The server's authoritative answers to a raise were dropped on the floor:
only the vitals pair (0x02E7/0x02E9) had parsers, so after any
RaiseAttribute/RaiseSkill/TrainSkill the client's stat model stayed
frozen at login's PlayerDescription — the root cause of #431's stale
derived skills and run speed. The GUI looked alive only because the
panel applies optimistic local raises.
New parsers with three-source-verified layouts (CA1 research doc §2.5/
§2.8): PrivateUpdateAttribute (0x02E3) and PrivateUpdateSkill (0x02DD —
the wire's ushort ranks + hardcoded adjustPP=1 pair and f64
lastUsedTime preserved exactly). WorldSession dispatches both as typed
events; LiveSessionEventRouter routes them into the J4 character owner's
LocalPlayerState like every other private update. The vestigial
PrivateUpdateSkillLevel (0x02DF) is deliberately unparsed — ACE has no
producer (verified).
OnAttributeUpdate now fans out to the derived-value observers, mirroring
retail's live-at-inquiry model (CACQualities::InqSkill 0x00592660 —
Set* writes raw, Inq* recomputes, notification carries no value): an
Endurance write notifies the Health AND Stamina vital observers (ACE
pushes only a Health record and its own comment says the client must
refresh both), Self notifies Mana, and every attribute write notifies
character-sheet consumers whose formula contributions just changed.
OnSkillWireUpdate preserves the login FormulaBonus — the wire record
carries no attribute contribution; CA3 replaces the cached field with
the live computation.
Also corrected while in the neighborhood: PropertyString.cs's comment
claimed opcode 0x02DD for PrivateUpdatePropertyString; ACE's enum says
0x02D5/0x02D6 (doc-only — nothing dispatched on either).
Conformance tests cover both layouts (including holtburger's golden
skill fixture with adjustPP=1), truncation/wrong-opcode rejection, the
Endurance/Self/Quickness fan-out contract, and FormulaBonus
preservation. Full hermetic suite 15,333 passed / 0 failed (one
load-sensitive transport flake observed on the first run, passed alone
and on the clean re-run — filed as #439 rather than chased).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two parallel research passes assembled with a hand-verification ledger on
every load-bearing claim.
Wire (ACE + Chorizite + holtburger, field-for-field agreement on all six
inbound layouts): PrivateUpdateAttribute 0x02E3, PrivateUpdateVital 0x02E7
(always Max-family vital ids 1/3/5), PrivateUpdateAttribute2ndLevel 0x02E9
(always current-family ids 2/4/6 — a parser must NOT treat the two as one
id space), PrivateUpdateSkill 0x02DD (ushort ranks + the hardcoded
adjustPP=1 pair, f64 lastUsedTime), PrivateUpdatePropertyInt 0x02CD
(AvailableSkillCredits=24) and Int64 0x02CF (AvailableExperience=2).
Ordered action->response chains for all four raise/train actions,
including the retail quirk that an Endurance raise pushes only a HEALTH
full-vital record and the client is expected to refresh stamina from it
too. Specialize/untrain/reset have NO dedicated opcode — item-Use plus a
confirmation round-trip reusing the same update messages. 0x02DF has no
ACE producer (verified); CA2 skips it.
Recompute (named-retail + live Ghidra): retail computes skills, vitals
maxima and run rate LIVE at inquiry time — Set* are raw-storage writes,
InqSkill re-derives from the attribute formula every call (verified in
the decompile, including the z==0 early-out that IS the attribute-less
Salvaging handling and the +10 augmentation adds), InqRunRate runs every
motion tick, and UI refresh is a value-less observer notification.
Two corrections to our own tree surfaced: PropertyString.cs's comment
claims 0x02DD (it is 0x02D5 — doc-only, nothing dispatches on it), and
SkillSnapshot.FormulaBonus is frozen at PlayerDescription parse — the
stale-cache half of #431 that CA3 replaces with the live computation.
RetailSkillFormula.TryCalculate already ports 0x00591960 exactly, so CA3
reuses it rather than porting anew.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The owner widened #431 into the full advancement family: real-time
refresh of vitals maxima (attribute AND direct vital raises), derived
skills, run speed under Quickness, attribute-less skills like Salvaging,
and the untested train/specialize/respec flows. The promotion survey
pinned the root cause: every outbound raise action (0x0044-0x0047) is
wired — which is why the GUI 'works' — while the inbound private
attribute/skill update family is parsed nowhere (only the vitals pair
0x02E7/0x02E9 is), so the server's post-raise truth never reaches
LocalPlayerState and no recompute ever triggers. Plan doc carries the
oracle targets (message family from ACE/Chorizite/holtburger, retail's
recompute chain in named-retail, specialization/respec semantics) and
five slices ending in a user-driven connected gate. #430 tooltips are
explicitly sequenced after, on the #409 tooltip system.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
#438 records the design agreed with the owner: launcher-owned opt-in WER
LocalDumps key (HKCU, minidump, capped count), crash bundle assembled on
the next launch from the dump + log tail + version + capability report,
and an explicit NO-auto-upload line — dumps can hold the plaintext
session password, so sharing stays a user action until there is real
infrastructure and a consent flow. The owner's own machine is already
armed manually for the #422 hunt; this productizes it for alpha users.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Attacking with no valid target has told the player nothing since Campaign V
slice V11 orphaned the DebugVM toast the message was wired to (#434 found
the drop; this closes it retail-faithfully).
Ground truth from the Ghidra decompile of
ClientCombatSystem::ExecuteAttack (0x0056bb70): retail writes
"You must select a valid combat target before attacking" via
ClientSystem::AddTextToScroll(..., 0x1A, true, 0) — the ClientLocal
SpewBox channel this codebase already routes every other client-local
refusal through. And retail has ONE message, not the two we carried:
attacking outside melee/missile modes is silent (ExecuteAttack is
unreachable there), so the invented "Enter melee or missile combat first"
text is deleted rather than rerouted, and the invented "No monster
target" is replaced by the retail string, which joins ClientTextRefusals
with its decomp citation.
Wiring: CombatFeedbackSlot gains the sibling BindOwned session-lifetime
shape, and SessionPlayerComposition.CompleteSessionPlayer binds it to
RuntimeCommunicationState.AddText(ClientLocal) with session-owned
teardown — a torn-down session's slot returns to its silent unbound
state. A binding-seam test
(CompleteSessionPlayerBindsCombatFeedbackToTheClientLocalSpewBoxRoute)
inspects the compiled composition for the BindOwned call and its
AddText-routing lambda, so the slot can never again pass its unit tests
while production leaves it unbound — the exact failure mode that hid
this defect. The two tests that pinned the invented strings now pin the
retail contract (exact string; silence for the unsupported-mode case).
Full hermetic suite 15,325 passed / 0 failed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The part-2 commit message and the callout said 40; the 8 reclassified
rows had left the Temporary table, so the true count is 31 (and the
previous commit's '40 temporary probes remain' line is corrected by this
note).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Part 1 deleted probes whose owning issues were closed. These 14 named no
issue at all, so each was traced to its introducing commit
(git log -S) instead of guessed at. Attribution split them three ways:
DELETED (7, investigations closed): ACDREAM_A8_DUMP_PV and
ACDREAM_DUMP_LIVE_SPAWNS (Phase A8), ACDREAM_DUMP_CLOTHING (#37),
ACDREAM_DUMP_EDGE_SLIDE (#32), ACDREAM_DUMP_STEPUP (L.2.3d-f),
ACDREAM_DUMP_VENDOR (the vendor campaign, 25 call sites across 8 files),
ACDREAM_DUMP_VITALS (#5, four independent read sites). VendorDiagnostics.cs
went entirely.
RECLASSIFIED (8, tools misfiled as probes): the DUMP_CELLS/DUMP_GFXOBJS
fixture-extraction family (replay-harness tooling with a roundtrip test),
PROBE_CELL (standing cell-transit tracer, pair of the permanent
PROBE_RESOLVE), DUMP_SKY and HIDE_PART (generic isolation tools), and
DUMP_STEEP_ROOF — which looked like an L.4 relic but observes LIVE
divergence-register row AD-56; deleting it would have removed the only
runtime lens on an active divergence. All moved to Permanent diagnostics
with their attribution recorded.
RESTORED (1): ACDREAM_DUMP_MOVE_TRUTH was deleted and un-deleted the same
day. It is not a probe — the canonical nine-stop soak
(run-connected-r6-soak.ps1) hard-fails every destination without its
'move-truth OUT' records, with a message that would misdirect the next
operator. Under the no-workarounds rule the gate's mechanism is restored,
not left broken with an IOU (#437, closed). Process lesson recorded on
both issues: a closed owning issue is NOT sufficient to delete a probe —
grep tools/ and the contract tests for consumers first.
Also lands the owner-requested default-off invariant: every diagnostic in
the codebase is inert until its env var is explicitly set. Exactly four
flags default ON and none is a diagnostic — RETAIL_CHASE, CAMERA_COLLIDE,
CAMERA_ALIGN_SLOPE, RETAIL_CLOSE_DEGRADES are retail behaviors wearing an
A/B off-switch. That set is now FROZEN by
LaunchOptionsDocumentationTests.OnlyTheFourRetailBehaviorFlagsDefaultOn;
docs/launch-options.md's Conventions and CLAUDE.md state the rule, and
CLAUDE.md now binds future probes to a documented row in the same commit.
