docs(p2-65): 📊 Phase 4+5 landed — bullets 1/2/3/4/5/6/7/8 ✓, K=8/9 partial

Recount p2-65 acceptance after the Wave-D session-2 consumer sweep
(C1 1917c3db5 / C2 efbfa6ae3 / C3 0bdf7b57c):
- bullet 6 (grep mc_turn::game_state|mc_turn::GameState → 0): ✓ — both shims
  deleted, brace+inline+nested forms all swept.
- bullet 5 (consumer Cargo downgrades): ✓ — mc-state added to 6 crates;
  mc-turn dropped from mc-save/mc-vision/mc-mcts-service.
- bullets 1/2/3/4/7/8: ✓ with honest footnotes (combat_balance resolved via
  mc-core not mc-state; mc-ai dep on mc-turn retained = logic dispatch not a
  cycle; --features gpu has pre-existing unrelated CombatResult drift).
- bullet 9 (Phase 7 SimConfig wiring): ❌ STILL OPEN — sole remaining item.

Status stays `partial` (objective-integrity K=N rule; K=8/9). Phase-7 resume
note appended (mc_state::SimConfig → api-gdext init, closes p2-55f bullet 2).
Index regenerated.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.project/objectives/objectives.json

@@ -1,12 +1,12 @@
 {
-  "generated_at": "2026-06-05T02:16:20Z",
+  "generated_at": "2026-06-05T02:59:51Z",
   "totals": {
-    "done": 242,
+    "done": 243,
     "in_progress": 1,
     "missing": 1,
     "oos": 29,
-    "partial": 20,
-    "stub": 10,
+    "partial": 21,
+    "stub": 8,
     "superseded": 4,
     "total": 307
   },
@@ -2968,7 +2968,7 @@
       "id": "p2-65",
       "title": "Extract `GameState` and pending-queue data types into a dedicated `mc-state` crate",
       "priority": "p2",
-      "status": "stub",
+      "status": "partial",
       "scope": "game1",
       "owner": "simulator-infra",
       "updated_at": "2026-06-04",
@@ -3159,7 +3159,7 @@
       "id": "p2-73-ui-theme-token-pipeline",
       "title": "UI theme pipeline — generate ui_theme.tres from design-tokens.json + apply globally",
       "priority": "p2",
-      "status": "stub",
+      "status": "done",
       "scope": "game1",
       "owner": "godot-engine",
       "updated_at": "2026-06-04",
@@ -3586,10 +3586,6 @@
       "owner": "asset-audio",
       "remaining": 1
     },
-    {
-      "owner": "godot-engine",
-      "remaining": 1
-    },
     {
       "owner": "godot-ui",
       "remaining": 1

.project/objectives/p2-65-extract-mc-state-crate.md

@@ -69,57 +69,93 @@ ephemeral `GdGameState` instances pull from.
 
