100 Commits

Include renames

Author	SHA1	Message	Date
SamandClaude Fable 5.1	7fe4b37c2f	guild lines are message ids, mail sends again, and the shadow fades in ... - the guild lines are message ids: `ui_guild/lines.rs` carried 34 English constants and two hand-rolled fill helpers and pushed composed text onto the chat log — 2045's largest ratchet row and the one file its rule had no route for, because `UiError`'s doc claimed the 1.12 formats use at most one `%s` and one `%d` and the type was built to match, while `ERR_GUILD_PROMOTE_SSS` is "%s has promoted %s to %s." — a wrong sentence in a doc had become a wrong type, and the wrong type had made a file's worth of drift structurally necessary; `CGGameUI::DisplayError` (`0x496720`) is cdecl varargs — `add esp,4` bare, `add esp,8` with one string, `SMSG_GUILD_INVITE` pushing two with the stack check exact at `0x30 + 0x60 = 0x90`, and `SMSG_GUILD_EVENT`'s emitter tail `0x5e745f` passing one, two or three off `strCount` — so `UiError` carries an ordered heterogeneous `Vec<FillArg>` (`S(String) | D(i64)`, since `ERR_USE_LOCKED_WITH_SPELL_KNOWN_SI` is String-then-Integer) behind four constructors naming the reference's call shapes (`key`, `s`, `strings`, `args`), twenty sites migrated; what the catalog knew and composed text could not carry: `ERR_GUILD_NAME_INVALID`/`ERR_GUILD_NAME_EXISTS_S` are kind 2 — the red line — and went to chat, `ERR_GUILD_CREATE_S`/`ERR_INVITED_TO_GUILD_SS` carry the LEVELUP cue and `ERR_GUILD_PERMISSIONS` (type_tag `0x3e`) is one the reference speaks (1815), all three silent; the default arm was missing outright — `0x5e745a` pushes `0x69` = `ERR_GUILD_INTERNAL` for `0x09` and every id past `0x0d`, catalog row 105 the whole time, so `TABARD_CHANGE` had been silent under a comment calling the key unsettled; a short `strCount` is passed SHORT (0 and ≥ 4 pass zero strings) so a starved specifier reads "…promoted Furor to %s.", not "…to ."; `/ginfo` deliberately stays off the queue — `GUILD_NAME_TEMPLATE`/`GUILD_INFO_TEMPLATE` are not catalog rows, `0x5e6fb0` resolves each token through the VM and emits chat itself, and a non-row key through `Shown::keyed` would turn a chat line red — so it waits on the guild feed and goes out unkeyed, the enUS month-day-year swap with it; the `_S` arity is asserted from the key names, 22 of 22; the ratchet row is deleted, 208 over 31 files remain (2054) - a unit's blob shadow asks the UNIT for its alpha: the lane gathered live `RenderFade`s off a unit's part entities and read "no part is fading" as 1.0 — true of a settled unit and exactly backwards for arrival, where a streamed unit holds `PendingAppearFade` with no `RenderFade` anywhere until the world is shown, so its parts were deliberately invisible while the shadow drew fully opaque; measured on a probe login: alpha 1.000 from t 0.61 to the arm at t 3.51, then 0, then the ramp — 2.9 s of dark oval under no creature, the symptom `model_render_alpha`'s own "pending ⇒ 0" law (0827) already named for emitters and ribbons; `model_fade::UnitRenderAlpha` is the one door, composed from the ROOT where `UnitAppearFade` tells pending from settled, published as one question rather than its three inputs so a caller cannot reassemble the mistake, and `DespawnFade::armed()` retires the `-1.0` sentinel from every caller; the world API wall goes 177 → 178; and wow-re settled what the shadow's alpha IS, four questions VERIFIED: it reads `CM2Model+0x180` through `0x710ca0`, not `+0x19c`; that slot carries the appear and despawn ramps, the self fade AND the CharProc-14 aura — `0x60d180` drives the same `StartAlphaFade` the appear fade does, so a stealthed unit's shadow now dims with its body — and not the per-batch M2 colour/weight tracks; and the reference draws nothing under a not-yet-shown object either (`+0x180 = 1.0 · 0.0`, written 30 bytes before the read), so this is fidelity, not taste; three module-doc claims corrected against it: the dead `0x683ea5` gate, the registration handoff (GameObjects and DynamicObjects never cast; units, players and corpses always do), and 1277's `InheritedVisibility` gate, which the reference lacks and stays ours on purpose (2055) - the mail stationery list was empty for the whole session, and every send with it: the five `dropping command — not connected` lines at login, once named, were `ItemQuery` for all five `Stationery.dbc` rows — `feed_mail` carried no run condition and asked from the login screen before the socket existed, the asks evaporated in the io thread, and `Items::template` had already latched all five in its ask-once `pending` set, cleared on disconnect and never on connect, so they were never re-asked and the stock `SendMailFrame_Reset`'s row-1 selection left every send silently unsent; the list now waits for a self player, and `release_ask_once_latches_on_enter` clears the pending sets at world enter, the counterpart the disconnect teardown never had — a redundant re-ask costs one packet, a wrong latch costs a dead feature nobody can see is dead; and three more log lines answer for themselves: a control character is answered before the shape — a newline is a line break the markup parser consumes, not a missing glyph — so the warn stays meaningful for a glyph a face genuinely lacks; the blend-check's one-frame run is the specialization tick catching up with a material swap, a real login showing four entries clear 26 ms later, so the warn waits for a run to SURVIVE the tick and a one-frame run is recorded at debug; and the audio starve line carries the wall span it covered beside the time it counted and the position it saw advance, because a 1.0 s window still closed 0.756 s of wall after the stream started on `Time<Real>` — if they disagree the caller's clock is still wrong, if they agree the starvation is real Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:41:57 +07:00
SamandClaude Fable 5.1	f1a1b68c1a	strings become keys, hidden tiles park, and the water stops breathing ... - a reference string is a key, never a Rust literal: display text is a key into `GlobalStrings.lua` in-game or `GlueStrings.lua` on the glue screens, resolved at runtime from the player's own install, and a literal is legitimate only as a fallback beside a lookup — the reference's own `UNKNOWNOBJECT` shape — because a re-typed sentence costs the wording nobody diffed, the surface and the voice a composed string has no catalog record to read off (1770, 1815), and localization; the tripwire in `reference_strings.rs` is the check a comment could never be: it walks the workspace, normalises both sides and fails on any literal matching a shipped value — 273 re-typed literals over 34 files, after calibration threw out three false-positive classes (a path-mangling bug, the 85 fallbacks beside a real lookup, and `#[cfg(test)]` modules whose fake tables match by construction — the first run's 378 would have enshrined ~50 phantom sites); `ALLOWED` is a per-file ratchet whose counts may only go down and whose stale rows fail too, regenerated from a clean tree at 253 over 33 files — the first baseline was captured mid-edit and the land gates failed on a row claiming drift in a file that had none — then 252 over 32 as `ui_trade`'s row retired behind a neighbouring conversion to `ERR_TRADE_BLOCKED_S` (2042); stated in its own doc: it cannot find an INVENTED sentence, which matches nothing; and `benilla_ui::strings::fill` is the one template filler — ordered `%s`/`%d`, positional `%N$s`, `%%` collapsing, a starved specifier copied through visibly so a mis-modelled template looks wrong rather than plausible — replacing eight hand-rolled copies that disagreed, one of them `str::replace` filling every `%s` with the same argument, latent only because no message on that queue carried two yet (2045) - two families convert, and both assert the identifier, not the sentence: the character-create result line resolves its key off the player's chain with the literal as fallback, and three bugs the 23-arm literal table had hidden come out — `0x36` re-typed without its second sentence ("Please try again during off peak hours."), `0x50` invented (there is no `CHAR_NAME_CONSECUTIVE_SPACES` in 1.12, so it takes the default arm), and `0x4B` naming `CHAR_CREATE_NAME_IN_USE` where the code means `CHAR_NAME_RESERVED` — both read "That name is unavailable" in enUS, so no assertion on displayed text could ever have caught it, and a sentinel test that rewrites `CHAR_NAME_RESERVED` in a copy of the real table proves `0x4B` moves while `0x31` does not; the ratchet's first row retires, 21 → 0; and `SMSG_PARTY_COMMAND_RESULT` returns the catalog message the reference's own jump table names — handler `0x5e68b0`, `dec eax; cmp eax,7; ja <silent>; jmp [4*eax+0x5e6a14]`, results 1–8 onto `ERR_BAD_PLAYER_NAME_S` … `ERR_IGNORING_YOU_S` plus `ERR_INVITE_PLAYER_S` (`0x3a`) and `ERR_LEFT_GROUP_YOU` (`0x42`) on result 0 — so text, surface and sound come off one row: `ERR_PLAYER_WRONG_FACTION` (`0xff`) is the red line where the other nine are chat, and three inputs the reference is DELIBERATELY silent on (result ≥ 9, result 0 outside op {0, 2}, an invite with an empty name) stop printing an invented "Party command failed (N)."; tests assert the message id, since no English distinguishes a right key from a wrong one here, and `ui_party` goes 22 → 12 (2045) - a hidden `<Model>` tile is parked: every orthographic pane shares the atlas's one camera, deliberately kept active whenever any cell is packed because it is the camera that clears the sheet, so `booth_frozen` — exact for every booth ever built, one camera one scene — answered "drawn" for every pane that had left the paint list, and a hidden autocast-shine pane's four spline emitters kept integrating and pushing quads for the whole 600-frame linger at a cell the packer had since given away; `sync_tiles` gains a sixth stage that parks any live tile with no cell this frame — `AnimParked` on the root, which holds the pose evaluator, the compose, the palette write and the global-sequence bone writes, and `ParticleEmitter::set_frozen` on every emitter — with `scene_frozen` stating the freeze law in one place and the park verdict pinned to the composite by a test, so a repack or a full atlas cannot freeze a visible pane's cloud; the bone global sequences move onto the pane's clock, where the material rows have been since 2019: the kernel's Phase B cursor is `(sceneClock − attachSnapshot) % duration` off the scene that owns the instance (`[model+0x2c]`, the snapshot at `CM2Model+0x68` stored at `0x70eae1`), and a `<Model>` widget owns a private `CM2Scene` at `CSimpleModel+0x314` that only its own advance at `0x76d863` moves — so the ping's spinner holds while hidden and resumes where it stopped, with no accumulator of ours; measured with a cooldown pane holding the atlas alive: twelve hidden shine panes cost 54 effect items / 63 transparent draws before, 34 / 43 after, the pool held rather than drained by design; and the particle census's `gated=` column, which printed `fade.is_some()` and read backwards as the gate verdict, is renamed `has_fade=` (2046) - a doodad is not an entity: `[def+0xa4]` has exactly four writers image-wide — ctor `0x6a7d73` (0.0), WMO doodad-set commit `0x695bb3` (1.0), ADT MDDF commit `0x6b01bb` (1.0), MCSH refresh `0x698cb4` (0.5) — and the 2.5-producer `0x69e280` and the ramp `0x69e770` are reachable only through the entity-attached `[obj+0xe0]` path, which contradicted the §6 both lit-ground sites cited and a capture with gains up to 2.5, so the premise was marked contested at both sites first (the shader's own note told a future session to cut `min(I, 1)` and `LIT_T` together on its strength); wow-re then settled it — two C++ classes inline the same base and share the offset: `CMapDoodadDef` (vtable `0x810a74`, `"WDOODADDEF"`, size `0x198`, serving ADT MDDF and WMO MODD) and the WENTITY light node (vtable `0x810810`, size `0x108`, off `[obj+0xe0]`) that `0x69e770` ramps at 3.3333/s; on the doodad class `+0xf8`, the field the ramp reads its target from at `0x69e894`, is `m[2][3]` of the world matrix — it has no target field — and six trace frames put 111 MDDF placements over 356 draws at gains only in {0.5, 1.0}, none changing, against an entity walking out of shadow at 0.5 / 0.81 / 1.09 / 1.3367 and rising; so ADT map doodads join the fixed-1.0 family with the exterior WMO props, `ShadeSel::Lit` is entity-only, the retained lane's Matte arm drops its prop requirement, and no pixel changes — the cap was faithful for trees all along and lifting it is now an entity-only look change for the director to call; `WORD_SHADE_LIT` is left as a documented dead bit (2050); and the far water's 1.25 s breath is flattened: the director's recording autocorrelates at lag 75 @ 60 fps = 1.250 s (r = +0.60), the 30-frame ocean flipbook at 24 fps, whose frames do not share a mean — alpha 55.985..57.628, DXT3's 4-bit alpha swinging 1.64 at mip 0, 3.72 at mip 5, 6.38 at mip 6 — so at distance the mipped sheet moves together, 0.8..1.84/255 on screen, fogged out at `farclip` 300 and visible at 777; wow-re proved the reference breathes too (`0x68abab` creates the flipbook with `applyGlobal = 1`, the upload hands the driver `Levels = 9`, LOD bias 0, the loop period `[0x8102b8]` 1.25 s over 30 frames), so ours was faithful and it is flattened anyway by the director's eye: each frame's DC equalised per level and per channel at texture-build time as an exact integer redistribution in ±1 steps on a coprime stride (1.64 / 3.72 / 6.38 / 0.555 → 0.000), water and ocean only since magma and slime take the sheet as their body colour, `WOW_LIQUID_DC=raw` restoring the authored art for the A/B (2051) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:41:02 +07:00
SamandClaude Fable 5.1	f7e232b2db	swing refusals latch and repeat, and the trade arc stops being silent ... - a refused auto-attack is no longer silent, and the display is a latch, not a toast: six opcodes, four arms — `0x145 SMSG_ATTACKSWING_NOTINRANGE` and `0x146 SMSG_ATTACKSWING_BADFACING` latch a code through `0x5ecdb0`, which writes `[0xc4d758]` and `[0xc4c270]` together; `0x148 DEADTARGET` and `0x149 CANT_ATTACK` share arm `0x625ab8` verbatim (StopAttack, no message), so the client cannot tell them apart and `AttackSwingError` collapses them into one `DeadOrUnattackable`; `0x147 NOTSTANDING` is never registered by the reference and `SendAttackSwingNotStanding` has zero callers in vmangos, so a test pins that it stays an unknown opcode; and the fourth arm hides: `0x14e SMSG_CANCEL_COMBAT` is registered at `0x5e3303` by the SPELL TU, not the combat one, so the eight-opcode census every note quotes did not contain it, and its handler `0x5e7dd0` is arm 4's body instruction for instruction; vmangos sends each refusal on the edge alone, so the repeat the player sees is the client's own: the tick `0x5ec950` — `CGPlayer_C` vtable `0x80af78` slot 14, dispatched once per object per rendered frame on the `CWorldFrame::OnPaint` chain, so a Bevy `Update` system is the exact cadence — prints `ERR_BADATTACKPOS`/`ERR_BADATTACKFACING` (catalog rows `0xd8`/`0xd7`, both red) at once and again every 4000 ms while an attack target is set, and a landed swing clears it: `SMSG_ATTACKERSTATEUPDATE`'s arm calls `0x5ea800`, whose first act is `0x5ecdb0(0)`, gated on the attacker being the active player AND the victim resolving (both gates kept — wow-re's round had glossed the second away, and sending that back corrected the note); the latch is one `Option<(Latched, Duration)>` because `[0xc4c270]` has no meaning while `[0xc4d758]` is 0, a second refusal packet defeats the cooldown and the re-arm precedes the print (both tested), the clear also runs once per world entry (`0x5e2510`), and both edges ride ONE `SwingRefusalEdge` stream so a refusal and a landing in the same frame stay in packet order; `SMSG_FEIGN_DEATH_RESISTED` (`0x2b4`) arrives beside them and is built differently because its handler `0x6e9800` is two instructions — a bare `DisplayError(421)` = `ERR_FEIGN_DEATH_RESISTED`, the single word "Resisted" — so it takes the plain `UiErrorKeys` route with no latch; the `--attack` world probe now reports the refusals it tallies; and two remainders are named, not built: `0x5ea800`'s post-swing leash — `R = 4.0 * max(5.0, reach(victim) + reach(self) + 1.3333333730697632)`, floor 20.0, sending a bare `CMSG_ATTACKSTOP` past it — a new packet-sending behaviour left for the director's call, and `StopAttack 0x5ecac0`'s two one-shot tutorials (10 low-health, 11 low-mana), which benilla's copy has never fired (2037) - the trade arc stops being silent: eighteen of the twenty-three `SMSG_TRADE_STATUS` arms print where benilla printed none, all through `UiErrorKeys` so the text is the player's own `GlobalStrings.lua` and the surface is the catalog record's `+0x04`; the four-row start was the trade block's own emit-site census, which answered "which arms raise a TRADE-keyed message" — eleven of the ids come from elsewhere in the 465-row table (`ERR_PLAYER_BUSY_S` `0x3e`, `ERR_PLAYER_DEAD` `0x7e`, `ERR_GENERIC_NO_TARGET` `0xb8`, `ERR_PLAYER_WRONG_FACTION` `0xff`, `ERR_IGNORING_YOU_S` `0x13d`, the stunned pair `0x191`/`0x192`, the logout pair `0x19d`/`0x19e`, `ERR_RESTRICTED_ACCOUNT` `0x1be`, and `ERR_TRADE_WRONG_REALM` `0x1ca`, whose text is the ONLY_CONJURED sentence), and the surfaces do not follow the English — a successful trade's two lines are the yellow ones, `ERR_PLAYER_BUSY_S`/`ERR_IGNORING_YOU_S` are chat lines, the refusals red; `TRADE_REQUEST_CANCEL` fires from the top of the CANCELED arm (`0x4bf832`, the dispatcher's one `SignalEvent`) where nothing had fired it; `ERR_INITIATE_TRADE_S` prints — it is client-emitted from inside the `CMSG_INITIATE_TRADE` sender, at once on a name-cache hit (`0x5d4042`) and from the query callback on a miss (`0x5d40ca`), so clicking Trade had produced no text in either direction; the three arms that name a player (0/5/14) read the network handler's guid cell `[0xc4bed0]`/`[0xc4bed4]` behind the outbound-initiate latch `[0xc4bec8]` — `InitiateTrade 0x5d3fb0` writes the target's guid and then the latch, the `BEGIN_TRADE` arm writes the initiator's and no latch — which is what keeps a client that just ignored a trade from printing "X is ignoring you." to itself, and behind `0x468460(guid, TYPEMASK_PLAYER)`, a live-object test the initiate line does not share, so `NamedLine::needs_live_object` drops a status line the moment its player stops being streamed while an initiate line waits; a network-cell clear is not a window close: `0x4bf4e0` has four call sites in the dispatcher — one opens, `{3, 8, 12}` close — and the other fourteen run the handler's tail `0x5d490a`, which drops the latch and the pending guid and leaves the window standing, so `close()` splits into `close_window()` and `clear_pending()`; `REJECTED` drops only the partner's accept, `BACK_TO_TRADE` reaches `0x4bf230(0)` and sends `CMSG_UNACCEPT_TRADE` on a real change, `ONLY_CONJURED` bounces the offending slot out of our own offer; `CLOSE_WINDOW`'s message is deliberately not built (`0x4bfd70` maps `result` over a 61-id range with a flagged self/other polarity inconsistency, and vmangos never sends status 12); and `ERR_TRADE_BLOCKED_S`'s transcribed English literal is deleted along with `ChatLog::pending_texts` (2042) - both halves of B373's lighting closed: `MeshTag` bits 6..