Home Never in std.types; move kind into v2.std

[gunbai-bot]

Summary

• Home Never beside Unit in std.types; delete it from v2.std.cardinality; retarget every importer including v2.extdeps.languages.rust (RustNever).
• Move std.kind to v2.std.kind (constructs v2.std.node); delete dag/std/kind.dag. That is the whole compile_cli CrossTreeDenied set that originated in std.kind.
• Acquisition door: acquisition_lookup_refusal_text_formats_a_dag_to_v2_denial is formatter-only. a_forbidden_dag_to_v2_acquisition_refuses_without_reading submits a dag→v2 import through acquisition_step and asserts AcquisitionLookupRefused / ReferenceCrossTreeDenied with empty reads. a_permitted_v2_to_v2_acquisition_advances is the same step with a known-memo v2→v2 import so a refuse-everything step fails.
• Not in this PR: skip / fold_list / length / list_map (algebra child adhoc-798d5122-dcd); gunbc witness/ratchet/census (clever-crane-830).

Test plan

• Dag import of a v2-only module still refuses on the acquisition door (executed claim-run below).
• No import v2.std.cardinality { Never }; no module std.kind; no dag/std/kind.dag.
• Zero CrossTreeDenied originating in std.kind on v2.compiler.compile and v2.cli.compile_cli.
• std.judgment_contract still imports v2.std.algebra { skip } until the algebra child homes skip.

Executed receipt (tree 9054db1)

Host cgroup memory.max=25769803776 (24 GiB, already bound; /sys/fs/cgroup is read-only so no nested leaf). Nested mkdir refused: mkdir: cannot create directory ‘/sys/fs/cgroup/royal-raven-317-exec’: Read-only file system.

Build:

RUSTC_WRAPPER= CARGO_TARGET_DIR=$PWD/target PATH="$(dirname "$(rustup which cargo)"):$PATH" cargo build --release -p v1-compiler --bin gunbc

Outcome: Finished release profile [optimized] target(s) in 3m 59s. Artifact target/release/gunbc 38301952 bytes, executable, timestamp 2026-10-13 05:59.

Green (unmutated acquisition_verdict_step), every test fn in v2.test.claim.module_acquisition:

./target/release/gunbc run --source-root dag --source-root src/v2 --entry src/v2/test/claim/module_acquisition.dag --claim-run

Outcome: exit 0. Includes:
PASS acquisition_lookup_refusal_text_formats_a_dag_to_v2_denial
PASS a_forbidden_dag_to_v2_acquisition_refuses_without_reading
PASS a_permitted_v2_to_v2_acquisition_advances

Red (scratch mutation, not committed): CandidateRefusedBy arm of v2.compiler.module_acquisition acquisition_verdict_step rewritten to AcquisitionAdvanced { state: looked.state }. Then:

./target/release/gunbc run --source-root dag --source-root src/v2 --entry src/v2/test/claim/module_acquisition.dag --claim-run --function a_forbidden_dag_to_v2_acquisition_refuses_without_reading --function a_permitted_v2_to_v2_acquisition_advances --function acquisition_lookup_refusal_text_formats_a_dag_to_v2_denial

Outcome: exit 1.
PASS a_permitted_v2_to_v2_acquisition_advances
PASS acquisition_lookup_refusal_text_formats_a_dag_to_v2_denial
FAIL a_forbidden_dag_to_v2_acquisition_refuses_without_reading

Mutation reverted with git checkout -- src/v2/compiler/module_acquisition.dag. Head remains 9054db138ff.

[gunbai-bot]

review 79147: v2.extdeps.languages.rust still imported Never from v2.std.cardinality after the home move. Fixed in 6d50fb8 — import std.types { List, Never }; the cardinality import is gone; type RustNever = Never still binds. No remaining import v2.std.cardinality { Never } in the tree.

[gunbai-bot]

Rehome vs the Manager-2 cut (royal-bee-157)

The cut: rehome route/ratchet/debt/progress-guard (and roster) into v2.workflow.*, census fold into a v2 home; then invert so v2.compiler.compile imports none of those. Layout path follows the module name. gunbc.instruments.type_declaration_use_census_reading stays dag.

What the mergeable diff is

HEAD of this PR does not contain that rehome. de21d1ca (native-route/frontier → v2.workflow) and 68327eba (census fold → v2.compiler) were both reverted in the same branch (7bfb16ec, a222f7d4). The open file list is Never in std.types plus std.kind → v2.std.kind. The PR body is accurate: “Not in this PR: gunbc witness/ratchet/census.”

compile.dag on HEAD still has:

import gunbc.witness.v2_native_route { … }
import gunbc.native_frontier_ratchet { native_frontier_verdict }
import gunbc.source_layout_namespaces { layout_namespaces }
import gunbc.instruments.type_declaration_use_census { … }

Those are still Dag→V2 when natively acquired.

The reverted rehome commits, judged against the cut

de21d1ca + 68327eba were the right rename+repoint: one name per module, files under src/v2/workflow/ and src/v2/compiler/type_declaration_use_census.dag, emit use-paths retargeted, reading/seed-growth left in gunbc, roster moved with debt (so a dag roster would not import v2.workflow.native_frontier_debt).

They did not invert. After those commits compile.dag still imported v2.workflow.v2_native_route, v2.workflow.native_frontier_ratchet, and v2.compiler.type_declaration_use_census — same-tree, so the Dag→V2 edges from those modules die, but the compiler entry still reaches a witness, a ratchet and a census. That leftover is the inversion slice, not a defect of the rehome itself.

Restoring the rehome on this PR (or a follow-up that is only the rehome) should revive de21d1ca+68327eba, not a parallel rename. Inversion (compile.dag imports none of those four) waits until that rehome is actually on main.

[github-actions]

HealPublished repository=gunb-ai/gunbc branch=session/royal-raven-317 parent=f8ca1e234974142d54935ee9892e7b56f804ea02 published_head=140b43973515aa007096cc5b9bf9b4e637abbf47 tree=c594a30b58865704c665f08c23a8f631c52a16cb paths=src/v1/stage0/src/std_types.rs