The client reads 137 environment variables (161 at audit start); 40
temporary probes remain, every one attributed. Full hermetic suite 15,322
passed / 0 failed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Each of these was temporary apparatus added to chase one bug, and each was
supposed to be deleted in the commit that fixed it. Fourteen closed issues
later they were still here: #337's support/wire-mesh trio, #171's sticky
timeline, #119's viewer and entity dumps, #113's phantom probe, and a dozen
more. 3,493 lines removed; the client now reads 144 environment variables
instead of 161, and 47 temporary probes remain instead of 64.
This is not only tidying. Every probe leaves a branch on its hot path when
unset, several re-read the environment per call rather than caching, and
the volume buries the diagnostics that are actually load-bearing. It is
also a headless correctness matter: HeadlessStaticStateAudit reflects over
PhysicsDiagnostics' flags to refuse a multi-session host when any is set,
and cannot see probes that live outside that owner.
Four files went entirely — WalkMissDiagnostic.cs, CollisionMeshWireframe.cs
and two test files whose only subject was a deleted probe.
TransitionTypes.SetContactPlane also sheds its CallerMemberName /
CallerLineNumber parameters, which existed solely for #337's cpSrc=
attribution and carried the instruction to strip them with the probe
family; no call site passed them, so no behavior changes. F2's collision
overlay survives and reverts to its proxy-cylinder form, which is what
removing the ACDREAM_WIRE_MESH upgrade means.
LaunchOptionsDocumentationTests earned its keep here: it refused the
deletion until docs/launch-options.md moved the 17 rows into Retired and
the frozen direct-read counts came down (PhysicsEngine.cs to zero,
TransitionTypes.cs 3 to 2). The documentation could not drift during a
cleanup this wide.
The 14 probes that name no owning issue are deliberately NOT deleted.
Nothing records when they became safe to remove, and guessing is how a
future investigation loses apparatus it needed; #435 stays open for their
attribution.
Full hermetic suite 15,321 passed / 0 failed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DebugPanel and DebugVM have not been constructible since Campaign V slice
V11 removed the ImGui frontend that hosted them: nothing in src/ ever
called their constructors, only a test did. Two consequences, both fixed
here — 35 environment reads inside them were unreachable, and roughly forty
XML doc comments across the diagnostics owners promised a runtime checkbox
that no longer exists. A flag documented as runtime-toggleable when it is
startup-only sends the next investigation down a path that cannot work.
Deleted DebugPanel.cs (340 lines), DebugVM.cs (548) and DebugVMTests.cs
(327). Corrected the surviving claims in PhysicsDiagnostics,
RenderingDiagnostics, CameraDiagnostics, PhysicsEngine and GameWindow to say
what is actually true: these flags are set from the environment at startup
or by direct assignment.
The one real dependant was CombatFeedbackSlot, whose binding target was
DebugVM. It now takes a plain Action<string>, which removes the dependency
without changing behavior — and makes visible that there is no behavior:
nothing binds the slot, so the combat refusals it carries ("No monster
target", "Enter melee or missile combat first") have been discarded all
along. Filed as #436 and pinned by a test, rather than papered over with an
invented chat message; the retail text and channel need the oracle first.
Deliberately untouched: F1's AcdreamToggleDebugPanel binding, which
GameplayInputCommandController consumes as a documented no-op so the key
does not fall through to a lower input scope; and the
DebugVmRenderFactsPublisher / DevToolsRuntimeSources chain, which is still
wired into production composition and deserves its own dead-code pass
instead of being pulled into this one.
Full hermetic suite 15,333 passed / 0 failed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The client reads 161 ACDREAM_* environment variables across 79 files. Only
about 25 were written down, and the audit found the documentation drifting
in both directions: CLAUDE.md still advertised ACDREAM_RUN_SKILL /
ACDREAM_JUMP_SKILL (deleted; skills are server-authoritative now, and the
jump fallback is 300, not the documented 200), while flags with real
side effects had no description at all.
docs/launch-options.md documents every one by lifecycle — production,
command line, measurement, automation, permanent diagnostics, temporary
probes, deprecated, retired — with a mandatory side-effects column. That
column is the point: #432 cost three days of taxed measurements because
ACDREAM_AUTOMATION_ARTIFACT_DIR reads like an output path and also builds
a per-frame diagnostics referee, and ACDREAM_STREAM_RADIUS silently
measures a streaming window production never uses. Rows now say so. Other
surprises the audit surfaced and recorded: ACDREAM_DUMP_SCENERY_Z swaps in
a duplicate scenery-placement path rather than only logging,
ACDREAM_PROBE_VIS silently also enables ACDREAM_PROBE_ENVCELL, and
ACDREAM_DUMP_ENTITY's id list doubles as an unrelated probe's watchlist.
LaunchOptionsDocumentationTests enforces it, because a hand-maintained list
of 161 flags is stale within a week: an undocumented flag fails, and so
does a documented row whose read site was deleted. It scans string literals
rather than GetEnvironmentVariable call shapes — the startup path reads
through an injected delegate, so a call-shaped pattern silently missed
ACDREAM_LIVE, ACDREAM_PAK_PATH and every other production flag. A third
test freezes per-file direct-read debt by exact count (20 files outside the
owner classes), so structure rules 4 and 5 can be paid down but not
regressed.
CLAUDE.md's 94-line env-var section becomes a 16-line pointer, and its
stale test-character paragraph is corrected.
Also fixed, all doc-vs-code mismatches the audit proved:
- RenderingDiagnostics.FrameProfEnabled described a GPU-query self-disable
that Campaign V slice V11 deleted.
- Two comments named ACDREAM_RENDER_BACKEND as a live co-requisite; it died
with the OpenGL backend.
- EnvCellRenderer.CollectCellAuditLines and its ACDREAM_A8_AUDIT doc: the
method had no caller anywhere and its documented caller never existed.
Filed rather than fixed, to keep this a documentation change: #434 (the
DebugPanel/DebugVM surface is never constructed, so ~40 "runtime-toggleable"
comments are false and 35 env reads are unreachable) and #435 (17 temporary
probes outlived their closed investigations; 14 more name no owner).
Full hermetic suite 12,202 passed / 0 failed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Second instance of the #429 (ad695589) boxing-comparer defect class, this
time in the measurement harness rather than production:
ACDREAM_AUTOMATION_ARTIFACT_DIR (with retained-UI screenshots) constructs
CurrentRenderSceneOracle, whose presence as partition observer re-enables
the G5-retired legacy InteriorEntityPartition every frame with per-entity
fingerprinting. Complete() then sorts one fingerprint per RESIDENT entity
(~60k across the streaming window), and the comparer's first key bound
x.ProjectionClass.CompareTo(y.ProjectionClass) to Enum.CompareTo(object),
boxing both operands. The 3-value enum almost always ties, so the boxing
ran on essentially every comparison: a measured ~6.2 MB and ~14 ms per
frame, everywhere — not town-specific and not view-triggered, which is
also why it appeared to "latch" (the resident set drives it, not the view).
Comparing the underlying integral value keeps the identical order. Hermetic
gate: one warmed observed partition of 20,000 entities allocated 15,876,088
bytes before, and passes a <64 KiB bound after
(OracleObservedPartitionAllocationTests).
Ordinary play never constructs the oracle, so no player-visible behavior
changes; what changes is that captures taken with the automation artifact
directory set are no longer taxed. The #429 acceptance data is unaffected
(owner drives and the deciding A/B arms ran with the artifact dir null).
The temporary [pview-alloc] attribution probe that localized this is
retired in the same commit; the gate test now guards the defect.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The directional-shadow topology rebuilt on every streaming-churn frame
and was the measured body of the run-hitch stalls (701 of 708 baseline
stalls alloc-correlated):
- The draw sort comparer's enum-vs-enum CompareTo bound to
Enum.CompareTo(object) and boxed BOTH operands on every comparison —
a constant ~38.9 MB of garbage per topology rebuild (~4M boxes),
handing the GC a forced gen0 collection mid-frame. The full ~100k-draw
sort is replaced outright: draws hash-group by exact DrawKey in one
O(n) pass over retained chained-index arrays, and only the
few-thousand DISTINCT group keys sort (order-preserving packed
material|cull|firstIndex|baseVertex + count|slot|layer|foliage keys,
first-appearance tie-break) — bit-identical emission order to the old
stable sort, near-zero allocation, and no per-draw comparisons at all.
- The caster frame sorts 4-byte indices keyed on SortKey.Value instead
of shuffling multi-hundred-byte records through a boxing comparer.
- Owner-approved pipelining: on a frame whose shadow inputs just changed
(the same frame already paying frame-view/landscape rebuilds), the
caster-frame and prepared-draws topology rebuilds defer to the next
quieter frame, capped at two consecutive deferrals — inside the GPU
fence depth, so retained draws never reference a released arena range.
First build, generation change, caster BuildSequence change, and
journal overflow force the immediate path; deferred refreshes skip
identity-mismatched journal rows.
