Delete dead emit_resolve_transitively_fn stamp from generated main

[gunbai-bot]

Summary

• Generated main already compiles through cli_run::compile_emission. emit_resolve_transitively_fn / emit_extract_import_paths_fn stamped an unused import-line walker (resolve_transitively_with_seen) with no callers. Deleted rather than rewritten (review 38602).
• Stacked on compiler_tests resolve_source_closure: close through the compile-closure authority #13518.

Receipts on 9b25ca28f7

• //gunbc/instruments:self-host: held. exit_status=0 warning_count=0 wall_s=873. Driver refused the poison specimen (tokenize_lex_e1_unrecognized_char).
• //gunbc/instruments:emitted-crate-workspace under a 48GiB cgroup leaf (memory.max=51539607552): still EmittedWorkspaceClosuresDisagree outside_dag=v2_std_layer,v2_std_refinement,dry_run (not_emitted empty, 240 files). Same refusal after the stamp was deleted.
• Same three outside_dag names on origin/main aa3f1dddbc and on compiler_tests resolve_source_closure: close through the compile-closure authority #13518 ea1a22c3c1. This disagreement pre-exists Delete dead emit_resolve_transitively_fn stamp from generated main #13520 (seed emit is compile_entry_emission / both-closure; the plan is emitted_compiler_closure_module_dag over dependency_resolution_facts_live). Not a land blocker for this deletion.
• Without a cgroup, the instrument path is HostBudgetUnreadable; EstimatedMemory cannot invent a cap.

Two causes in that outside_dag triple

• v2_std_layer and v2_std_refinement are DAG modules the seed emit includes (both-closure) and the plan omitted (ImportBearing on dependency_resolution_facts_live). Fix is Producer fix: dependency_resolution_facts unions reference edges for import-bearing files #13526 (producer union in dependency_resolution_facts), not a per-consumer reroute of emitted_workspace_plan.
• dry_run is separate: emit_dry_run_module in 05_emit_rust.dag writes src/dry_run.rs when has_services. It is not a .dag module. That exception also lands in Producer fix: dependency_resolution_facts unions reference edges for import-bearing files #13526 so the instrument integration control can clear completely. Closed Route emitted-crate-workspace plan through compile-entry both-closure #13528 duplicated that work.

Test plan

• ctrl-build --remote -- cargo clippy --all-targets -- -D warnings on 9b25ca28f7
• self-host on 9b25ca28f7
• emitted-crate-workspace cgroup A/B: this head, compiler_tests resolve_source_closure: close through the compile-closure authority #13518 base, main
• Do not merge from this session.

Made with Cursor

[gunbai-bot]

review 38602 is right on both findings.

The import_only_emitted_resolve mutant was a hardcoded string and never called emit_resolve_transitively_fn. resolve_transitively_with_seen had no call site: retained-host compile already goes through cli_run::compile_emission, same census main.rs recorded when those helpers lost their callers.

Deleted emit_resolve_transitively_fn and the orphaned emit_extract_import_paths_fn stamp (and the string tests) instead of rewriting them. 9b25ca28f7.

— sent from quiet-eagle-374

[gunbai-bot]

Verified review 77332 (claude/opus APPROVE) against 9b25ca28f7.

Call sites: emit_resolve_transitively_fn / emit_extract_import_paths_fn are gone from both 05_emit_rust.dag (emit_main_pipeline_fns now concatenates collect/extract_module_path/dep_pool/compile_target_helpers only) and v1_compiler_emit_rust.rs. Tree grep for those names hits only the namespace-cut plan sentence that records the deletion. Remaining extract_import_paths( calls are cli_run / entry_resolve, not the deleted stamp.

The two resolve_transitively_with_seen mentions stay. src/v1/stage0/src/main.rs is already past-tense census of the handler deletion (were this handler's private … machinery). real_ingest_test.dag is a 2026-07-20 TEMPORARY note about pool-pull; rewriting it is that dissolution, not this stamp deletion. Not touching them here.

— sent from quiet-eagle-374