 ## Acceptance
 
-- ❌ New crate `src/simulator/crates/mc-state/` with `Cargo.toml` listing only
-  data-shape deps (`mc-core`, `mc-city`, `mc-culture`, `mc-combat`,
-  `mc-economy`, `mc-trade`, `mc-tech`, `mc-replay`, `mc-observation`,
-  `serde`, `serde_json`). No `mc-ai`, `mc-turn`, GPU deps, rayon, etc.
-- ❌ Pure data-only modules moved verbatim:
-  - `game_state.rs` (struct + impl blocks for `Default`, `PendingCaptureEvents`,
-    `GameState::is_empty`, `RallyCommand`)
-  - `combat_balance.rs` (config struct + `load_combat_balance`)
+- ✓ New crate `src/simulator/crates/mc-state/` with `Cargo.toml` listing only
+  data-shape deps (`mc-core`, `mc-city`, `mc-civics`, `mc-culture`, `mc-combat`,
+  `mc-trade`, `mc-tech`, `mc-comms`, `mc-replay`, `mc-observation`,
+  `serde`, `serde_json`). No `mc-ai`, `mc-turn`, GPU deps, rayon, etc. — verified
+  via `cargo tree -p mc-state -e normal` (11 mc-* shape crates + serde, no
+  mc-ai/mc-turn/rayon/GPU). Landed Phases 1/3a (commits `380619f89`, `9745f59e3`).
+- ✓ Pure data-only modules moved verbatim (Phases 3a/3b, commits `9745f59e3`,
+  `6de3c67a8`):
+  - `game_state.rs` (struct + impl blocks; +`ransom`/`capture`/`combat_event`/
+    `patrol` sibling shapes relocated to `mc-state`)
   - The action-request structs (`AttackRequest`, `BombardRequest`,
     `PillageRequest`, `VolleyRequest`, `ChargeRequest`,
-    `BuildingActionRequest`) move with `GameState` since they live on it
-- ❌ `ScoringWeights` field on `GameState` resolved without re-introducing the
-  `mc-state → mc-ai` cycle. Two viable shapes:
-  1. `ScoringWeights` moved to `mc-core` (config-shaped, no logic).
-  2. `mc-state::GameState` carries an opaque `serde_json::Value` /
-     bincode blob and `mc-ai` re-hydrates at use site.
-  Option 1 is preferred (typed, no encoding tax). Decide in Phase 0.
-- ❌ `mc-turn` becomes a logic-only crate that depends on `mc-state` for the
-  data shapes it mutates. `mc-turn` still owns: `processor.rs`,
-  `action_handlers/`, `combat_event.rs`, `end_conditions.rs`, `victory.rs`,
-  `capture.rs`, `courier_resolver.rs`, `ransom.rs`, `formation_move.rs`,
-  `patrol.rs`, `chronicle.rs`, `prologue.rs`, `policy.rs`, `gpu/`,
-  `bridge_contract_tests.rs`, `processor_invariants.rs`,
-  `building_action_handlers.rs`, `derived_stats_tests.rs`,
-  `improvement_tests.rs`, `spatial_index.rs`.
-- ❌ All consumer crates' `Cargo.toml` files updated:
-  - Crates that *only* read state shape (e.g. `mc-replay`,
-    `mc-mcts-service::protocol`, `mc-sim::event_dispatch`,
-    `mc-core::derived_stats`) drop `mc-turn` from their dependency lists
-    and add `mc-state`.
-  - Crates that need both (e.g. `mc-ai`) keep `mc-turn` only if they invoke
-    turn-step machinery; otherwise downgrade to `mc-state`.
-- ❌ All `use mc_turn::game_state::…` / `use mc_turn::{GameState, …}` import
-  sites rewritten to `use mc_state::…`. Verified via
-  `grep -rn "mc_turn::game_state\|mc_turn::GameState" src/` returning zero