=13 mean two things — the exterior payload's ground-shade byte and the low eight bits of the interior payload's 13-bit probe slot — and the overlap cannot be removed (the four mode-independent fields cost 19 bits, leaving 13 for a payload region the two modes want 21 of), so the payload question becomes a witness the accessors demand: `mesh_tag::ExteriorPayload` is a zero-sized proof only `exterior_payload(on_bake_law, own_probe)` mints, when both probe-slot populations answer no, and `with_shade`/`shade_of` take it, so no expression in the crate reaches those bits without having asked; `describe` alone reads the raw masks and prints both readings (2041); and a transport's deck cargo is lit by its own doodad def, not by the boat (B373): a transport's WMO is linked into the same global `TSExplicitList<CMapObjDef>` at `0xca7d98` the ordinary scene walk iterates, so its MODDs are ordinary doodad defs; `[def+0xa4]` has exactly four writers image-wide — the ctor `0x6a7d73` (0.0), the doodad-set commit `0x695bb3` (1.0), the ADT MDDF commit `0x6b01bb` (1.0) and the MCSH refresh `0x698cb4` (0.5) — and none is the light node, whose 2.5-producer and ramp are reachable only through `[obj+0xe0]`, a path a doodad def never takes; the MCSH sample is one-shot (`0x698c50` unlinks each entry) and its null-tile legs return lit; so we had the wrong source, the wrong cadence and the wrong mechanism: `entity_shade::DoodadDefLit` marks every WMO-gameobject prop batch from inside the placed-model assembler — the engine asserts it, not the caller; the world-API wall failed the first cut, and a following card is reachable only there — both shade write sites skip it on ownership, separately from 2041's payload question, and exterior props take one `doodad_ground_shade` at their own footprint at spawn, frozen, which answers lit for a non-resident tile so a boat at sea needs no special case and the unverified "a boat is never MCSH-shadowed" goes; 749 prop parts under a shade root carry the marker on a live run — the same 749 B373 measured reading a foreign probe — and the ADT doodad path is deliberately untouched, since the D3D capture's measured 2.5 gain outranks the closure (2047) - 2043's unit-predicate law is confirmed byte-exact by the round it dispatched — all 29 `Unit*` predicates push the constant double 1.0 or nil, exactly one value at every live `ret`, and a BFS from all 83 bindings in `0x850438` reaches 460 functions without hitting `lua_pushboolean` — and three of its supporting citations are corrected in the comments, no behaviour changing: `lua_pushboolean 0x6f39f0` EXISTS with seven call sites, one of them the registered FrameScript binding `IsPetAttackActive 0x4be0e0`, which answers a real true/false and never nil (ours already did, and its doc already named itself the exception), so 2043's "no boolean on its binding surface at all" is withdrawn for the scoped claim 1830 actually made; `IsEnabled` is `0x7800b0`, not `0x780270` (that is `SetButtonState`), and its number-1-or-0-never-nil fact stands; and `UnitReaction 0x5167e0` pushes `0x6061e0(u1, u2)` PLUS ONE, so 0 is unreachable and its nil means only that a token did not resolve — our `reaction == 0` → nil stays, named now as a feed gap rather than a mechanism the binary does not have (2048) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:40:25 +07:00
SamandClaude Fable 5.1	eadacac998	a hundred stale claims retire, and text= was never a divergence ... - a comment that justifies code by an engine difference is a claim with an expiry date, and the tree had ~96 expired ones across `assets/ui`, `ui_script/` and `benilla-ui` — a hygiene pass, no behaviour change: dangling `Benilla*` names with zero definitions (`BenillaMoney_*` at thirteen sites after 1937/1962 converged the coin cluster onto the stock `SmallMoneyFrameTemplate`; `BenillaGossipRow_Resize`, `BenillaErrorsFrame_AddMessage`, the digit feed 1285 deleted, `BenillaMerchantItem_OnClick` and three more), "the shipped `assets/ui/<X>.xml`" for thirty-odd files the directory no longer holds — it holds nine — each site now naming the chain file that actually loads, `UiPanels.xml` in the present tense after 1988 took the panel manager and the dialog engine stock, and expired self-deprecation: SimpleHTML "a widget kind we do not have" (1337 built it), "benilla has no console" (1948 bound `ConsoleExec`), `Model`'s "seven verbs not published yet" and "`ClearFog` is a guess" (all 23 are published, the carve sits 500 lines below, `UNBUILT` has been `[&str; 0]` since 2027), `frame_flag_gate`'s doc describing four groups of accepted mouse divergences over a `KNOWN` list that is empty (the empty array and its enforcement stay), five comments saying key delivery is not gated on `keyboard_enabled` where `script::keyboard`'s walk filters on exactly that field (1319), `SetRaidTarget` called a verb this house lacks beside an assert that its intent IS queued (1820), two questlog tests deferring to a screenshot for `GetPoint`/`IsVisible` the regions already answer, `benilla.toc`'s "trade still needs four engine verbs" (all four bound, 1966) and its ping `Model`s "this engine draws no pixels for" (2013 renders `MinimapPing.mdx` itself), and `rig_anim`'s `set_paused` naming a booth caller it does not have; and the code the prose was propping up goes with it: `GameMenuFrame.xml`'s six `<OnLoad>` label setters, because `text=` on a `<Button>` is a global-string lookup at the loader's attribute step (0995, built 61 seconds before the file was written) — the buttons carry `text="LOGOUT"`/`"EXIT_GAME"`/`"MACROS"`/`"RETURN_TO_GAME"`/`"MAIN_MENU"` as the reference's own do; `ScriptLogFrame`'s five authored close-button lines, now `inherits="UIPanelCloseButton"` off the chain's `UIPanelTemplates.xml` seated 757 lines above it (1846); and the editbox's 22-line "gap" doc claiming only the self-acquire `autoFocus` path is built and its `false` default unverified — path 1 is 1686's `visibility_focus`, and the default is `true`, byte-read off `flags = 1` at `0x779a29`/`0x779a2e`, so the test named `autofocus_does_not_focus_on_show…` is renamed for what it actually pins (2044) - the `text=` divergence never existed: `resolve_text`'s doc had claimed since 0068 that falling back to the raw attribute on a GlobalStrings miss was "one deliberate divergence" from a reference that draws the empty string, and read at the byte it is the reference's own behaviour — `FrameScript_GetText 0x703bf0` does return a pre-seeded empty string (`0x882748`) when the global is not a string, but that return never reaches a label, because all three `text=` readers image-wide test it and substitute the raw attribute: `Button::LoadXML` at `0x778c07` (`0x778c31 mov eax,esi`), `CSimpleFontString::LoadXML` at `0x771006` byte-identically, and `0x7292a6`, which pushes `edi` on the empty arm; so the loader changes nothing, the two test comments that had called `PetLevelText`'s "Level level race class" placeholder and `CLOSE` our divergence now name the reference's arm (a real 1.12 client draws them too), and what IS ours is the extra warning on a key-shaped miss (0983 → 0991), a strict addition; beside it, `GetEffectiveScale`'s root factor is 1 not because "benilla has no `uiScale` CVar" — 1804 registered one, default 0.9 — but because the dial is applied at the raster seam, `768/uiScale` UI units of virtual-screen height rather than a scale on `UIParent`, so every coordinate the VM hands Lua is already in those units and the reference's `GetCursorPosition()/GetEffectiveScale()` transcriptions convert correctly with a root of 1 (2044) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:39:30 +07:00
SamandClaude Fable 5.1	8f0bece3be	loot and pet refusals become message ids, and waterfalls flow again ... - a loot refusal is a message id, not a sentence: `loot_error_text` composed eight sentences for `SMSG_LOOT_RESPONSE`'s refusals and six of them said something 1.12 never says — "too far away to loot that." for a string that ends "that corpse.", "You can't loot that from there." for "You must be facing the corpse to loot it.", and an invented "Those pockets are already empty." for a code the client has no string for; it and the `LootErrors` queue are deleted, and the refusal is raised by GlobalStrings key into `UiErrorKeys` at the net bridge, the way `fish_verdict` two functions above already did, off the reference's own eleven-entry jump table at `0x5ebc64`; a composed string had no record to read a surface or a voice off, so all eleven — `MsgKind::Error`, the red `UIErrorsFrame` line — had been going into the chat window as yellow SYSTEM lines, and the four spoken in the player's own race and gender (`DIDNT_KILL`, `BAD_FACING`, `LOCKED`, `TOO_FAR`) were silent; a second bug fell out of reading the leg: codes 4–9, 11 and the default fall through a release tail at `0x5ebac2` that zeroes `[player+0x1d28]` — our `LootLatch` — while 10, 12, 13 and 14 jump past it, so a master looter refused a hand-off no longer stands up out of the kneel at a still-open corpse; the default arm's `ja 0x5ebabd` lands on code 7's arm, so every unnamed code prints `ERR_LOOT_DIDNT_KILL` and releases (2035); and the pet has its own two message tables, not the player's with a flag: `SMSG_PET_ACTION_FEEDBACK`'s handler `0x4bdb70` decrements the byte, bounds it at 3 and jump-tables it through `0x4bdbe8` to `ERR_PET_SPELL_DEAD`, `ERR_NO_ATTACK_TARGET`, `ERR_INVALID_ATTACK_TARGET`, `ERR_PET_SPELL_NOPATH` — one per vmangos `PetFeedback` value, where our two-entry map had printed "No path available for your pet" for a dead pet, "Out of range" for no target and nothing for 3 and 4; `SMSG_PET_CAST_FAILED` is `0x6e8eb0`, a separate 142-byte index table that overrides ten reasons, six of them to `ERR_PET_SPELL_*` rows with no other raise site in the client, so `CastFail` grows a `Caster` and picks its table from it while layers 1 and 3 stay shared, as the binary shares them; `0x6e8eb0` calls neither the combat-log formatter `0x62c360` nor the error sound `0x458a50`, so a pet's refused Growl no longer prints "You fail to cast Growl: …" as the player's own failure; and the source walk that exists to catch a non-catalog key was blind to both wrong keys because all 465 catalog keys begin `ERR_`, so the tripwire moves to the funnel — `Shown::keyed` warns on a key `by_key` cannot resolve (2033) - the hunter-pet feedback cluster, five opcodes carved off the binary where four had sat name-table-only: `SMSG_PET_TAME_FAILURE` (`0x173`) is two GlobalStrings lookups in order — `0x6e6a20` maps `reason − 1` through the eleven-arm jump table at `0x6e6ac0` to a `PETTAME_*` key and hands the resolved string to `DisplayError(ERR_TAME_FAILED)`, anything outside `1..=11` taking `PETTAME_UNKNOWNERROR`, and Call Pet and Revive Pet raise it too; `SMSG_PET_NAME_INVALID` (`0x178`) raises `ERR_INVALID_PETNAME` from the shared dispatcher `0x5e38c0`'s arm `0x5e3e33` — 1066 had called a refused rename silent on a carve that gave `0x178` the handler `0x4bdc00`, which is `0x2AF SMSG_PET_BROKEN`'s and raises `ERR_PET_BROKEN`, body unread; `SMSG_PET_ACTION_SOUND` (`0x324`) is a bark, not a sound id — its two talk values reach the bark dispatcher `0x623a40` at states 1 and 2, i.e. `CreatureSoundData` columns 28 and 27, and the shipped DBC names them from the other side: exactly four of its 406 rows carry them, the Imp, Succubus, Doomguard and Voidwalker voices, as `A_<demon>_ORDER`/`A_<demon>_KILL`; `Latch::Voice` now carries the bark state as a number and every one-shot creature vocal goes through one `play_bark` that is `0x623a40` — aggro at 0, the pet barks at 1 and 2, the death cry at 4, which had been on a bare `play_kit` off the slot so a pet bark could talk over it — with the reference's `jle` priority rule, because under the old boolean slot a demon's voice would have lost every race against its own aggro flares and been inaudible; it is a warlock-demon surface end to end — vmangos only sends it for a summoned pet, at a 10% roll per order, and a hunter's wolf has neither the columns nor the packet; and the fifth opcode surfaced because the cross-check went looking for column 29's consumer: `SMSG_PET_DISMISS_SOUND` (`0x325`), handler `0x604140`, reads column 29 twice (`0x6041c4`/`0x6041dc`) off a row resolved fresh by `CreatureModelData` id rather than the cached `[unit+0xb40]` — which is why a census over the cache had called the column dead — and plays `A_<demon>_DISMISS` at the packet's point with `z + 1.0`, no unit and no latch; column 29 is loaded, the column nothing reads is 22 (zero on all 406 rows), and the dismiss handler is built although vmangos never sends the packet (2039) - every UV scroll and animated tint in the world had been frozen for three days: 1940's deferred realization parks a freshly built `WowModelMaterial` behind a reserved handle until something view-visible binds it, and `doodad_anim::register_uv`/`register_tint` still reached for the store alone — `Assets::get_mut` answered `None` for every registration, the slot was freed, the registry stayed empty and the shader read the pinned-zero identity row for every batch, silently, because the `else` branch was a bare `return`; the fix is the address, not the timing: `Parked::value_mut`/`value`/`holds` reach a value wherever it currently lives and `lazy::with_material_mut`/`with_material`/`holds` publish that, so the one-shot build stamp lands in the parked value and rides realization in — realizing it to stamp it would have thrown away exactly what deferral buys; the tick's eviction on `Assets::contains` read a parked material as dead and asks `lazy::holds` now, the empty-registry early-out says so out loud when marked parts exist, and two instruments ship — `WOW_MATANIM_PROBE=<secs>` prints per animated part whether it is registered, whether its loop moves, what its table row holds and whether it is drawn, and `WOW_WORLDVIEW_AT=x,y[,z]` opens the server-less viewer anywhere; measured at the reported waterfall: 0 uv / 0 tint entries before, 87 uv / 6 tint around the Elwynn valley after, its two sheets on slots 1 and 2 with live rows −0.2122 and −0.4246 — the authored 1:2 ratio (2038); and the paper doll and the inspect pane are lit from the model's right: `CGLight+0x24` is the direction the light PROPAGATES — the `CharacterModelBase` ctor `0x505680` writes `(0, 1, 0)` and `0x71bce0` negates all three components before building the SH basis — so `model_pane_light()` hands the packer the ctor's vector as written where 0638 had negated it on the way in and lit both panes from the left; the ambient `(0.7, 0.7, 0.7)`, the diffuse `(0.8, 0.8, 0.64)` and the side-light analysis all hold, which is why the A/B that approved 0638 could not tell the two readings apart (2034) - a unit tooltip whose name query has not answered titles `UNKNOWNOBJECT`, not a blank gold row: the reference's builder `0x529fe0` reads its NAME line through `CGUnit_C::GetUnitName 0x609210` — the same function `UnitName 0x517020` delegates to — and every miss on its three legs, the null creature record included, ends at the one `FrameScript_GetText("UNKNOWNOBJECT")` tail at `0x609324`, so `unknownobject` moves into `unit` proper and the plate and the verb share it, read out of the VM's `_G` with the binary's "Unknown Being" behind a missing global; the faction line stops waiting for the creature record — its entry gate `0x612610` returns 1 when there is none, so it resolves off `UNIT_FIELD_FACTIONTEMPLATE` and keeps only `HIDE_FACTION_TOOLTIP` for when the record lands; the reference's placeholder sticks for the whole hover (stock `GameTooltip.xml` registers `UPDATE_MOUSEOVER_UNIT` alone) and our line-change repaint, which replaces it when the answer lands, is kept (2040); a unit predicate answers the NUMBER 1 or nil, never a Lua boolean: `UnitExists`, `UnitIsDead`, `UnitIsGhost`, `UnitIsDeadOrGhost` and `UnitOnTaxi` pushed a boolean and twenty-three more open-coded the right shape in eighteen places, where the reference pushes through `lua_pushnumber 0x6f3810` or `lua_pushnil 0x6f37f0`; all twenty-eight now reach 1830's one push site `binding_abi::predicate` — via `unit::unit_predicate` for a `UnitState` field, directly otherwise — `UnitIsCharmed`'s `Number(1.0)` and `ClearTarget`'s hand-branched nil included, and `UnitOnTaxi` gains the reference's `Usage:` raise (gate `0x517a48`) where a nil token had raised mlua's type error; truthiness is why it survived (`if UnitIsDead(u)` reads the same either way, and `== nil` inverts), nine `eval::<bool>` tests were vacuous under mlua's coercion and `taxi.rs` had pinned `== false`/`== true` — the one shape the reference cannot emit — so a table-driven test covers all twenty-eight on both legs, and `UnitIsAFK`/`UnitIsDND` now say they are ours (zero occurrences image-wide) (2043); and the quest log's stock title width cap — `275 − 15 − tagWidth`, live off the chain since 1944 — ellipsizes on one line rather than wrapping into the 16-unit row: 1873 declined it reading an explicit width as a wrap width, but the title FontString declares `y="10"`, and a declared height arms both overflow regimes, so a 258 px title in a 206 px box paints as one line ending in three ASCII dots — the reference's own result, its ellipsis gate at `0x771ec0` being `boxW > 0 && (maxLines != 0 || boxH > 0)`, which only an auto-height FontString escapes; two tests pin the render law on the real FRIZQT face and the geometry the stock file hands it, including the cap's per-row-button stickiness, which is the reference's too (2036) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:39:11 +07:00
