decl_ref constructors move to std.decl_ref: dissolve the parallel-lane fork that reds shared closures

[gunbai-bot]

What

Two parallel lanes each minted an identical fn decl_ref / fn decl_field_ref constructor pair on the same day (2026-08-01): std.primitive_identity (#7549) and std.roster_frontier. Because v1-seed fn names are not module-scoped, any resolve closure containing both modules refuses every bare decl_ref reference as an ambiguous decl_ref — this is the current-main floor red that has been failing #7571's runs through no fault of that branch (6 ambiguous refusals in the primitive_identity_join_witness_test closure).

This PR dissolves the fork at its §3 home: the constructors move to std.decl_ref, beside the DeclarationRef type they construct, with decl_ref_constructor_authority_note recording the incident. Both duplicates delete; 13 importers repoint from std.roster_frontier to std.decl_ref.

Net −70 lines (+47/−117 across 17 files) — a de-fork, as it should be.

Why it reds main today

• dag/std/primitive_identity.dag (merged Primitive identity join D0: five-surface census and join refusals #7549) and dag/std/roster_frontier.dag both declared top-level fn decl_ref(module_path, decl_name) and fn decl_field_ref(...).
• v1-seed toplevel fn names are not module-scoped: two same-named fns in one resolve pool → every call site refuses as ambiguous.
• The join witness file imports both modules' worlds transitively, so its whole closure reds; the same landmine sits under any future closure that pulls both std modules.

Witness receipts (by execution, release binary, this branch)

dag/test/claim/primitive_identity_join_witness_test.dag — all 15 test fns green:

• w_open_denominator_counted_receipt · w_derived_surface_census_population · w_interpreter_measurement_rows_marked_measurement_only · w_derived_rows_marked_derived_from_authority · w_map_keys_joins_all_five_surfaces · w_fold_list_spelling_fork_named_finding · w_fold_list_registry_gap_proves_open_denominator · w_index_by_registry_algebra_gap_named_finding · w_unknown_realization_refuses · w_duplicate_semantic_definition_refuses → true
• w_host_unspecified_traversal_classified_for_map_projections · w_red_missing_identity_on_joined_surface · w_red_surface_joins_two_identities · w_red_cross_authority_identity_fork · w_red_incompatible_invocation_assignment → true

dag/test/claim/roster_frontier_annotation_witness_test.dag — 5/5 green (census_closure_frontier_rows_count_is_twenty_one, rows_well_formed, and the three frontier_expiry fns).

Compile-clean: all repointed importer entries compile with 0 blocking diagnostics (gunbc compile --target dag).

Unblocks

#7571's floor: its recurring red is this main-side fork, not its own diff. Once this merges, #7571's next floor run is meaningful.

🤖 Generated with Claude Code

[gunbai-bot]

Addressed the REQUEST_CHANGES from review 46502: the duplicate constructor at dag/gunbc/host_standup.dag is deleted in 8a74095. A full-tree sweep confirms std.decl_ref is now the only definition site of fn decl_ref / decl_field_ref (disposition_redundancy's decl_ref_eq is a distinct comparator, not a copy). host_standup is an import-free module, so no import was added — its bare decl_ref calls resolve from the same pool source its DeclarationRef type references already use. Receipts: all 27 host_standup spine + assimilation-deduction witness fns green by execution; compile-clean 0 blocking; the module is not in the emitted stage0 set, so the regen fixed point is untouched. — sent from sleek-tern-275

[gunbai-bot]

Addressed review 46764: fresh regen on the merged tree lands in fd94366 — src/v1/stage0/src/std_roster_frontier.rs drops the constructor pair (+1/−22), so the seed matches its .dag source and std.decl_ref is the sole definition site in both representations. Root cause of the miss: my first regen ran before the main merge brought #7571's regenerated seed back in, and a second regen attempt used a stale pre-#7585 regen_stage0 binary that produced a phantom unicode-escape drift in v1_compiler_infer.rs — rebuilt the bin, reran, and the only remaining drift was the honest one, now committed. — sent from sleek-tern-275