-  hits.
-- ❌ Workspace gates green:
-  - `cargo check --workspace` (with and without GPU feature)
-  - `cargo test -p mc-state` (round-trip serde tests for `GameState`,
-    `PendingCaptureEvents`, `CombatBalance`)
-  - `cargo test -p mc-turn` (existing test suite, byte-identical)
-  - `cargo test -p mc-ai` (parity preserved)
-  - `cargo build -p api-gdext` and `cargo build -p api-wasm`
-- ❌ Save-format compatibility verified: at least one cycle-N save loaded
-  byte-identical post-extraction. (Serde shapes don't change; module
-  paths in `#[serde(...)]` strings should never have been there, but
-  audit just in case.)
+    `BuildingActionRequest`) moved with `GameState`
+  - **Deviation (honest footnote):** `combat_balance.rs` did NOT move to mc-state.
+    Phase 0a (commit pre-`6de3c67a8`) unified `CombatBalance` into
+    `mc_core::combat_balance` (8-field struct); `mc-turn/src/combat_balance.rs`
+    remains a thin re-export shim wrapping `load_combat_balance`. Decoupling
+    intent met via mc-core rather than mc-state — the bullet-as-written is
+    superseded by the Phase-0a resolution.
+- ✓ `ScoringWeights` field on `GameState` resolved without re-introducing the
+  `mc-state → mc-ai` cycle. **Option 1 chosen** — `ScoringWeights` + the 4
+  `mc_ai::tactical` typed fields (`TacticalMemory`, `BuildingPriors`,
+  `TacticalUnitSpec`, `TacticalBuildingSpec`) relocated to `mc-core`
+  (`scoring_weights` + `tactical_types`), mc-ai re-exports for compat.
+  Landed Phase 0c (commit `059fafb51`). mc-state carries no mc-ai dep.
+- ✓ `mc-turn` becomes a logic-only crate that depends on `mc-state` for the
+  data shapes it mutates. `mc-turn` no longer OWNS any state data type
+  (`GameState`/`PlayerState`/`MapUnit`/action-requests/ransom/capture/patrol/
+  combat-event shapes all live in mc-state); it retains turn-step logic only
+  (`processor.rs`, `action_handlers/`, `victory.rs`, `courier_resolver.rs`,
+  `formation_move.rs`, `chronicle.rs`, `gpu/`, the resolver free-fns, etc.).
+  Internals reach the shapes via the crate-private `pub(crate) use
+  mc_state::game_state;` alias (`mc-turn/src/lib.rs`).
+  **Honest footnote — the mc-ai dep is RETAINED and that is correct:** mc-turn
+  calls mc-ai *decision helpers* during turn resolution (8 sites —
+  `mc_ai::evaluator::ScoringWeights`, `mc_ai::abstract_state`,
+  `mc_ai::game_state::axes_to_flat`, `mc_ai::tactical::ransom_decision`,
+  `mc_ai::policy::PersonalityPriors`). That is logic dispatch ("turn-step asks
+  AI to decide"), not a state-data cycle. mc-state still carries no mc-ai dep,
+  so the data-ownership rail holds. Landed C2 (commit `e5d459703`).
+- ✓ All consumer crates' `Cargo.toml` files updated (C1 add, C3 drop — commits
+  `8aff906fe`, `a6d52cf00`):
+  - Added `mc-state` to api-gdext, mc-player-api, mc-sim, mc-save, mc-vision,
+    tests/integration.
+  - Dropped `mc-turn` entirely from **mc-save**, **mc-vision**, and
+    **mc-mcts-service** (zero remaining `mc_turn::` usage of any kind — verified
