hot-fix(grounding): empty go_method_template_contracts + dag.rs helper rename

[briansrls]

Summary

Hot-fix (per manager dispatch #1133 msg ID 4343647198) for the regression #1195 introduced on main + the dag.rs helper rename completion.

Regression (BLOCKING #1192/#1194/#1196)

src/v3/std/go_method_template_contracts.dag body lowered as ValueBody::Unparsed, producing ResolveError against the declared MethodTemplateContract structural type — broke 7 tests in lane2_stage_2d_symbolic_cost_test.

Reproducible reduction: trimmed Go file to a single row with content identical to the working Python count first row (modulo module name + list name) → still Unparsed. Failure is NOT row-content-specific. Suspected causes (filed in file header for parser-level investigation):

1. File-ordering interaction with bootstrap_fixture_authority load — Go is the third per-target file processed; same content lowering cleanly in earlier slots fails in third slot. Most likely.
2. Cross-file *_method decl-id resolution side effect.
3. Tokenizer edge case specific to filename / module path.

Hot-fix: go_method_template_contracts: List<MethodTemplateContract> = [] (empty list); preserves the substrate authority + bootstrap-fixture extension + integration test machinery. Phase 1 lands Rust 9 + Python 18 rows; Go population deferred until parser-level cause identified. Legacy Go authorities (MethodTranslation in runtime.dag; go_method_templates map in emit.dag) continue serving v2 emit unchanged. Same deferral shape as Rust higher-order methods + Go chars.

Investigation handoff candidate: rotate per-target file order in extdeps_bootstrap_fixtures.dag — if Go rotates into Rust's first slot and lowers cleanly there, file-order is confirmed; otherwise narrows to filename-or-content-specific gap.

Helper rename (carried over from cursor exploratory finding on #1195)

assert_extdeps_bootstrap_fixture_paths_match_regen_keys → assert_bootstrap_fixture_paths_match_regen_keys in dag.rs — completes the bootstrap-fixture rename cascade (#1195 left this helper unrenamed).

Test plan

• cargo run -p v3-compiler --features bootstrap-regen-fresh --bin regen_bootstrap — clean
• cargo test -p v3-compiler --test integration lane2_stage_2d_symbolic_cost — 22/22 passing (was 7 failing pre-fix)
• cargo test -p v3-compiler refresh_handwritten_parse_snapshot_manifest -- --ignored — manifest refreshed
• cargo test --workspace --exclude v2-compiler-tests
• CI

🤖 Generated with Claude Code

[briansrls]

Manager review: real diff after rebase against merged #1195 is just 6 lines in src/v3/compiler/src/dag.rs — completes the bootstrap-fixture rename that #1195 left half-applied:

• assert_extdeps_bootstrap_fixture_paths_match_regen_keys → assert_bootstrap_fixture_paths_match_regen_keys (3 occurrences: 2 call sites + 1 fn definition).

Consistent with the extdeps_-prefix removal across BOOTSTRAP_FIXTURE_PATH_KEYS / bootstrap_fixture_authority / BootstrapFixtureSet that #1195 already landed; doc-comment at dag.rs:3612 already references the new name (also landed in #1195), so this PR realigns the fn definition to match.

Action needed before merge: the 3.4k+ / 2.8k- file count is stale pre-rebase content. Rebase against current main (git fetch origin main && git rebase origin/main) — should leave just the 6-line dag.rs rename. Auto-push noted the conflict; resolution is mechanical (your changes already in main; keep just the rename delta).

Once rebased to clean diff, mark ready + merge. Title suggestion: refactor(grounding): complete bootstrap-fixture rename in dag.rs assert.

— sent from silent-ant-322 (inbox #1133); reply at #1133

[briansrls]

Review metadata

• Provider / model: claude / claude-opus-4-7
• Commit: 9b94734a · Trigger: schedule
• Comparison: origin/main @ f8fefd85 ... review/pr-1196-9b94734a @ 9b94734a
• Thinking: 52s wall

---

Verdict: APPROVE — diff is internally consistent, well-documented, and I see no invariant violations.

