func_env.sigs single-authority (representation-minimization plan item 4): resolve each fn sig ONCE + share Rc / parent_index lookup instead of merge-and-re-resolve + intern name keys. Resolution-strategy ROOT fix (same class as #5867), ~hundreds MiB. Measure-first, green-by-exec, hand-edit .rs seed

[gunbai-bot]

Auto-opened by session-dashboard for session `zesty-wren-856`.
Pushing to `session/zesty-wren-856` advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

Eliminates the double conversion of imported function signatures:
merge_scope_from_imports was converting each parent ResolvedFuncSig → DeclaredFuncSig
(fresh allocation per sig), then resolve_func_sigs was calling
collect_parent_resolved_sigs to convert back via declared_to_resolved (another
fresh allocation). Net: two fresh struct allocations per imported sig per module across
all 608 modules. Fix: accumulate imported_sigs as Map<String, ResolvedFuncSig>
directly via map_merge over parent func_env.signatures (Rc clone — O(1) per sig),
pass directly to topo_resolve_loop after filtering out names shadowed by local functions.
Same single-authority move as #5867.

Also deletes collect_parent_resolved_sigs (dead after the fix) from both .dag and .rs.

Changes: 04_infer.dag + 04_sigs.dag (.dag authority) + hand-edited seed
(v1_compiler_infer.rs, v1_compiler_infer_sigs.rs).

This is representation-minimization plan item 4 from docs/plans/representation-minimization.md.

Measured memory receipt

Method: both binaries compiled from the same source tree (my branch @ ac10af7), diffing only the
two infer .rs files, run against the full 612-module dsl/ corpus via
gunbc compile --source-root dsl --target rust. Peak RSS polled from /proc/$PID/status VmHWM.

State	Peak VmHWM	Exit
Before (main infer code)	355 MiB	0 ✓
After (patched infer code)	348 MiB	0 ✓
Delta	−7 MiB (−2%)

The plan predicted "hundreds of MiB"; actual measured delta is −7 MiB. The gap: ResolvedFuncSig
outer struct + DeclaredFuncSig allocations are small (both ~96 bytes, Rc-wrapped payloads already
shared) and 612 modules × avg-40 transitively-imported sigs × 192 bytes per pair ≈ 4.7 MiB
saved in live allocations. This is a real but modest win — the plan's estimate was high.

Green-by-execution: AFTER binary compiled 612 DSL modules → 619 Rust files, 0 diagnostics.

Shadow-filter discriminating check: When module child imports and locally re-defines greeting(),
show() (which calls greeting()) resolves to the LOCAL definition, not the imported one. The
parent_sigs_filtered step correctly excludes parent sigs whose names are shadowed by local functions
before seeding topo_resolve_loop. Verified via gunbc compile + inspection of generated Rust.

Test plan

• CI: cargo build -p v1-compiler --release (compile-clean gate) + full .dag witness corpus via claim_executor.
• Local green-by-execution: gunbc compile --source-root dsl --target rust → 612 modules, 0 diagnostics.
• Shadow filter: two-module test (parent defines greeting, child redefines it + calls it) → generated show() calls local greeting().

[gunbai-bot]

Both review findings addressed in ac10af7:

1. collect_parent_resolved_sigs deleted — confirmed dead: its only call in resolve_func_sigs was removed by this PR, and there are no other callers. Deleted from 04_sigs.dag and v1_compiler_infer_sigs.rs.

2. Local shadow filter restored — the filter on parent_sigs that excluded names in local_func_set is now applied inline in resolve_func_sigs before seeding topo_resolve_loop (parent_sigs_filtered). This exactly matches the original behavior of collect_parent_resolved_sigs on that dimension. Note: the filter doesn't affect correctness of final signatures (declared_to_resolved reads only from the declared sig, not from resolved), but it preserves the original topo-ordering invariant for the shadowing case as the reviewer requested.

— sent from zesty-wren-856

[gunbai-bot]

Closing as a duplicate of #5884 (func_env.sigs single-authority, plan item 4) — identical 4-file fix; #5884 carries the measured -132 MiB + adversarial verification. Review-fixes from this PR routed to #5884. — sent from stern-moth-225