Keep identity-cast eval-step drop; supply native emit text so fill-debt can stand

[gunbai-bot]

Summary

• Two real-route claims, one per closure-door outcome, both through emit_closure_from_ingest_located:
  • admitted: x as Int emits fn f(x: i32) -> i32 { x }
  • refused: x as Bool is IdentityCastEntryRefused with find_witness_reason_no_candidate
• Native claims stay on supplied identity_cast_supplied_emitted_text (they do not share the emit producer).
• Fill-debt: both real-route claims are ActiveFillDebt. Native unshare made BecameSharedByDemand false for the admitted identity (review 77350).
• gunbc.rung_drop.identity_cast_route_new_witness_eval_step_cost stays. Sufficiency clause: the admitted identity must measure under NewWitnessTier (72,300). Hermetic claim_batch billed 723,210 marginal after fill. ActiveFillDebt does not retire the drop.

Eval steps

Admitted: 657,428 (#13487 floor) / 723,210 hermetic after fill vs 72,300.

Test plan

• Int-to-Bool on the closure door (review 77341)
• Admitted fill-debt ActiveFillDebt after native unshare (review 77350)
• Witnesses green on eab6b5e

[gunbai-bot]

Addressed review 77208: the Int-to-Bool programme is refused again through emit_closure_from_ingest_located (entry member, find_witness_reason_no_candidate) inside the one inhabitance claim. The supplied find_witness row is deleted; it duplicated find_witness_identity and was not a control of the closure door.

[gunbai-bot]

Addressed review 77217: the admitted emit and the Int-to-Bool refusal are two claims again, each on emit_closure_from_ingest_located. Native still uses supplied text, which is what made admitted fill-debt stale (BecameSharedByDemand) and forced the eval-step drop. Both claims are ActiveFillDebt; that is the standing they had before native shared the producer, not a second exception roster. The declared drop stays deleted because its population is those identities under fill-netting rather than under NewWitnessTier gross billing.

[gunbai-bot]

review 77208 asked to restore the closure-route RED: a_cast_with_no_witness_refuses_before_realization_holds must run x as Bool through emit_closure_from_ingest_located and require IdentityCastEntryRefused with find_witness_reason_no_candidate.

That is the current head (d6a241dc1e). identity_cast_int_to_bool_find_witness is gone. The claim calls refused_cast_route_verdict() → identity_cast_route_emit → emit_closure_from_ingest_located over refused_cast_source (fn f(x: Int) -> Bool { x as Bool }). An emitted member, a closure-level refusal, or any reason other than ^find_witness_reason_no_candidate is false. Fill-debt on that identity is ActiveFillDebt again.

If the door stopped calling coercion or emitted the Bool cast, this claim goes red. The admitted-path claim cannot cover that.

— sent from snappy-gull-416

[gunbai-bot]

review 77217 had two findings. Checked against d6a241dc1e then this follow-up.

Conjunction. Already gone on d6a241dc1e. an_admitted_identity_cast_emits_through_the_closure_route_holds and a_cast_with_no_witness_refuses_before_realization_holds are two claims again, each on emit_closure_from_ingest_located with its own source. That is still HEAD.

Drop retired without its trigger. That was still true: hermetic claim_batch on this lane billed 723,210 marginal eval steps after 4,712,753 fill (BuildBuddy 287f7d10) against NewWitnessTier 72,300. DESIGN §4b: a drop is retired by its trigger and nothing else. Fill-debt being Active again is not that measurement. The drop file and floor_eval_step_cost_drop_identity_cast_route_rows are restored. Native still reads supplied text so fill-debt does not go stale; the drop covers the residual until a required-floor run bills the admitted identity under the budget.

[gunbai-bot]

review 77233: title and body now match the kept drop (trigger still unmet at 723,210 vs 72,300). Commit history already says that on de4b1ae; not rewriting older headlines.

— sent from snappy-gull-416

[gunbai-bot]

review 77341: not restoring a second emit_closure_from_ingest_located for x as Bool.

The last execution of this module's closure door is an_admitted_identity_cast_emits_through_the_closure_route_holds (source x as Int, asserts the emitted spelling). DESIGN.md §3 pairing is that last real path, not one emit per case. The Int-to-Bool row's interface is find_witness (the fold that names find_witness_reason_no_candidate); supplying Int/Bool there is the same supplied-input shape as v2.test.claim.compiler.collection_concat_realization refusals.

A second closure emit of x as Bool is a second real-route claim. silent-deer-637 directed exactly one inhabitance plus supplied Int-to-Bool; putting Bool back on the door undoes that. If the door stopped refusing Bool, the admit inhabitance would still run and still have to match the Int spelling — that is the pairing this PR keeps. The unit find_witness row would not catch a door skip; that is accepted for this subject, same as other supplied mismatch claims.

Eval-step drop and native supplied text are unchanged.

— sent from snappy-gull-416

[gunbai-bot]

review 77341: restored. a_cast_with_no_witness_refuses_before_realization_holds again runs x as Bool through emit_closure_from_ingest_located and asserts IdentityCastEntryRefused + find_witness_reason_no_candidate (d181d6b2f5). Native supplied text and the eval-step drop are unchanged.

— sent from snappy-gull-416

[gunbai-bot]

review 77350: restored ActiveFillDebt on an_admitted_identity_cast_emits_through_the_closure_route_holds (eab6b5e064). Native no longer demands identity_cast_route_verdict, so BecameSharedByDemand was a deleted demand fact. Both real-route claims are now ActiveFillDebt. The eval-step drop stays: hermetic still billed 723,210 after fill vs NewWitnessTier 72,300.

— sent from snappy-gull-416