fix(render): correct foliage-wind classification, receiver/caster desync, and frame binding (Campaign VM VM6 review)

Opus dual-lens review of the three VM6 commits (0930c35d, 39e8408c,
6cc5e183) found two blockers and two should-fix issues; all landed here
along with the review's nits and documentation corrections.

Blockers:
- A1: the procedural-scenery classifier tested bit 31 alone instead of
  the full top nibble (0xF000_0000 == 0x8000_0000), so it also matched
  LandblockStaticEntityIdAllocator's 0xC... namespace (fences/gates/
  building shells with a cutout subset), the 0xDA11_D0xx paperdoll id,
  and the 0xFFFF_FF01 portal-tunnel id as procedural scenery — all
  three would have swayed. ProceduralSceneryIdAllocator.IsInNamespace
  now does the exact top-nibble test; FoliageWindClassification
  delegates to it.
- A2: GroupKey (the receiver's instance-batching key) did not carry
  FoliageFlags while the caster's dedup key already did, so a scenery
  instance and a non-scenery instance sharing a mesh subset coalesced
  into one receiver InstanceGroup whose flags were last-writer-wins —
  disagreeing with the correctly-keyed caster. GroupKey now carries
  FoliageFlags, computed before key construction and set exactly once
  at group creation; the imperative re-stamp is gone, and CachedBatch's
  now-redundant FoliageFlags field is removed.

Should-fix:
- A3: the world receiver pass bound UniformAtmosphericFrame only by
  accident (leftover from the caster pass, which runs first each
  frame, since Vulkan binding state isn't reset between passes).
  DirectionalShadowFrameBinding now carries the caster's exact
  AtmosphericFrameBufferBinding and BindDirectionalShadowReceiver binds
  it explicitly.
- A4: a Setup-composed tree's opaque trunk part never got the trunk
  flag because HasCutoutSubset is cached per GfxObj part, not per
  entity. FoliageWindClassification.ComputeEntityHasCutoutSubset now
  ORs HasCutoutSubset across an entity's resolved sibling parts once
  per entity, threaded into ClassifyBatches/AddDirectionalShadowBatches
  via a new optional override parameter.

Nits: A5 hashes the per-vertex flutter seed relative to the instance
origin instead of absolute world XY (fp32 sin() precision loss at far
landblock corners), mirrored in both foliage_wind.glsl and
FoliageWindModel; A7 documents the max(maxHeight, 0.5) divide-guard as
a deliberate pseudocode divergence; A8 switches FoliageWindExclusions'
construction to ToFrozenSet() and softens the "never stale" doc
comment to "no slower than one frame behind."

Tests added: top-nibble classification (0xFFFFFFFFu now correctly
false), GroupKey inequality across entity-driven scenery/landblock-
static classification, a caster-batch test proving the same pairing
never coalesces, ComputeEntityHasCutoutSubset unit + end-to-end
two-part-Setup tests, the caster→receiver AtmosphericFrame binding
carry-through, flutter-hash translation invariance relative to
instance origin, and a Storm-wind mid-height displacement floor
guarding against a "no motion" regression.

Docs: plan VM6 body corrected to the five-row WeatherKind table, "bits
1 and 2", "all four" caster shaders, and top-nibble wording throughout;
the owner gate checklist's Rain/Storm step; the stale v1-only shader-
interface compatibility entry; semantic-bindings-v1.md's v2 members
folded into the main 192-byte block; the IA-25 register row's top-
nibble wording; AtmosphericFrameInputs.cs's ABI size reference.

foliage_wind.glsl's A5 change recompiled exactly the five shaders that
include it (mesh_atmospheric.vert, the four directional_shadow_world_*
casters) plus the manifest; no other .spv changed.

Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,041/0 failed (no environment-specific failures this run).
RenderPackValidator 30/30. Full hermetic-filtered solution: 15,269/0
failed across 15 projects.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

