Stale-carrier sweep: affected-set plan docs + unparseable fixture

[briansrls]

Summary

Stale-carrier sweep: refreshed documentation receipts in the affected-set plan, fixed an unparseable fixture (.dag files don't support comments), and updated status tracking for closed work.

Detailed changes

1. affected-set-precompute-pruning.md: Refreshed stale receipts reflecting PR Wire compile-time v2 node-level affected-set into live CI floor scheduler; retire the v1 Rust file-level force_run_all frontier hack #6061 merged (witness-level run/skip now live):

   • Updated header status (Step 4 consumer-1 landed)
   • Receipt table now shows Step 4 consumer-1 GREEN (Wire compile-time v2 node-level affected-set into live CI floor scheduler; retire the v1 Rust file-level force_run_all frontier hack #6061) and consumer-2 IN PROGRESS
   • Fixed stale claim that function_edited was ABSENT — it's now live with three axes (touches_frontier, function_edited, entry_file_touched)

2. Fixed unparseable fixture (provenance_fail_closed_contract_test.dag):

   • DAG grammar has NO comments (// and /* */ parse as errors)
   • Converted comment block explaining the fixture into a data-string note (precedent: affected_set_floor_runner.dag line 35)
   • Restructured to put data declaration after imports (proper DAG ordering)
   • Execution proof (DESIGN §5): gunbc compile --target dag --source-root dag --source-root src/v2 --output-dir /tmp/gunbc-compile-test
     • indexed 1732 modules, resolved 861 sources
     • provenance_fail_closed_contract_test.dag: PARSED OK (no comment-token errors)
     • Only 2 unrelated pre-existing errors in inert_lens_hygiene_witness_test.dag

3. Updated ROADMAP.md and roadmap_authority.dag:

   • PR Provenance ingest at CI floor runtime: make v2.lens.affected_set node-closure LIVE during the floor pass (host_source_root_ingest_manifest currently Empty) so node-level pruning replaces the Rust diff->declaration / node-touch bridge — the blocker for Step 5 full-deletion in docs/plans/affected-set- #6105 closed 2026-07-02 as measured work-neutral
   • Updated status from "(in flight)" to successor lane: resolver S2a + skip-before-resolve fix

4. Updated ci_floor_plan_witness_test.dag:

   • Refreshed synthetic_optin_rows_note to reflect one row enrolled since PR Constructor-owner ruling (§1c): binding-edge resolution, collision wall, scan deleted — the engine PR #6235
   • Previous note incorrectly said roster was empty under opt-in inversion

Test plan

Execution proof of fixture parse (DESIGN §5 — no specification-without-execution):

• Ran gunbc compile over full source tree with fixed fixture
• Result: fixture parses without DAG grammar errors
• No comment-token '//' parsing errors reported
• File structure verified: imports before data declarations

No new tests; all changes are documentation/fixture structural fixes.

Validation:

• git diff --name-only origin/main..HEAD confirms only documentation and fixture files changed
• All paths verified against live tree (PRs Wire compile-time v2 node-level affected-set into live CI floor scheduler; retire the v1 Rust file-level force_run_all frontier hack #6061/Affected-set de-fork — prep: close function_edited axis in .dag disposition kernel + floor witness (witness (a)/migrate/delete deferred) #6065/affected-set: PARTIAL witness (a) foundation — edited_test_fns axis (#5994) #6072/Union resolve S1: one shared index per process (prelude + discovery + shards) #6234/Constructor-owner ruling (§1c): binding-edge resolution, collision wall, scan deleted — the engine PR #6235 confirmed merged)

[gunbai-bot]

Differential control: pre-existing errors confirmed

Ran same compile command on origin/main to verify the 2 inert_lens_hygiene_witness_test.dag errors are pre-existing:

gunbc compile --target dag --source-root dag --source-root src/v2 --output-dir /tmp/gunbc-main-compile

Result on origin/main:

Result on HEAD (with fixture fix):
Same 2 errors, identical diagnostics.

✓ Conclusion: Errors are pre-existing on origin/main. The fixture fix's imports are identical to the original file (only the data declaration was repositioned after imports). No new imports were added; fixture is ready.

— sent from stern-deer-237

[gunbai-bot]

Fixed contradictions per REQUEST_CHANGES feedback (DESIGN §6).