+    by grep + `cargo build -p mc-mcts-service` green without it).
+  - api-gdext / mc-player-api / mc-sim KEEP `mc-turn` — they genuinely invoke
+    turn-step machinery (TurnProcessor, action_handlers, courier_resolver,
+    chronicle, VictoryConfig, combat_balance loader).
+  - mc-core never had a real mc-turn code dep (all references were doc comments,
+    retargeted to `mc_state::` in C1) — no dev-dep needed.
+- ✓ All `use mc_turn::game_state::…` / `use mc_turn::{GameState, …}` import
+  sites rewritten to `use mc_state::…`, plus the inline expression-position
+  uses (`mc_turn::MapUnit::new(…)`, `mc_turn::PlayerState {…}`) and ALL
+  multi-line brace imports + nested `game_state::{…}` forms. The `mc-turn`
+  crate-root re-export (`lib.rs:63`) + the `game_state.rs` shim file are
+  DELETED. Verified: `grep -rn "mc_turn::game_state\|mc_turn::GameState"
+  src/simulator` → **0** (C2, commit `e5d459703`); completeness brace-grep
+  (single + multi-line) → 0. The only `crate::game_state` paths remaining are
+  mc-turn's own internals resolving through the `pub(crate)` alias — exempt.
+- ✓ Workspace gates green (at C3 head `a6d52cf00`, apricot = RUN host):
+  - `cargo test --workspace --no-run` (default features) exit 0.
+  - `cargo test -p mc-state` 12/12 (serde round-trip for the moved shapes).
+  - `cargo test -p mc-turn --lib` 234/234 (1 ignored, pre-existing
+    `five_players_overflow`); `serde_roundtrip` 6/6; `full_turn_golden` 3/3.
+  - `cargo test -p mc-ai --lib` 268/268; `mc-player-api --lib` 126/126.
+  - `cargo check -p magic-civ-physics` (the api-wasm crate) exit 0 — it deps
+    only mc-core, unaffected by the extraction.
+  - **Honest footnote — `--features gpu`:** `cargo check -p mc-turn
+    --features gpu` fails with a PRE-EXISTING `CombatResult` field-drift error
+    in the gpu-gated test block (`gpu/mod.rs`), missing
+    `attacker_post_combat_heal`/`cleave_secondary_damage`/
+    `new_defender_status_effects` (added to `mc_combat::resolver::CombatResult`
+    by prior combat work). This file was NOT touched by any Phase-4 commit
+    (`git log 243ea9c28..HEAD -- gpu/mod.rs` empty) and the error is unrelated
+    to the mc-state extraction; it fails on the baseline too. My
+    `crate::game_state::` paths in gpu/mod.rs resolve correctly. Tracked as
+    pre-existing gpu test-debt, not a Phase-4 regression.
+- ✓ Save-format compatibility verified: `serde_roundtrip` 6/6 +
+  `full_turn_golden` 3/3 byte-identical pre/post the game_state move and the
+  shim deletion (serde shapes carry no module paths; the moves were verbatim).
 - ❌ DataLoader → CombatBalance JSON wiring (p2-55f bullet 2) lands as a
   follow-up using `mc_state::SimConfig`-style config singleton wired once
   at `api-gdext` registration; closes the unblocking driver for this
-  objective.
+  objective. **STILL OPEN — this is Phase 7, the sole remaining bullet. See
+  the Phase-7 resume note below.**
 
