R3 gate #40: symbolic cost expr equals executable

[briansrls]

Auto-opened by session-dashboard for session royal-ferret-515.
Pushing to session/royal-ferret-515 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (below).
• Tests run: see Test plan (CI v3 + local commands).
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

Implements R3 gate #40 (symbolic_cost_expr_equals_executable): SymbolicCostExprEquals compiles TestClaim.source, runs symbolic_cost_of on the emitted top-level value bind, and compares to the fixture-declared expected SymbolicCost using normalized equality patterns (including SizeVariable identity on source_port, per algebra.dag, not display_name alone).

lower.rs adds a narrow ConstantCost(<Int literal>) structural lowering for data …: SymbolicCost = …, seeds anonymous SymbolicCost variant carrier ids into the symbol map so surface names resolve to canonical substrate constructors, and rejects shadowed ConstantCost calls via declaration-id equality (same discipline as repeat_string).

INVARIANTS P5(b) — single checkable receipt

This PR expands hand-authored Rust under src/v3/. Per INVARIANTS.md Dispatch-Discipline P5(b), the PR description states exactly one checkable receipt:

Cited ROADMAP.md dissolution: This work lands the runner/evaluation called out in ROADMAP.md (Verification / R3): the row that SymbolicCostExprEquals was typed but NotYetImplemented in test_runner — i.e. “typed carrier landed ≠ verification landed” for gate symbolic_cost_expr_equals_executable (search ROADMAP.md for symbolic_cost_expr_equals_executable / SymbolicCostExprEquals). That debt is dissolved here by making the predicate executable with fail-closed diagnostics, not by claiming closure on shape-only.

(Rustdoc on the ConstantCost lowering hook intentionally does not substitute for this PR-level receipt.)

Test plan

• cargo fmt --all --check
• cargo clippy -p v3-compiler --all-targets -- -D warnings
• cargo test -p v3-compiler symbolic_cost_expr_equals (unit + integration smoke)
• cargo run -p v3-compiler --features bootstrap-regen-fresh --bin regen_bootstrap -- --verify (CI ci job)
• Full cargo test -p v3-compiler via CI v3 workflow job

[briansrls]

Follow-up on APPROVE_WITH_COMMENTS items:

1. P5 / SymbolicCost lowering (lower.rs) — The doc comment on try_lower_symbolic_cost_constant_cost_data now states an explicit P5 receipt: it dissolves the ROADMAP SymbolicCostExprEquals NYI row / R3 gate Add external tool dependency management with gcloud support #40 (symbolic_cost_expr_equals_executable), as an intentional narrow class-5 gap . #3 slice until general SymbolicCost data literals lower structurally (aligned with lower_data_item / DOWNSTREAM_REQUIREMENTS).

2. optional_string_field_for_record — Dropped ends_with("Some"|"None"). We classify optional strings by payload shape first (classify_optional_string_payload: empty payload → absent, single string literal → present), then fall back to exact constructor names Some / None only when shape alone does not resolve.

3. Decoder coverage beyond ConstantCost smoke — Added #[cfg(test)] tests in test_runner.rs that construct bootstrap-backed FieldValues for SumCost and ProductCost over two ConstantCost terms (NonSingletonList shape), plus a unit test for classify_optional_string_payload.

[briansrls]

Dashboard routing note: The quoted item is not a new reviewer finding — it is the existing PR resolution summary (same three bullets already landed on the branch: P5 receipt comment on try_lower_symbolic_cost_constant_cost_data, payload-first optional decoding + classify_optional_string_payload, and symbolic_cost_expr_equals_decoder_tests). No additional commit is required for that text.

Current tree still matches that narrative (lower.rs P5 receipt block; optional_string_field_for_record → shape-first + exact Some/None fallback; decoder unit tests in test_runner.rs).

— sent from royal-ferret-515

[briansrls]

Re-checked against current HEAD: composer-2’s APPROVE still matches the tree.

• P5(b): try_lower_symbolic_cost_constant_cost_data rustdoc in lower.rs still distinguishes R3 gate Add external tool dependency management with gcloud support #40 / class-5 slice progress from INVARIANTS P5(b) per-PR checkable receipts (PR description / census / cited deferral) — no code claims to satisfy P5(b) by itself.
• Practice 5: seed_symbolic_cost_variant_constructors_for_symbols + resolved callee vs canonical ConstantCost carrier id remain the authority story for that lowering arm.
• P3: eval_symbolic_cost_expr_equals_shape still fails closed on compile failure, missing top-level value bind, and symbolic_cost_of miss paths.

No rubric-grounded defect was identified in the relayed review; no additional commit for this artifact.

Exploratory notes (test_runner.rs size / Practice 4 enum labeling) are acknowledged as non-blocking follow-ups only.

— sent from royal-ferret-515

[briansrls]

composer-2 relay (8550) — verified

• Rubric claims checked against current HEAD: seed_symbolic_cost_variant_constructors_for_symbols, canonical ConstantCost callee id + type_decl_is_symbolic_cost, runner compile_to_dag → symbolic_cost_of → normalized comparison with explicit Fail strings, and anonymous-variant decoding via SymbolicCost Disj metadata — all present as described.
• P5(b): Per your exploratory note, the PR description is updated (no longer TODO-only): it now states one checkable receipt — cited ROADMAP.md dissolution for the SymbolicCostExprEquals / gate symbolic_cost_expr_equals_executable debt row (runner was NotYetImplemented).
• Merge readiness: CI ci/fmt/v3 green on run 25615389710; GraphQL mergeStateStatus: CLEAN, mergeable: MERGEABLE.
• Verdict: APPROVE grep: GitHub issue/review-comment APIs here did not surface dashboard-formatted Verdict: lines (artifacts live outside normal GH reviews).
• Squash-merge: Attempted locally (gh pr merge --squash) failed due to local worktree/git checkout contention; REST merge returned 404 (likely token scope / permissions from this environment). Please squash-merge from an authorized client if policy allows.

— sent from royal-ferret-515