test umbrella dissolution: corpus sweep A (dag/test)

[gunbai-bot]

Summary

First half of the umbrella corpus sweep for proud-wren-892. Depends on mechanism PR #7270 (U1–U3). Dissolves aggregators under dag/test/** and retargets related enrollment rows. Split so gh pr diff stays under GitHub's 300-file cap (MODELING-COHERENCE).

Stack

1. test umbrella dissolution: mechanism (U1–U3) #7270 — mechanism
2. this PR — dag/test sweep
3. sweep B — src/v2/** + wire check_orphan_helpers_or_err into the naming walk

Test plan

• gh pr diff succeeds (<300 files)
• CI floor green on affected set
• Follow with sweep B before considering the lane closed

Made with Cursor

[gunbai-bot]

Addressed review 43055 on the rebuilt tip (8247804):

1. types_arch_collapse_witnesses — deleted the dangling aggregator that called removed architecture_dissolution_witnesses(); leaf test fns remain the enrollment surface.
2. claim_witness_corpus_ci_runner keystones — retargeted to *_test.dag entry paths with function: "" (U3 file-grain expand).
3. uri transport — listed all four uri leaf test fns as ClaimRuns (ClaimRun argv path does not expand empty function).
4. U2 orphan deferral — kept unwired on this PR by design; dissolve-on test umbrella dissolution: corpus sweep B (src/v2 + orphan wire) #7274.

Also rebased onto current main (merge conflict clear) and fixed regen/fmt drift via seed-retained intrinsic registration + emitted schedule U3 note.

— sent from proud-wren-892

[gunbai-bot]

Verified review 43070 against the tip: the five falsifier_substrate_long_lane_rows for fallback_arm_census_witness_test.dag (and the matching WitnessExclusionRow) were gone while the long-lane file and its in-file enrollment note remained — bare de-enrollment.

Restored all five rows + the exclusion on this tip. Per-PR planted coverage stays complementary, not a substitute.

— sent from proud-wren-892

[gunbai-bot]

Addressed review 43124: confirmed the known-RED gap.

• Promoted english_emit_add_ingest_round_trip_holds to test fn and re-enrolled it in known_red_probe_entries (the S2 discriminating control named in known_red_frontier_note).
• Dissolved the aggregator english_emit_add_emit_ingest_round_trip_holds.
• Promoted sibling serialize/grammar leaves to test fn so they stay live without the umbrella.
• Kept english_emit_add_prose_holds enrollment.

— sent from proud-wren-892

[gunbai-bot]

CI fix for UnexecutedDeferredWitness on tip:

Sweep A had already dropped the logic_truth_tables_all known-RED probe_red row, but the umbrella test fn remained in the excluded logic_ground_truth_test.dag. Discovery still deferred that leaf with no explicit consumer → admission refusal.

Deleted the aggregator; the three table leaves stay enrolled as known-RED controls. Frontier note / dissolve-on counts updated 4→3.

— sent from proud-wren-892

[gunbai-bot]

Addressed review 43160: confirmed. no_fake_anomalies_witness_test.dag still invoked dissolved roadmap_belt_actuate_witnesses. Retargeted the passing corpus run to leaf witness_uuid_deterministic. Broader stale-ref scan of deleted umbrella names found no other live call sites outside the swept files.

— sent from proud-wren-892

[gunbai-bot]

CI fix for the 3 discovery reds on 09649a9:

1. witness_roster_names_the_cited_incident — mechanical sweep had changed the roster length pin 5→4; restored to 5 (and the ratchet note) to match ip_link_carrier_read_probe.
2. red_rest_network_advertises_hermetic — was a RED helper the umbrella negated (!…); promoting it raw made discovery treat the intentional false as fail. Replaced with green leaf rest_network_is_not_hermetic.
3. emit_coverage_frontier_witnesses — wet template flipped emit_host_native_only_verdict_test to file-grain f: "", but backing enrollment still exact-matched the leaf name. schedule_entry_matches now accepts empty function as file-grain (all decls enrolled).

— sent from proud-wren-892

[gunbai-bot]

Addressed review 43179: confirmed. english_emit_add_matches_serialize_holds / english_emit_add_grammar_inverse_holds were promoted to test fn with no known-red/offline roster consumer (file stays OfflineLocalRecipe). Demoted them back to plain helpers (main's shape) and force-eval them from the enrolled english_emit_add_ingest_round_trip_holds known-red leaf so they keep an executing consumer without minting deferred orphans.

— sent from proud-wren-892

[gunbai-bot]

CI fix for batch-4 discovery reds on 81cde77:

commit_workflow had retargeted emit_host_classical_not_ingested_equals_eval_test and emit_host_native_only_verdict_test to file-grain check_fns: []. That expanded into the wet equals_eval / native-run test fns which these files deliberately keep off per-PR (mock-shape aggregates only; real compile/run is wet/nightly). Restored ingested_classical_not_emit_mock_shape_all_holds and emit_host_native_only_verdict_mock_all_holds.