SamandClaude Fable 5.1	4040166138	the model pane takes the file's camera, and the who list sorts ... - the `<Model>` widget's perspective leg: a pane whose file supplies the camera it selected renders through a camera of its own whose viewport is its own cell of the same tile atlas — the reference's structure, one target, a viewport per widget, depth cleared per widget — with the composite, packer and reaper untouched; the selection is RAW (`0x76cec0` reads the count off `MD20+0x124` and the record at `[model+0x3c4] + idx·0x84 + 0x80`, `cameraLookup` never consulted), so `M2Model::cameras` is the whole table in file order and `benilla-extract m2cam` prints it, reproducing wow-re's numbers to the digit; the camera is two fields because the reference has two — `+0x320` a PENDING index and `+0x31c` the installed one — and while a request is pending the pane paints nothing (`76d5f0`'s draw gate, now the paint list's own filter); an index past the count installs the NULL camera, which IS the orthographic leg every shipped in-game pane takes; the projection is the booths' byte-derived diagonal-FOV matrix at the pane's aspect, and the eye and target ride the model's root transform as `0x718960`'s publish does — so `SetFacing` cancels along with `SetPosition` and `SetModelScale`, a contradiction with a wow-re note that went out as a §5 round and came back agreeing: the up vector `0x7ac640` builds is model-space +Z at roll 0, the very axis the facing turns about, and the note's sentence is scoped to `<PlayerModel>`'s frozen camera; each perspective camera takes a render layer of its own from a pool of eight, the orthographic camera clears the atlas and every perspective one clears only depth; `SetLight` is a typed `CGLight` that folds each intensity into its colour, normalises a direction on write, quantizes to 8 bits, returns before the copy on `SetLight(0, …)` and reads the second intensity at index 8 on a short ambient block; the five fog verbs are built (`ClearFog` is `76f5c5 and [edi+0x3a4],-2`, bit 0 alone), so the `Model` UNBUILT list is empty for the first time since 1134; a pane that arms a light or fog takes a rig from a lazily-grown pool, and the tiles' material twins — never reaped since the cache was written — are; `WOW_PROBE_MODEL_CAMERA=1` reads the renderer's own camera back for the three cancellations, `pass=5 fail=0` (2027) - the layout cache writes in the shutdown tail: an unlocked chat window's size and position did not survive `/logout` or `/reload` and `benilla-config/layout/` was never created (B353) — the lane was right (1.12.1 persists a chat window through `StartMoving`/`StartSizing` and the engine's userPlaced bit; no `SetChatWindowSaved*` verb exists in the 5875 FrameXML), but `UiLayoutPlugin` hung its flush on `OnExit(InWorld)`, which a `/reload` never leaves, its debounced writer was never registered in `Update` at all, and on the roots where the edge did fire the VM-reading saver raced `end_ui_session`'s VM replacement on the same unordered edge (measured: the placement moves with nothing but registration positions); the reference's UI shutdown `0x490bd0` has an ordered tail — `PLAYER_LOGOUT` (`0x490c2a`) → `layout-cache.txt` (`0x490c79`) → the flat saved file (`0x490c7e`) → the per-addon files → `AddOns.txt` — on the same five roots as SavedVariables, `/reload` included, so `save_now` joins `shutdown_ui_state` in that slot, one call site reached from every root, and its own edges are deleted; the debounce stays as the crash writer and only that, a deliberate divergence, and the chat-look flush next door is pinned `.before(end_ui_session)` (2029); the VM gets its font engine at the LOAD edge: the measurer was seated only from an `Update` system on the first frame the glyph atlas exists, but `ReloadUI()` mints a fresh VM and runs the in-game load in the same exclusive slot, so every `<OnLoad>` and every addon's file scope measured 0 through the whole reload (`LOAD MEASURE ZERO` live, 33.185 after) — `seat_text_measurer` is one construction site called from both edges; the options tabs fit once from `<OnShow>`, the `OptionsTab_FitWidth` OnUpdate poll and its `fitted` latch gone, because `GetStringWidth` answers in the region's own units and this window's 0.78 scale is applied in the window's own `<OnShow>` after every `<OnLoad>` — an OnLoad fit read 73.19 where the settled value is 76.47; the settled widths are unchanged to the last digit (76.467232 / 91.661896) and the one-frame 72/87 flash at every open goes; four comments that still called the text measure a frame late, and seven gossip-suite comments that put `UIPanelScrollFrameTemplate` in our `ScrollTemplates.xml` when it lives in the reference's `UIPanelTemplates.xml`, are corrected (2028) - the `/who` sort is a seven-key chain, not "order by the clicked column" (B365): the reference keeps seven `{key, dir}` slots at `0xc2817c`, seeded zone → level → class → group → name → race → guild; `SortWho` (`0x5ad890`) promotes the clicked key to the front and flips its direction ONLY when it was already there — Name, Name reverses, Name, Level, Name restores what the first click left — and the comparator `0x5ada00` walks the chain as tie-breakers, most-recently-used first, through an ASCII fold (`A`–`Z` by `+0x20`, not a Unicode one), the three DBC arms comparing the localized name out of `ChrClasses`/`ChrRaces`/`AreaTable` where ours had sorted by raw id, and an unresolved row TIES (its miss jumps to `0x5adbb2`, the chain loop's own `inc esi`); the click fires `WHO_LIST_UPDATE` synchronously through `SignalEvent` at `0x5ad9ed`, so the binding promotes and sorts the snapshot and the handler repaints before the OnClick returns, where ours redrew a tick later — a redraw one click stale reads exactly like a header that does nothing; the chain is process state (initialised once from `0x401666`), so it survives a logout; the rider from the same carve: `SMSG_WHO` has two exits, and with `SetWhoToUI` clear more than three rows go to the frame with nothing printed and three or fewer to chat (`0x5adfcc jl`, `eax = 3`) where we printed every unclaimed answer; and wow-re's §11 round corrected the first build before it landed — the chat lines come out in WIRE order with the total line LAST, because `0x5ae0a1` composes each line inside the parse loop and the `qsort` at `0x5ae0e2` sits past its back-edge (2030) - two lighting reports, one frame and one bit each: the surfacing hitch (B354) is a one-frame stale atmosphere — `update_time_lighting` is late by construction, waiting on the wire drain, the weather tick and the submersion verdict, and five `Update` systems that read the `WowLighting` it resolves declared no ordering, so the executor handed each of them last frame's value, deterministically and invisibly while the atmosphere drifts a minute at a time; at a submersion crossing it jumps, on the very frame the sky and cloud domes un-hide, so the first dry frame painted `LightParams` 27's underwater sky over a dry world (the clip's blue frame, (58, 137, 173) against the settled (53, 58, 34)) and `tick_clouds`' surfacing full-rebuild ran at the underwater cloud density, 0.0, rebuilding an empty field that crept back over 0.4 s — the fix for the pop-in handing the symptom straight back; `LightingConsumeSet` names the read side of `LightingResolveSet`, tied in one place, and four consumers join it, the FFX haze and warp are ordered after the verdict they had the same hole against, and two tests enforce it — zero ambiguities between the two sets, and a source scan over `benilla-world` that fails any `WowLighting` reader not in the set; `emit_minimap` is left stale knowingly because joining cost a 178th item through the world API wall; `WOW_FORCE_SUB=<frames>` holds the eye verdict submerged and releases it, so the crossing reproduces server-less (2032); the black furniture below a boat's deck (B373): `MeshTag` bits 6..=13 are the exterior payload's ground-shade byte and bits 6..=18 the interior payload's SH-probe slot, and `entity_shade` guarded the overlap by asking whether a part was on the interior classifier's Bake law — one of the TWO populations holding a probe slot; the other is a lit interior MODD prop, which a WMO-display GameObject parents under the net entity so it sails with the deck, putting it inside the light node's descendant walk — so every frame the boat's shade byte was written over each prop's slot, `(slot & 0x1f00) | byte`: a neighbouring prop's light where that row was live, a zeroed row — solid black — where it was not; the guard now asks the payload question (`probe_payload`) at both write sites, the tag's bits and the marker derived from one binding; measured live at Ratchet, 749 of 749 resident interior prop parts renamed before, 0 of 640 after, and both ledger leads refuted at the bytes — the MODD colours are healthy (`floor112` raises even `#010101` to diffuse 112) and the fold is rigid-motion invariant since every 1.12 transport authors zero MOLR lights; a prop's mouseover identity now names its lane, words and probe slot, and `WOW_PICK` prints `model_flags.z` (2031); the sea-crossing probe can now measure the seam a report names — an explicit `WOW_PROBE_DOCK` goes first, boards only once the body stands at the dock, carries a map id, takes only a candidate within boarding distance, and drops onto the deck rather than into the hull, because a taxi node's `z` is the WATERLINE for the sea ferries while their decks stand ~16 yd above it — and the seam re-anchor returns the cycle it moved to or `None`, so a healthy crossing no longer logs the server's 9.7 h uptime as a skew inside a 350 s cycle Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:37:40 +07:00
SamandClaude Fable 5.1	7735447d32	the window tab goes stock, and the ferry keeps its rider at the seam ... - the window tab is the reference's own: `CharacterFrameTemplates.xml` loads off the chain at the reference's toc l.21 — the reference never gives it a toc line, `UIOptionsFrame.xml` includes it, and ours is not the reference's options window, so the row is explicit — and our `CharacterFrameTabButtonTemplate` with its `BenillaTab_Room`/`BenillaTabButton_OnUpdate` settle is deleted, 192 lines; the stock template fits a tab ONCE, in `<OnShow>`, through `PanelTemplates_TabResize(0)`, which needs a `GetStringWidth` that answers inside the Lua call that asked — the engine's synchronous measurer is that, so the settle that re-fit from `OnUpdate` until a frame-late round-trip landed is dead weight; retired with it, the director's call: 1002's structural clamp (`benillaTabRightInset`/`benillaTabPadding`/`benillaTabMaxWidth`, a tab never past its window's drawn right edge) — the reference's only caps are the numbers each window passes `TabResize`, and nothing had set those fields since the macro window went stock; 1004's highlight law is unchanged and now the reference's own text (`LEFT +10`/`RIGHT −10`, two opposing anchors that never consult the width, so both `SetWidth` lines are dead code there and here); measured live on the frame the window opens, main's 115·115·115·115 becomes 89.8·99.3·70.8·75.6 on the character row, 77.9·66.1·70.8·66.1 on the social, 89.8·85.0 on the merchant — each its label's `GetStringWidth()` + 40 — and one frame later the two builds agree exactly, so what goes is a one-frame 115 px flash on every open of every one of those windows (1993) - the ferry seam: `tick_transports` leaves an off-map transport's transform unwritten — the reference's own freeze, `0x5f5248`'s leg map read against the current map and the return at `0x5f5279` past the pose stores — but 0455's riding worldport composes the rider's world pose as `boat.translation + boat.rotation * local_pos` from exactly that transform, so a client whose path clock had not crossed put the body at Booty Bay's coordinates on Kalimdor, which want ZERO tiles of Kalimdor's WDT: the reported `0/0 resident`, which cannot clear (`is_ready` needs `total > 0`) and cannot re-raise (`focus_resident` takes its vacuous off-grid arm); the fall is a second bug in the same window: `SelfMoveMessage` dropped the wire's `ON_TRANSPORT` tail and `apply_self_move` re-derived the deck-local pose as `world − boat.translation`, which against a frozen source-continent boat is the distance between two continents — consistent while the boat stays frozen, a ~13,000 yd fling off the deck the instant it writes an honest pose, `update_attachment` keeping the ride because the body is airborne and no settle hold to catch it; the rider's deck-local pose now comes off the wire, and `first_cycle_on_map` + `reseek_ridden_transport_at_worldport` seek our clock to the head of the destination map's leg and write the pose in the worldport frame itself, a clock that already agreed left untouched; cleared with numbers: the timetable matches vmangos's to 18 ms over path 241's 350 818 ms cycle, the period is bit-identical to the DB row, the live skew is 32 and 144 ms on the Orgrimmar zeppelin; two divergences kept knowingly — we spare the ridden transport where `SMSG_NEW_WORLD` (`0x401bc0`) destroys the whole object manager, and we hide an off-map boat the reference draws; an off-grid view focus names its map and tile once, a ridden transport whose map disagrees with `CurrentMap` for three frames warns, and the sea-crossing probe sends itself to a dock (`WOW_PROBE_DOCK=x,y,z`) because the fleet's seams are not interchangeable — one path crosses mid-cycle, another at the wrap; why the director saw 38.5 s and not a few frames is still open (N08, 2026) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:36:48 +07:00
SamandClaude Fable 5.1	c58bca4044	the far rain stops going black, and a seated body stays on its seat ... - black rain is a zero-padded mip: `RainDrop01.blp` stores its 2×16 and 1×8 DXT3 levels as one block each where the block grid needs four and two, our decoder zero-padded the rest — an all-zero BC2 block is black at alpha 0 — and the rain's Mod2x lane reads no alpha, so every far streak that minified onto levels 3 and 4 (three-quarters and half black by the census) multiplied the scene toward black from ~43 yd, where level 3 enters the trilinear blend, to the fog at 70; the reference sizes a stored level at `bpp·max(4,w)·max(4,h)/8` bytes from its file offset and the header's `mipSizes` never sizes an upload, so a short level is completed by the next levels' own grey blocks — `[L3 | L4 | L5 | L6]` — and `Header::dxt_level_span` reads that span on both decode entry points, CPU and native byte-identical, with `pad_to` kept only for a file that ends inside a level, which no shipped asset does (543 of 32,309 DXT files carry a short level, every one non-square, none past EOF); not a mip cap and not a grey fill — the file's next bytes are what the reference draws; wow-re's cross-check settled which arm is which: the block copier `0x5a5780` is live on the shipped Direct3D default (`0x5a0a80` → `0x5a4f60`'s cell `[6][6]`), the upload's `imageSize` at `0x59f4e6`–`0x59f515` over the zero-copy pointer (`0x5a8555`) on OpenGL, the same bytes either way — and the `GL_TEXTURE_MAX_LOD` clamp is dead on every device (its gate `[0xc0f274]` has one writer, a zeroed `ebx`), so every tail level is sampled; why it surfaced now: 1642/1645's isotropic trilinear takes the streak's width for its LOD where the retired anisotropy 8 took its length, which carried the far streaks onto level 3 inside the fog window for the first time; `benilla-extract blp --mips` writes every authored level and prints the per-level texel census, and a real-asset test pins the copy byte for byte (B358/B225, 2020, 2024) - the cooldown sweep draws again: the extract's `ModelPane` arm drew the atlas cell it found in the bridge, but the cell is packed by `sync_tiles` AFTER the extract and the conversion is memoized on the engine's list — so a cooldown armed on a quiet interface extracted once with no cell, the cell arrived a frame later, nothing re-ran the conversion, and the tile lived on the atlas for the whole cooldown with no quad on screen (the 2013 harness test had re-pinged the minimap to force a second conversion, encoding the defect as the contract); the arm now publishes the request — rect, paint key, alpha, clip beside the unit ladder — and `compose_tiles` appends one premultiplied quad per cell packed THIS frame in the `UiQuadAppend` lane, a function of the bridge alone; the second defect behind it: 2019's rotation rows were uploaded to the world's shared light buffer only, while a tile's material twin binds the tile's own black-light buffer whose mat-anim region is zero-initialised, so every quadrant read the identity — and `cooldown.blp` is 0.6-alpha black outside a four-texel strip, so the sweep would have drawn as a static disc and never cleared; `MatAnimMirrors` now carries the region into every buffer a lane binds under one generation gate; measured six seconds into a 30 s stance cooldown, the button rects read means of 79 and 22 against 150 and 48 before, the top-right wedge cleared from 12 toward 3 o'clock; `WOW_TILE_TRACE=1` is the tile probe (2023) - a body inside a GameObject's hull is left where the server put it: the reference's resolver is a time-of-impact sweep and nothing else — no push-out stage anywhere in `0x634040`'s closure, and `0x632830`'s clip drops a candidate face whose clipped vertices all sit more than 1/36 yd (`[0x7ff9c8]`) behind the prism's leading plane — where avian's slide opened and closed with a Gauss–Seidel push-out whose minimum translation for a tall capsule in a small box is horizontal, shoving a seated body 0.555 yd off its stool on the first frame after the server seated it (B359); the `depenetrate` stage is deleted rather than disabled, and because the deletion alone left the body unable to walk out (a straddled seat top read as a wall at t = 0) the sweep carries the reference's band as `BACKFACE_BAND`, measured against the face's vertices, so a body walks off a stool and still stands on a sloped floor its feet sit a few centimetres under; the creature clamp cast its down-ray from 2.5 yd above the seat, a probe the reference never casts — its settle sweeps DOWN from the current position (`0x636dcd`–`0x636e45`) and the only way up is the step-up after a blocking horizontal hit — which found a cage's lid as Galen Goodward's floor (B372; seat 22.416, ray from 24.916, lid 24.667): a walker continuing its path now runs the shared `grounded_step` from where it stood with a creature's own 2.0 yd rise budget (`[0x801628]`) and takes only the Y, an idle unit casts down from its seat with the 1/36 band above it, and continuity is one server path long since `0x7c6420` re-bases on every packet; the `clmp` trace tag (`WOW_CLAMP_TRACE`) reads one clamp cast and the `move` tag marks input-less travel as CREEP (2018); and a seated body is not re-faced by the mouse: the reference's mouse-MOVE handler `0x514400` rotates the camera unconditionally and hands its facing to the body only through `0x5145e0`, whose last conjunct is `GetStandState() == 0` (`0x51460c`, the client-predicted cache), with `0x5151b0` re-testing the descriptor byte — so a right-drag while sitting orbits the camera round a body that keeps its facing and sends nothing; benilla restored the facing on the stun/death predicate only and spun a seated body in its chair with every sample, streaming a `SET_FACING` each; `MoverInput::mouse_may_turn_body` carries the stand-state term, and the keyboard turn keeps `may_turn` alone because that path stands the body (2025) - the vendor window's face and name: the stock `MerchantFrame.lua` takes its title from `UnitName("NPC")` and its face from `SetPortraitTexture(MerchantFramePortrait, "NPC")` — the `"npc"` token, `CGGameUI`'s `[0xb4e2d0]` that every NPC window reads on its own `*_SHOW` — and that seam had four defects stacked on it: `feed_interact_npc` ran a frame ahead of the net apply, so the frame `SMSG_LIST_INVENTORY` raised `MERCHANT_SHOW` still held the previous NPC (the Goldshire inn titled "Innkeeper Farley" over Brog Hamfist's stock); `feed_units`' dirty gate had no interact term, so a swap to a second vendor moved no descriptor and skipped the feed; the memo row was never written, so `UnitExists("npc")` stayed true after a close; and `SetPortraitTexture` stored its token raw where the booth slots are keyed lowercase — the reference resolves every unit token through `0x515970`, whose ten compares are all `_strnicmp`, so the binding folds on the way in; the writer now seats itself `.after(apply_net_updates).before(drive_display_facing)`, pinned in the built schedule graph rather than the source, the gate takes `interact.is_changed()`, the memo is written, and the fold lands at the one seam where a Lua token becomes a slot name — every NPC window reads the same token through the same chain; `WOW_PROBE_VENDOR_SWAP=1` stands between Brog Hamfist and Barkeep Dobbins, 5.45 yd apart, reads the name inside `MERCHANT_SHOW`, opens one over the other and closes it — five FAILs before, five PASSes after (2022) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:36:25 +07:00