Owner-accepted in both presentation modes: stall frames 5.8/s -> ~0.45/s
uncapped (0.49/s capped), median stall 20.3 -> 13.7 ms, >25 ms frames
near zero, 275 fps uncapped baseline restored. Allocation gate: a warmed
topology rebuild must allocate <2 KiB (DirectionalShadowPreparedDrawTests).
docs/ISSUES.md carries the full evidence trail; the residual
content-proportional rebuild milliseconds are filed as the
incremental-topology successor, and the pre-existing town-view scaling
latch is filed as #432.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Local commit for the implementation handoff; push withheld per owner
direction until the fix session.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adds the TEMPORARY PlayerPresentationProbe (ACDREAM_PROBE_PLAYER_PRESENT)
per the #429 apparatus plan and records the measured two-arm evidence in
the issue. Probe strips with the fix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
User-directed enhancement ('I want the night sky to look very good'):
retail's star layer is one small texture stretched over a 10-poly dome
cap, so stars smear regardless of source-image quality. With the
Atmospheric render pack active, sky.frag now renders the star layer
(GfxObj 0x010015EF, identical in all 20 Dereth day groups) as a fully
procedural sky computed from the view direction: hash-derived stars on
a cube-face grid in three density tiers plus sparse diffraction-spiked
standouts, sized in SCREEN pixels via derivatives so they stay crisp at
any resolution and FOV, over the user-approved 0.4-1.3% cool mottle
(gen_starfield2.py seed 11, approved 2026-08-23). The draw is forced
additive; the day/night fade rides the star layer's existing retail
lighting product so the schedule matches the authored keyframes. Pack
inactive = retail look byte-untouched.
EnhancedNightSkyRuleTests pins the uParamA gate, the exact star-layer
id, the forced-additive draw, and the pack-runtime wiring; sky shader
SPIR-V recompiled and re-pinned. Hermetic App suite green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The aurora was never missing data — it was a missing mechanism plus a
misread. New decompile evidence closes the April-2026 contradiction:
retail plays the sky carriers' PES through the Setup's own DefaultScript
(GameSky::MakeObject @0x00506EE0 -> CPhysicsObj::makeObject @0x00513970
sets state|=0x80000; animate_static_object @0x00513DF0 ticks
ScriptManager + ParticleManager). The pes_id column stays dead — that
half of the April finding stands; the ids are byte-equal mirrors.
- SkyPesFrameController is now the production owner (ACDREAM_ENABLE_SKY_PES
deleted): script ids resolve from the Setup DefaultScript
(SkyObjectData.DefaultScriptId; the pes_id column is a one-time-logged
cross-check), slots persist by (index, gfx id, properties) per
CreateDeletePhysicsObjects @0x005073C0 — a day-group swap keeping the
carrier no longer restarts its emitters — and stale slots stop before
replacements claim the slot-derived owner id.
- RetailParticleFacing ports calc_draw_frame @0x0050DFA0: degrade mode 2
faces the viewer roll-free (set_vector_heading) instead of the camera
plane; modes 3/4/5 spin the authored frame around one local axis
(rotate_around_axis_to_vector) — Dereth authors 54 mode-5 emitters that
previously got no facing at all; 1,583 mode-2 emitters get the exact
law; authored/mode-1 paths are unchanged.
- The 2026-08-23 'whole-sky tint' was the Rainy-group lightning/thunder
PES playing at the debug anchor inside their 0.03-0.19 window, not the
aurora: the aurora is nine faint viewer-facing glows pulsing on
6.7/15/55-minute rebirth cycles, in every day group, all day.
Research: docs/research/2026-08-23-sky-default-script-port.md.
Register: AD-112 filed (camera-anchored synthetic owners vs sky-cell
physics objects). ISSUES #2 corrected (the playback ban is lifted by the
new evidence); #28 fix landed pending the connected night gate.
Tests: RetailParticleFacingTests (16), SkyPesFrameControllerTests (6);
hermetic suites App 6,076/0, Core 4,905/0.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two April stand-ins, neither registered, met at the horizon from altitude as
a hard line between the dome's rim and the fog-coloured clear:
1. sky.frag fogged every non-additive sky layer with a 0.2 floor. Retail's
GameSky::Draw @0x00506FF0 disables fixed-function fog around the whole
sky draw unless an AdminEnvirons fog override is active (SetFFFogEnable(
LScape::m_override_enabled ? 1 : 0)); additive layers stay unfogged via
SetFFFogAlphaDisabled(1) at D3DPolyRender::SetSurface 0x59c882. The sky
pass now sets ApplyFog only for (override active && !additive), with no
floor.
2. WorldRenderFrameBuilder overwrote the authored fog range with one derived
from the streaming window (538..2189 m always). Retail sets FOGSTART/
FOGEND straight from the keyframe's MinWorldFog/MaxWorldFog
(SkyDesc::GetWorldFog @0x00500CE0 -> SetFFFogProperties @0x005A2F70) with
no draw-distance scaling; zfar is a constant 4000 m. The builder now
leaves SceneLightingUbo.Build's values alone; ACDREAM_FOG_START_MULT /
_END_MULT are deleted from RuntimeOptions.
Guards: SkyFogRuleTests (source-level, the sky renderer has no hermetic
harness); sky.frag.spv re-pinned in VulkanShaderManifestTests with the
reason. Research note 2026-04-23-sky-fog.md carries a correction banner.
App hermetic 6,070/0, Core 4,707/0 (Release). Owner look gate owed: night
and rain fog are now retail's shorter authored ranges.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Holtburg windmill axle (GfxObj 0x010010CE, 8 polygons, all
Stippling.NoPos + SurfaceType.Base1Solid) extracted to a 0-vertex mesh.
NoPos ("NO_POS_UVS", acclient.h:7380-7388) means "this side has no
texture coordinates" — true of every solid-colour polygon, since
nothing samples them — not "there is no positive face". Extraction read
it as the latter and dropped the polygon entirely, client-wide, for
every untextured polygon on every object.
Retail's D3DPolyRender::DrawMesh (@0x0059d4a0, named-retail decomp
~line 426048) draws an untextured subset on an ordinary object exactly
like a textured one; the only retail cases that skip an untextured
subset are a building shell (RenderDeviceD3D::DrawBuilding @0x0059f2a0
sets ObjBuildingOrBuildingPart=1) or an EnvCell interior
(RenderDeviceD3D::DrawEnvCell @0x0059f170, arg4=1). The #119
investigation's "retail's skipNoTexture never draws them either"
conclusion was itself wrong as a general rule.
- MeshExtractor.PrepareGfxObjMeshData / GfxObjMesh.Build: emit the
positive side whenever PosSurface is a valid index, regardless of
NoPos; the existing UV-index-0 fallback already produces zero
texcoords for a NoPos polygon with no UVs on the wire.
- RetailUntexturedSurfacePolicy.IsUntextured(SurfaceType): the one
place that answers "is this surface textured"
((type & (Base1Image|Base1ClipMap)) == 0), replacing the old
`isSolid = NoPos || Base1Solid` (which also mis-classified a NEG-side
batch by the POS-side's NoPos flag).
- RetailUntexturedSubsetPolicy.Draws(isBuildingShell, isUntextured):
the shared draw-time gate wired into WbDrawDispatcher.ClassifyBatches,
.PackedOracle.ClassifyPackedBatches, and
.DirectionalShadows.AddDirectionalShadowBatches — one predicate so the
three walks cannot drift (Campaign VM VM6 lesson).
- CellMesh.cs / MeshExtractor.PrepareCellStructMeshData deliberately
KEEP their NoPos-gated skip for cell-wall geometry — retail's
DrawEnvCell really does skip untextured subsets there; register row
AP-234 documents the NoPos-vs-Surface.Type approximation.
- PakFormat.CurrentBakeToolVersion 4->5 (LauncherInstallRecordStore in
lockstep): a pak baked by an older tool is missing every untextured
face. No bake was run as part of this commit.
Also fixed: WorldBuilder's own upstream ObjectMeshManager.cs has the
identical NoPos bug (ObjectMeshManager.cs:959,984) — our port had
faithfully carried it over, and our own conformance test
(Build_NoPosFlag_OnlyEmitsNegSide) asserted the bug as correct WB
conformance. Renamed/reworded to Build_NoPosFlag_EmitsBothPosAndNegSide
with a citation for why retail decomp overrides WB here.
Issue119UpNullGfxObjDumpTests re-run against the installed DAT:
#119's own two objects (0x010002B4 9/9 polys, 0x010008A8 1/1 poly) now
gate DRAWS on every polygon instead of extracting to nothing.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The Holtburg windmill (Setup 0x020003E5, cycle 0x0300061B) is a 60-frame
quarter turn that loops by 4-blade symmetry: frame 59 and frame 0 are the
same picture but 88.5 deg apart. BuildBlendedFrame wrapped the cyclic
node's next-frame index to frame 0, so the seam slerped 88.5 deg backwards
inside one 33 ms interval - the blades visibly flapped back every two
seconds.
Retail never blends animation frames at all: CPartArray::UpdateParts
(0x005190F0) applies get_part_frame(floor(frame_number)), holding every
authored frame for its interval and hard-cutting at the wrap. The
render-side blend now holds the boundary frame at BOTH ends of a node's
window - the same rule the #61 link-tail fix already applied to one-shot
nodes - so every seam is retail's cut while interior frames stay smooth
(the owner's choice over dropping the blend, 2026-08-23). Register row
AP-233 records the blend as the deviation it has been since the R1-P5
cutover.
