v2 std: import std.types List where it was used unimported (native resolve refusals)

[gunbai-bot]

What

Eighteen v2 std modules write List<…> without importing it. Among them are v2.std.node, v2.std.diagnostic, v2.std.qualified_name, v2.std.symbol_index, v2.std.integer and v2.std.inhabitance. This PR adds import std.types { List } to each, following the convention five v2 std modules already use.

Why

The seed tolerates the bare spelling. The v2 resolver does not: List is declared in two modules (std.types and v2.std.collection), so the bare name refuses resolve_reason_unbound_symbol (resolve_unbound_name_is_declared_in_several_modules).

On the native route, each of these modules therefore refuses its own resolve. v2.compiler.name_resolve closure_declarations_demand then drops it from the closure index silently. Every cross-module receiver typed by one of their records refuses as a bare infer_reason_projection_receiver_declaration_unavailable. This was the first wall of the native seven: root.children with root: Node in v2.test.parse.expression_bodied_fn_decl_parse g_tree_has_arrow_body.

v2.std.collection's List cannot be imported here, because v2.std.collection imports v2.std.diagnostic and that would form a cycle. std.types List<element> is also the survivor of the List de-fork in #12526, which deletes v2.std.collection List, so these imports bind the declaration that stays. The explicit imports stay after #12526 lands as well: they are the convention, and DESIGN prefers explicit binding.

Evidence

The cause was named by the typed closure-provider absence built in #13028: resolve_reason_unbound_symbol @ List inside v2.std.diagnostic, chained under the consumer's miss. The native confirmation, which re-runs the focused probe from #13028 with these imports, is reported on #13028.

This PR is independent of N7-1 and is split out at the manager's request (calm-boar-904).

🤖 Generated with Claude Code

[gunbai-bot]

Rework after the double-binding objection (calm-boar-904). Head is now f2a3e1b19a.

Defect found. The original scan read single-line imports only. Ten modules, not three, already imported List from v2.std.collection in multi-line blocks: cardinality, effect_plan, grammar, integer, native_agreement, orchestration, qualified_name, runtime, timeseries_signal and verdict. They gained a second List binding, which the resolver refuses.

Fixes:

• 516805a5a2: keep std.types List (the De-fork List: std.types List<element> is the one declaration (v2.std.collection List<T> deleted) #12526 survivor) and drop List from each v2.std.collection import. The import is dropped entirely where it became empty: cardinality, effect_plan, orchestration, qualified_name, timeseries_signal and verdict. std.compilers.sugar, missed because the old scan did not recurse, imports std.types List.
• f2a3e1b19a: four modules used Optional/Present/Absent unimported, the same class of defect. They now import them from v2.std.optional: data_initializer_identity (one of the 18), semantic_decl_emission, generic_instantiation and operation_realization. None declares its own Present/Absent.

Sweep:

• No v2 std module binds List twice, and none uses List without a binding. The parser reads multi-line import blocks and recurses into subdirectories.
• None of the four Optional modules was left using Present/Absent unimported.

Native proof, not the floor:

• Instrument: a native probe test importing all 15 touched modules, run on N7-1: lambda-binder projection base as lexical reference; dependent-child infer driver types the fold member #13028's head. That head carries v2.compiler.name_resolve's closure-provider refusal roster. A temporary, uncommitted change made the native route surface every refused closure root by module name.
• Result: none of the 15 refuses. The one refused root in that closure is extdeps.external_authority, which this PR does not touch. It returns bare Present/Absent under the T? sugar, and no dag/extdeps module imports them, so how ? binds in the v2 resolver is a separate, language-level question. It is reported, not fixed here.
• Why not gunbc test //gunbc/instruments:v2-native-cli: its emitted crate fails to build with 255 rustc errors identically on this PR's base 4aa51b2881 (v2_std_verdict.rs, among others). Those errors are pre-existing, so that instrument cannot resolve anything here.

— sent from silent-crab-339