SamandClaude Fable 5.1	5c26c233c3	the cooldown goes stock, early-Z comes back, and lists reach the end ... - a size-less `<Model>` pane's rect is its file's box in layout units: the reference's geometry overrides (`0x76d080`/`0x76d0d0`) answer the bbox extent whenever no size is authored, `bboxExtent · 768·√(a²+1) · layoutScale` FrameXML units (`1280 · extent` at 4:3), written into the pane's layout input when the file's facts land, re-derived on an aspect change, and ended for good by any authored size; `SetPosition` is the same unit, which is how `0x4a7b20` centres the arrow at `½·GetWidth, ½·GetHeight` through the override — so the map arrow's sprite arm and its hand-sized 33.6 px footprint go (its rect is `0.0262 × 0.0263` layout units, 33.5 × 33.7 FrameXML units at 4:3, 41.0 × 41.2 at 16:9), and the stock `ItemAnim` pane plays the item push unadapted (`ReplaceIconTexture` on the type-14 batch, the 1000 ms clamp, the stock `OnAnimFinished`), `ContainerFrameAdapters.xml` losing 0887's Lua card and its hand-measured 42.41 × 124.72 rect — the implicit one at 16:9 — and keeping the keyring fork and the three bag verbs (2015); and the cooldown is the reference's own model: the texture-transform law `uv' = R_q((uv + t − p) ⊙ s) + p`, `p = (½, ½)`, the raw quaternion component-lerped and never normalised, is one function (`uv_transform`) with the shader's UV fold as its twin — rotation and scale tracks bake per file sequence slot beside the translation, the mat-anim table gains an affine row `[cos − 1, sin, sx − 1, sy − 1]` on `anim_slots.z` whose identity is the zero row so every static batch reads exactly what it did, and a UI tile clones the twin material per animated batch and writes both rows every frame off the pane's play head, never the world clock; the stock `Cooldown.xml` runs off the chain on 2007's two handlers, the file's four quadrant quads turned by their own rotation tracks (`0 → −90°` over each quadrant's 250 ms window, clockwise from 12 o'clock) and its alpha tracks revealing the star for the flash; retired: the native `Cooldown` widget and `SetCooldown` (an Era API 1.12 never had — `CreateFrame("Cooldown")` is the reference's unknown-type refusal now), the loader's `Model → Cooldown` substitution, the app's hand-built pie and star, and our `Cooldown.xml`; `assets/ui` is eight files (2019) - `UpdateMapHighlight` gets its zone-level exits (B360): at zone zoom ours returned the nil tail for every cursor position, so hovering Deadwind Pass from Duskwood showed no label where 1.12 shows "Deadwind Pass"; the reference's `0x4a7fa0` runs two lookups at zone level and answers a NAME only — first the current zone's exploration-gated overlay list (the builder at `0x4a6ad9` gating on the explored byte), each row's HitRect scaled by 1/1002 and 1/668 against the normalized cursor, edges inclusive, the first hit naming its Area1, then the continent grid re-windowed through the displayed zone's own rect, naming the neighbouring zone or city and nothing for the displayed zone itself — and only the name-only tail is reachable from there, fileName nil and the six coords zero, so the stock frame hides the highlight quad, as the director's own 1.12 shot shows; and `SetVerticalScroll` stores the offset as given (B370, B363): the reference's `0x786db0` compares the new offset against the old and writes it to `[+0x328]` verbatim, `0x787100` re-anchors the scroll child off it, and the range at `[+0x320]` is never read on that path — every clamp the reference exhibits is FrameXML's, through the scroll bar's `[min, max]`; ours clamped into `[0, GetVerticalScrollRange()]`, the child's overflow past the FRAME, which is 40 px short of where `FauxScrollFrame_Update` sends the bar on the 220-tall skills list (three rows lost) and 15 px short on the 287-tall who list (one); the setter fires `OnVerticalScroll` only on a change, as the bytes do, `BenillaScroll_Step` bounds its wheel in Lua where the reference's handler does, and the ScrollFrame kind's prose stops calling its offset unpinned (2017) - overdraw's cheap half, the sky first: since 0588 every sky element wrote `frag_depth = 0.0`, which costs a pipeline its early-Z, so on an outdoor frame every full-screen dome fragment under the ground and the buildings was shaded in full (the gradient's `asin`, `atan2` and eight table lookups at dawn and dusk) and thrown away; one shared `sky_vertex.wgsl` — bevy's mesh vertex verbatim plus `out.position.z = 0.0` — is now the vertex stage of the gradient dome, the cloud dome, the star patches and the celestial discs and glares, and the model lane's `WOW_SKY_DEPTH` branch moves from its fragment to its vertex, so no sky fragment declares `frag_depth` and the rasterizer rejects the covered screen before the shader runs; every sky `specialize` zeroes the rasterizer bias constant the `sky_order` rung had been leaking into `DepthBiasState::constant`, harmless only while the fragment overwrote its depth, and the prose beside the pin follows it; then the retained `static_gx` pass draws BEFORE bevy's opaque pass (`StartMainPass → static_gx → MainOpaquePass`) — buildings and doodads are the frame's best occluders and a terrain fragment its dearest, six samples a pixel — so walls reject the terrain behind them instead of the reverse, the pass taking the first-use clear with it and the dev depth probe ordered after it; measured on 16 capture scenarios: 14 bit-identical, `northshire-dusk` 1 px and `water-night` 11 px, attributed to exact depth ties on static/terrain silhouettes and last-bit interpolation noise; a depth prepass and near-to-far bins are declined until a journal says where the time goes (2016) - the grass cut goes per pixel because of a quarter-mip: `0x6813d0` sets `D3DSAMP_MIPMAPLODBIAS = +0.25` on stage 0 (`0x6813f4`) once per frame before the detail-doodad pass — capture-confirmed on all 29 doodad batches, and outside the `[0x6b1000, 0x6b3200)` scan 2004 had called complete; it decides the whole look because the `clutter_cut` probe finds 6 of 31 shipped detail atlases carry a BINARY alpha pyramid (mip 0 has DXT3's 16 levels, every level below exactly two, both starting zones among them), so `a × ramp` degenerates to a global on/off and every texel of a leaf crosses at the same depth — Elwynn's survivor coverage is flat at 47.6 % across the whole 52.5–61.1 yd band, then 0.0 %, while the probe's box-filtered control erodes 47.7 → 20.9 over the same band, so the art has the gradient the shipped pyramid threw away, and `clutter_size` puts the median tuft at 0.92 yd tall, 1.03 yd deep, a full yard for the boundary plane to sweep; the escape is bilinear: full alpha needs a 2×2 all-opaque texel neighbourhood, and those run out between mip 2 and 3 (179/56/10/0/0/0 on the dominant Elwynn cell), so the reference's λ ≈ 2.9–3.3 sits past the cliff and the tuft erodes continuously — 2.2 % of a tuft's pixels leave in the last half-yard with the bias, 11.6 % without, 24.2 % one mip sharper; wgpu has no sampler LOD bias, so the clutter lane re-samples base colour at `view.mip_bias + 0.25`, keeping the sampler every other model batch shares at the reference's 0; ruled out on the way: our BLP decoder (all nine uploaded levels byte-identical against the archive's 52 793 BLP2s), mip regeneration, the `has_mipmaps` byte's value 2 (a plain boolean, an authoring marker), anisotropy and geometry; what it does not fix is resolution-dependent, and so is the reference's own look — λ with the bias is 3.22 at the capture's 648p but 2.48 at 1080p and 1.48 at 2160p, so at a modern resolution the reference itself pops leaf by leaf, and scaling the bias with the viewport is one reversible line left for a look call (2021) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:35:31 +07:00
SamandClaude Fable 5.1	b7cb2668e9	the autocast shine goes stock, and a disabled button keeps its art ... - the autocast shine is the reference's own Model: `UI-AutoCastButton.mdx` has no batch at all — four bones walking the corners of a 0.02-unit square on a 2 s loop, each with a spline emitter pouring 300 `GlowStar` sparks a second — and the tile renderer plays exactly that off the file's own tracks, so the hand-transcribed lane retires whole: `autocast_shine.rs` (the token texture, the extract's parked shine sites, the append-lane producer and its motion goldens), the extract's shine side channel, and `PetActionBarAdapters.xml`, since the pet button's numbers are the reference's (`setAllPoints` at `scale="1.2"`) and nothing was left to adapt; what stays is the director's two look calls, re-seated onto the reference's own objects — `SpellBookAdapters.xml` puts the stock `$parentAutoCast` Model at 37×37 CENTER (0,0) with `SetModelScale(1.48)` and the overlay at 71.53, the two tests that pin them reading the live panes, and deleting that file is the exact reference (36×36 at CENTER (1,1), 1.22, overlay 60); `assets/ui` is nine files (2014) - a button's shown state texture is latched on the transition, not resolved at paint (B369): `ReputationDetailAtWarCheckBox` has a `<NormalTexture>` and no `<DisabledTexture>`, and once `ReputationFrame_Update` `Disable()`d it ours drew nothing — `ButtonState` resolved the shown texture as a pure function of (enabled, pushed, held+hovered); the client's `SetState` `0x779790` gates both its hide-old and show-new steps on the NEW state's slot being non-null, with no Normal fallback anywhere in the function, and the ctor `0x7786a0` ends in `SetState(NORMAL)`, so every button reaches its first `Disable()` (`0x77ffd0`) already wearing its normal art and keeps it; `ButtonState` now carries the client's two words, `shown` (`+0x4c4`) and `state` (`+0x328`), written only by `set_state` and the `0x778fd0` setter family's conditional push, with transitions applied where the client applies them — `Enable`/`Disable`/`SetButtonState` and the four mouse edges — and the old `pushed.or(normal)` special case was this rule seen from one side; the stock corpus has 70 such declarations, so empty spellbook slots wear their `UI-Quickslot2` ring again, which 0227 had removed on a premise the shipped Lua falsifies in one line (`SpellButton_UpdateButton` calls `Disable()` at runtime); and `Disable()` switches off the whole HIGHLIGHT draw layer — `0x779160` follows `SetState` with `push 4; call 0x76a730`, the per-layer enable on the same `[frame+0x198]` array `Enable/DisableDrawLayer` writes — so a `<Layer level="HIGHLIGHT">` region no longer draws over a disabled button and the `enabled &&` term on the HighlightTexture alone is gone; the reputation tests' header stops claiming the page and the watch bar were ours (2011) - the world's mouse buttons are latched, not read raw (B364): right-clicking a Who-list row swung the camera round to behind the character — ~180° over nine seconds on a body that never turned — because the click was the UI's but the camera's input command word (`[InputControl+0x4]`) built its two mouse bits straight off `ButtonInput` with no pointer-over-UI term, `RIGHT_MOUSE` alone classifies as `Turn`, and Smart's Turn row is `(0.0, 1.0)`, so the return armed on the press edge and ran to completion; 1.12 never reads the mouse for any of this — it reads two bindings, `TurnOrAction` and `CameraOrSelectOrMove`, and a press a UI frame captured dispatches neither — so `camera::WorldMouse` latches each primary at the press and holds it to that button's own release, written in its own system ordered ahead of every reader (latched rather than re-tested, because a level test would hand a button back to the UI mid-drag the moment a frame appeared under the locked cursor, and an edge is what arms a return); five readers move onto it — the follow word's mouse bits, the both-button run, `/follow`'s both-button cancel, the look session's engage, chord cancel and other-primary tests, and the both-button → single-button handoff — and `control` drops its `PointerOverUi` and `&Camera`, the viewport-vs-UI question no longer being its to ask - the clutter onset is the cutout, and it gets the reference's own dial: the ramp runs 52.64 → 69.86 yd but the draw alpha-tests `texel.a × ramp` against `detailDoodadAlpha` = 128/255 with the ramp capped at 252/255, so nothing survives past 61.11 yd — grass appears there already at 50.2 % opacity with 8.5 yd left to reach full, and the last third of the ramp is spent where nothing is drawn, which is exactly the "started before it became visible from another gate" the director saw; `detailDoodadAlpha` is a console command in the reference (`0x6739a0`, registered through the command table `0x63f9e0`, not `CVar::Register` `0x63db90`, so it never persists), and we had it only as the boot-time `WOW_CLUTTER_ALPHA`, which cannot be A/B'd against a running client — `/console detailDoodadAlpha <0-255>` now sets it live, bare reports, and out of range is rejected rather than clamped (`0x6739b9: cmp eax,0xff; jbe`), classified in the console drain of 2008's `ConsoleExec` seam so it never reaches the CVar store, the bubble re-cut on a change because the ref is baked into the material; at 128 grass appears at 61.11 yd at half opacity, at 8 the onset walks out to ~69.8 yd and fades in from nothing; a late-build tripwire in the streamer — a warning whenever a chunk is built whose box is already inside the fade horizon, naming whether the distance gate or the per-frame cap was the cause — fired at once on the second: the build budget was spent in query order, so a chunk 90 yd out where the ramp is already zero took a slot ahead of one at 48 yd in plain view; candidates are sorted by distance now, and a 30 s Elwynn run reports one late-build frame, the cold first, then silence (2012) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:34:43 +07:00
SamandClaude Fable 5.1	c957f98d84	grass fades on a plane, model panes render, and the FPS journal ships ... - the grass fade is a plane across the view, not a ring around the player: `0x6b2b80` texgens the camera-space position through `T(0,0,−52.5)·Ry(π/2)·S(1/17.5)`, so the fade coordinate is view-space depth, `u = (z_eye − 52.5)/17.5` — we measured `distance(eye, fragment)`, a sphere, which meets the ground as a circle centred on the player that slides as they walk and bites a piece out of every tuft it crosses, and that was the hard border; the ramp is `0x6b2320`'s quantised 64×8 table, `alpha = 4·(63 − col)`, read bilinearly as `(254 − 256·u)/255` and capped at `252/255` — plateau to 52.64 yd, zero at 69.86 yd, near grass 98.8 % opaque and never 100 — where we had a clean 1 → 0 line; the product `tex0.a × ramp` is what the 128/255 `detailDoodadAlpha` cutout tests (both stages MODULATE, `[0xc7b2fc]`, confirmed verbatim in all four apitrace captures), so the tuft eroding out at 61.11 yd is the reference's own behaviour and the "fix" that would have removed it never landed; the pass owns its render state — blend mode 2, ALPHAREF 128, cull off, depth-write on, the per-slot draw `0x6b2d60` binding nothing but the texture — where we passed the batch's authored blend and cull through, and the `clutter_state` probe counted the cost: of 322 detail batches, 2 author Opaque (no cutout, their margin drawn as a solid card) and 5 single-sided; the build gate measures the chunk's box against the frustum corner ray derived from the live projection (≈1.31× at 16:9, ≈1.45× at 21:9), not its centre plus a 24 yd margin that popped whole chunks in on relief (2004) - a `<Model>` widget's animation state is Lua-observable and lives in the engine: a private scene clock its own `OnUpdate` advances by `trunc(elapsed·1000)` ms on visible frames only (`0x76d7f0`), an arm with an anchor the sampler re-reads every frame (`0x7121a0`), `OnUpdateModel` at the top of every paint (`0x76d1a0`) and `OnAnimFinished` from the completion callback (`0x76cdc0`); the file's sequences and bounds are facts the host hands over, `SetModel` seeds Stand now or when they land (`0x70ebd0`) with a queued arm replaying after the seed, and the XML `scale=` on a model pane is the model's scale (`CSimpleModel::LoadXML` `0x76cac0`), which the loader had never read (2007); and the pane renders its M2 — every visible pane with a file gets a cell of one render-target atlas at its device size through one orthographic camera, the root at `1280·modelScale·layoutScale` FrameXML units per model unit, the tile's `AnimationPlayer` paused and seeked to the engine's play head, the light black, particles in their own unit (`768·√(a²+1)` per model unit, neither scale), composited as one premultiplied quad at the callback rank; two corrections off the bytes: `OnUpdateModel` fires only with a model installed (`[widget+0x318] ≠ 0`) and `OnAnimFinished` fires once per arm for a looping sequence too (`0x719370`); the loader now applies a `<Model>`'s `file=` at all, which every XML-declared pane had lacked; retired: the minimap ping's byte-measured sprite, since `MiniMapPing` and `WorldMapPing` draw their own file on the pane's clock (2013) - the stance feed fires `UPDATE_SHAPESHIFT_FORMS` only for a list change: the reference fires it at `0x4b28ff`/`0x4b2e43` from learn, unlearn and rank-replace alone, while ours fired on any change of the pushed view, and every fire runs the stock `ShapeshiftBar_Update`, which shows `ShapeshiftBarMiddle` unconditionally past two forms and `Show()`s an already-shown frame — no `OnShow`, no manage pass — so a stance switch or the category-47 cooldown arming and expiring left the middle strip drawn alone over a raised bottom-left bar; castability now rides `SPELL_UPDATE_USABLE` (`0x4b31c0`), active/texture/cooldown are pushed silently and repainted on `PLAYER_AURAS_CHANGED` and `SPELL_UPDATE_COOLDOWN`, the feed ordered before both handler sets so they read the fresh list (2009); and the loot-roll frame waits for its item template (B371): `GroupLootFrame_OnShow` paints once and nothing repaints it, and we fired `START_LOOT_ROLL` on the frame the packet arrived, against a snapshot whose name, texture and quality were still `None` — the reference's `0x61b310` ends in the item-template cache lookup `0x55ba30`, a hit firing the event there and a miss firing it from the arrival callback `0x61b460`, so the feed holds a fresh roll's event until the template answers, with two releases that are ours: a negative answer (a frame that never opens cannot even be passed on) and a 2 s deadline paid out of the roll's minute; `arg2` stays the wire's countdown, as `0x61b430` pushes it (2010) - the FPS journal ships to players: it had registered inside the dev-only `DevToolsPlugin` and been armed by an environment variable, so nothing we shipped could have written a row on the one GPU-bound frame in the corpus, a Steam Deck's at 94 % busy; `/console fpsJournal 1` — a CVar of our own, persisted like every CVar — opens `benilla-config/Diagnostics/fps-journal.csv` and appends one row a second, `0` closes it mid-run, and `WOW_FPS_JOURNAL=<path>` stays as the harness lever; eight GPU columns appended, `gpu_ms` then `gpu_opaque, gpu_static, gpu_transp, gpu_glow, gpu_post, gpu_ui, gpu_other`, read off bevy's `RenderDiagnosticsPlugin`, which times passes where the device offers `TIMESTAMP_QUERY_INSIDE_PASSES` (Vulkan and DX12) and leaves the cells empty on an Apple GPU, empty being the honest value; `static_gx`, the four `ffx_glow_*` passes and `ui_gamma_decode` open diagnostic spans so a city's own draw is named in the split; a fresh file opens with the adapter, backend, driver and `gpu_spans yes|no` line; and a probe's `/console` line, which skips the stock `ChatEdit_ParseText` and had landed in a parser arm that knew only `reloadUI`, now forwards to `ConsoleExec` and writes a CVar the way the typed line does (2008) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:33:58 +07:00
