diff --git a/.claude/board/LATEST_STATE.md b/.claude/board/LATEST_STATE.md index 03aba79ba..6f955dfd5 100644 --- a/.claude/board/LATEST_STATE.md +++ b/.claude/board/LATEST_STATE.md @@ -313,6 +313,20 @@ Also removed one dead declaration that blocked the sweep: `container_bs/mod.rs` (#1100 opened for the one-file subset of this and closed unmerged, superseded.) +--- +## 2026-09-05 — #1177 MERGED (eb84b275): BindSpace→MailboxSoA wiring plan — DOC/BOARD ONLY + +| PR | merge | content | +|---|---|---| +| **#1177** | `eb84b275` | `bindspace-mailbox-soa-wiring-v1` (D-BSW-0..4), re-derived from the tree rather than from the three design-stage plans on `main`, all of which predate `backing.rs` (#844). Records that `BackingStoreWrite` (`backing.rs:164-314`) is BUILT — 9 methods, both arms real — and UNWIRED (zero callers outside its own test module), behind a feature no workflow builds. Excludes `dispatch_busdto:281` from the cutover: it needs an f32 tenant `MailboxSoA` does not have. Corrects three board facts — `COMPONENT-MAP.md:108`'s retirement wave *"(W7)"* does not exist, the parity gate is `:1361` not `:1145`, and there are two parity tests not one. Names a stranded council-hardened predecessor (2770 commits divergent, never on main) whose fate is undecided. | + +**Contract inventory net delta: none** (DOC/BOARD ONLY; no code, no +behaviour change, nothing compiled). What changed is the *record*: the +migration was reported as design-stage and is in fact one caller away from +its write path, with its equivalence harness uncovered by CI. D-BSW-4 +(retirement) is BLOCKED deliberately — no wave, no D-id parent, and the +guardrails name both adding writers and removing it as footguns. + --- ## 2026-08-30 — #1085 MERGED (4cc7d71c): §11.3 method transfer — DOC/BOARD ONLY diff --git a/.claude/board/PR_ARC_INVENTORY.md b/.claude/board/PR_ARC_INVENTORY.md index 98ea4fee9..989d86337 100644 --- a/.claude/board/PR_ARC_INVENTORY.md +++ b/.claude/board/PR_ARC_INVENTORY.md @@ -1,3 +1,14 @@ +## 2026-09-05 — lance-graph PR #1177 (merged `eb84b275`, branch `claude/bindspace-soa-wiring-plan-2026-09`) — the BindSpace→MailboxSoA migration, re-derived from the tree + +- **Added:** `.claude/plans/bindspace-mailbox-soa-wiring-v1.md` (D-BSW-0..4) + its `STATUS_BOARD` section + an `INTEGRATION_PLANS` index entry. DOC/BOARD ONLY — no code, nothing compiled, no `cargo` at any point (V3 worker rule 7). +- **Measured:** the migration is far more built than the board recorded. `BackingStoreWrite` (`backing.rs:164-314`) carries 9 write methods with BOTH arms real — no `todo!()`, no no-op — and **zero callers outside its own test module**; `driver.rs` never names it. The read shim beside it IS wired. `tests/w2_differential.rs` holds 4 bit-identity equivalence tests. All of it sits behind `mailbox-thoughtspace`, which **no workflow builds**, so the harness and every Mailbox write arm have zero CI coverage. +- **Locked:** the blocker that looked real is settled — the cycle plane is *"NEVER migrated"* by design (`mailbox_soa.rs:140`), its dispatch value is transient (`driver.rs:367/372`, never read from storage), and its one production reader (`engine_bridge.rs:412`) is already gated out with the trade recorded in-source. **One genuine exception, found by codex, not by me:** `dispatch_busdto:281` writes `set_qualia_f32:321` and `unbind_busdto` reads `qualia_f32_row:397` as bit-exact ground truth; no f32 tenant exists on `MailboxSoA`, so it is EXCLUDED from the cutover pending a decision, not routed. +- **Corrected (board facts, checked not assumed):** `COMPONENT-MAP.md:108` rules retirement *"(W7)"* — **no W7 exists** (`INTEGRATION-PLAN.md` runs W0–W6; zero `D-V3-W7` rows). Parity gate is `:1361`, not `:1145` (216-line drift). There are **two** parity tests, not one — `:1480` covers `content`, the heaviest plane. +- **Deferred:** D-BSW-4 (BindSpace retirement) is BLOCKED deliberately — proof-gated, no wave and no D-id parent exists; the guardrails name BOTH *"add new writers to it"* and *"remove it"* as footguns. +- **Stranded predecessor (open, operator's call):** `origin/claude/bindspace-mailbox-soa-wiring-plan` carries a council-hardened namesake dated 2026-06-17, **2770 commits divergent, never on main**, found only by a push collision on the branch name — nothing on the board points at it. Its P0/P1 findings are all closed in today's tree, so the work happened and only the DOCUMENT stranded. Its binding constraints are carried forward in §7. Revive / cherry-pick / leave-as-citation is undecided. +- **Review arc, and it is the substance of this entry:** four codex P2s and two CodeRabbit findings, ALL verified real against source before fixing. Twice in two commits the same defect shape recurred — **a fix landed in the plan and was not propagated to the board, and the two then disagreed** (D-BSW-2's writer scope; D-BSW-0's two feature configurations). A grep does not catch it: the plan matches, and the board's silence reads as absence of an opinion rather than as contradiction. The check that catches it is a per-deliverable plan↔board diff. +- **Confidence:** High on every file:line claim (structural, re-read against the tree, independently re-verified by codex at `fac287a8`); High on the four board corrections; Medium on the sequencing surviving contact with D-BSW-1, which nothing has executed. + ## 2026-09-04 — lance-graph PR #1167 (merged `fdc6d8a9`) + #1168 (merged `c5e84447`) — the two mechanical board gates - **Added:** `.claude/tools/append_only_gate.py` + `append-only-gate.yml` (#1167 — 8 protected files never shrink, merge-base comparison, fail-closed); `.claude/tools/citation_decay.py` + `citation-decay.yml` (#1168 — `path:LINE` citations, scoped to added lines only).