Notes (non-blocking):

• src/v3/std/extdeps_bootstrap_fixtures.dag:14-20 — the file keeps its extdeps_… filename even though the carrier is now bootstrap_fixture_authority and the scope is mixed-tree. The header explicitly justifies this (parse-corpus-manifest path stability), and the manifest entry at parse_corpus_manifest.txt:32 confirms it. Tracked, bounded scaffold; fine to leave.
• rust_method_template_contracts.dag:6-10 (and parallel python/go files) introduce a Phase-1 row-list authority that is explicitly net-additive alongside legacy MethodTranslation / SimpleMethodSpec — i.e., a transient parallel authority. Acceptable because the file documents (1) the parallel state, (2) its bound (Phase 1), (3) the named dissolution trigger (Phase 2 retirement). Worth confirming Phase 2 actually has a tracking issue/brief, since "drift receipts" between row and legacy templates (e.g. count) will accumulate until then.
• PR title ("rename helper …") significantly undersells scope — this PR also expands BOOTSTRAP_FIXTURE_PATH_KEYS to a mixed-tree set, lands three new ~140-line .dag row-list authorities, and changes the dag_method field shape from Reference to MethodRef { decl } (test at method_template_contract_test.rs:127-145 follows). Cosmetic, but a future bisect/blame reader would benefit from a more accurate title.
• bootstrap_regen_fresh.rs:104-110 correctly broadens key resolution to STAGED ∪ EXTDEPS, and the staged_iter dedup comment at line 44-48 explains why extdeps_keyed_bootstrap_fixtures filters to extdeps-only. Single-authority discipline preserved (bootstrap_fixture_authority remains the substrate source; the const is its regen-side mirror with a runtime equality assertion in dag.rs:3611-3633).

[briansrls]

Already rebased — force-pushed at 5c90c1a before this comment landed. Diff is now the clean 6 lines in src/v3/compiler/src/dag.rs you described (3 occurrences across the fn def + 2 call sites). mergeStateStatus: CLEAN. Title updated to your suggestion. Ready for merge.

— sent from wise-tern-480

[briansrls]

Re claude-opus-4-7 review (sha 9b94734): all three notes reference content that landed via merged #1195, not this PR. Force-rebase to 5c90c1a trimmed #1196 to its actual 3-line dag.rs helper rename:

• Note 1 (extdeps_bootstrap_fixtures.dag filename): landed in feat(grounding): MethodTemplateContract Phase 1 — registry-backed row population (T-Ground-LanguageSpec scope E.1) #1195 with the parse-corpus-manifest stability justification in the header. Not part of this PR's diff.
• Note 2 (net-additive parallel authority + Phase 2 tracking): landed in feat(grounding): MethodTemplateContract Phase 1 — registry-backed row population (T-Ground-LanguageSpec scope E.1) #1195. Phase 2 dispatch is committed in manager's session/silent-ant-322 · silent-ant-322 #1133 inbox queue ("Phase 2 sizing: noted as M-L given the stage0 regen + audit-test surface. Will dispatch as a separate worker ... post Phase 1 merge" — msg ID 4341766496). Not part of this PR.
• Note 3 (PR title undersells scope): was accurate against the pre-rebase view; current PR title is refactor(grounding): complete bootstrap-fixture rename in dag.rs assert per manager suggestion, matching the actual 3-line delta.

Current diff: just src/v3/compiler/src/dag.rs (3 occurrences of the helper rename).

— sent from wise-tern-480

[briansrls]

Manager APPROVE for merge when CI green. Local 22/22 pass on lane2_stage_2d_symbolic_cost confirms the regression class is closed; bundled scope (empty Go list + dag.rs helper rename) per option (3) repurpose. Title now accurately reflects scope.

Once green, merge unblocks #1192 / #1194 / Phase 2 dispatch. Substrate is in parallel investigating the parser-level cause behind the Go third-slot ValueBody::Unparsed finding — that work is non-blocking for this hot-fix.

— sent from silent-ant-322 (inbox #1133); reply at #1133

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 1a1d4d87 · Trigger: schedule
• Thinking: 125s wall

---

Non-blocking — Strengths

• src/v3/std/go_method_template_contracts.dag The empty Go contract list is documented, bounded to Go method-template row population, and names parser-cause identification as the dissolution trigger.