Test: Advance_CyclicSeamHoldsLastFrameInsteadOfBlendingIntoFrame0 (fails
on the previous code at ~45 deg, passes held at 90 deg). Core 4,696/0,
App 6,068/0 hermetic (Release).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
VM3 and VM6 user-accepted live at 2560x1440 (High); #424 and #425 fixed
during the gate. Roadmap entry flipped to shipped; plan ledger VM7 row
SHIPPED; #422 recorded as carried by owner decision.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Live Holtburg at 2560x1440: the Low preset needed 67,368,164 resident bytes
(screen-sized HDR/depth/ray targets are 44 MB of that) against an absolute
64 MiB ceiling that had only been validated at 1080p, so Options -> Apply
fell back to the default path; every later Apply was then refused by the
controller's failure memo, which treated the user's deliberate choice like
automatic re-activation.
RenderPackResidentBudget.Effective scales the declared 1080p figure by the
viewport's pixel-count ratio (never below 1), still capped by the hardware
MaxPackResidentBytes; both pack graphs use it and the performance-matrix
tool judges its resident column by the same rule (contract test updated).
RenderPackController.Request gains explicitUserChoice, which clears the
memo for that selection; RenderPackSelectionBinding passes it on every
display edge (Apply, including resolution changes) and keeps the memo for
the startup request.
Tests: RenderPackResidentBudgetTests (1080p/720p keep the declared
ceiling, 1440p = 16/9x, 4K = 4x, hardware cap wins, zero extent rejected);
controller explicit-retry; the binding test now proves the user's next
Apply activates once the cause is gone. App hermetic lane 6,068/0 (Release).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
GLFW auto-iconifies an exclusive-fullscreen window on focus loss; for one
frame the window size reads 0x0 while the swapchain is still created, so
GameWindow.OnRender's PrepareFrame guard let a zero-area RenderFrameInput
through. The retail path tolerated it silently; the render-pack controller
correctly rejects a zero activation extent, which surfaced the latent frame
as an unhandled ArgumentOutOfRangeException during the owner's VM6/VM3 gate.
RenderFrameOrchestrator.Render now returns RenderFrameOutcome.ZeroArea
before BeginFrame when either dimension is <= 0 (no GPU frame, phase,
measurement, diagnostics or recovery runs) and GameWindow skips
NoteFrameClosed for it. Test: ZeroAreaViewport_SkipsTheFrameBeforeAnyGpuWork.
Verify: App hermetic lane 6,059/0 (Release).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The fault is pack-independent (fired on the retail preset, 1080p, 45 s,
first launch after a build), ~1 in 57 offline runs today, never under a
debugger (46 runs), no WER/event trace on this machine. The pre-campaign
binary cannot be measured with this gate (no in-process close verb).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
-ExeOverride runs another AcDream.App.exe through the SAME isolated
settings/scene/capture path so two binaries compare without mixing tool
versions (VM0's recorded fieldOfView trap). -DebuggerScript launches the
client under cdb so ntdll's debug heap validates every free and a heap
corruption breaks at the corrupting site with a stack.
#422 (1-in-8 STATUS_HEAP_CORRUPTION exit after a pack-on offline capture):
16 runs attached + 24 runs under the debug heap on 1f151242, all clean
(P ~ 0.5 % at the filed rate). Recorded as not reproducible on the current
binary, carried as a watch item for the owner's VM7 matrix with the loop as
the first step if it recurs; no longer a merge blocker. Evidence under
docs/research/evidence/vm7 alongside the final pack-off-vs-base+normals
robust diff (5-10 px, lifestone band only).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Round-5 narrow re-review of 754d59d9: APPROVE (verified to the SPIR-V
disassembly; shadows-on path proven opcode-identical). Four
non-blocking nits.
N1 (test robustness): T3 (ReceiverVertexShadersFallBackToThePlainLightDirectionWhenTheShadowFlagIsClear)
was satisfied by comment prose -- round 5's own explanatory comments
quote the exact plain-pipeline direction substring
("-uLights[i].dirAndRange.xyz, ... matched bit-for-bit"), so the
substring-only assertion passed even with the code mutated. Fixed:
comment lines are now stripped (new StripLineComments helper) before
any assertion, and a new ordered regex per shader asserts the exact
branch SHAPE -- shadowGatedOff ? -uLights[...] : normalize(uShadow...)
-- not just substring presence. Mutation-tested locally against the
rewritten test: (1) swapping the ternary's true/false operands --
FAILED (previously passed); (2) deleting the fallback entirely,
collapsing to the pre-round-5 buggy expression -- FAILED (previously
passed). Original file restored and reverified passing after each
mutation.
N2 (doc accuracy): "numerically the plain pipeline" overstated the
round-5 fix in three places (plan doc, mesh_atmospheric.vert,
terrain_atmospheric.vert). The direction expression is bit-for-bit;
the SUM is not, because the atmospheric shaders' split ambient+point
vs directional accumulation (and terrain's two varyings vs the plain
pipeline's one) reassociates float summation order by ~1 ulp -- which
is exactly the measured mean |Delta| 0.007 the coordinator's own
pixel-proof evidence already recorded (well under the 65 px noise
floor). All three rewritten to say the receiver "matches the plain
pipeline to within float summation-order rounding (measured mean
|Delta| 0.007 on the offline scene)."
N3 (coverage): RenderPrepared's own cascadeCount == 0 exit (the F2
fix) had no direct test even though RenderPrepared already has 8
direct call sites in this file. Added one: ResidentMaximumReachMeters
at/below CameraNearMeters, passed straight to RenderPrepared with an
otherwise-fully-valid environment (so the fitter, not the environment
gate, is what returns zero cascades), asserting IsBindableFor true /
IsValidFor false.
N4 (latent): RenderPrepared's OWN "if (!environment.ShouldRender)"
exit is a fourth bufferless-disabled path -- unreachable via Render
(whose own gate already validates ShouldRender first) but the same
shape, and RenderPrepared is called directly by tests and any future
caller. Took the preferred fix: publishes the disabled binding there
too, via the same helper, so every exit on a frame that draws the
world publishes when the pack's AtmosphericFrame is bound. Also made
EvaluateGateAndPublishDisabledBinding self-contained: it now resets
_currentFrameBinding to Disabled on its own entry instead of relying
on Render having done so first (idempotent with Render's own reset).
Regenerated SPIR-V: mesh_atmospheric.vert and terrain_atmospheric.vert
are comment-only changes (N2), so only the manifest's source hashes
changed -- compiled .spv bytes are unchanged, consistent with round 3's
precedent for comment-only shader edits.
Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,055/0 failed. Full hermetic-filtered solution: 15,283/0 failed
across 15 projects.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Closes the automated half of slice VM6 at 754d59d9 and records how it was
proven. The first pixel method (clock pin 0 s vs 3 s, wind-off pair as the
control) is recorded as confounded: the treeline silhouette carries a
bimodal 0-or-~280 px rasterisation churn between runs of the same binary,
and two conclusions drawn from it were retracted. The replacement
apparatus (same pin, wind on vs off, two captures per arm, robust mask =
AND of the four cross pairs minus both repeat pairs, shadows off so only
geometry can move) found the round-4 defect (wind welded to the shadow
gate: 1 m wind = 49-65 px vs a 22 px floor while the CPU state was
correct) and, after round 5, gives 14,993 robust wind pixels vs a 65 px
floor, every one on a treeline tree, hillside tree or shoreline bush, with
the wind-off frame matching the pre-round-4 plain pipeline at mean |d|
0.007 (round 4's eec95535: 1.77, the overhead-sun regression the narrow
review caught).
tools/vm6/wind-pixel-proof.py is the analysis command the closeout names;
evidence overlays under docs/research/evidence/vm6. Plan: VM6 ledger row
CODE-COMPLETE 2026-08-23, owner visual gate owed (section VM6 Acceptance).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Round 4 (eec95535) fixed wind but introduced a new lighting bug: the
receiver VERTEX shaders source the sun direction from the shadow block,
not only the shadow visibility term. mesh_atmospheric.vert's
accumulateLights read uShadowLightDirectionAndSource unconditionally for
every directional light; terrain_atmospheric.vert did the same for its
single sun term. Round 4's PublishDisabledReceiverBinding writes
direction (0,0,1) into that block on every shadow-gated-off frame (user
sun-shadow-strength 0 in daylight, indoor/portal cover, night), so every
such frame was lighting outdoor terrain and objects from straight
overhead instead of the authored sun. Publishing the environment's real
direction would not have restored parity either -- the celestial shadow
source direction (sun/moon disc) is not the authored light direction.
F1 (BLOCKER): fixed in the shaders themselves, exact parity with the
plain pipeline. Both receiver verts now branch on the same flag bit
acdreamDirectionalShadowVisibility already reads
((uShadowTextureAndFlags.w & 1u) == 0u) and, when clear, use the EXACT
plain-pipeline expression instead of the shadow block's direction:
-uLights[i].dirAndRange.xyz in mesh_atmospheric.vert (matching
mesh_modern.vert, hoisted out of the light loop as a uniform branch);
-uLights[0].dirAndRange.xyz in terrain_atmospheric.vert (matching
terrain_modern.vert's sunDir/-sunDir form). The (0,0,1) word in the
disabled block stays as the documented normalize()-cannot-NaN guard; its
comment now says so explicitly since it is no longer read as a light
direction when the flag is clear.