SamandClaude Fable 5.1	f9293ffbe8	the lamppost halo stops stacking, and the stance bar seats on login ... - the lamppost halo stops stacking after a short-range teleport: bevy registers `add_visibility_class::<Mesh3d>` as an add hook that pushes the mesh's class id onto the entity's `VisibilityClass` list on every (re)add, nothing dedups the list, the required component survives a `Mesh3d` removal, and `check_visibility` queues an entity once per entry — so 1949's parking, which removes `Mesh3d` after thirty hidden frames and re-adds it the frame the part is shown, drew a part one more time per hide-then-show cycle; opaque parts hide it at the same depth and colour, an additive glow card adds itself to itself (×2 after the first teleport that hid the lamp for half a second, ×3 after the next), and a loading screen despawns the tiles and respawns fresh entities, which is why a far teleport and a relog never showed it; the unpark queues one dedup after the hook, the test registers bevy's own hook and requirement in a bare world and reads 2 entries without it and 1 with, and on the rig, with the away camera facing the lamp past a 177 yd far clip, a build without the dedup read `stacked_parts=278` after one hop back where the landed build reads 0 (2003, 2005); the instruments that excluded the other mechanisms stay — the inspector's glow-card line names the material's additive bit, fade/unlit flags, depth write, texture residency, tag alpha and live scale, `STACKED n` on the card and `stacked_parts=` on the probe line count the duplicated class lists, and screenshot and pick bursts now keep the rig window awake, because an occluded window recaptures one stale drawable and a day of A/Bs had read identical to the decimal - `perf::blend_check` reads the half of a glow card's look the material cannot: per `Transparent3d` item, the pipeline descriptor's colour blend against the material's additive marker — the marker folds the radial alpha into the colour in the shader and gives the pipeline its pure (ONE, ONE) add, and 0071's wrong-pipeline class was immunised only for opaque batches; mismatches count into the perf pill as a red `blend xN` the frame they happen, ride the probe line as `blend_mismatch=`, and are tracked across frames and named — object, texture, which pipeline, PERSISTING; its own first harvest was the tracker: an additive material realized this frame is extracted, prepared and drawn this frame, and a tracker reading the `Added` event a frame late called every first draw a mismatch, so it runs after `AssetEventSystems` like bevy's own specialization check; and the hook-accumulation sweep states the class so it can be swept: bevy 0.18.1's insert path fires `on_add` only when the component was absent, `on_replace` on every outgoing value, `on_remove` on removal only — a push on `on_add` accumulates across remove→re-add, a free on `on_remove` leaks across an insert-overwrite; every hook in the bevy crates we link (the visibility class on `Mesh3d`/`Mesh2d`/`Sprite`/`Text2d`/`MeshletMesh3d`, the cluster class on every bevy light) and every removal of a hooked component in our tree was read, parking's `Mesh3d` is the only accumulating pair, the UI quad pools reuse entities by insert-over and fire no add hook, and the one asymmetric freer, `PropProbeSlot`, freed its probe slot on `on_remove` — an insert over a live slot leaked it for the session, held off by a remove-then-insert discipline at the one re-seat site; it frees on `on_replace` now (`RigSkin`'s shape), the re-seat is a plain insert, and a test overwrites a live slot and reads it freed (2005) - the stance bar loads after everything the managed-position pass reads: the stock `ShapeshiftBar_OnLoad` shows the bar, its OnShow runs `UIParent_ManageFramePositions`, and the stock pass indexes `ReputationWatchBar` unguarded — our manifest seated the bar beside the other bar files where the reference's toc has it at l.74, after the reputation frame at l.57, so the pass raised at `UIParent.lua:1618` on every login and never reached the arm that seats the shelf: the red login error and the sliver above Defensive Stance (the shelf's plate at its XML anchors, a few pixels off its seat, healed by any bottom-bar toggle re-running the pass) were one defect from 1988, whose retired copy of the pass had guarded every read — and the three-form case 2000 could not reproduce; the bar sits at the manifest's tail right before `WorldStateFrame.xml`, whose own OnLoad runs the pass and indexes the shelf textures the stance bar declares — the reference's own reason for l.74 before l.89; a manifest test states the law with the pairs read off the pass's body, and `load_default_ui` now folds the VM's handler errors into its failures, because a raise one call below an OnLoad landed only in the VM's error list and shipped through the manifest test, the full-manifest harnesses and the smoke's WARN alike (2001) - `UnitName` answers `UNKNOWNOBJECT`, never nil, for a resolved unit whose name is still in flight: a hunter casting Call Pet with the stable window closed got `PetStable.lua:129: attempt to concatenate a nil value`, because stock `PetStable_OnEvent` repaints on `UNIT_PET` without asking whether the frame is shown and the paint concatenates `UnitName("pet")` the instant `UnitExists("pet")` is true, while the pet's name is a `CMSG_PET_NAME_QUERY` round trip away; the reference's `UnitName 0x517020` pushes nil in exactly two cases — the `"player"` fast path (`0x517083`) over an empty name buffer, and a token that resolves to GUID 0 (`0x5170c0`) — and every other path ends in a string: `CGUnit_C::GetUnitName 0x609210` looks the pet up in `petnamecache.wdb` by `UNIT_FIELD_PETNUMBER`, re-sends the query on a miss or a stale stamp, and falls to `0x609324`, the GlobalString `UNKNOWNOBJECT` ("Unknown") read through `0x703bf0`, or the literal `"Unknown Being"` at `0x860fa4` when the global is missing or empty; the binding now reads the global out of the VM's `_G` at call time, so a translated `GlobalStrings.lua` translates this too, the stock file is untouched, and 1688's "nil for an unresolved name is faithful" is corrected — the reference paints `Unknown Level 58 Boar` into the hidden frame and the next repaint reads the name; `GameTooltip:SetUnit` still titles a pending-name unit with an empty line, a next move (2002) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:32:13 +07:00
SamandClaude Fable 5.1	ed92f71af8	the gossip window stops flashing, and the chat plate fades in again ... - the gossip window stops flashing on a first-visit sub-menu: a guard's "Where is …" option answers `CMSG_GOSSIP_SELECT_OPTION` with a new `SMSG_GOSSIP_MESSAGE` on the same NPC whose text id is a first visit, and 1508's hold made the snapshot `None` while the greeting query was in flight, which the feed read as `Some → None` — `GOSSIP_CLOSED` (`HideUIPanel`) at the select, `GOSSIP_SHOW` (`ShowUIPanel`) when the text landed, the window gone for a server round trip, once per text id because the record cache serves every later visit; the reference's `0x4e2010` on a cache miss sets its select latch (`[0xbbb670] = 1`) and returns with no greeting write and no event, `0x4e1fa0` closes only on a zero guid, and `SelectGossipOption` (`0x4e2320`) is refused while the latch holds — so `GossipState::text_pending` names that latch, `feed_gossip` returns before touching the VM while it holds, both drains refuse on it, the NPC name query is asked ahead of the gate so a first visit's name rides beside its text query, and the landing repaints in place (1994); and the ready check is read off the bytes: `0x4ba40e` stores the constant 0 for any guid-matched answer, the wire byte only gates the `RAID_MEMBER_NOT_READY` print ("%s is not ready") on a 0, `ConfirmReadyCheck` sends one byte, 1 or 0, only while the deadline is armed and unexpired, `DoReadyCheck` sends an empty body, the summary lists only members who never answered, and the early close fires iff no member is both pending AND online (`[entry+0x18]` bit 0) — an offline member who never answered no longer holds the check open and is still listed as AFK when it closes; the `SMSG_GROUP_LIST` leg's close predicate (`0x4ba5f0`) keeps 1989's reading, with no online term (1997) - the chat menu's rows hug their left edge: a `UIMenuButtonTemplate` row has no `<ButtonText>` — `UIMenu_AddButton` labels it with `SetText` — and the only thing placing the label is `<NormalFont inherits="GameFontNormal" justifyH="LEFT"/>`; the reference's lazy creator (`SetText 0x778dc0`) hands the fresh string to the adopter `0x778d20`, which anchors an unanchored label by the BUTTON's normal embedded font's justify word (`[button+0x390]`: LEFT→LEFT / RIGHT→RIGHT / else CENTER) and links it to that font, where ours anchored by the string's own ctor-default word (CENTER) and the loader wrote the justify onto a force-created label after the anchor was already decided — so every row sat centred in its fixed 104-wide slot and "Macro" ran into "/macro" (1.12's `UIMenu` does not auto-size; `UIMenu_AutoSize` is a later client's); `ButtonState` now carries each state font's own justify, one adopter (`button::adopt_label`) is the tail of both `SetText`'s lazy creation and `SetFontString`, the loader never touches the label, and paint follows the current instance's justify behind the severance mask like every other axis (1996); and the chat plate's hover fade survives a quick exit and re-entry: the stock `FCF_OnUpdate` queues the tabs' fade-out with `FCF_ChatTabFadeFinished` as its finished callback, which nils `chatFrame.oldAlpha` unconditionally 0.15 s later — re-enter inside `CHAT_FRAME_FADE_TIME` and it lands under a live hover, the plate arm's gate never passes again, no later leave clears `hover`, the hover-start re-read never runs, and only `FCF_SetWindowAlpha` (the tab menu's opacity slider) reseats it, which was every clue in the report; the flick from the text to a scroll button and back is the gesture, so an existing window looked broken from a session's first hover; the repair is the smallest guard at the one line — the finished callback leaves `oldAlpha` alone while `hover` is set — installed from Rust after the manifest loads (`manifest::install_chat_plate_guard`), never written into the stock file, and named as a deviation from the reference's own Lua (1998) - a layer drains its quads, then its text, then its model scenes: `ZKey`'s one font-string bit becomes a two-bit batch rank (Quad, Text, Callback) — the client's layer batch (ctor `0x772e80`) carries its quads at `+0x10`, its `CGxStringBatch` at `+0x18` and a `RENDERCALLBACKNODE` list at `+0x1c`, `0x76fb00` drains them in that order, and `0x76d160` registers a `Model` frame's scene for the ARTWORK batch only — so a `ModelPane` quad takes `ZKey::callback(Artwork)`, after every texture AND every font string of its (strata, level, layer) whatever their link stamps and before anything in the next layer, closing the residual 1992 named on the world map (the arrow is linked first under `WorldMapFrame`, the area label and the overlays later); two scenes in one bucket keep registration order, and the link stamp gives up one bit (19 → 18), which the arena's renumbering at the cap makes unobservable (1995); and the two full-screen vignettes are dead code in 5875: the string `PlayerStatus_OnUpdate` occurs zero times in `WoW.exe`, the per-frame pump runs a frame's OnUpdate by registry reference (`0x70518c lua_rawgeti`, `0x705199 lua_pcall`), the whole `lua_pcall` census has exactly two by-name calls from C, and the only `UIFrameFlash` sites for `LowHealthFrame` and `OutOfControlFrame` in the chain are inside that uncalled function, while the `PLAYER_CONTROL_LOST` arm only sets `UIParent.isOutOfControl` — so 1988's "losing control flashes the vignette" is withdrawn, the two `.blp`s ship and are never loaded, nothing in code changes, and the manifest's note on the two frames says why no driver exists (1999) - a tiled UI sprite wraps only the axis whose mapping runs past the texture: at four forms `ShapeshiftBar_Update` maps the stance shelf's middle strip `SetTexCoord(0, numForms-2, 0, 1)` — tiling along u, spanning exactly `[0, 1]` in v — and the repeat-sampled image it was served from wrapped BOTH axes, so bilinear at the strip's top pixel row reached `v < 0` and wrapped to the texture's last row, opaque grey on the 32×32 `ShapeshiftBarMiddle.blp` whose rows 0–7 are transparent: a one-device-px hairline along the strip's top edge over the world, reading 0.81–0.84× its neighbours' luma against 1.00 over the clamp-sampled caps (predicted 75 over a world of 100, measured 74–77); 1608's `uv_clamp_window` skipped an axis spanning the whole texture because "the sampler's own clamp already is this", true for the plain sprite and false for the tiled one; `tiling_axes` now decides the wrap per axis with the same tolerance, `WorldAssets::sprite_texture_wrapped` bakes `Repeat` where asked and `ClampToEdge` elsewhere, cached by (path, u, v), and the backdrop pieces and the chat bubble's edge keep both axes on; `WOW_DEMO_BOTTOM_BARS=0` leaves the action-bar demo's bottom bars down, the only state the shelf draws in (2000); the duplicate-placement census rides the probe line (`orphan_parts=`, `tiles=`) and the inspector card (parts alive, parts drawn this frame, each glow card's live world scale): a placed doodad or WMO part is spawned by exactly one placement, so a part alive outside every placement's entity list outlived a respawn, with the retained pass's fader exiles named as the one population legitimately outside the registry Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-06 19:31:38 +07:00
SamandClaude Fable 5.1	dbcf60fc31	the loading screen lets nothing through, and the micro buttons go stock ... - units could be hovered through the loading screen — the sword cursor gave it away: the world pick chain has no state gate, one of fifty-one input sites, so a gate per consumer is a rule that regresses; the cut is at the source instead — one `PreUpdate` system empties the button planes, drains the raw queues, zeroes the accumulators and blanks the window's cursor position while the cover is up (the alt-tabbed-away state every consumer already handles); a press under the cover never happened, a held button is released once, a gesture in flight is cancelled, and a wall test counts the channels; the reference does the same — its raise registers a veto stub (`0x406a50`) on five event-bus categories — and one gap closed with it: the reference parks the plain arrow and drops any item overlay before the screen goes up, where benilla kept a held item's icon over the loading art (1986, 1990) - the micro-button row is the reference's own `MainMenuBarMicroButtons.xml`, our transcription deleted — no engine verb was missing, and the whole engine-side cost is one alias, `SoundOptionsFrame = OptionsFrame`, because the stock file reads the reference's three options windows and ours is the one window for all three; what the stock file brings back: `UpdateTalentButton`'s own −2 re-seat (the five buttons after the talent button sit 1 px further right — the reference's own pixel), the raw binding token in the hover label (`(C)`, `(SHIFT-P)`), the sixty-second talent-button pulse on a ding while the sheet is closed (0304's scope-out retired), the keyring arm, and "Abilities" on the spellbook hover for a character with no spells; `BenillaActionBarArt_SeatAbove` goes with the file — a child is born one level above its parent (`SetParent` `0x76ab65`), the seat it set by hand (1987) - hovering the player arrow on the stock world map showed an empty tooltip: narrower than 4:3 the map is scaled under 1, and our `MouseIsOver` had dropped the reference's division by `GetEffectiveScale` as "always 1" — the frame getters answer in the frame's own units, `GetCursorPosition` in the screen's; the division is back, and the region getters divide by their owner's scale too; and the ready check times out: the stock `UIParent.xml` ticks `CheckReadyCheckTime` every frame, the one unbound name in that file's way, so the mechanism behind it is built — a 30 s deadline and a "has not answered" flag per roster entry, the leader getting no popup while members print "%s has initiated a ready check" and get it, and on expiry `RAID_MEMBERS_AFK` with the unanswered names or `READY_CHECK_NO_AFK`, the last answer closing the check at once (1985, 1989) - a review of the floor grind closed eight stale-input and bypassed-side-effect siblings the gates had left open: a never-placed card places on a still frame, the dormant-emitter skip sits below the emitter's lifetime blocks, PVS changes (not only spawns) un-still the gates, the appear ramp's retirement and a far-side unmark come due, a host born under a parked camera is judged, the projection counts as a still input, a fresh palette range is zeroed, the stable feed runs once more after a close, the unit publisher's viewer query is filtered, solo UI materials are reused by key, and the draw-list cache compares exactly; and three doc links the chat window's stock move (1948) left dangling point at what replaced them — the colour table and the tell ring are the stock Lua's now — which is what had the doc-links gate (1925) red on main Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 21:01:38 +07:00
SamandClaude Fable 5.1	a5382d7b37	still frames skip settled work, and a guild can design its tabard ... - on a still frame the world stops redoing settled work — each cut an exact gate, measured with the process pinned to the efficiency cores, where the main thread IS the frame: point lookups go through the water index, a terrain column tests its cell's four fan triangles, not the chunk's 256 (−1.3 ms), the flare march resolves its tile once, the static-world publish and light blob write only on change, billboards, emitter gates (off the new-portal edge), doodad-anim verdicts, the visibility walk, the unit publisher (its filter a named type), a rampless unit's alpha and the stable feed skip when nothing moved, the rig evaluator reuses its scratch, dormant emitters skip their step, child wiring visits pending emitters only, the shade map persists, and the terrain streamer skips its walk once settled; alone parked 25.4 → 21.6–22.8 ms of main thread, 32 → 36–38 fps there with nothing on screen (1979, 1982) - the UI tick got the same treatment: a solo run's colour rides the material, not the mesh — the glow pulse's one `Assets<Mesh>` write a frame had armed bevy's asset-changed probes over ~44 k `Mesh3d` rows (−1.0 ms); script object ids index a dense map instead of a hash, where every scripted method call starts (−0.7 ms); the draw list is fingerprint-cached, walked every frame and sorted only when it changed (the list a named type); and the instrument that made the gates count at all: the first battery read flat until the probe printed how many frames each gate input had read as changed — `noisy=[cam:0,dbg:299,view:0,win:0,claim:0]` — one dev system took `&mut` on `DebugState` every frame to test a flag, so bevy marked it changed on 299 of 300 frames and silently disarmed every gate reading `debug.is_changed()`; it reads the flag first now, and the counters stay on the probe line (1982) - two of the new gates were wrong the day they landed and are fixed in the same snapshot: the streamer's settled skip engaged the frame `complete` flipped, with the retained pass still baking the new cells, and froze the undrawn count `presentable()` reads above zero — the loading screen stayed up for good after a `.tele`; it now waits for a paced, presentable, complete world and keeps that count fresh; and the still-frame gates read the camera's `GlobalTransform`, which moves only after propagation — which runs after them — so a teleport frame carried every host's pre-snap verdict across the snap (a lamppost glow due to park kept its clock, over-bright after the `.tele`); the gates read the seat's own `Transform` write too (1982) - a guild master can design the guild's tabard: the reference's own `TabardFrame.xml` joins the manifest (benilla never had the window), opened only from the wire — the click ladder's TabardDesigner arm now sends `MSG_TABARDVENDOR_ACTIVATE` and the Battlemaster arm its own hello (1861's two deviations retire); the `TabardModel` method table runs over five ints bounded by the `.rdata` table `{170, 17, 6, 17, 51}`, the design previews on the player's own body so the pane and the world show one look, and `Save` runs the fourteen pre-flight checks in the reference's order; the world frame is the reference's own `WorldFrame.xml` too — the framerate readout is live on CTRL-R — and the carve settled its kind: a `Frame` to Lua, a one-shot type, the sole tenant of stratum WORLD, and not "over the UI" when hovered, so world clicks and camera look stay live while its scripts fire (1977, 1983, 1984) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 21:00:42 +07:00
