NFR burn-down: src/v2/compiler — translate-serialize-lowering (enumerate 33 typed-census wildcards)

[gunbai-bot]

Typed NFR census (artifact/nfr-typed-census-2a1dd41, checker #12610) burn-down for src/v2/compiler/, group 3 of 4.

Each _ => arm whose scrutinee is a closed coproduct is replaced by explicit arms for the remaining variants, each with the wildcard's body verbatim (the language has no or-patterns, so bodies are cloned). Where needed, the newly named constructors are added to the file's import list. No behaviour change and no refactor. No roster rows added; wildcards kept: none.

Files: src/v2/compiler/06_translate.dag src/v2/compiler/body_lowering_fold.dag src/v2/compiler/program_assembly.dag src/v2/compiler/program_partition.dag src/v2/compiler/target_serialize.dag

Verification: this PR's floor CI compiles and executes the claims (no local whole-corpus build per session policy).

🤖 Generated with Claude Code

[gunbai-bot]

Addressed review 72754 in 19d4379:

• body_lower_is_core_substrate: the four-way disjunction now lives once, in body_lower_is_core_substrate_leaf(node), and each enumerated arm calls it.
• conj_optional_named_child: rather than add a new helper, each arm now calls the existing target_serialize.translate_grammar_inverse_not_realized_diagnostic(emitted: parent), which builds the identical literal. That helper is imported, so the diagnostic has one authority instead of two. conj_named_child in target_serialize.dag had the same copied literal and now calls it too.

— sent from valiant-heron-154

[briansrls]

APPROVE at 3f8cb5b. The compiler partitions preserve the prior fallback results, the atom-kind probes are equivalently routed through node_atom_identity_optional, and repeated substrate behavior is delegated rather than duplicated. All four checks are currently queued or running, including floor, so this is a code approval only; landing remains conditional on the full check set completing green.

[gunbai-bot]

Re review 72905, addressed in 9622625:

• The mechanism was misstated, and I've corrected it. I read v1.compiler.emit_rust. The tail-call match renderer (emit_rust_tco_match / emit_typed_tco_match_arm) does not drop arms. It emits each nested variant under an Rc field as a match guard (collect_pattern_rc_variant_guards), and rustc does not count guarded arms toward exhaustiveness. The ordinary renderer avoids this by grouping those arms into one nested match (rc_grouped_arm_plan, Emitter flattens nested constructor patterns: classify_arrow_body_form's Conj arm is unreachable in emitted Rust — a silent wrong answer, not a warning #8570); the loop path never consults that grouping. All arms are present and none selects wrongly. The failure is a loud E0004 at the target compiler, not §4b silent wrongness. The old comment in the code said "drops"; it now states the actual mechanism.
• The class is now filed: dag/gunbc/recurring_failure_mode/tail_call_match_lowering_skips_rc_arm_grouping.dag. The row records the mechanism, why it was dormant (every such match ended in _ =>, which gave rustc an unguarded arm), rung and ceiling, and the next-rung trigger: emit_rust_tco_match consumes the same grouping authority, proven by an enumerated nested arm set in a tail-recursive fixture that emits and builds.
• Why the step helper stays, rather than a seed-emitter fix in this PR. The helper is the form neat-boar-16's ruling for this lane prescribes for self-calling functions: delegate the decision to a helper and keep the self-call in the function. It is not a workaround that routes around a line-stop. It remains correct after the lowering is fixed; the row names the condition under which the need for it dissolves. The fix belongs in the v1 seed emitter plus regeneration of the stage0 mirror, which is outside this mechanical lane. I'm handing it off to the lane owner with the row as its brief.
• Exposure elsewhere: the only exposed site in this PR's closure was this one; emit-build reported a single E0004. The other NFR PRs' emit-build runs will show any further tail-recursive sites, and each will get the same treatment.

— sent from valiant-heron-154

[gunbai-bot]

In 3e8fbb2, per a ruling from silent-koi-310 on the tail-call emitter bug: grammar_tokens_fold_list_targets_into is back to its shape on main, wildcard included. The delegation helper existed only to avoid E0004, which makes it a §5 workaround. The root fix is a separate lane (work item adhoc-6e41e4c1-f65), and this site gets enumerated after it lands. The RFM row tail_call_match_lowering_skips_rc_arm_grouping stays and now records that disposition. This resolves the scaffold concern in review 72905.

— sent from valiant-heron-154

[briansrls]

APPROVE at 3e8fbb2, conditional on the current emit-build completing green. The compiler enumerations and helper extractions are sound. One accounting correction is required for the joint activation: src/v2/compiler/target_serialize.dag::grammar_tokens_fold_list_targets_into deliberately retains its _ arm under the silent-koi-310 ruling because the tail-call renderer emits nested Rc-field arms as guards and rustc refuses E0004. The PR body therefore cannot say wildcards kept: none; carry this exact site as a retained NFR row in #12610's stack, with the recorded tail-call/Rc grouping dissolution trigger.

[gunbai-bot]

Consolidated into integration PR #12951 at the operator's request; this branch is merged there unchanged in its own merge commit. Closing here; the branch is kept. — sent from deep-ferret-305