F2: RenderPrepared's cascadeCount == 0 return is a third bufferless-
disabled path reachable from a frame that already passed Render's own
two gates (the cascade fitter can still find zero usable cascades) --
publishes the same disabled binding now, via the same
PublishDisabledReceiverBinding helper (re-signatured to take a bare
AtmosphericFrameBufferBinding so all three call sites -- Render's two
early-outs plus this one -- share it).
F3: removed a stray duplicated " -- Closeout and merge" fragment under
the plan's VM7 heading.
F4: corrected the false "the flag bit makes it numerically the plain
lighting sum" claim in the plan's round-4 paragraph and in
WbDrawDispatcher.DirectionalShadowReceivers.cs -- the flag bit alone
only fixed the shadow VISIBILITY term (already correct before round 4);
it took both that AND round 5's light-DIRECTION fallback to actually
match the plain pipeline.
T1: extracted Render's gate prologue (environment evaluate -> two
early-outs -> PublishDisabledReceiverBinding) into internal
EvaluateGateAndPublishDisabledBinding(frame, in input, out environment,
out environmentGateTicks), behaviour-preserving, called by Render before
it touches world/terrain -- the ArgumentNullException.ThrowIfNull(world)/
ThrowIfNull(terrain) calls keep their exact position relative to the
gate. No test in this suite constructs a real WbDrawDispatcher +
TerrainModernRenderer pair (still true), so this extraction is what
makes the gate itself testable; two new tests drive it directly with
PlayerInsideCell: true and with ResidentMaximumReachMeters <=
CameraNearMeters, asserting TryGetCurrentFrameBinding true / IsValidFor
false for both.
T2: proves the actual composition WbDrawDispatcher.PipelinesFor and
TerrainModernRenderer both use -- TryGetCurrentFrameBinding feeding
ShouldSelectReceiverPipeline -- selects the receiver pipeline for the
atmospheric world pass once a disabled binding is published, and still
refuses a non-atmospheric pass name.
T3: shader-source guard (same style as AtmosphericPostProcessGraphTests'
existing shader-text tests) pinning that both receiver verts contain the
flag-gated fallback and reference the same uLights expression the plain
verts use, so a future edit that drops the fallback fails this test
instead of only showing up in a pixel capture.
T4: the (0,0,1) test's doc comment and an inline assertion comment now
say the value is a NaN guard, not a light direction.
Regenerated SPIR-V: mesh_atmospheric.vert and terrain_atmospheric.vert
recompiled to different bytes this time (a real code change, not a
comment); manifest updated to match.
Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,054/0 failed. Core.Tests 4,695/0 failed. RenderPackValidator 30/30.
Full hermetic-filtered solution: 15,282/0 failed across 15 projects.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The reviewer's offline pixel apparatus found a real design defect, not a
test artefact: foliage wind was welded to "directional shadows rendered
this frame." Evidence: offline High preset, sun-shadow-strength=0,
wind-strength 2 + lean/branch 1 m — wind-on vs wind-off at the same
pinned clock differed by only 49-65 px, inside the apparatus's own 22 px
run-to-run noise floor (no measurable motion). A CPU probe independently
confirmed ResolveFoliageWind was correct (first advance snaps to Clear
0.25/0.15, gate 1, one graph) — the correct uniform never reached the
world pass.
Root cause: DirectionalSunShadowRenderer.Render's two early-out paths
(!environment.ShouldRender, ResidentWindowUnavailable) left
_currentFrameBinding at its pure Disabled (no-buffer) default.
WbDrawDispatcher.PipelinesFor and TerrainModernRenderer's matching
selection logic only choose the atmospheric receiver pipeline
(mesh_atmospheric, the only pipeline that #includes foliage_wind.glsl)
when TryGetCurrentFrameBinding returns true; with no buffer it always
returned false, so the world pass silently fell back to the plain
mesh_modern pipeline, which has no wind code at all. Because the shadow
gate is ActiveDayGroupMultiplier = dayGroupPolicy x elevationResponse x
strength, this killed wind every night (elevation response -> 0), at
user sun-shadow-strength 0, and under the portal/login cover.
Fix (decouple, not patch): DirectionalShadowFrameBinding gained
IsBindableFor ("a real current-frame allocation exists") separate from
IsValidFor ("...and it is Enabled with real shadow content" -- kept
exactly as VolumetricShaftRenderer's own gate needs it).
TryGetCurrentFrameBinding now returns IsBindableFor. When the built-in
pack supplies an AtmosphericFrame binding (declared packs never do, so
their receiver shaders -- which never declare set 3 binding 5 -- are
unaffected), Render's two early-out paths call a new
PublishDisabledReceiverBinding: it allocates one real ring slice and
writes a DISABLED DirectionalShadowUniforms block -- every matrix
Identity, every control/bias term zero, TextureAndFlags all zero (bit 0
clear is exactly what directional_shadow_receiver.glsl's
acdreamDirectionalShadowVisibility already reads as "no shadow, full
visibility" via its existing early return 1.0), and a unit light
direction (0,0,1) so a fragment shader's normalize() can never produce
NaN. BindDirectionalShadowReceiver and TerrainModernRenderer's
shadow-buffer bind now check Buffer is not null instead of Enabled, so
the disabled block actually gets bound once it is selected.
PublishDisabledReceiverBinding is internal (not private) specifically so
it is testable without standing up a real WbDrawDispatcher/
TerrainModernRenderer pair -- no test in this suite constructs either.
New tests: (a)/(b) PublishDisabledReceiverBinding is bindable-not-valid
with a bound AtmosphericFrame and a genuine no-op with an unbound one;
(c) BindDirectionalShadowReceiver emits both UniformDirectionalShadow and
UniformAtmosphericFrame binds for a disabled binding; (d)
VolumetricShaftRenderer's gate still reports NoCurrentDirectionalShadow
for a disabled binding. ShouldSelectReceiverPipeline itself is untouched
and its existing tests (parametrized directly on bindingValid) remain
valid; no existing test asserted the old "disabled shadows -> plain
pipeline / no binding" behaviour in a way this fix invalidates -- every
existing caller either bypasses Render (calls RenderPrepared directly)
or uses a stale-serial binding IsBindableFor still correctly rejects.
Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,050/0 failed. Core.Tests 4,695/0 failed. Full hermetic-filtered
solution: 15,278/0 failed across 15 projects.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Narrow re-review of a82959f1: APPROVE, with follow-ups. All items landed.
N2 (structural): (a) extracted the ONE shared InstanceGroup-from-key
construction seam, WbDrawDispatcher.CreateGroupFromKey(key, registration,
frame) — before this there were two near-identical `new InstanceGroup
{ ... }` initializers (GetOrCreateInstanceGroup and GetOrCreatePackedGroup)
that had already drifted once (the round-2 F1 bug). Both routes call it now;
CreateGroupFromKey's own `new()` is the only production InstanceGroup
construction site repo-wide, same precedent as AppendPackedInstance. (b)
GroupKey.FoliageFlags lost its `= 0u` default and moved before CullMode in
the declaration (CullMode keeps its default, C# requires optional params to
trail required ones), so a `new GroupKey(...)` that omits it is a compile
error. Fixed every real construction site the reorder/requirement touched:
the 2 production sites, ToKey (a reconstruction from InstanceGroup the
review didn't count but the reorder broke), and 5 test sites (one more than
the review's "4" — InstanceGroupClearTests had a second, implicit
target-typed `MakeKey` factory the original count missed). Verified by a
full solution build.
N1: added CreateGroupFromKey_CopiesFoliageFlagsFromTheKey
(InstanceGroupClearTests) — a key carrying FoliageFlags 0x2 in, the created
group's FoliageFlags 0x2 out. That test plus N2b's required field are what
actually guard the round-2 F1 blocker; reworded PackedDispatcherOracleTests'
existing test comment to say what IT proves (the classification-to-
BuildIndirectArrays-to-BatchData.flags path), not that it guards the
classifier.
N3: corrected the plan's round-2 paragraph — folding FoliageFlags into the
G2/G3 digest is correct and symmetric, but CompareClassifiedOutput only
runs from RenderScenePViewFrameProductController.BuildAndCompare, which has
no production caller anywhere in src/AcDream.App/, and both of
RenderScenePViewFrameProductTests's own callers construct the controller
without the optional dispatcher argument — so the fold catches nothing
until that oracle is wired to an actual caller.
N4: the plan's F6 note now names both classification caches — the classic
route's EntityClassificationCache.EntityCacheEntry (self-heals per entity
on its own next eviction) and the packed route's
PackedProjectionClassificationEntry/PackedClassifiedBatch.Key
(PackedProjectionClassificationCache.BeginFrame clears its entire cache in
one shot on a RenderSceneGeneration change) — and notes neither mechanism
is keyed to a pack switch specifically.
N5: deleted the now-unused single-generic ComputeEntityHasCutoutSubset<T>
overload; its 4 test call sites now use the two-generic, zero-alloc
overload with an unused int context and a static (_, value) => value
lambda, so there is exactly one ComputeEntityHasCutoutSubset to keep
correct.