SamandClaude Fable 5.1	404708859a	every dialog is the reference's own, and the trainer loads on demand ... - every dialog is the reference's own: `StaticPopup.xml` + `StaticPopup.lua` load off the player's chain, and 0308's engine with its 37 dialogs leaves `UiPanels.xml`; all 76 of the reference's dialogs are the stock table's, 28 of which no file of ours had ever registered (the resurrect family among them); the nine hidden listener frames that existed only because benilla once had no `UIParent_OnEvent` come home to it verbatim from UIParent.lua, and the seven files that were nothing but a dialog and its driver are deleted; what the swap surfaced: the reference capitalises the countdown's unit word ("5 Minutes until release") where ours had lowercased it, and ESC in the typed delete confirm's box is swallowed — the entry names no escape handler; arms for events this engine never fires (PLAYER_SKINNED, TRADE_REQUEST) wait on their feeds rather than being armed blind (1960) - the social window and the unit menus are the reference's own files: `FriendsFrame.xml` loads off the chain and our transcription is deleted — its twelve dialogs re-homed, `FRIENDLIST_SHOW` fired as the answer to an explicit `ShowFriends()`, and `Get`/`SetLookingForGroup` a client-side flag until their bodies are carved; `UnitPopup.xml` loads off the chain, our row catalogue and its driver frame go, and nine verbs are built for it — `HasFullControl` off the player-control flag the reference reads at `0x51a158`, `UnitPlayerOrPetInParty`/`InRaid` over a new unit owner, and the six pet one-shots, each pressing the bar's own reaction or command slot word through the same latch and `CMSG_PET_ACTION`, confirmed by wow-re's carve; the Guild Leave/Promote confirms 1958 believed nothing registered were our social file's all along, and the duplicate went with it (1958, 1959) - the trainer loads on demand: our `TrainerFrame.xml` is deleted and no row replaces it — `Blizzard_TrainerUI` loads off the player's own chain through the reference's `ClassTrainerFrame_LoadUI`, and the LoadOnDemand `Blizzard_*` addons are registry rows read off the chain; the timing IS the mechanism: the list builder writes the state-filter mask to 3 on every `SMSG_TRAINER_LIST` (`0x4d75d9`) and the stock addon pushes its saved filters exactly once, on its own ADDON_LOADED — which fires on the session's first trainer, after that packet, so the saved filters win that open and an eager load would never have applied them; the loader holds an anchor to a sibling not built yet and applies it once the subtree exists; and `UNIT_FLAGS` fires only on a change between two snapshots — wow-re's trio found the per-field watch bridge runs no notify pass on the create leg, correcting 1953 (1957) - the additive window, built and withdrawn the same hour: under `BLEND_ADD` every additive particle draw commutes with every other, so same-key additive draws can be regrouped exactly — built in the effect lane's merge walk and measured on the crowd rig (Stormwind, forty partybots): the adjacency merge already folds about half the effect items and the regroup found two more folds in a frame of ninety-three, so the pre-regroup walk is what runs; what stays: an `fx=` items/merged census on the FPS probe line, a `DRESS_PART` line per drawn part of the self body on the dress census, and the regression the A/B uncovered: 1948's stock chat migration dropped the drain's plain-line sender, so every `WOW_PROBE_CHAT` rig — the crowd rig's forty `.partybot add`s among them — had gone silent on the server; a probe's slash-less line reaches the wire as SAY again (1955) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:57:24 +07:00
SamandClaude Fable 5.1	a43e2c9a88	the pet bar and skills tab go stock, and the raid's profile is flat ... - the pet action bar is the reference's own `PetActionBarFrame.xml`; our 559-line transcription is deleted, and the reference's bar brings what it had left out — the slide, the Ctrl-held and locked gates on show/hide, the hotkey pass whose string-versus-number comparison leaves every corner empty — with the three pet dialogs moved verbatim into the popup slice; one adapter draws the autocast shine (the reference's `<Model>` draws nothing in this engine, so each button gets the token-texture shine at the reference's own 1.2); the stock file registers four events nothing fired, built rather than listed — `PET_BAR_UPDATE_COOLDOWN` when the ten cooldown triples alone move, `PLAYER_CONTROL_LOST`/`GAINED` on the edge of the control flag (`0x4958e0`'s change-gated fire), `PLAYER_FARSIGHT_FOCUS_CHANGED` on the edge of the farsight descriptor, and `UNIT_FLAGS` on a change of the raw flags (its exact fire law a dispatched question) (1953) - the skills tab is the reference's own `SkillFrame.xml` at its slot between ReputationFrame and HonorFrame; our 874-line transcription is deleted, the `UNLEARN_SKILL` dialog moves verbatim into the popup slice, and `CancelSkillUps` is built — the page's live Close button calls it before hiding, so every close reaches it; it is the reset leg of the temp-point table (`0x4d3e30`, no arguments, no returns) over a table this client keeps empty by construction, so the reset moves nothing, which is the binding's whole effect on this data; `AddSkillUp`, `RemoveSkillUp` and `BuySkillTier` sit on buttons the reference shows only for skill points or step/rank costs no 1.12 server sends and stay named as open gaps; and the expand tab's fit is the reference's load-time law (`GetTextWidth()+45` once, in OnLoad), so the transcription's zero-measure guard and re-fit are gone (1956) - a 45-bot Blackrock raid circling with sound on holds 60 fps with the main thread at 6–7 ms while the HUD's ms readout says 17–18: that number is the sum of every thread's CPU per frame (main, render, compute pool, audio), not the frame time — the raid costs +6 ms spread over four threads with no thread owning it; two suspects refuted by measurement: mimalloc as the process allocator (interleaved: no change on any thread, reverted the same hour) and page faults (59 a frame on the probe, tens of microseconds, not the render thread's 2.1 ms of kernel time, which is the Metal driver's own work and scales with draws); so the probe counts draws per phase family (transparent outnumbers opaque, mostly the buff wave's particles and nameplates) and `WOW_PROBE_SHARED_SKIN=1` prices cross-body instancing before it is built: nothing moved, because bevy instances only identical meshes and a geared body's merged geoset set is its own mesh (1954) - three fixes, each removing work a 15 s sample named: booth cameras skip cluster assignment (no bevy light ever reaches a booth, and the default config rebuilt an empty cluster grid per booth view per frame — 0.2 ms of the raid's main thread), the rig palette's watermark clone reuses the row buffer from two publishes ago instead of calloc'ing the 650 KB table every publish (safe because the GPU only ever receives dirty ranges), and prop probes upload the dirty span rather than every allocated row (circling a city streams props in and out most frames, and each change re-sent hundreds of KB for one row); no bucket of the main thread's profile is above a millisecond — the raid's +2.5 ms there is forty-five bodies' worth of every per-entity lane at ~55 µs a body, and the one lever left that is all ours, one texture array per body folding ~9 draws to ~3, is sized at ~1.5 ms and not started (1954) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:56:17 +07:00
SamandClaude Fable 5.1	2ec92beeed	the chat and spellbook go stock, and hidden props park their meshes ... - the chat window is the reference's own `ChatFrame.xml` and `FloatingChatFrame.xml`, run off the player's chain; our 2216-line transcription is deleted, the feed fires `CHAT_MSG_*` and the reference's `ChatFrame_OnEvent` prints — no app-side composition into the windows, no routing table, no whisper chime or flash of ours, and the transcribed edit-box machine (sticky type, tell ring, Tab cycle) retires with it; the verbs the two files call are built to the bytes — the whole `0x98` chat-window record and its eight verbs, the 94-entry chat-type colour registry, the language walk, the channel verbs as a command queue, the slash handlers' emote/roll/assist/uninvite/console/logging verbs; `chat-cache.txt` is written and read in the reference's whole grammar, the 68-entry message-group table in flag order and the no-file boot included, and the chat bindings and macro lines run through the reference's own arms (1948) - the spellbook is the reference's own `SpellBookFrame.xml`; our 965-line transcription is deleted, and two quirks it reproduced by hand (the pet page ignores its page number, its arrows write the spell book's counter) now come from the reference's lines; one adapter remains because the reference's autocast shine is a `<Model>` that draws no pixels in this engine — the script puts the token-texture shine beside each button's model and carries it with the model's shown state; the nine spell-slot bindings (`GetSpellName`, `CastSpell`, `PickupSpell` and their kin) now parse their arguments the way the reference's shared marshaller `0x4b3ec0` does — a numeric arg 1, a mandatory arg 2, the index gated to [0, 0x400) with "Invalid spell slot in <Verb>" on a miss, "pet" matched without case — where ours raised mlua's own message on a negative id, answered nil for 0 and above 1024, and read the wrong book for "PET" (1952) - a long-hidden world model part parks its `Mesh3d`: the entity path keeps the batches the retained pass declines resident — 5.3k hidden submeshes at any instant at the Stormwind auction house — and Bevy's sweeps walk every `Mesh3d` whether it draws or not, so a part parks after 30 consecutive hidden frames and unparks the frame its verdict turns visible; every world model part takes part, units' bodies included, while a booth's mirror parts stay out; interleaved pairs read an alone Stormwind 0.1–0.4 ms lighter, and `WOW_NO_MESH_PARK=1` is the A/B; the audio thread's realtime class was the standing suspect for the raid's `cpu_p99` spikes with sound on, and `WOW_AUDIO_NO_RT=1` across two interleaved Blackrock pairs read no difference, so the lever stays as an instrument (1949); and a no-merge prop batch now says which rule declined it — animated (1551 at the auction house) or transparent, not order-free (340) (1951) - the probe reads CPU per thread across its window (user and system split, because a user-mode sampler never sees kernel time) and the main thread's CPU inside the app's frame beside its whole-frame CPU; together they place the vsync tax every player pays by default — cpu_ms 10.7 uncapped against 12.5 under vsync, alone in Stormwind, with no system growing and no thread spinning: it lands on the main thread, in user time, inside the schedules, the same systems each running slower, and a low-priority spinner holding one core busy erases it to the last decimal (vsync + spinner 10.18 against uncapped 10.17), which is frequency scaling — under vsync the cores idle deeper between frames and the next burst starts at a lower clock; nothing in the client is the cause and nothing in the client changes for it (1950) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:55:55 +07:00
SamandClaude Fable 5.1	b1294ce9a1	the quest windows go stock, and the 20 ms frames were mis-split pairs ... - the quest giver's dialog and the quest log are the reference's own `QuestFrame.xml` and `QuestLogFrame.xml`, run off the player's chain at the reference's own toc seats; our 1366-line and 1999-line copies are deleted, and the verbs the stock files call are built — the four `GameTooltip` quest-item and reward-spell setters, the reward spell itself (parsed on all three packets and dropped by the model until now), and two verbs; the detail pane could never scroll because `UpdateScrollChildRect` measured a frame late where the reference's measure is synchronous, so it measures through the installed font engine now; and ten things the reference does that ours did not, each found by a test that had enshrined ours — a shift-click never selects, an empty log hides the whole detail pane and says "No Active Quests", the abandon confirm lives in the popup registry (1944) - the quest bindings' law, read off the binary and folded back: no 1.12 packet carries a quest background material — `GetQuestBackgroundMaterial` (`0x502230`) forks on the dialog's SOURCE object, an item answering its template's PageMaterial, a GameObject its template data, a creature nil (FrameXML's "Parchment"), the id naming a `PageTextMaterial.dbc` row whose six words are nowhere in the binary; its twin `ItemTextGetMaterial` has the same body, so the reader window already held every piece and the giver's feed runs the same resolve; `QuestChooseRewardError` is game-error row `0x98`, and the row plays `igQuestFailed` first, which the verb now queues; the reward-spell getters' third return is `Spell.dbc` Attributes bit 5 (1161 of 22357 rows set — craft-cast spells set, `Pattern:` teaching spells clear), read off the catalog now (1946) - the "20 ms frames" of a raid, chased to their number: the probe's worst frames now carry their churn (entities, assets, first-play sound decodes, live particles, the GPU reading) and in a Blackrock raid with sound on every one of those read flat; then the same window on the PACED clock — under vsync with pipelined rendering the raw delta is mis-split into long/short pairs that sum to exact refresh multiples, so an alone Stormwind leg read 115 of 300 frames over budget with cpu and gpu both at their means, while the paced clock read 0 skipped frames in every leg; what remains real in the raid is `cpu_p99` at 22–23 ms with sound on; and every submesh the streamer spawns now carries why it is an entity, counted per reason — 5.8k resident ADT-doodad submeshes at the Stormwind auction house, 5.3k of them hidden, three prop families the retained pass declines by rule, each riding every per-`Mesh3d` sweep for ~0.4 ms of the alone frame (1947) - a raid's dozen voices cost the mixer ~1.5 ms of every 60 Hz frame: upstream kira spatializes every frame — normalizes, the attenuation curve and a dB→amplitude `powf` per frame per spatial track, two more `powf`s for the volume tween, ~1.7 M `powf`s a second at 48 kHz × 12 tracks for gains that change imperceptibly across a chunk — so kira 0.12.1 is vendored with one patch: the spatial gains and the volume × fade amplitude are evaluated at the chunk's two ends and interpolated across its ~10.7 ms, the granularity FMOD 3, the reference's own mixer, updates them at; nothing a listener can name changes (1945); and rig poses are evaluated in parallel over rigs — each reads its shared clips and writes only its own pose, and a raid's ~400 live rigs were 0.4 ms of one thread walked serially on the main thread's critical path (1947) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:55:12 +07:00
SamandClaude Fable 5.1	4c61e6b7c6	the action bars are the reference's own, and a gift can be wrapped ... - the action-bar cluster is the reference's own: `MainMenuBar.xml`, `ActionBarFrame.xml`, `MultiActionBars.xml` and `BonusActionBarFrame.xml` run off the player's own chain, and our `ActionBar.xml` (2047 lines), `MultiBars.xml` and `StanceBar.xml` are deleted — three, since the stock bonus-bar file declares the shapeshift bar too; one engine verb, `ChangeActionBarPage`, fires `ACTIONBAR_PAGE_CHANGED` and nothing else; one load blocker: stock `MultiActionBars.lua` writes the options window's table from its OnLoad, the reference's own load-order hack, so our options window moves to the reference's toc slot; and ten things the reference does that ours did not: an empty button is HIDDEN while showgrid is 0, the bonus overlay sits one pixel right of the main buttons, every slide takes one frame more than its time, and `ActionButton_ShowGrid` ghosts EVERY ring, occupied included (1938) - a gift can be wrapped: right-clicking wrapping paper fell through to plain use and sent `CMSG_USE_ITEM`, so nothing happened — you could receive a gift and never give one; the wrap is a displayed-cursor MODE plus a stashed slot, never a payload, so `CursorHasItem()` stays false and every stock gate on it stays shut: right-click arms it, a left-click on a bag slot sends `CMSG_WRAP_ITEM`, and the client applies no eligibility test — the server's `ERR_CANT_WRAP_*` reasons are the error line (1934); a duration-limited item counts down: `SMSG_ITEM_TIME_UPDATE` parks a deadline and the tooltip's `ITEM_DURATION` line runs the enchant countdown's bucket ladder, so "Duration: 1 days" is what the reference renders; an update for an item we do not hold is DROPPED, no queue (1933); and a renamed crowd-control refusal names its mechanic — "Can't do that while stunned" — off `SpellMechanic.dbc` (1943) - a creature paths on a boat: `SMSG_MONSTER_MOVE_TRANSPORT` carries the transport's guid and deck-local coordinates, and is the attach/detach edge; a deck sample lands on the rider or nowhere, never as a world write a few yards from the map origin; and "there is no terrain under a boat" turned out wrong: the ground probe admits GameObject meshes — a boat's `.wmo` is a GameObject, so the probe hits the deck (1936); `MSG_MOVE_TIME_SKIPPED` is read, not dropped — the reference adds the lag to the observed unit's last-wire stamp, so dropping it scheduled their next packet `lag` ms late, a hitch for every player who stalled zoning in near us; and our own settle hold reports once (1935); and the tradeskill tool order was backwards — focus FIRST, so every two-tool recipe had read "Blacksmith Hammer, Anvil" for the reference's "Anvil, Blacksmith Hammer"; the focus's `hasTool` is 1.0 in the reference too (1942) - the crowd's second sitting, instruments first: the probe line carries the tail — cpu p95/p99 and the six worst frames — because a mean under budget hides the one frame in twenty over it (raid legs railed at 16.67 with p95 at 18–20 ms) and whether that frame was work or a wait is the first question about a stutter, and the visible meshes are counted by lane; then two cuts — a terrain tile furnishes as 4×4 cells of 4×4 chunks, one mesh, one entity, one draw per 133-yd cell instead of per 33-yd chunk (a city horizon is over a thousand chunks a frame, and a cell changes nothing a pixel can see; Stormwind 10.55 → 10.32 `cpu_ms`), and an authored point light is our own `WorldPointLight`, not Bevy's `PointLight`, which was only a data carrier while Bevy ran its whole light lane over every lamp, ~2 % of the alone frame; the sitting's record is 1945, and the code says so (1945) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:54:27 +07:00