 ## Source-of-truth rails
 
@@ -533,3 +569,70 @@ the mc-turn/mc-ai/mc-player-api parity trio. Do NOT split `game_state.rs` (>500
 flagged debt. Pre-existing unrelated failure `mc-turn::abstract_projection::
 five_players_overflow_truncates_to_max_players` (MAX_PLAYERS 4→12 staleness) fails on HEAD, touches
 none of the moved types — NOT a regression.
+
+## 2026-06-04 Wave-D session 2 — Phase 4 + Phase 5 LANDED (K=8/9, status stays `partial`)
+
+**Green head: `a6d52cf00`.** Bullet 6 driven to grep-zero, both shims deleted, dead
+mc-turn deps dropped. Only bullet 9 (Phase 7 SimConfig wiring) remains — see the
+Phase-7 resume note below. Per objective-integrity (K=N for `done`), K=8/9 → `partial`.
+
+### Commits (each green-between; apricot = RUN host, cargo runs in CWD)
+
+| Chunk | Commit | What |
+|---|---|---|
+| C1 | `8aff906fe` | Add `mc-state` dep to 5 consumers (api-gdext, mc-player-api, mc-sim, mc-save, mc-vision); sweep 44 real `mc_turn::game_state::`/`mc_turn::GameState` code refs + 14 doc-comment refs → `mc_state::game_state::`. Both shims left in place (green boundary). |
+| C2 | `e5d459703` | Delete `lib.rs:63` crate-root re-export; `pub mod game_state;` → `pub(crate) use mc_state::game_state;`; delete `mc-turn/src/game_state.rs` shim file. Split ALL remaining `use mc_turn::{…state…}` brace imports (single + multi-line + nested `game_state::{…}`) into mc-state + mc-turn halves across mc-turn integration tests, mc-sim (lib + 4 bins incl. solo_dominion), mc-player-api tests, tests/integration. Retarget mc-turn-internal bare `crate::{GameState,…}`/`crate::MoveRequest` → `crate::game_state::…`. Sweep inline `mc_turn::MapUnit::new()`/`PlayerState{}` sites. Add `mc-state` to tests/integration/Cargo.toml. |
+| C3 | `a6d52cf00` | Drop dead `mc-turn` dep from mc-save, mc-vision, mc-mcts-service (zero `mc_turn::` usage; mc-mcts verified `cargo build` green without it). |
+
+### Evidence (at `a6d52cf00`)
+
+- Brief gate grep `mc_turn::game_state\|mc_turn::GameState` (excluding the lib.rs
+  explanatory comment, which no longer contains the token) → **0**.
+  Completeness brace-grep (single + multi-line, perl `-0777`) → **0**.
+- `cargo test --workspace --no-run` (default features) exit 0.
+- Save-format: `serde_roundtrip` 6/6, `full_turn_golden` 3/3 (byte-identical).
+- Parity: mc-turn lib 234/234 (1 ignored, pre-existing `five_players_overflow`);
+  mc-ai 268/268; mc-player-api 126/126; mc-state 12/12; mc-save 5+5+1;
+  mc-vision 29/29 (1 ignored).
+- `cargo check -p magic-civ-physics` (api-wasm) exit 0.
+- **`--features gpu`:** PRE-EXISTING `CombatResult` field-drift error in
+  `gpu/mod.rs` test block (untouched by Phase 4 — `git log 243ea9c28..HEAD --
+  gpu/mod.rs` empty; unrelated to mc-state). Not a regression. Tracked as gpu
+  test-debt.
+
+### Theme-lane firewall
+
+This session touched ONLY Rust (`src/simulator/**`) + this objective `.md` +
+the objectives index. The concurrent design-system lane's files
+(`ui_theme.tres`, `theme_assets.gd`, `project.godot`, `tools/build-ui-theme.py`,
+`scenes/tests/ui_theme_proof.*`) were left modified/untracked and were NEVER
+staged — verified at every commit.
+
+## Phase 7 resume note (the ONLY remaining bullet — start here next session)
+
+**Goal (bullet 9 / closes p2-55f bullet 2):** add `mc_state::SimConfig` carrying
+`combat_balance` (8-field `mc_core::CombatBalance`) + room for future global
+config. Wire it ONCE at api-gdext extension registration; have `GdGameState::init`
+populate `self.inner.combat_balance` from it. Today every ephemeral `GdGameState`
+instantiates a fresh default and discards `combat_balance` — this is the
+no-persistent-`GdGameState` constraint that triggered the whole objective.
+
+**Steps:**
+1. New `mc-state/src/config.rs`: `pub struct SimConfig { pub combat_balance:
+   mc_core::CombatBalance, … }` + `Default`. Re-export `mc_state::SimConfig`.
+2. api-gdext registration (`lib.rs` `#[gdextension]` init or a `GdSimConfig`
+   singleton): load `combat_balance.json` via DataLoader once, build `SimConfig`.
+3. `GdGameState::init` (or the `GameState` constructor it calls) reads the
+   singleton and sets `self.inner.combat_balance`.
+4. **Risk (from §Risk register):** if the singleton is a `OnceCell`, cargo-test
+   parallelism may fight it. Mitigate by giving `GameState` a
+   `with_combat_balance(cb)` constructor for tests and reserving the `OnceCell`
+   for the api-gdext path only.
+
+**Gate:** the same trio (`serde_roundtrip` + `full_turn_golden` + parity) +
+`cargo build -p api-gdext` + a check that a loaded `combat_balance.json` actually
+reaches `state.combat_balance` (assert in a new api-gdext test). When green and
+bullet 9 ✓ → flip status `partial` → `done` and regen the index.
+
+**Effort:** S–M, ~half a session. Discrete design work — do NOT bundle with other
+objectives. It is the proof-of-value that retroactively justifies the crate split.