A6 (reviewer-filed): ResolveFoliageWind's _windMean/_windGust started at 0
and always eased toward the weather target by clock delta, with no
distinction for a graph's first-ever advance. A pinned clock
(ACDREAM_SKY_PHASE_SECONDS, the offline pixel gate's determinism pin) never
advances between calls, so the wind reached only whatever fraction the
first (1-second-clamped) step produced and sat there forever; live, the
first 10 s after a graph is constructed (pack selection / login) spun up
from dead calm even though the weather already IS what it is. Fixed at the
root: the first advance (_windFrameSerial == -1, the constructor sentinel)
now snaps _windMean/_windGust straight to the target; every later advance
eases over WeatherSystem.TransitionSeconds exactly as before. Added a
SetWindClockSecondsOverrideForTesting seam (_windClockSecondsOverride is no
longer readonly) so a hermetic test can advance the pinned clock by an
exact amount between two resolves without a real-time Thread.Sleep; two new
tests prove the first-advance snap is exact and a second advance still
eases at the normal rate. The three existing indoor/wind-disabled/amplitude
gate tests pass unchanged.
Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,046/0 failed. Core.Tests 4,695/0 failed. Full hermetic-filtered solution:
15,274/0 failed across 15 projects.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Narrow re-review of 43e3abed found every A1-A5/A7/A8 item resolved but
one new blocker in the production packed classifier.
F1 (BLOCKER): RetailPViewPassExecutor.DrawPackedProductionRoute — the
route production world geometry actually draws from — never computed
FoliageFlags at all. WbDrawDispatcher.PackedOracle.cs's
ClassifyPackedBatches built its GroupKey with the field defaulting to
0u, and GetOrCreatePackedGroup never copied it onto the created
InstanceGroup, so production BatchData.flags bits 1/2 were always zero
for every scenery entity: the world geometry never swayed even though
the independently-classified shadow caster did, so shadows visibly
swayed under rigid trees. Both classifier call sites now compute
FoliageFlags via the identical FoliageWindClassification.Classify call
and entity-scoped HasCutoutSubset OR the classic (non-packed) path
uses, and GetOrCreatePackedGroup copies it exactly like
GetOrCreateInstanceGroup always has. The G2/G3 classified-output
digest (AddOpaqueSubmissionGroup/BuildTransparentSubmissionDigest) now
also folds GroupKey.FoliageFlags into its hash — present in the key
since round 1 but never actually read by either digest function, so a
content-level (not just group-count-level) classic-vs-packed
divergence is now caught.
F2 (medium): the delayed-alpha replay path (PrepareDeferredAlphaDraws)
hardcoded Flags = 1, dropping bits 1/2 for any group replayed through
it — a trunk instance promoted into the alpha-blend group mid-fade
(the #188 translucency-promotion case) would stop swaying for the
duration of its fade. Now 1u | key.FoliageFlags.
F3 (nit): ComputeEntityHasCutoutSubset's three call sites (classic,
caster, and the newly-fixed packed classifier) each allocated a
closure over _meshAdapter per Setup entity per frame. A new
context-taking overload passes the mesh adapter as an explicit
argument to a static lambda instead, letting the compiler cache one
delegate for the method's lifetime rather than allocating fresh ones.
A3 test gap: WbDrawDispatcher.BindDirectionalShadowReceiver is now
internal so DirectionalShadowGpuTests can drive it directly with a
bare RecordingGpuDevice pass encoder, proving it emits
UniformAtmosphericFrame with the exact buffer/offset/size a
DirectionalShadowFrameBinding carries — paired with the existing test
proving that binding carries the caster's real bind forward untouched.
F1's missing test: PackedDispatcherOracleTests chains
FoliageWindClassification.Classify (called with the packed
classifier's exact argument shape) for a real 0x8... scenery entity id
through BuildIndirectArrays — the same shared, already-tested
production step both classic and packed group lists feed into BatchData —
proving the resulting flags word carries bit 0x2. Driving
ClassifyPackedBatches/GetOrCreatePackedGroup directly was not a "cheap
test": both are private instance methods reachable only through the
full RetailPViewPassExecutor route, which needs a real IGpuDevice,
world-pass scope, mesh manager, and compiled pipelines to construct —
no test anywhere in the App test project stands one up.
Nits: F4 corrects foliage_wind.glsl's header comment from "bit 31" to
the top-nibble test; F5 documents at the receiver bind site that the
caster's own AtmosphericFrameBufferBinding has its seven ABI v1
members zero/Identity by construction (only the two v2 wind members
are valid) — safe today because mesh_atmospheric.vert reads that
binding solely for wind displacement, flagged as a footgun for a
future v1-reading addition to that shader; F6 notes in the plan
(rather than fixes) that EntityCacheEntry does not proactively
invalidate when FoliageWindExclusions changes on a pack switch —
harmless with the pack off, self-heals on the entry's next natural
eviction.
foliage_wind.glsl's F4 comment-only change updated the SPIR-V
manifest's source hashes for the five includers (mesh_atmospheric.vert
+ four directional_shadow_world_* casters); the compiled .spv bytes
are byte-identical since comments do not affect bytecode.
Verify: Release build 0 warnings/0 errors. App hermetic-lane filter
6,043/0 failed. Core.Tests 4,695/0 failed. RenderPackValidator 30/30.
Full hermetic-filtered solution: 15,271/0 failed across 15 projects.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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>
Found at the VM6 spot-check; pre-existing from Campaign AR; values were
owner-accepted so it is filed for its own gate rather than changed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
FoliageWindByDayGroup / FoliageWindDayGroupPoint(int ActiveDayGroup, ...)
becomes FoliageWindByWeather / FoliageWindWeatherPoint(string WeatherKind,
...) in AtmospherePolicyDeclaration (Plugin.Abstractions is BCL-only, so
the key is the exact member name of AcDream.Core.World.WeatherKind rather
than the enum itself). The raw activeDayGroup index carries no weather
meaning by itself; WeatherState.cs already classifies each day group's
authored DAT name into one of five real weather kinds, and that fact was
already threaded through AtmosphericFrameInputs.Weather / uAtmosphereWeather.x
— this reuses it instead of guessing an index-to-category mapping.
Built-in table (BuiltInAtmosphericRenderPack.AtmospherePolicy()): Clear
0.25/0.15, Overcast 0.60/0.35, Rain 0.85/0.60, Snow 0.35/0.20, Storm
1.00/0.75 — all five WeatherKind members declared, the invented "Cloudy"
row dropped. RenderPackAtmospherePolicyEvaluation.FoliageWind now takes a
WeatherKind and matches by weather.ToString() (ordinal) against each
declared point's name; a kind absent from the table falls back to the
declared Clear row, then to (0,0) if Clear itself is undeclared. The
delta-seconds EMA interpolation (EaseTowardTarget) is unchanged.
AtmosphericPostProcessGraph.ResolveFoliageWind and its two callers
(RenderPostProcess via inputs.Weather; RenderDirectionalShadows via
foundation.Atmosphere.Kind) now pass WeatherKind instead of the day-group
int.
RenderPackValidation.ValidateAtmosphere (runs for every pack declaring an
AtmospherePolicy, not gated to Tier2/shadow packs) now rejects an unknown
or non-exact-case weather-kind name and a repeated kind, mirroring the
existing ActiveDayGroupMultiplier duplicate-key check.
Tests: RenderPackAtmospherePolicyEvaluationTests rewritten for the
kind-keyed API (all five kinds resolve to their declared row, an unlisted
kind falls back to Clear, ordinal exact-case matching, null-table
handling); RenderPackSpirvValidatorTests gains four descriptor-validation
cases (unknown name, wrong case, duplicate kind, the five-kind table
accepted); AtmosphericPostProcessGraphTests' three foliage-wind cases now
select WeatherKind.Storm via `with` instead of an assumed day-group index.
Spot-check (per the coordinator's ask, not changed here): yes —
ActiveDayGroupMultiplier / EvaluateDayGroupPolicy (pre-existing, Campaign
AR/VM3-era — BuiltInAtmosphericRenderPack.AtmospherePolicy()'s three rows
`new ActiveDayGroupMultiplier(0, 1.0), (1, 0.35), (2, 0.20)`) key the
sun-ray/shadow/volumetric day-group strength multiplier by the same raw
activeDayGroup index with an undocumented assumed meaning (0=brightest ...
2=dimmest), the identical class of issue this commit fixes for foliage
wind. Left unchanged per instruction; flagging for the coordinator to file.
Full solution Debug and Release builds green. App hermetic filter
6024/6026 — the same 2 pre-existing failures as VM6a/VM6b. Both were
re-run in isolation per the verification ask: both still fail alone (not
a load-flake in this environment) — confirmed via git stash earlier this
session that both already fail on the unmodified pre-VM6 baseline, so
they are pre-existing and unrelated to this change. Core.Tests hermetic
4697/4697. RenderPackValidator.Tests 30/30. No shader/spv changes in this
commit (pure C#/docs fix).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Procedural-scenery foliage (trees/bushes — entity ids in the
ProceduralSceneryIdAllocator's 0x8XXYYIII namespace) sways with weather in
mesh_atmospheric.vert and all four directional_shadow_world_* caster vertex
shaders, both calling the identical new foliage_wind.glsl include so the
shadow moves with the leaf by construction.