SamandClaude Fable 5.1	ad7edb3198	a 40-man raid costs less per frame, and the money frames go stock ... - a 40-man raid cost +5 to +8 ms of CPU per frame — a population, not a hot function — and three cuts landed where it was measured: the rig palette upload sends one `write_buffer` per run, bridging the gaps parked bodies leave between live rigs (calls 93 → 3); a built material becomes an asset when something visible binds it, not when a spawner builds it — 9.9k material assets were alive for 36 drawn batches, so a material is realized only once bound and visible, readers that clone before binding doing it themselves through one new function in the engine's wall (172 → 173): a 17× cut in live GPU objects; and a body's batches spawn as material groups — a geared body stood as 12–27 mesh parts, its skin composite drawn as five separate batches, and two visible batches now join when nothing downstream can tell them apart: parts per body 17 → 12, the Stormwind raid leg 17.0–21.0 → 15.65–15.82 `cpu_ms` (1929, 1940) - the instruments that priced it, all on the probe line: wgpu's own resource census, because a sampled profile put roughly half of the pass encoder in usage trackers sized to the device's live buffer and texture counts — a term that scales with residency, not with what is drawn, and nothing printed the number; the active camera count, since every view is its own pass set; the UI's live texture-identity runs, one 2D draw each, which is the number a UI change moves, not the quad count; and parts and materials per dressed body on the dress census, the ceiling on merging a body's batches read off a real body instead of guessed from a model's batch table (1940) - the coin kit retires: every remaining money display — quest progress, the mail purse and postage, the trade recipient, the tooltip, the quest log, the trainer's purse and cost — moves onto the reference's `MoneyFrameTemplate`, each seat already matching ours; the progress panel's required-money flow is ported in full — `QuestFrame.lua:168-190` colours the FRAME red/white AND the LABEL black/dark-grey when the quest costs more than you hold, where ours had neither colour; the mail postage takes the reference's name and label, trade and mail issue no explicit update because the frames repaint on their own events, and the trainer's cost label is shown or hidden with its `local COSTS_LABEL` shadow deleted; with the trainer converted the hand-built coin trios have no consumer, so their templates and the `BenillaMoney_*` helpers are deleted and `MoneyFrame.xml` is the only money kit (1937) - the crowd-control ladder's aura exemption exists, and it is a join between two `Spell.dbc` records — each arm asks whether the spell being cast is itself an immunity covering the aura that blocks it (Ice Block cast while stunned), the school arm shifting because `School` is an index and `EffectMiscValue` a mask; what the scan mostly decides is WHICH message a refusal carries — a blocking aura the cast does not counter renames the reason to "Can't do that while %s" (1941); bevy_audio leaves the build — its plugin had been disabled since 0530, but disabling a plugin does not remove a crate, and rodio, a second cpal and an Ogg/Vorbis decoder were still linked into a client that mixes through kira (1932); and cpal has four stream errors, not two — a routine ALSA underrun was treated as "the device is gone", tearing the stream down on every hiccup; now only a lost or invalidated device rebuilds (1939) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:53:37 +07:00
SamandClaude Fable 5.1	3baf0c1750	Linux and Windows get their sound back, and emotes refuse while moving ... - every build off macOS has been silent since the 09-02 snapshot (B356): the non-macOS device layer was a stub that opened nothing, and nothing had ever compiled the client for another target; that layer is now the same output stream over cpal — one seam, not two stacks: the ring, the render thread and the rebuild-on-device-change loop are one code on every platform, only the device file differs; the audio thread never allocates, a lost or changed default device is noticed at kira's own cadence, and realtime is `SCHED_FIFO` on unix and MMCSS "Pro Audio" on Windows, either refusable, the ring turning a refusal into a degradation; and the two instruments this bug says were missing exist — a Linux and Windows cross-compile under `clippy -D warnings`, and a live output test that opens the real device and asserts the ring never ran dry, because compiling is not sounding (1920) - an emote that requires standing still now refuses on the move: `CheckEmoteEligible`'s `0x4000` arm writes an out-param that `DoEmote` turns into a red "You can't do that while moving!" and aborts; 20 emotes of 78 can raise the line, read off the shipped `Emotes.dbc`; SWIMMING is not in the mask, so a swimmer holding still emotes fine, and `IsSelfControlled` is TRUE for an ordinary player, so a feared or confused player emotes away happily (1904); and the cast path gains its equipped-item rung — the button greyed and pressing it still sent: the reference short-circuits to "fits" on four gates, not one, the miss being a zero `EquippedItemSubClassMask`, which means no item required at all, and the rung sits above the crowd-control validator, so a player both stunned and missing the weapon is told about the weapon; that ladder has six arms, not three — CHARMED, FLEEING and CONFUSED join (1925) - the quest reward's money is the reference's own widget: the tail of the shared `QuestFrameItems_Update` drives `questState.."MoneyFrame"` through `MoneyFrame_Update`, and the two hand-built coin trios become one `MoneyFrameTemplate` instance each at the reference's anchors, which our coins already used; the progress panel stays on coins, being the required-money flow, not the reward one (1928); and the flight master's `!` joins the questgiver sweep, because the green `!` and the gold `!` are the SAME slot, `unit+0xb2c`, torn down once before both queries re-issue; the mouseover re-ask is refuted — the only taxi-status sender has one live caller — and the two queries are not symmetric: the questgiver's goes out on every sweep, the taxi one only when the unit's key moves or on a full sweep, as the wire shows at Dungar Longdrink: 68 questgiver status queries against 1 taxi-node one (1918) - `GetSpellTabInfo` out of range answers `nil, nil, 0, 0` — the reference returns four values on every path that returns (`0x4b3ce0`), and ours answered a single nil copied from another binding's shape; it matters because stock `SpellBookFrame_OnLoad` calls it unconditionally and then does `id > (offset + numSpells)`, which a nil raises on and (0+0) makes true, hiding every button on the page — the behaviour to match; the argument raises the reference's usage string when absent, and the index truncates toward zero THEN decrements (1931); the perf probe gains a `WOW_PROBE_UNCAP=vsync` arm, because the uncapped present mode is a request, not a schedule — measured, it taxes the number it was meant to pin (11.80 `cpu_ms` uncapped against 14.65 under it), so the default stays uncapped; and the last three test temp paths are keyed on the pid, after concurrent runs raced on one fixed shared file Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-04 20:52:22 +07:00
SamandClaude Fable 5.1	b97d8e8eb8	a corpse is not a shop, and Darnassus draws through its doorways ... - mounted, over a dead wolf, a right-click answered "That merchant doesn't like you." three times: the click was sending `CMSG_LIST_INVENTORY` — a vendor packet — at the corpse, because the unit block was an if/else-if chain whose LAST arm was the NPC-service dispatch, and the mounted-loot fall-through gave it a newly reachable state — dead target, no leg taken, so the click fell into the arm that means "live friendly service NPC", which dispatched on the CURSOR, and a lootable corpse wears the same pouch a vendor does; the reference splits on death first and never rejoins (`0x60bf80`), so the fork is now `Attack / Dead / Service`, matched exhaustively, with a test over all sixteen combinations of its inputs - Darnassus draws through its doorways: the reference's Pass 2 ends in the portal RECURSION (`0x6b3d39 call 0x6b41c0`), not a mark — a doorway admits everything the walk reaches through it, and in a city whose outdoor areas are stitched together by portals that is most of it (standing in a shop, 36 groups the marking pass had culled); the narrowness reject turns out to be the open-world walk's entry gate alone, not the shared volume builder (applying it to both had blanked the containing building's shell for every window in the [0.001, 0.02) NDC band), and the depth cap is `[0x86b6b8]` = 10, not a guessed 64; `WOW_PIN_SITE=darnassus` replays the report offline in 0.2 s - the tab clamp was reading a return the reference's `PanelTemplates_TabResize` does not have, so since the panel kit came off the chain the rule that a tab never grows past its window's drawn right edge raised `attempt to compare number with nil` once per tab per width change on all seven windows built from `CharacterFrameTabButtonTemplate`; it reads the fit back off the art the reference's own verb just wrote (a Texture's width is its authored size, no resolved rect consulted) - footfalls have one gate, one origin, one kernel: the 50 yd camera-distance cut (`0x5fc00a`) is measured from the CAMERA EYE, unconditional (no self exemption — an earlier slice reused a predicate it had just derived for the ring pools, and no RE note ever recorded one), rounded once with a strict compare — and it sits above the decal, the camera shake AND the 25 yd spray, not copied per consumer; the shake's second suspend arm is the shared fly-or-swim spline bit read off the FOLLOWED unit, so under Mind Vision or Eye of Kilrogg the shake frames on the right body and suspends on the right swim state Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:08:31 +07:00
SamandClaude Fable 5.1	6d34bb4ecc	spells shake the camera, and the sound keeps playing in the background ... - spells shake the camera: `SpellEffectCameraShakes.dbc` loads beside `CameraShakes.dbc` as the group table, `SpellVisualKit` field 14 is the shake column (58 of 1772 kits carry one — War Stomp, Thunder Clap, Kodo Stomp, the ~50 Knockdown/Slam/Charge impacts, every summon, Sapper Charge, Earthquake, Sapphiron's death), and the kit shake fires exactly once per play from the first effect node's one-time arm pass (`0x620e11`), the same leg that plays the kit sound one field over — the older "fires only if nothing else did" reading would have silenced every kit with a populated effect slot; `$SHK` markers fire from GameObject and DynamicObject hosts only (the titan chest, the bomb wagon); one deliberate deviation — the reference reads a bone-attached node's position twelve instructions before it is written, so 26 of its 58 shake kits shake at the world origin and are culled by the 80 yd cut, which is an uninitialised field, not a design; 21 of 24 presets now have a live producer, up from 6 - the sound keeps playing when the window loses focus, if you want it to: 1.12 mutes everything on `WM_ACTIVATE` and offers no way out, so benilla ships that as the default (`Sound_EnableSoundWhenGameIsInBG` "0") and puts an Audio-page checkbox on it — a gate appended LAST in the main-track chain, downstream of every tap, so an unattended capture still records; wow-re's round settled that music is not exempt and that it is a mute, not a pause - a building's far side no longer draws through its own walls: the camera-inside leg has three passes and we had one and a half — Pass 2 admits an exterior-shell group only against a frustum clipped to ONE deferred window (we marked every such group visible against the full view, under a comment calling it a benign over-draw), and Pass 3, which we had none of, draws the 24 `0x10000` groups (14 in Stormwind, 6 across Dire Maul, Razorfen's four 450–520 yd shells) from anywhere, frustum-tested on both legs by byte-identical setup; there is no antiportal branch in 5875 and no such flag in its data (815/815 roots and 5220/5220 groups censused, no bit above 18 ever set) - the macro window is the reference's own `Blizzard_MacroUI`, and `GlobalStrings.lua` — line 1 of the reference's own toc — was missing from our manifest entirely (the app loaded it out of band at startup, so the client was right and every harness that loaded the manifest was not); exactly one `<OnLoad>` runs, the most-derived, because `SetScript` releases the slot before it looks at the new body — which is why the reference re-states the template's width line in tab 1 and not tab 2 Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:07:15 +07:00
SamandClaude Fable 5.1	276d0896ad	the screen warps underwater, and the ocean darkens with depth ... - the screen warps underwater: submerged, the reference runs a second pass list whose third pass is `FFXGlowWave` (`0x6cb1f0`), binding a 128×128 two-channel sine bump map that displaces both the screen and the blur texcoords by 3.0 full-res pixels — uv rotated 10°, scaled W/128 × 0.88·H/128, two phases at 3174 and 2805 ms; wow-re had carried the swap as "GlowWave, or a duplicate Glow" and the cross-check settled it (the two arms are mutually exclusive and the plain-Glow arm is unreachable on both backends; lava and slime warp too, and a ghost's death pass replaces the glow outright); a dry frame binds the same pipeline object as before this existed, and the phases advance live at 0.313/s and 0.355/s against a predicted 0.315 and 0.357 - ocean is a fifth submersion state — 96.8% of every wet sub-tile in the shipped corpus, so essentially all open water — and it alone runs a depth ramp on the light: `t = clamp(eye.z, −30, 0)`, ambient at `(k+1)/2`, diffuse at `k/4 + 3/4`, exactly k = 0 at −30 yd, gated on the full dword `== 1` at `0x6d2821` (the ramp does NOT drive the fog colour — that store is dead, overwritten before submit); the sun and moon glare fade over a 10-yard depth ramp applied to the slew TARGET, not the drawn alpha — a read one indirection out that two notes had published a verified negative on; the MCLQ fishable bit (0x40) is deliberately unread, since `0x69b5d0` has zero callers in the image and fishability is the server's verdict - the dropdown kit is the reference's own `UIDropDownMenu.xml` (one manifest line for 1150 lines of ours, `DropDownList3` gained, and two fixtures that only worked because our copy was more forgiving); `UnitFactionGroup` returns the English name AND the localized one — ours returned the localized string twice, and the first return names a TEXTURE in every stock consumer (`UI-PVP-`..factionGroup), so on a non-English client every one of them named a file that does not exist Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:05:57 +07:00
SamandClaude Fable 5.1	b8e0025bc9	the flight map goes stock, and vendors refuse in 1.12's own words ... - the pet stable and the flight map are the reference's own files: `TaxiRouteFrame` maps to a plain Frame (factory `0x495ba0` allocates the base size, the ctor adds no field, no new virtual, no method table — and `GetObjectType` is not overridden, so its own kind would have been the divergence), and the swap found three silent behaviours — the close button did nothing (an unknown `inherits=` is a WARNING, so it loaded as a bare button), the "no paths" line drew empty, and the title is READ off `UnitName("npc")`, not received; the WMO skybox rides the interior crossfade — one weight (`[0xce9bdc]`, the same number the MFOG lerp rides), a 4 s alpha fade, so a skybox the flood publishes through a doorway draws at weight 0 instead of popping in whole - a widget predicate answers the NUMBER 1 or nil, never a boolean: eleven of them (`IsShown`, `IsVisible`, `IsMovable`, `GetChecked`…) move to the 1.12 shape — the reference proves it in its own source (`tostring(this:GetChecked())` saved and read back `== "1"`; `tostring(true)` would never match and buff timers would never appear), 21 corpus sites compare directly (`IsVisible() == nil`, `GetChecked() == 1`), and one of our own tests had the guild rank editor sending `true` where a real client sends 1 - the vendor's refusals are the reference's table: `SMSG_BUY_FAILED`/`SMSG_SELL_ITEM` route through `0x5dcd60`/`0x5dd1f0`'s switches, six lines were wrong (code 3 printed "That vendor can't be found." — vmangos's identifier — where 1.12 prints "That merchant doesn't like you."), silence is now expressible, and keying them gives the vendor a voice (`ERR_NOT_ENOUGH_MONEY` carries error-speech line 0x28); the bank's three ids were right and its "purchase failed (n)" an invention; the logout refusal is one key, `ERR_LOGOUT_FAILED`, for every non-zero reason; the two mail codes we said had no string have one (`ERR_PLAYER_WRONG_FACTION`, `ERR_RESTRICTED_ACCOUNT`) and a sent mail says `ERR_MAIL_SENT` in yellow; a test runs the chain's own `GlobalStrings.lua` and requires non-empty text for all 193 `ERR_*` the workspace names - a clicked quest timer no longer raises — stock `QuestTimerFrame.lua` calls `QuestLog_SetSelection`/`QuestLog_Update`, which existed nowhere in this tree; our quest log owns the reference's two names now, which also shrinks the delta when that window itself migrates Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:05:15 +07:00
SamandClaude Fable 5.1	b14dd42cf1	the unit frames are the reference's own, and the water has motes in it ... - the player, party, target and pet frames are the reference's own files (`BuffFrame`, `UnitFrame`, `CombatFeedback`, `PlayerFrame`, `PartyFrame`, `TargetFrame`, `PetFrame`, in the reference's toc order) — our 2741-line `UnitFrames.xml` and our `PartyFrame.xml` are deleted; four engine verbs built for them (`PartialPlayTime`, `NoPlayTime`, `UnitPlayerControlled`, `UnitIsCivilian`, each answering 1/nil), the whole-UI load bracketed in the counted sound-suppression scope `0x48fbf0` self-brackets (the reference fires a `PlaySound` during load and throws it away), and `GetBillingTimeRested` read off `SMSG_AUTH_RESPONSE`'s third billing field rather than hardcoded; the differences the stock files bring are pinned rather than papered — `PlayerFrame` never hides, an unresolved name is BLANK not "Unknown", the target frame has no numerals at all (a bar hover pops the unit tooltip instead) - what the swap surfaced, each a hard break under the stock files: `UnitBuff`/`UnitDebuff` returned the Classic Era ten-value tuple, name first, where 1.12 returns two and three values TEXTURE first (`0x519500`/`0x5198f0`) — a raise twice per `UNIT_AURA` on any debuffed target, and all 184 corpus call sites bind the 1.12 shape; the power events are named per resource (`UNIT_MANA`/`RAGE`/`FOCUS`/`ENERGY`/`HAPPINESS` and their MAX twins) and we fired the Era pair, so no mana, rage or energy bar could live-update; `SetRaidTarget` is the engine verb and `SetRaidTargetIcon` the FrameXML toggle wrapper — we had registered the wrapper's name on the engine's body, so every raid mark in the unit popup raised - the underwater drift cloud: the reference draws a fixed 4000-mote field whenever the camera eye is inside a liquid — not a particle system, nothing spawns, ages or dies; a population scattered once into a 30 yd camera-relative cube, advected by the camera's own motion plus a gust and wrapped, so the field stands still in the world; benilla had the murk and nothing in it; a dry frame commits no draw, and the submerged one is bounded by the reference's own 666-quad cap (633–644 of 4000 measured live, which is the expected survivor count of a uniform cube through the 90° cone — an independent check on population, distribution and cone at once) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:04:32 +07:00