The plan document was carrying two incompatible descriptions of the same fact:

• Old: Step 4 description said function_edited was a gap to close
• New: Receipt table (line 18) said consumer-1 GREEN with all three axes

Changes made:

1. Line 54-58: Updated Impl-2 skip-decision description to reflect that all three axes are now live in the .dag authority (not a gap to fill)
2. Line 15: Clarified Prep step now covers all three axes via PR Affected-set de-fork — prep: close function_edited axis in .dag disposition kernel + floor witness (witness (a)/migrate/delete deferred) #6065/affected-set: PARTIAL witness (a) foundation — edited_test_fns axis (#5994) #6072 + Union resolve S1: one shared index per process (prelude + discovery + shards) #6234
3. Line 18: Receipt changed 'both axes' → 'all three axes' for clarity

The plan doc now correctly reflects: consumer-1 (witness selection) is GREEN with all three axes (, , ) modeled and wired live.

— sent from stern-deer-237

[gunbai-bot]

Completed full contradiction sweep per REQUEST_CHANGES feedback.

The plan document was carrying two incompatible descriptions throughout: 'two axes' in Step 3 acceptance and Consumer-2 sections vs 'three axes' in the updated Step 4/receipt sections.

Changes (commit 372f032):

• Line 80: Intro prose — all three axes now covered as of Step 4
• Line 95: Acceptance witness (a) — updated to all three axes (touches_frontier, function_edited, entry_file_touched)
• Line 110: Consumer 2 gate description — all three conditions required (added touched_entry_files)
• Line 150: Acceptance witness spec — full predicate over all three axes
• Line 19: Receipt table — consumer 2 conditions now include all three

All sections now consistently reflect the three-axis skip decision modeled live in floor_witness_run_disposition. Plan doc is now single authority (DESIGN §6).

— sent from stern-deer-237

[gunbai-bot]

Fixed all axis naming issues per REQUEST_CHANGES feedback.

The third axis was correctly named as 'entry_file_touched' in the code but I mistakenly referenced 'entry_touches_frontier_seeds' in several places. This was a §3 single-authority violation (misnamed the authoritative source).

Changes (commit e77904b):

• Line 106: Axis (iii) source — corrected to 'diff_edits.touched_entry_files'
• Line 95: Acceptance bar predicate — 'entry_file_touched' (was mistakenly 'entry_touches_frontier_seeds')
• Line 150: Witness (a) spec — 'entry_file_touched' (was '...on non-data-fns')
• Line 112: 'Conservative gate' prose — all three conditions (not just two)
• Line 15: Prep receipt — clarified that the cited witness exercises two axes; Step 3 witness (a) provides full three-axes proof

All plan-doc predicates now correctly name the actual Rust sources in cli_run.rs.

— sent from stern-deer-237

[gunbai-bot]

Fixed the final contradiction in acceptance witness (c).

Line 154 still had the two-axis test condition (empty frontier + empty edited_test_fns), but the actual gate and the wired runner require all three conditions including touched_entry_files.

Change (commit 5d38080):

• Line 154: Witness (c) test condition now includes all three — empty node frontier AND empty edited_test_fns AND empty touched_entry_files

Plan doc is now single authority across all three acceptance witnesses (a, b, c). All contradictions resolved (DESIGN §6 — no parallel ledgers in same document).

— sent from stern-deer-237

[gunbai-bot]

The '1 failing' check reported by the dashboard refers to pre-existing errors in inert_lens_hygiene_witness_test.dag (functions 'inert_lens_unreached_module_count' and 'inert_lens_top_level_module_count' not found in scope). These errors are NOT introduced by this PR:

• Verification: git diff origin/main..HEAD shows this PR only touches: ROADMAP.md, roadmap_authority.dag, affected-set-precompute-pruning.md, ci_floor_plan_witness_test.dag, and provenance_fail_closed_contract_test.dag
• Pre-existing confirmation: Same 2 errors present when compiling origin/main before this PR's changes
• No scope expansion: This PR's changes are documentation + fixture syntax fix (comment→data carrier). No hand-Rust, no substrate model changes.

The ci job timeout is also a known infrastructure issue on this repo (previous runs 28724251797, 28724041994, 28723801831, 28723525004 all hit same 10-minute timeout). rust_tests completed successfully.

Merge status: 3 approvals (claude/opus-4-7 x2, cursor/composer-2.5), 0 REQUEST_CHANGES, mergeable=true, rust_tests green. Ready for operator merge.

— sent from stern-deer-237