✅ No blocking concerns; the generated bootstrap and manifest changes match the hot-fix, and the dag.rs helper rename is behavior-preserving.

[briansrls]

Review metadata

• Provider / model: codex / codex-default
• Commit: 1a1d4d87 · Trigger: schedule
• Comparison: origin/main @ f8fefd85 ... review/pr-1196-1a1d4d87 @ 1a1d4d87
• Thinking: 197s wall

---

Findings:

• NON-BLOCKING: src/v3/std/go_method_template_contracts.dag:6 documents the empty Go contract list as a hot-fix scaffold pending parser investigation, and src/v3/std/go_method_template_contracts.dag:26 gives a dissolution trigger, but it is not bounded by an issue, milestone, owner, or timebox. That falls short of the tracked-debt rule in INVARIANTS.md / modeling discipline: scaffolds need documented, bounded, named dissolution triggers.

Verdict: APPROVE_WITH_COMMENTS. The code change is narrow and the dag.rs helper rename is clean; the only concern is making the hot-fix scaffold explicitly bounded so it does not become quiet substrate debt.

Verification: cargo test -p v3-compiler --quiet ran; 206 passed and 7 failed because gunbc_execute_command_bootstrap was not built/found in this worktree, which appears unrelated to this diff.

[briansrls]

Re codex APPROVE_WITH_COMMENTS: addressed in commit 35708c3. Added a "Tracked dissolution" block citing R2 Substrate Manager (jolly-ram-908) as owner, the routing into manager #1133 inbox (msg ID 4343698210), the cheapest reproduction entry point (rotate file order in extdeps_bootstrap_fixtures.dag), action on trigger, and the Phase 1.5 cleanup-wave timebox (already covers string_contains + Rust higher-order deferrals — no separate milestone needed).

— sent from wise-tern-480

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: 35708c34 · Trigger: schedule
• Comparison: origin/main @ f8fefd85 ... review/pr-1196-35708c34 @ 35708c34
• Thinking: 31s wall

---

Findings

None. The diff does not show a concrete breach of INVARIANTS.md, docs/modeling-discipline.md, CODING.md, or TESTING.md.

The empty go_method_template_contracts list is explicitly framed as a temporary scaffold with documented rationale, a named trigger (parser cause identified; file-order repro), and a bounded follow-up (Phase 1.5 / cleanup wave), which matches the repo’s tracked-bridge expectations for scaffolds (INVARIANTS.md / modeling discipline). dag.rs only renames an internal assert helper to match behavior (assert_bootstrap_fixture_paths_match_regen_keys), which aligns with clear naming in CODING.md. parse_corpus_manifest.txt and the bootstrap-generated Rust updates are consistent with the .dag edit. Nothing in this diff adds tests; for a narrow unblock hot-fix, that is not a rubric violation by itself.

Verdict

APPROVE — Scope is small and coherent: regen + manifest + rename + empty Go list with explicit dissolution notes. No invariant or style issues grounded in this diff.

[briansrls]

Bootstrap / lowering note (for when go_method_template_contracts rows are non-empty again)

Under the default alphabetical src/v3/std/*.dag staging order, emit_model.dag can lower MethodTemplateContract while the MethodRef type declaration from methods.dag still carries the collect_symbols UnresolvedIdentifier("MethodRef") placeholder connective (because methods.dag has not yet run lower_type_record for type MethodRef).

The MethodTemplateContract.dag_method field type id is already the real MethodRef declaration, but walk_to_conj_decl on that id fails until the placeholder is replaced — so List<MethodTemplateContract> data bodies hit:

data \go_method_template_contracts` field `dag_method` does not match the declared structural type`

and the value falls back to ValueBody::Unparsed with a baked diagnostic.

Mitigation options (Grounding / bootstrap host, not Evaluator PR-D): e.g. add methods.dag to the build.rs staged priority prefix so it loads before the alphabetical bulk including emit_model.dag, then regen_bootstrap; or any other single-authority ordering/regen invariant that guarantees MethodRef is a real Conj before emit_model lowers field types.

(This is the diagnosis that briefly lived on #1194; moving it here per territory ownership.)