docs/architecture/retail-divergence-register.md

@@ -62,7 +62,7 @@ accepted-divergence entries (#96, #49, #50).
 | IA-22 | **Filed 2026-08-13 (#391, user-directed: "we should only support modern resolutions. Not any old format").** The Config Resolution dropdown offers a CURATED list — the monitor's real mode enumeration filtered to modern widescreen families (16:9/16:10/21:9/32:9, ≥1280 wide, fitting the desktop; `DisplayModeCatalog.Curate`) — and its Defaults value is the desktop's own mode. Retail offered the adapter's complete enumeration including 4:3 legacy modes and authored `800x600` as the row default (`gmConfigUI::InitOptions SetDefaultValue(0x03200258)`; `gmClient::Init @0x004047af` `Device::ForceDisplayResolution(1, 0x320, 0x258)`). | `src/AcDream.App/Rendering/DisplayModeCatalog.cs`; `src/AcDream.App/UI/Layout/ConfigOptionsPageController.cs` (Resolution row); fixture fallback `src/AcDream.UI.Abstractions/Panels/Settings/DisplaySettings.cs` (`AvailableResolutions`, 800x600 removed) | Explicit product direction. **Amended 2026-08-16 (#407, Campaign CC gate round 1):** the dropdown now offers `DisplayModeCatalog.WindowedResolutions` — the curated hardware modes UNIONed with the static modern-ladder sizes that fit the desktop — because a WINDOWED pick is a plain Size write needing no video mode, and remote/RDP virtual displays advertise almost no modes (the live RDP display exposed only 1920x1080 + the 2056x1290 desktop, starving the dropdown). The original "an offered mode is supported by construction" invariant now holds for the FULLSCREEN half only: the fullscreen apply still validates against the hardware `Resolutions` list plus `GlfwDisplayModeSwitcher`'s monitor-mode-list hard guard, so a fullscreen pick of a windowed-only entry refuses safely (log-and-stay, #388; the #392 apply-result seam is that family's open follow-up) — "Graphics mode not supported" crashes remain unreachable from the dropdown. | A user wanting a genuine legacy 4:3 mode cannot pick it; retail-parity comparisons of the Config tab's list/default will show the deviation. | decomp sites in the Divergence column; ISSUES #391 |
 | IA-23 | **Filed 2026-08-17 at the night-round review fix round (F8).** `gmHouseUI::DisplayPurchaseTimeText @0x004a3110`'s not-yet-expired branch renders `"You may buy another landscape house at " + strftime("%c", localtime(timestamp + 0x278d00)) + ". This restriction does not apply to apartments."` — byte-decoded from raw pushed literals at `@0x004a3265`/`@0x004a321d`/`@0x004a3235` (all three text pieces confirmed; a prior filing had wrongly called this "unrecoverable"). This port renders the SAME three pieces, in the same order, with the same expiry-timestamp math, but formats the middle date/time piece with .NET's culture-default `DateTime.ToString()` (no explicit format string) rather than the C runtime's `strftime("%c", ...)` — the two engines do not share a format table, so the RENDERED SHAPE of the date/time differs (e.g. .NET's short numeric date+time vs the CRT's `Ddd Mon DD HH:MM:SS YYYY`-style locale string) even though both express "the process's own locale's full date+time" and use the SAME underlying instant (local time, matching retail's `localtime()`). | `src/AcDream.Runtime/Gameplay/RuntimeHouseState.cs` (`Recompute`'s not-expired branch) | Both are "whatever the process locale says" full date+time strings; no game-logic reads or parses this text back, it is pure chat-scroll presentation, so a differently-shaped (but equally legible) date string carries no functional risk | A retail-side-by-side visual comparison will show a differently formatted date/time (not a byte-identical `strftime("%c")` reproduction) — cosmetic only | `gmHouseUI::DisplayPurchaseTimeText @0x004a3110`; `strftime`/`localtime` CRT calls at `@0x004a322c`/`@0x004a3216` |
 | IA-24 | **Filed 2026-08-22, Campaign AR.** An explicitly selected atmospheric render pack adds cascaded real-time directional shadows from terrain, trees, buildings, players, monsters, and other retained outdoor casters. The one shadow direction follows the visible authored sun, then the dominant haloed moon (`0x01001F6A`), then the secondary moon (`0x01001F67`); a moon supplies direction only while colour/energy remains retail's single interpolated `SkyTimeOfDay.DirColor × DirBright` channel. Retail renders none of these real-time object-shadow maps and does not expose a second moon light. | `src/AcDream.App/Rendering/Packs/AuthoredCelestialShadowSource.cs`; `src/AcDream.App/Rendering/DirectionalSunShadowRenderer.cs`; pack-only receiver shaders; evidence `docs/research/2026-08-22-dereth-celestial-shadow-sources.md` | This is the user-requested headline graphics enhancement and is strictly opt-in. The retail path remains the default and authoritative fallback; pack-off does not build/select shadow work or change `SceneLighting`. One selected source reuses one cascade array, so moon support does not multiply shadow resources. | Pack-on output intentionally differs from retail. A wrong celestial identity/transform or stale source transition would visibly misalign shadows from the sky; pack-off output changing would violate the campaign's primary safety contract. | `SkyDesc::GetLighting @0x00500A80`; `GameSky::UseTime @0x005075B0`; installed Region `0x13000000`; cited research note |
