wiring-liveness whole-tree: underscore-prefix dead-wire sweep + inherited drift fixes

[briansrls]

Auto-opened by session-dashboard for session sunny-fox-408.
Pushing to session/sunny-fox-408 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

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[gunbai-bot]

Re review 32568 findings:

Finding 1 — transport src/v2-only scope vs documented dsl/ trajectory

Verified: dsl/ currently has 46 dead wires (run wiring_liveness_whole_tree --source-root dsl to reproduce). Enrolling it now would break the gate. The src/v2-only scope is the correct current boundary; dsl/ enrollment is blocked on those 46 fixes.

On the scaffold marker: .dag files have no comment syntax (deliberate parse wall — // and /* */ are unparseable by design). The dissolution trigger is carried in two places already:

• The dissolve-on: doc comment on check_wiring_liveness_streaming in coproduct_reflection.rs names the Realization carrier for wiring_reach_saturate as the trigger
• The wiring_liveness_intentional_dead_wire_fixture named constant in the transport carries the why for the exclusion via its identifier

The dsl/ enrollment trigger is: fix the 46 dead wires surfaced by running the gate against --source-root dsl. The gate itself enforces src/v2 won't regress while that work proceeds.

Finding 2 — WholeLivenessCorpus fail-open clarification

Valid read gap. This variant is being replaced by DiagnosticsCollector in the upcoming rebase onto #5826 (smart-wolf's carrier PR, sole authority per parent §3 pin). The clarification will be moot once the rebase lands — DiagnosticsCollector is the surviving variant and its doc comment already describes the collect-and-continue semantics. No separate fix needed on this head.

Finding 3 — coproduct_reflection.rs parallel-rep debt

Acknowledged and tracked per the existing dissolve-on: marker. No action.

— sent from sunny-fox-408

[gunbai-bot]

Addressing review 32573 findings (all non-blocking per APPROVE verdict):

Finding 1 — parallel-rep debt (check_wiring_liveness_streaming vs .dag reachability fold)

Tracked. The dissolve-on: marker at coproduct_reflection.rs:826-834 names the exact trigger (Realization carrier for wiring_reach_saturate landing, collapsing the O(n²) interpreter fold). The suggestion of a fixture-level cross-check (Rust path dead-wire count == .dag lens dead-wire count for a known fixture) is a good hardening item — filing as follow-up work after the dissolve-on trigger lands. Until then the two paths must stay in manual lockstep; the comment is the authority.

Finding 2 — WholeLivenessCorpus silent corpus shrinkage

The eprintln! of modules_resolved + modules_excluded at wiring_liveness_whole_tree.rs:87-94 already surfaces this to stderr on every run (visible in CI floor logs). A hard assertion of the form modules_resolved == discovered - excluded would require a two-pass discovery (first scan for the total count, then whole-tree resolve), which is a non-trivial change. The WholeLivenessCorpus non-blocking mode is by design: per-module fail-closed (#5760) means a broken module emits a located diagnostic and is excluded, not a global abort — the floor stays live. The logged count is the observability knob for now; a hard assertion gate is reasonable follow-up if corpus shrinkage becomes a real detection gap.

Finding 3 — bootstrap.dag _child rename (diff line 1916, file line 261)

Pre-existing semantics, confirmed. My _-prefix sweep in this PR just made the intent explicit. The reviewer correctly identifies it as the surface a future fix should target. Not actionable here.

— sent from sunny-fox-408

[gunbai-bot]

HOLD — merge after #5826

ResolveTypecheckGate::WholeLivenessCorpus in this branch is a §3 carrier fork of DiagnosticsCollector from smart-wolf's #5826 (both non-blocking, both drive compile_to_resolved; DiagnosticsCollector is a strict superset — also returns the diag set, which this path ignores). Per the anti-fork pin, single authority = DiagnosticsCollector.

After #5826 merges to main, I will:

1. Rebase wiring-liveness whole-tree: underscore-prefix dead-wire sweep + inherited drift fixes #5774 onto main
2. Delete WholeLivenessCorpus variant + duplicate typecheck_gate param add
3. Switch wiring_liveness_whole_tree.rs to pass ResolveTypecheckGate::DiagnosticsCollector (ignore returned diags with _)

Everything else in this PR is approved and stays unchanged.

— sent from sunny-fox-408

[gunbai-bot]

Addressing review 32582 (REQUEST_CHANGES):

Finding 1 — orphan transport referencing deleted binary

Valid and fixed. wiring_liveness_whole_tree.rs and its Cargo.toml entry were already on main (landed in #5743 wave-3). My gate-flip strip incorrectly deleted them with git rm, then added wiring_liveness_transport.dag pointing at the now-deleted file — producing exactly the phantom-target §5 concern the review flagged. Fix in 919de50485:

• Restored binary source + Cargo.toml entry from main (net-zero vs main for those files)
• Deleted wiring_liveness_transport.dag (it was orphaned from the moment the binary was accidentally deleted)
• Removed ensure_wiring_liveness_whole_tree_built from host_prelude.dag
• Removed the orphaned roster row from realization_vocabulary_containment.dag

Finding 2 — dead code (check_wiring_liveness_streaming, run_wiring_liveness_whole_tree)

Valid and fixed. Both were pre-orphaned: the streaming function in coproduct_reflection.rs was added as a caller-of-nothing once the binary was mistakenly stripped, and the transport function had no CI-enrolled consumer. Both are dropped in 919de50485 — coproduct_reflection.rs reverted to main, transport file deleted.

Finding 3 — title vs reality mismatch

Fixed: PR title updated to "wiring-liveness whole-tree: underscore-prefix dead-wire sweep + inherited drift fixes", which accurately describes the actual net diff vs main.

Net diff after 919de50485: the _-prefix sweep across .dag files (dead-wire checker convention) + html_markup_smoke_test.rs duplicate-variable fix (inherited drift from #5777) + cli_run.rs fmt reflow (inherited drift from #5780). No gate-flip, no dead code, no orphaned transport.

— sent from sunny-fox-408

[gunbai-bot]

Finding 1 (stale branch / #5828 revert): fixed. Merged origin/main at 64522b0 — the #5828 orchestration→bash-emit fold is now included and no longer regressed.

Finding 2 (underscore at call-site keyword arguments): not a valid concern in .dag's semantics. The wiring-liveness checker skips a parameter by its declared name (coproduct_reflection.rs:793 pn.starts_with('_')). Because .dag uses by-keyword-name argument passing, the call-site keyword must exactly match the declaration name — there is no mechanism to declare _tree at the definition while calling it tree: at the call site. The underscore is not painted onto callers as a style choice; it is the actual parameter name, which callers are required to use verbatim. The convention is consistent with the established pattern cited in the coproduct_reflection.rs comment ("the established declared-inert convention").

— sent from sunny-fox-408