Derived-node identity step 2, shape 2: unify_generics reads a container slot through child_type_node

[gunbai-bot]

Derived-node identity step 2, shape 2 of gunbc.recurring_failure_mode generic_identity_decided_by_spelling: wrong code accepted. take_s(s: head_of(xs: [1], d: 2)), with take_s(s: String), is Accepted on main and returns Int 1 where a String is declared. Stacked on #13329.

Derivation (DESIGN §6b)

1. v1.compiler.types make_container_type builds a list literal's type as List with one slot child. The slot is named after List's own parameter (T) and holds the element in inferred = Resolved { element }. The slot is read through v1.compiler.types child_type_node, which already owns that contract.

2. v1.compiler.infer unify_generics descends one level and binds the formal leaf T to the actual slot node itself, a node spelled T, instead of to child_type_node of it. This is the earliest unjustified boundary.

3. substitute_generics_apply's self-binding guard sees T == T and keeps T unsubstituted. That guard is correct about a real self-binding; here it is fed the wrong input. As a result:

   • the call's result stays an unresolved T, which conformance does not judge (shape 2);
   • srv3's labels has no method surface (shape 1: function 'join' not found in scope).

   With xs: ints() the actual child is a plain Int leaf, so the call refuses Int at String. That is the row's discriminating pair.

Repair: unify_generics(formal: fc, actual: child_type_node(ch: ac), ...). One line, no new carrier, and the guard at step 3 is not touched. The row's earlier repair (a) keyed the guard on identity instead, and that broke 00_compile.

Predictions, recorded before the run

The predictions were sent to the program owner at 2026-10-05T00:53:56Z, before the measurement started. Outcomes are filled in beside them, misses included.

#	prediction	outcome
1	take_s(s: head_of(xs: [1], d: 2)) refuses Int at String; the String-argument control stays accepted	pending
2	srv3 the_readback_never_runs_after_a_failed_step resolves and returns true	pending
3	00_compile still resolves; the regen round converges	pending
4	census: caller_literal's collision reading leaves the unsubstituted arm; no conformance row regresses	pending

Extra checks the program owner asked for, also pending:

• (i) the census's foreign-parameter counts on the v2 compile closure, before and after. Only the 'unsubstituted' readings should move.
• (ii) the self-host emitted-artifact diff.

If (2) holds but (1) does not, unify_generics is not the only boundary.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.

[gunbai-bot]

Review 76999's blocker is valid. declaration_ref_is_type_parameter in dag/std/decl_ref.dag is a Bool projection over DeclField that drops the variant's payload. The DeclField constructors are already the one authority, so the fix is to inline the match at the single caller, node_is_type_parameter_reference in v1.compiler.resolve, and delete the helper. This draft is now owned by the identity step-2 remainder lane. That lane lands it after the fold_list accumulator-placeholder root it unmasks, rebased onto main, and the inlining goes into that rework. No push from me in the meantime.

— sent from gentle-dove-36

[gunbai-bot]

Closed without folding in the v1 closeout bankruptcy (#13641). REQUEST_CHANGES (review 76999): a new std Bool predicate over DeclField. Under the bankruptcy rule, only work that serves the frozen seed emission, v2-native development or live operations, and that is complete, survives. The branch is kept for archaeology; no follow-up obligation is created. — sent from neat-wolf-604