fix(ci): align phase1 nat_semiring rung gate to §2.1 parse-only + §2.4 SKIP

[briansrls]

Summary

Aligns scripts/v4-phase1-nat-semiring-rung-gate.sh to the §6 follow-up called
out in docs/planning/v4-ladder-rung-specs-2026-05-30.md — the spec ratified
rungs 0–2 acceptance predicates but the gate script (landed via #3953) still
used cargo check for R0-rust-parse and go build for R0-go-parse, both of
which the spec §2.1 explicitly forbids, and lacked the §2.4 SKIP cell
vocabulary.

§2.1 parse-only receipts:

• R0-rust-parse now probes rustc -Z parse-only per emitted .rs. When the
  pinned toolchain rejects -Z (stable rustc), the cell is SKIP with receipt
  phase1/nat_semiring/rung0/rust_parse_driver_unavailable per the spec's
  ship_disposition: GAP carve-out ("until Compiler Spine ratifies a stable
  parse driver").
• R0-go-parse uses gofmt -e -l (the §2.1 allowed parse-only receipt). The
  previous go build proxy is removed.

§2.4 PASS|FAIL|SKIP cell vocabulary:

• Verdict slots default SKIP (not FAIL) until a receipt-bearing command
  actually runs.
• R1-rust-typecheck executes only when R0-rust-parse=PASS; otherwise SKIP
  with blocking_receipt: upstream_blocked:R0-rust-parse. The spec forbids
  marking R1 FAIL when R0 did not execute.
• R2-rust-compile executes only when R1=PASS (rung 2 Rust ⊇ rung 1, §2.3).
• R2-go-compile executes only when R0-go-parse=PASS.
• Emit-unavailable (no Cargo.toml / no .rs / no .py / no .go) yields
  SKIP with *_emit_unavailable per §2.1, not FAIL.

Row aggregate: PASS iff every cell is PASS, otherwise FAIL. All-SKIP
rows are FAIL (there is no row-level SKIP), per §2.4 single-authority
vocabulary.

No headline-aggregate surprise is expected: §6 baseline already ratified
all-rung FAIL under STRICT=1 as expected substrate gap signaling; this PR
makes the per-cell receipts honest under the spec contract.

Test plan

• bash -n scripts/v4-phase1-nat-semiring-rung-gate.sh (syntax check) — local PASS.
• CI invokes the gate under STRICT=1; per §6 baseline the expected outcome
  is rung0/rung1/rung2 FAIL with blocking_receipt: upstream_blocked:R0-rust-parse
  on stable rustc (no -Z parse-only support) — substrate gap, not a regression.
• Spec-authority cross-check: docs/planning/v4-ladder-rung-specs-2026-05-30.md
  §2.1 (parse-only receipts), §2.4 (cell vocabulary + prerequisites), §6
  (script follow-up scope).

Authority: docs/planning/v4-ladder-rung-specs-2026-05-30.md §2.1, §2.4, §6
(Ladder/Fixture Manager, this session).

🤖 Generated with Claude Code

[briansrls]

Review metadata

• Provider / model: openai-pro / gpt-5-5-thinking
• Commit: 1f83fe51 · Trigger: manual
• Comparison: main @ 887f0f2e ... session/wise-seal-69 @ 1f83fe51
• Conversation: View conversation

---

1. Story of the diff

This PR reworks scripts/v4-phase1-nat-semiring-rung-gate.sh from a coarse “fail until proven pass” gate into a §2.4-style verdict matrix where each cell can be PASS, FAIL, or SKIP, while the row aggregate still fails unless every cell passes. It splits R0 parse receipts away from later compile/typecheck receipts: Rust R0 now tries a rustc -Z parse-only path and skips when that driver is unavailable, Go R0 uses gofmt -e before go build, and Python keeps py_compile as the shared parse/compile receipt. It also adds host-tool discovery for the new parse drivers (rustc, gofmt) and updates receipt attribution so unavailable emit artifacts or upstream prerequisites become explicit *_emit_unavailable / upstream_blocked:* receipts rather than being silently conflated with target failures. chatgpt-review-e53f1bbf-f701-45…

2. Invariant categories

1. LAYER MODEL — N/A. The diff is implementation-only shell CI/gate plumbing; it does not introduce substrate types, Dag-carried facts, reusable .dag variants, or cross-pass substrate modeling.
2. INVARIANTS.md + modeling-discipline.md — Finding.

scripts/v4-phase1-nat-semiring-rung-gate.sh:183, scripts/v4-phase1-nat-semiring-rung-gate.sh:197, and scripts/v4-phase1-nat-semiring-rung-gate.sh:308 invoke host toolchain commands directly instead of through the gate’s timeout wrapper, even though the script declares V4_PHASE1_NAT_SEMIRING_TIMEOUT_SECS as the timeout “per toolchain check” at scripts/v4-phase1-nat-semiring-rung-gate.sh:22. This violates the host-process boundary / fail-closed shape: a hung rustc -Z parse-only probe, per-file Rust parse, or gofmt -e parse can now hang the gate rather than producing a bounded PASS|FAIL|SKIP receipt. The existing run_step pattern is still used for compile/build checks, so the new parse-only paths should be brought under the same bounded execution contract.

1. CODING.md — Compliant. The script remains data-and-functions shell style rather than adding object-like structure; the new row_aggregate helper at scripts/v4-phase1-nat-semiring-rung-gate.sh:337 removes repeated row-rollup loops and makes the PASS-only row contract explicit.
2. TESTING.md — Compliant. This PR is itself a CI boundary gate change rather than product library behavior; the changed checks are expressed as executable gate predicates with concrete toolchain receipts, and the row aggregation preserves behavior-driven outputs (rung0, rung1, rung2, blocking_receipt) rather than asserting on implementation internals.
3. LOCKED DESIGN DECISIONS — Compliant. The PR does not alter locked substrate or bootstrap design authority; it aligns this interim host script with the stated Phase 1 rung-gate authority and keeps the script as a bounded transport rather than turning it into a new modeling authority.
4. TRACKED vs UNTRACKED DEBT — Compliant. The Rust parse-driver gap is explicitly surfaced as SKIP with phase1/nat_semiring/rung0/rust_parse_driver_unavailable at scripts/v4-phase1-nat-semiring-rung-gate.sh:189-190, and the row aggregate still fails when any cell is not PASS, so the gap is visible rather than accepted as success.

2.5. Top-down PM intent review

Compliant. The highest-level intent here is to make the Phase 1 nat_semiring gate report the true acceptance state without overclaiming success: parse-only cells must be separated from compile/typecheck, SKIP must remain visible, and row success must require all cells to pass. The diff preserves that intent by making all verdict slots default to SKIP, assigning FAIL only when a predicate actually rejects, emitting upstream-blocked receipts, and aggregating every row as PASS only when all constituent cells are PASS at scripts/v4-phase1-nat-semiring-rung-gate.sh:336-347. It does not dilute the Pure Bootstrap direction into permanent hand-written authority; it remains an interim host gate over emitted artifacts.

3. Verdict

REQUEST_CHANGES. The semantic shape is directionally correct, but the newly introduced parse-only host-process invocations bypass the declared per-toolchain timeout contract. Fixing those three calls to run through the same bounded run_step/timeout path should make the PR approvable.