— sent from proud-wren-892

[gunbai-bot]

Rebased session/proud-wren-892-sweep-a onto current main. Resolved roadmap_belt_actuate_witness_test.dag (kept attempt-grain probed_at UUID arity from main + leaf test fn promotion; dropped dissolved umbrella). Also retargeted main's new expectation_frontier_witness_test (#7276) off dissolved umbrella names to surviving leaves.

— sent from proud-wren-892

[gunbai-bot]

Rebased again onto latest main (8d543888). Resolved falsifier_workflow_witness_test.dag (promoted main's new timeout-basis leaf; dropped dissolved umbrella). Retargeted claim_batch unit-test fixtures off commit_workflow_witnesses. Tip is mergeable=MERGEABLE with main as ancestor.

— sent from proud-wren-892

[gunbai-bot]

Conflict notice looks stale: tip 9ef6e31d92 already has main (8d543888) as an ancestor. GitHub reports mergeable=true / mergeable_state=blocked (checks, not conflicts). Local git merge origin/main is already up to date; falsifier_workflow_witness_test.dag and roadmap_belt_actuate_witness_test.dag have no conflict markers.

If the PR UI still shows the conflict banner, refresh — the resolve was in the rebase that landed this tip.

— sent from proud-wren-892

[gunbai-bot]

Rebased onto latest main (862bb773). Resolved design_palette_witness_test.dag: kept main's Optional Present/Absent API shapes for contrast/HSL/grade_pairing and promoted those leaves to test fn (dropped dissolved keystone umbrella).

— sent from proud-wren-892

[gunbai-bot]

CI fix for compiler_tests_rust_blobs_are_all_rostered: rostered the unmarked ct_render_rust_applied_type_qualified_base_test blob in language_source_scaffold_index (declared 27 vs rostered 26 — this PR's touch pulled the live-tree coverage witness into the affected set).

— sent from proud-wren-892

[gunbai-bot]

MODELING-COHERENCE UNAVAILABLE was not a .dag modeling defect — #7274 had been squash-merged into this branch (1084e8c), pushing the PR to 373 files so gh pr diff returns HTTP 406 too_large (fail-closed).

Restored tip to pre-merge f07e96d (198 files); gh pr diff works again. Sweep B re-opened as a stacked follow-up (same content as #7274). Combined tip kept on session/proud-wren-892-sweep-ab-combined.

— sent from proud-wren-892

[gunbai-bot]

Verified review 43304: the file-grain admission expand is new LOC in HAND_MAINTAINED cli_run.rs, but it is not a new seed surface — it closes a false-refuse gap under the existing interim text-scan for Phase 0(b) admission keys.

Landed the missing checkable §7 receipt (same form as peer CLI_RUN_*_SCAFFOLD_MARKERs):

• marker cli_run_witness_admission_source_scan
• ROADMAP lane 5-dissolve-patches + module-identity-storage-binding Phase 1 (b)
• DELETE WHEN host consumes v2.workflow.witness_admission.witness_admission_explicit_consumer_manifest (emitted), retiring the scan + unit RED
• explicit not-a-census-shrink note (counted interim debt until that dissolve)

Discriminating unit tests green locally (file_grain_bin_wet_expands_admission_consumer_keys, witness_admission_source_scan_scaffold_marker_is_declared).

— sent from proud-wren-892

[gunbai-bot]

Verified review 43319: confirmed. The receipt rg cli_run_witness_admission_source_scan … == 1 was false (4 hits: scaffold header, receipt line, string literal, unit test).

Tightened to a declaration-grain pin that does not self-match the comment: split IDENT / TYPE_ANN across lines, then rg -F "${IDENT}${TYPE_ANN}" — verified count 1 (the const declaration only). Bare-token multi-hit called out as intentional.

— sent from proud-wren-892

[gunbai-bot]

Addressed review 43336 on tip 25cb0911f5:

1. falsifier_native_cache_cold_entries — verified: dissolving emit_host_meet_join_equals_eval_all_holds had left only emit_host_meet_equals_eval_holds. Added a peer self_host_wet_entry for emit_host_join_equals_eval_holds (same entry file). Cold-roster note updated.

2. proactive_verification_ledger — added ^family_equals_join citing emit_host_join_equals_eval_holds; retitled the meet row so the family tracks both peer primaries. Enumeration witness updated 9→10 family receipts / 12→13 total rows.

— sent from proud-wren-892

[gunbai-bot]

Addressed review 43345 on tip b5964bba90:

Verified in language_source_scaffold_index.dag: three identical ct_row_ct_render_rust_applied_type_qualified_base_test declarations and three roster slots (main already had two; this PR had added a third). Deduped to a single declaration and a single roster entry. Registration of the blob remains; the parallel copies are gone.

— sent from proud-wren-892