Classification (FoliageWindClassification, AcDream.App.Rendering.Wb): two
new BatchData.flags bits, computed once per (entity, subset) from four
inputs — entity id (bit 31 for procedural scenery), the pack's declared
FoliageExclusions membership, the subset's TranslucencyKind, and
ObjectRenderData.HasCutoutSubset (computed once per mesh at build time, not
per frame). Bit 1 marks an alpha-cutout leaf subset; bit 2 marks an opaque
trunk subset (only when its own mesh also owns a cutout subset, so rocks
stay still). WbDrawDispatcher.ClassifyBatches (world receiver) and
AddDirectionalShadowBatches (caster) call this with the same four inputs, so
casters and receivers classify identically without needing to share state.
Retail's mesh_modern/terrain_modern/mesh_detail pipelines never read these
bits, so pack-off output is unaffected.
Motion model (foliage_wind.glsl, mirrored bit-for-bit in the new
FoliageWindModel for hermetic CPU tests): height-squared-scaled slow lean
for every foliage subset, plus branch swing and per-vertex-hash-decorrelated
flutter for cutout subsets only. AtmosphericPostProcessGraph.ResolveFoliageWind
resolves the wind block once per frame.Serial — advanced by whichever of
RenderDirectionalShadows (which runs first) or RenderPostProcess is called
first that frame, with the second reading the already-advanced state, which
is what keeps the caster and receiver reading byte-identical clock/strength
values. The per-day-group mean/gust target (AtmospherePolicyDeclaration.
FoliageWindByDayGroup, keyed by the same day-group index convention
ActiveDayGroupMultipliers already established: Clear/Cloudy/Overcast/Rainy)
eases toward its target over WeatherSystem.TransitionSeconds (10s) so a
weather change never snaps; wind-enabled off or indoor instead gates the
OUTPUT to an exact zero (not an asymptotic approach) so a settings toggle or
cell transition is immediate. The wind clock is a Stopwatch started at graph
construction (monotonic, session-relative magnitude for GPU sin() accuracy),
overridable by the same ACDREAM_SKY_PHASE_SECONDS pin SkyRenderer already
uses, for deterministic offline gates.
New settings: wind-enabled, wind-strength, wind-direction-degrees (225°
default — no authored retail wind direction exists to read),
wind-lean-metres, wind-branch-metres, wind-flutter-metres (0 on Low),
wind-canopy-height-metres.
Register row IA-25 files this as an intentional, strictly opt-in divergence:
retail applies no per-vertex wind displacement to any geometry. Known,
accepted limitation: classification is per mesh-subset (one BatchData.flags
word per indirect-draw batch), not per entity instance, so the rare case of
one mesh subset being reachable from both a procedural-scenery and a
non-scenery placement would classify all of that subset's instances alike.
Tests: FoliageWindClassificationTests (the full classification matrix),
FoliageWindModelTests (identity on non-foliage/calm-wind/base-vertex,
canopy-top displacement bound, z-never-increases, trunk has no flutter
term), RenderPackAtmospherePolicyEvaluationTests (exact day-group lookup,
no interpolation across day-group ids, easing convergence without overshoot
or discontinuity), AtmosphericShaderAbiTests (each of the five shaders calls
acdreamFoliageDisplace exactly once; mesh_modern/terrain/mesh_detail call it
never), and four AtmosphericPostProcessGraphTests additions (indoor/disabled
exact-zero gating, settings-to-UBO wiring, same-frame-Serial idempotency —
the last proxies the caster/receiver agreement invariant without needing
this hermetic harness's WbDrawDispatcher/TerrainModernRenderer dependency
chain to exercise RenderDirectionalShadows directly).
App hermetic filter: 6015/6017 (the same 2 pre-existing failures as VM6a,
confirmed unrelated). Core.Tests hermetic: 4697/4697. RenderPackValidator.Tests:
30/30. Full solution Debug and Release builds green. Shader recompile
touched exactly the 5 edited files' .spv (plus manifest); the retail oracle
set and every other pack shader are byte-identical.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
AtmosphericFrame (set 3/binding 5) grows additively from 160 to 192 bytes:
two appended vec4 members, uAtmosphereClockWind and uAtmosphereWindAmplitude,
carry the foliage-wind clock/weather and amplitude inputs VM6b's shader
displacement will read. RenderPackShaderAbi renames the old constant to
AtmosphericFrameSizeBytesV1 (160), adds AtmosphericFrameSizeBytesV2 (192),
keeps AtmosphericFrameSizeBytes pointing at the current (v2) size, and adds
ShaderAbiVersion = 2. RenderPackSpirvValidator.ValidateAtmosphericFrame
accepts either the v1 (seven-member, 160-byte) or v2 (nine-member, 192-byte)
shape and rejects anything else naming both — this is why the frozen
external sample packs under samples/*/Shaders/*.spv, whose GLSL sources are
not in this tree, need no rebuild: a v1 shader bound to the 192-byte buffer
still reads correctly, since a bound range only needs to be >= the block's
own declared size.
DirectionalSunShadowRenderer's caster pass now binds AtmosphericFrame too
(both the multiview and per-cascade sites), through a new
AtmosphericFrameBufferBinding the graph owns and supplies via
DirectionalSunShadowRenderInput. AtmosphericPostProcessGraph.RenderDirectionalShadows
builds its own 192-byte ring allocation for this, separate from the world
receiver's frame block, because the caster pass runs before RenderPostProcess
constructs that block within the same frame. The four world caster pipeline
variants (opaque/cutout, base/multiview) are now allowed to declare binding
5 in the validator; terrain casters are untouched.
This commit is plumbing only: the two new members are always written but
never read by any shader yet (zero placeholders), so pack-on and pack-off
output are both pixel-identical to before. VM6b wires the real weather-driven
values and the shader-side displacement.
App hermetic filter: 5972/5974 (2 pre-existing failures unrelated to this
change, confirmed against the unmodified baseline). Core.Tests hermetic:
4697/4697. RenderPackValidator.Tests: 30/30. VulkanShaderManifestTests
(retail oracle set): 7/7, byte-identical.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Opus narrow re-review of 51178f7c: APPROVE. Residuals closed: the AR plan
no longer says '<=1 LSB' unqualified (99.99% of pixels; 95 foliage-
silhouette pixels up to 73 LSB, 58 isolated); the campaign doc says the
same; AtmosphericColorPipelineTests now read the SHIPPED exposure/vignette
defaults from BuiltInAtmosphericRenderPack.Descriptor and the graph's named
DefaultVignetteStrengthFallback instead of literals.
Measured for the gate (offline Holtburg hillside, High defaults vs pack
off): mean luminance -17% noon, -44% dusk, p95 unchanged, clip 0.06% both;
neutral High vs pack off: 110,561 px at |d|=1, 95 at >=5 (foliage edges).
Evidence images under docs/research/evidence/vm3/.
#422 filed: one High-default offline capture exited with
STATUS_HEAP_CORRUPTION after a clean managed shutdown; 1 in 8 runs, never
under validation layers. VM7 gate item.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fix round from Opus review of 87677f9c (APPROVE WITH FIXES):
1. (A1/A3) Hardened the atmospheric_filmic.frag / atmospheric_bloom_
downsample.frag shader-source pinning test: asserts the exact decode
call count (6 - three in lowFusedScene, three in main's non-fused
branch), that the 2.2 display-gamma exponent and 1/2.2 inverse in
atmospheric_common.glsl are formatted FROM AtmosphericColorPipeline
.DisplayGamma (so shader literal and CPU-tested value cannot drift),
that the 0.18 contrast pivot in atmospheric_filmic.frag is formatted
from AtmosphericColorPipeline.LinearMidGrey, and pins
AtmosphericPostProcessGraph.BloomKneeLinear/BloomThresholdLinear as
exact literals (0.73f / 1f). Also removed a stray trailing "()" from
an existing comment in atmospheric_filmic.frag that was inflating the
decode-call count to 7.
2. (B1) Re-derived the "vignette-strength" default for the linear-light
post stack. The vignette multiply now happens on linear colour before
the final encode, so a corner factor of (1 - strength) displays as
(1 - strength)^(1/2.2), not (1 - strength) directly. The accepted
look was strength 0.12 under the OLD gamma-space pipeline: a 12%
on-screen corner darkening. Under strength 0.12 in the new linear
pipeline that same 0.88 corner multiplier would only display as
0.88^(1/2.2) ~= 0.9435 (5.6% darkening - visibly weaker). Solving
(1 - strength)^(1/2.2) = 0.88 gives strength = 1 - 0.88^2.2 ~= 0.245,
which reproduces the accepted 12% corner darkening. Since
RenderPackSettingValueCodec requires every declared default to be
step-aligned from the minimum and 0.245 is not a multiple of the old
0.01 step, the step also moves to 0.005 (a finer slider, not
coarser) so the exact derived default is a valid grid point -
verified by running the ExternalTierTwoPackCanRenameEveryOwnedId
AndShaderAsset validation test, which failed with "invalid default
value" before this correction. Also updated the matching fallback in
AtmosphericPostProcessGraph.FromDescriptor (0.12f -> 0.245f) for
consistency, and added
AtmosphericColorPipelineTests.VignetteDefaultReproducesTheAccepted
TwelvePercentCornerDarkening pinning encode(1-0.245) ~= 0.88.