-| IA-25 | **Filed 2026-08-22, Campaign VM VM6.** An explicitly selected atmospheric render pack sways procedural-scenery foliage (trees/bushes — entity ids in the `0x8XXYYIII` `ProceduralSceneryIdAllocator` namespace) in `mesh_atmospheric.vert` and the four `directional_shadow_world_*` caster vertex shaders, driven by a weather-table lean/branch/flutter vertex displacement (`foliage_wind.glsl`, `FoliageWindModel` CPU mirror) whose mean/gust strength is looked up per DAT-classified `AcDream.Core.World.WeatherKind` (Clear/Overcast/Rain/Snow/Storm — the same classification `WeatherState.cs` already derives from the active day group's authored name, not the day group's raw index, which carries no weather meaning by itself) and eases toward its target over `WeatherSystem.TransitionSeconds` (10 s) so a weather change never snaps. Retail's fixed-function renderer applies no per-vertex wind displacement to any scenery mesh — Dereth's trees are static geometry. Wind direction (`wind-direction-degrees`, default 225°) is a plain pack default: there is no authored retail wind direction to read (no wind data exists in retail at all). Render-only: `WorldPicker` picks the undisplaced mesh, so a swaying leaf can be up to `lean + branch` metres from its pick volume at the moment of a click; foliage subsets are cosmetic scenery, not interactable in retail either. | `src/AcDream.App/Rendering/Shaders/foliage_wind.glsl`; `src/AcDream.App/Rendering/Wb/FoliageWindClassification.cs`; `src/AcDream.App/Rendering/Packs/FoliageWindModel.cs`; `src/AcDream.App/Rendering/Packs/AtmosphericPostProcessGraph.cs` (`ResolveFoliageWind`); `src/AcDream.App/Rendering/Packs/BuiltInAtmosphericRenderPack.cs` (wind settings + `FoliageWindByWeather`) | Explicitly opt-in graphics enhancement — the retail path (`mesh_modern`, `terrain_modern`, `mesh_detail`) never reads `BatchData.flags` bits 1/2 and is pixel-identical with the pack off. The classification never touches Runtime/Core physics — the collision BSP is the trunk, and picking against the undisplaced mesh has no gameplay consequence since foliage is not interactable. | Pack-on output intentionally differs from retail (moving foliage where retail has none). A wrong classification bit would sway a non-foliage object or leave a real tree still; a caster/receiver clock or amplitude mismatch would visibly misalign a leaf's shadow from the leaf itself. Pack-off output changing would violate the campaign's primary safety contract. | None — retail applies no vertex wind displacement to any geometry; `ProceduralSceneryIdAllocator` (bit 31 entity-id namespace, existing acdream mechanism, not retail) |
+| IA-25 | **Filed 2026-08-22, Campaign VM VM6.** An explicitly selected atmospheric render pack sways procedural-scenery foliage (trees/bushes — entity ids in the `0x8XXYYIII` `ProceduralSceneryIdAllocator` namespace) in `mesh_atmospheric.vert` and the four `directional_shadow_world_*` caster vertex shaders, driven by a weather-table lean/branch/flutter vertex displacement (`foliage_wind.glsl`, `FoliageWindModel` CPU mirror) whose mean/gust strength is looked up per DAT-classified `AcDream.Core.World.WeatherKind` (Clear/Overcast/Rain/Snow/Storm — the same classification `WeatherState.cs` already derives from the active day group's authored name, not the day group's raw index, which carries no weather meaning by itself) and eases toward its target over `WeatherSystem.TransitionSeconds` (10 s) so a weather change never snaps. Retail's fixed-function renderer applies no per-vertex wind displacement to any scenery mesh — Dereth's trees are static geometry. Wind direction (`wind-direction-degrees`, default 225°) is a plain pack default: there is no authored retail wind direction to read (no wind data exists in retail at all). Render-only: `WorldPicker` picks the undisplaced mesh, so a swaying leaf can be up to `lean + branch` metres from its pick volume at the moment of a click; foliage subsets are cosmetic scenery, not interactable in retail either. | `src/AcDream.App/Rendering/Shaders/foliage_wind.glsl`; `src/AcDream.App/Rendering/Wb/FoliageWindClassification.cs`; `src/AcDream.App/Rendering/Packs/FoliageWindModel.cs`; `src/AcDream.App/Rendering/Packs/AtmosphericPostProcessGraph.cs` (`ResolveFoliageWind`); `src/AcDream.App/Rendering/Packs/BuiltInAtmosphericRenderPack.cs` (wind settings + `FoliageWindByWeather`) | Explicitly opt-in graphics enhancement — the retail path (`mesh_modern`, `terrain_modern`, `mesh_detail`) never reads `BatchData.flags` bits 1/2 and is pixel-identical with the pack off. The classification never touches Runtime/Core physics — the collision BSP is the trunk, and picking against the undisplaced mesh has no gameplay consequence since foliage is not interactable. | Pack-on output intentionally differs from retail (moving foliage where retail has none). A wrong classification bit would sway a non-foliage object or leave a real tree still; a caster/receiver clock or amplitude mismatch would visibly misalign a leaf's shadow from the leaf itself. Pack-off output changing would violate the campaign's primary safety contract. | None — retail applies no vertex wind displacement to any geometry; `ProceduralSceneryIdAllocator` (top-nibble-0x8 entity-id namespace, existing acdream mechanism, not retail) |
 
 ---