_emit_anchor for program-mode emit_rust. The layered-model / single-authority / Shape A vs Shape B framing in the review matches the tree.

Verdict: APPROVE — findings hold post-merge; no fix commit.

Stdout-only artifact (8488): The routed excerpt stops at harness bootstrap (diff upload / prompt fill / typing) — it does not include a separate findings block beyond the full [api-review] comment above.

Merge readiness: N/A — #2447 already squash-merged (2026-05-09T23:14:12Z). Nothing left to gh pr merge.

— sent from wise-cat-637