3. (A4) Renamed VolumetricShaftFrameParameters.LinearSunColor ->
AuthoredSunColor in VolumetricShaftQuality.cs (internal, 2 references,
both in that file - safe). Left LightSource.ColorLinear unrenamed:
grep shows 13 files depend on it (GlobalLightPacker, SceneLightingUbo,
LightBake, LightManager, EnvCellRenderer, RenderingDiagnostics, and
several Core tests) across the shared retail default-path lighting
UBO pipeline - renaming it is out of VM3's pack-only scope and would
touch the mandatory-unchanged default path. Added a pointer comment
on the field in LightSource.cs (and a one-line note at its
WorldRenderFrameBuilder.cs call site) documenting the same
display-space-not-linear fact and explaining why the rename is
deferred to its own default-path colour-space pass.
4. (B4) Added a citation beside acesFitted in both atmospheric_filmic
.frag and its C# mirror (AtmosphericColorPipeline.AcesFitted):
Krzysztof Narkowicz, "ACES Filmic Tone Mapping Curve" (2016). The fit
takes linear scene light in and returns linear display light in
[0,1] - it does not itself gamma-encode. Evidence: acesFitted(0.80 *
decode(0.46)) = 0.2064 un-encoded versus the accepted 0.51 on screen.
5. (B2) Rewrote the VM3 section of docs/plans/2026-08-22-visualmaster-
campaign.md with the shipped truth in place of the pre-implementation
guess: exposure stays 0.80 (at exposure 1.0 the linear pipeline maps
gamma-0.5 to 0.6017, essentially the same 0.6163 the owner called too
bright), bloom threshold stays 1.0 (a fixed point of both exponents),
knee moves 0.45 -> 0.73, vignette-strength moves 0.12 -> 0.245. Added
the old-vs-new curve table at exposure 0.80 across ten gamma inputs.
Replaced the acceptance criteria's "new automated test on the
recording RHI" with the CPU mirror + shader-source pins actually
used, and recorded that the real-frame masked capture WAS run
(retail vs High-with-every-effect-neutral, artifacts/vm3):
independently re-verified by re-running the pixel diff against the
checked-in screenshots - 110,561 px at |delta|=1 and exactly 95
pixels at |delta|>=5, confined to foliage-canopy silhouette edges
against sky with nothing on any ground/building/water surface. Noted
the Stage-1 luminance table re-capture is still owed at the owner
gate.
6. (B3) Corrected docs/plans/2026-08-21-atmospheric-rendering.md's VM3
summary sentence: the bloom intermediate is already linear after
extraction (no separate "bloom read" decode), and the neutral-preset
claim is now phrased as a measured numerical identity (<=1 LSB on a
real frame) rather than an unqualified "is" statement.
7. (A5) Corrected toolchain attribution: tools/compile-shaders.ps1 used
the managed Silk.NET.Shaderc path (shaderc_shared.dll) to compile in
both this round and the original VM3 commit - a Vulkan SDK glslc was
detected and its path recorded, but the managed compiler is what
actually ran. Regenerating this round only changed the atmospheric_
filmic frag stage's manifest hash (comment-only edits); the compiled
.spv bytes are unchanged, and every retail-oracle shader
(mesh_modern, terrain_modern, mesh_detail, etc.) remains untouched.
8. Replaced an invented motive in the atmospheric_filmic.frag contrast-
pivot comment ("rounded up for a stronger gamma-space contrast
feel") with the actual reason: the previous 0.5 was simply the [0,1]
midpoint of the standard contrast formula, not a deliberately chosen
value; in linear the perceptual mid-grey is 0.18.
Verify: Release build 0 warnings / 0 errors. App hermetic-filter tests:
5970 passed / 0 failed / 0 skipped. VulkanShaderManifestTests: 7/7 pass
(retail-oracle SPIR-V byte-identical; only the atmospheric_filmic frag
manifest hash changed, no .spv bytes changed).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Closes review finding F4 (docs/research/2026-08-22-campaign-ar-review.md):
retail's main-world colour, sun rays and volumetric shafts are all
gamma-encoded display-space values (the 2013 client has no linear
lighting pipeline), but bloom thresholding, ACES (Narkowicz fit),
Rec.709 luma saturation, the contrast pivot and the vignette were all
operating directly on those gamma values, then writing the result to
the UNORM swapchain without re-encoding.
- atmospheric_common.glsl gains acdreamDecodeDisplay/acdreamEncodeDisplay
(pow(c, 2.2) / pow(c, 1/2.2)). 2.2 is the retail-era CRT/early-LCD
display-gamma assumption, deliberately not the sRGB piecewise curve,
which would claim a precision retail's authoring pipeline never had.
uAtmosphereSunColor's comment is corrected from "authored linear rgb"
to "authored display-space rgb (retail has no linear pipeline)".
- atmospheric_bloom_downsample.frag, atmospheric_filmic.frag (both the
fused-Low and non-fused paths) decode every world/ray/volumetric read
before summing/thresholding; atmospheric_bloom_blur.frag is unchanged
(it already reads the now-linear bloom buffer); atmospheric_sun_rays.frag
and atmospheric_volumetric.frag are documented as writing display-space
colour that the consumers decode.
- The contrast pivot moves from 0.5 (a gamma-space midpoint) to 0.18
(linear mid-grey, the standard 18%-grey-card exposure convention).
The final filmic output is clamped in linear, then re-encoded before
the UNORM write.
- Bloom threshold/knee are re-derived for linear light: the pre-VM3
gamma-space pair was threshold 1.0 / knee 0.45, i.e. a soft range of
[0.55, 1.0] in gamma. Decoding both ends with the same 2.2 assumption
gives decode(1.0) = 1.0 (threshold unchanged) and
decode(0.55) = 0.55^2.2 ~= 0.27, so linear knee = 1.0 - 0.27 ~= 0.73.
Replaced the inline 0.45f literals with named constants
BloomThresholdLinear = 1f / BloomKneeLinear = 0.73f on
AtmosphericPostProcessGraph. bloom-strength's 0.65 default is
untouched.
- Exposure stays at its accepted 0.80 default: in linear,
encode(acesFitted(0.80 * decode(0.46))) ~= 0.50, reproducing the same
accepted midtone the old gamma-space pipeline produced as 0.51 for the
same 0.46 input (0.46 * 0.80 fed straight into acesFitted, no
decode/encode). Highlights now retain more (gamma 0.9 input moves from
~0.74 to ~0.85 through the full pipeline) and blacks deepen slightly
(gamma 0.1 moves from ~0.09 to ~0.05) — the owner's visual gate judges.
- Added AtmosphericColorPipeline, a CPU mirror of the GLSL decode/encode/
ACES/grade/filmic math (line-for-line, with a header comment requiring
it stay mirrored), and AtmosphericColorPipelineTests: neutral-preset
identity within half an 8-bit step for a 0..255 grey sweep (proving the
neutral preset is numerically the pack-off image), decode/encode
round-trip within 1e-6, monotonic-in-exposure, the pinned midtone/
highlight/shadow numbers above, and the bloom-knee derivation.
- Added a shader-source pinning test so a future edit cannot silently
drop the colour-space conversions: atmospheric_filmic.frag must
contain exactly one acdreamEncodeDisplay( call in main()'s output,
atmospheric_bloom_downsample.frag must contain at least three
acdreamDecodeDisplay( calls.
- Regenerated SPIR-V (tools/compile-shaders.ps1, glslc from the
installed Vulkan SDK). Only atmospheric_bloom_downsample.frag.spv and
atmospheric_filmic.frag.spv changed in bytes; every other pack shader
that includes atmospheric_common.glsl recompiled to a byte-identical
binary (the new decode/encode helpers are unreferenced dead code for
them). VulkanShaderManifestTests' retail-oracle SHA-256 set
(mesh_modern, terrain_modern, mesh_detail, etc.) is untouched and
still passes — the retail default path did not change.
- Docs: noted the linear-light move in the AR plan's Slice 1 section,
and added a "Colour space" section to the render-pack ABI doc
(docs/render-packs/semantic-bindings-v1.md) naming which inputs are
display-space and pointing at atmospheric_common.glsl as the
reference implementation. No ABI version bump — the binding layout
is unchanged.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Fixed step-count volumetric ray marches (Low/Medium presets, e.g. 40 steps)
sample the same t values at every pixel, so a shadow-cascade edge crossing
lands at the same march step across a whole neighbourhood of pixels and
reads as visible concentric banding around light shafts. Offset each
pixel's march phase with interleaved gradient noise (Jimenez 2014,
fract(52.9829189 * fract(dot(gl_FragCoord.xy, vec2(0.06711056,
0.00583715))))) instead of the fixed 0.5 sub-step center: t = (step + ign)
/ steps. This decorrelates the step boundaries across pixels, turning the
banding into fine per-pixel noise that the existing quarter/half-res
upsample already averages away. Pack-on only — atmospheric_volumetric is
never used on the retail default path; extinction, colour, and early-outs
are unchanged.
Regenerated SPIR-V via tools/compile-shaders.ps1: only
atmospheric_volumetric.frag.spv and shaders.manifest.json's source hash
for that stage changed.
Also files #421 (docs/ISSUES.md) for Campaign AR review finding F7: the
directional-shadow pass's DirectionalShadowTransformBufferSet duplicates
the main pass's N.5 instance SSBO instead of binding it, to be resolved
once the planned GPU-culling step for shadow cascades lands. Updates the
VM5 ledger row and outcome note in
docs/plans/2026-08-22-visualmaster-campaign.md.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