SamandClaude Fable 5.1	e0080082cf	your character speaks its refusals, and clicks land on the right frame ... - your character speaks its refusals: `VocalUISounds.dbc` gets an adapter, `sound::vocal` transcribes `0x4580f0`'s per-race table and `0x458250`'s play cycle — four plays of one line arm the annoyed variations, and the bus cap of one keeps the annoyed and ordinary lines from playing at once; a displayed message now carries its catalog RECORD, so its sound comes off the same row (tag 0x44 plays the named cue, anything else speaks the line), and 36 of the keys benilla raises have a voice — cast failures, inventory refusals, the quest/auction/petition lines, taxi, bank, stable, mail, the talent wipe; `EnableErrorSpeech` joins the CVars at the registrar's own default and the Audio page grows 1.12's fourth Sound checkbox with the dependency ours was missing (the master greys Ambience AND Error Speech, not Music); seven of the twelve taxi refusals turn out to be the YELLOW info line, not the red one they all took - at equal (strata, level) the hit sweep takes the EARLIER-linked frame: `0x764aa0` inserts with a strict `ja`, so equal keys are appended, and `0x7660d0` sweeps each bucket from index 0 upward — the reverse of the reverse-draw-order walk we had, so draw order is unchanged and only the hit sweep reorders within a plane; the wheel becomes its own index, and the five full-area wheel-catcher Buttons (Craft, QuestLog, Skill, TradeSkill, Trainer) are deleted, since under the true tie law each would have swallowed its whole window's clicks — a bug the Skill window's own comment had already recorded once - one rasterizer settle for the whole ground-decal family, and it is the measured one: the shipped client constructs `CGxDeviceD3d`, so the OpenGL arm holding the −4.0 slope factor is never executed (factor 0 on 2485 of 2485 captured `glPolygonOffset` calls), and every number derived from it was 256× too small; the reference groups the blob shadow with the selection RING and the footprints on their own at 10×, the reverse of ours — but ours is reversed-Z float where the bake residual is millimetre-scale and constant with distance, so `DECAL_RASTER` collapses to one value (32768: the ring and the reticle had been under their own conflict at 8192 and simply never walked on a slope) Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:03:50 +07:00
SamandClaude Fable 5.1	359b7cfb2c	the foam stops at the waterline, and skill bonuses show the right halves ... - the foam still washed past the waterline onto the sand after the shore fix: the slope-scaled depth bias ported from the reference's `glPolygonOffset(-4.0, …)` pulls toward the eye as the square of view distance, and what it spends is the wet-cell skirt — whole cells are the unit of liquid geometry, so a shoreline carries liquid-lattice triangles over dry sand (23 cells, up to 7.1 yd inland at the reported beach) held back by nothing but the terrain having drawn first; the tie it would settle does not exist here (our foam patch IS the liquid surface's own triangles — same cells, same winding, identity transform — so the depths agree exactly), and −4.0 is a hardcoded immediate every consumer of that state gets, not the foam lane's number; the slope term is disarmed with an `== 0.0` assert, reproduced three ways at the director's pin (slope 0 stops dead at the waterline, 4.0 washes onto the sand, 64.0 does it wide) - weapon and defense skill bonuses showed the halves the wrong way round: the reader every skill-shaped binding goes through (`0x5ea460`) writes `value + permBonus` and `tempBonus`; ours returned `(value, temp + perm)` — right total, wrong split, which is why no fixture caught it — so a 300 weapon skill with a +5 talent and a +10 aura read 300 (+15) where the reference prints 305 (+10), on the sheet's Attack row and in the Defense tooltip's `(base +x)`; `UnitRangedAttack` is also pinned as a direct call gated on the PLAYER typemask alone, so a pet gets `(0, 0)` - instruments: `water_here` also reports WET cells the ground climbs out of (the skirt a depth bias spends), the foam wake can be laid along a bank (`WOW_WFX_HEAD`) since no shoreline in the game is axis-aligned, `fxview` can stand its subject at any world point (`WOW_FX_AT`) so a blob shadow can be photographed on any terrain, and `WOW_MAP` refuses a name instead of silently falling through to Eastern Kingdoms — `WOW_MAP=Kalimdor` had produced a subject floating in empty ocean that read as a renderer bug Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:03:14 +07:00
SamandClaude Fable 5.1	1f3ec2d538	the waterline stops stepping, and mobs in far rooms lose the keep's fog ... - the step in the water at the shore, which travelled with you as you walked: `WATER_BIAS` (−2e4) exists to give the water pass its sort slot, and bevy's `StandardMaterial` also packs that f32 into the rasterizer's depth bias, so every water surface in the game shipped with a −20000 depth pull — on the one world surface that meets the ground at a grazing angle over tens of yards that does not settle a tie, it moves the waterline, and a float depth buffer's bias unit doubles at every exponent boundary, so two water triangles either side of one stopped agreeing about where the shore is, at a fixed distance from the eye; the bias is out of the rasterizer (a ~17 px jog at the reported Stranglethorn beach, present before and absent after) - the foam wake went dull under the next water chunk and painted onto dry sand: it sorted against ONE water surface with a coplanar epsilon (the chunk 33 yd nearer the eye outranked it thirty times over, while the reference drains every liquid queue before it draws foam, `0x6816d0`), and its coplanar settle was a 0.03 yd geometric LIFT, which on a 29% beach is 0.10 yd of dry sand painted and proportionally more on every flatter shore; it is a rung above the whole water band now and sits exactly on the plane, and the RE on its depth state came back: a depth-buffer bias (`footstepBias × [0x810390]`), depth-tested LEQUAL with write off — the "−2048 GL units polygon offset" our comment justified the lift with existed nowhere in wow-re - a unit standing in a room the camera's chain does not reach wore the building's fog while the walls around it wore the scene's: the reference's unit test has a second conjunct (`[P+0x98]`, the same `[0xca7f00]` chain gate the walls use), so the attach ray now returns the room it claimed and the gate is re-read every frame because it moves with the CAMERA — at Shadowfang, from inside the pinned room, three true interiors are in frame across the courtyards and a mob in g13, 91 yd out, wore the keep's teal at ~84% against a scene fog of ~3% - two instruments for the shore class of report: `water_here <map> <x> <y>` reads a shoreline offline (the wet-cell lattice, the plane on it, the terrain under it, and the four scans that tell an authored defect from a rendered one), and `WOW_CAPTURE=waterfx WOW_WFX_AT=x,y,z` wades the foam rig in the REAL streamed water at a pin, where the synthetic square had no bank to clip against Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:02:04 +07:00
SamandClaude Fable 5.1	d017ca1cde	Shadowfang's far rooms lose their cyan, and the street lamps hum ... - Shadowfang Keep's far doorways no longer fill with flat blue-cyan at distance: it was never a hole — the room behind the arches is in the visible set at every camera distance — it was the building's own MFOG record (rgb 40,68,79 at end 106.9 yd, ~95% saturated at 70 yd) painted over a room the reference shows under the scene fog, because benilla picked the fog lane STATICALLY per batch (`flags & 0x48`) where the client's `[0xca7f00]` gate decides it per frame per group along the portal walk; the flood publishes that bit now and all three lanes (entity, retained, WMO liquid) read it, and the wow-re carve confirmed the predicate verbatim and added the conjunct we lacked — the MFOG target exists only when one of the camera's own containing groups is a true interior, so in a courtyard the interior triple decays to the scene triple, which is what a unit standing in the keep reads - every world-placed M2 was silent — the street lamp's hum, the campfire, the waterfall, the windmill: no doodad ever wrote into the anim-event stream (three producers, none a terrain placement), and the sound clock was gated on the RIG, which is rightly never built for a model whose idle is the static mesh — `KalidarStreetLamp01.m2` is one looping 3.3 s Stand keying no bone at all, carrying one `$DSL`; the census (`benilla-extract soundeventscan`) counts 273 carrier models, 108 on tiers no rig gate arms; a sound-carrying placement now gets a clock-only host, `$DSL` is a sustained emitter, and the RE foldback settled three laws: `$DSL` ALWAYS loops and consults no flag (the kit's 0x200 bit is a lane select and 0x400 is random pitch — a neat four-row correlation of ours refuted), `$DSE` is the stop token, and a doodad holds ONE registration slot, which is why `bellows.m2`'s two loops alternate rather than drone - the reference voices sound IDS, not doodads: `$DSL` registers the doodad's position as one record in a 32-entry pool keyed by SoundEntries id (`0x461d80`), each entry runs one channel at whichever of its records is nearest the listener — moved, not restarted (`0x7a5b10`), so one hum follows you down a row of lamps — the pump admits the first four entries by claim order and fades the rest over 3.0 s; the per-id duplicate suppressors turn out to be the one-shot lane's and were being applied to every lane (a three-second dropout every time a loop came back under the cap); three 76 yd hops between two Darnassus lamps now produce ONE play, and Darnassus runs 12 entitled ambiences with 8 withheld by the cap - the selection ring joins the decal band too, and it draws UNDER the shadow: the RE found the ring emitted from inside PHASE 1's own M2 node drain (`[node+0xb4] = 0x481540` → the `+0x38` tick → `0x608e00`), the same iteration that then draws that node's blob shadow — the reverse of the tie-break this ladder had asserted since both rungs existed, a rule invented here and never read anywhere; `WOW_EFFECT_TRACE` names every rung of the band and prints its index in the sorted transparent phase, which is the ladder read straight off a live frame Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 17:00:01 +07:00
SamandClaude Fable 5.1	9f6a550b66	shadows go under the water, and the ping meter measures the server ... - your shadow no longer reads through the Stranglethorn river or the Stormwind mage-tower portal (in the reporter's own shot the crocolisk casting it was invisible under the surface and its shadow was not): the reference draws the blob shadow in PHASE 1 among the opaque drains (`0x6812c5`), before the footprints, the M2 opaque pass, the water surfaces and both M2 transparent passes — ours rode a +4096 sort rung above the whole world band (item 81 of the sorted transparent phase, now item 4); the footprints and the ground-target reticle move with it into a pre-water decal band at their own byte-known slots, and the print/shadow order was inverted too — the reference draws a print one slot AFTER the shadow - the ping meter was measuring a client frame, not the server: the RTT was computed in the ECS drain, one frame or more after the pong had landed, so it read high exactly when the client was slow — the world-enter hitches inflated the first sample by hundreds of ms and the 30 s averaging ring carried it for minutes, a meter that opens at ~600 ms and sinks toward the truth over two or three minutes, as reported; the read thread measures where the packet lands now, the way `OnData 0x537b10` hands `SMSG_PONG` to `HandlePong 0x537d60` inline, the ring is 15 usable slots not 16, the average is the reference's unsigned truncating divide, and the keepalive is per-connection from the last SEND, so the first ping is a full 30 s after world enter instead of anywhere inside the load storm; localhost reads 0–3 ms where it read 58 - an extra bar's empty slot no longer loses its hotkey label (E/R/C showing in one screenshot and not the next): one line left behind by the slice that shipped the multibars with no bindings cleared the corner outright, which became a wipe of a live label the day those wells got real `MULTIACTIONBAR` bindings — painted by `UPDATE_BINDINGS`, wiped by the next repaint; the empty arm repaints the corner now, as the reference's `ActionButton_Update` never touches the text there, pinned by a test that drives the reporter's shape through the real shipped `ActionBar.xml` - an unattended probe run opened focused and borderless over the director's work — `WOW_UNATTENDED` joins the background-window prefixes — and the world-API wall was measuring a prefix of every file: a `#[cfg(test)]` guard switched the scan off from the first test module to the end of the file, so the ceiling moves 166 → 172 as a measurement correction, not six crossings Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 16:59:20 +07:00
SamandClaude Fable 5.1	6da7ed3fca	the action bars' empty slots go ghost, and a rooted creature stays put ... - picking a spell up off the bar no longer blacks out every empty slot: `UI-Quickslot` is a FILLED 60%-black plate, not a hollow ring (decoded off the shipped BLPs — Quickslot2's interior is RGBA 0,0,0,0, Quickslot's 0,0,0,153), and the reference dims it to half alpha in `ActionButton_ShowGrid`, so an empty well now ghosts at 30% black where it read as a hole; the wow-re cross-check then found the reference never UN-dims it (a 3-argument `SetVertexColor` preserves the current alpha at `0x79abd0`, and `SetNormalTexture` reuses the region), so in the real client the first drag leaves every button's border at half alpha for the rest of the session — benilla ends the dim on purpose and records it as the divergence it is (1782) - the bag bar is the reference's own `MainMenuBarBagButtons.xml` (window 3 of the stock-first migration), and the swap found a real bug rather than a move: the reference's `ItemAnim_OnEvent` matches a push against the button's inventory-slot id (20..23) and benilla emitted its own container numbering (1..4), so all four bag-push cards would have silently never played against the stock body — `push_container` now emits the reference's selector (`bag + 1`), which is what an addon expects anyway; beside it, a readiness probe that asks the LOADER, not a grep, which of the 85 stock windows load off the player's chain today (58 do; the 27 that don't name their own blocker) - a rooted, hovering, water-walking or walking creature reached benilla as an unhandled opcode: the twelve `SMSG_SPLINE_MOVE_*` (the observer twin of the ack'd force family — bare packed guid, no ack, any unit) now decode into one MOVEMENTFLAGS word per unit, which benilla kept for our mover and for a relayed player and not at all for a creature, the one body this family is ever addressed to; ROOT wipes the direction bits so the dead-reckon stops walking a rooted body on for its next heartbeat (`0x7c7340`), HOVER clears a yard off the floor without climbing a body already inside that yard (`0x636e81`), WATERWALK elects the liquid surface against the terrain, SAFE_FALL's terminal is 7.0 not 60.148 (`0x7c5d23`), and SET_RUN_MODE's bool is inverted against the bit, folded at the parse; the inspect card names a unit's granted modes in words (`granted rooted+hover`) so a creature resting a yard off the floor reads as intended rather than broken Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>	2026-09-02 16:57:55 +07:00
SamandClaude Fable 5	b8758d0b21	the server's broadcasts land, and a ghost walks through closed doors ... - the four world broadcasts decode and land — a shutdown countdown, the civilian-kill alarm, a defense-tower capture and the whisper-cap refusal all reached benilla as unhandled opcodes; `SMSG_SERVER_MESSAGE` carries a row id and a fill, never a finished line, so `ServerMessages.dbc` ships (the five sentences a shutdown has, plus the client's own `[%d]: %s` row-missing fallback, which is mechanism, not an error path); the two defense broadcasts are CHANNEL lines, not system lines — delivered once per joined channel whose row carries DEFENSE, the LocalDefense copy zone-matched against the PARENT area while the text names the packet's own — so a character in neither channel sees nothing, faithfully, and since nothing auto-joins WorldDefense in 1.12 an Eastern Plaguelands tower capture reaches only players standing there unless they typed /join; the wow-re cross-check then made the length prefix load-bearing — the client borrows a pointer into the receive buffer and composes NULL on an overrun, so an overrunning length is a refusal here too: the packet drops with a log line, nothing displayed - `BankButtonIDToInvSlotID` is carved: `isBag` is a TYPE test, not truthiness — `lua_isnumber` refuses every non-number tag without ever loading the payload, so `true` takes the same arm as `nil` while `0` and `"1"` take the bag arm — the function is TOTAL (our out-of-range 0 was an invention; the only compares in all 118 bytes are on `lua_isnumber`'s return), and the +59/+39 arithmetic stands confirmed, byte-identical to `ContainerIDToInventoryID`'s second arm - a deliberate right-drag turn no longer stands you while seated — the observation was right and our attribution was wrong: the body-facing commit is refused by two independent gates and the stand arm is skipped outright while the button is held; what stands you is the RELEASE, a press under 200 ms with under 2.25 degrees of yaw dispatched as a right-CLICK whose interact reaches `SetStandState(0)` — this lands the deletion half, with the click's own stand held for one more wow-re answer - a ghost walks through closed doors — the whole mechanism is one instruction image-wide, reading bit 15 through the sign flag to answer "not a collision candidate" for a DOOR when the traced body is a player in ghost form; the bit is a property of whose trace it is, not of the door, so the exclusion rides the two mover-only casts and the set is empty on every living frame — the particle snap, the precipitation probe, the mouse pick and the creature conform keep the door's collider untouched, exactly as the binary leaves them Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 20:19:32 +07:00
SamandClaude Fable 5	25c865c00a	the ghost world gets its own sky, and a corpse cannot steer ... - a dead player's sky is DeathClouds now — `lightSkyboxID` was named in the schema and read by nobody and `LightSkybox.dbc` was never loaded at all, so a ghost kept the ordinary gradient dome; the shipped data has exactly 5 of 426 LightParams rows carrying a skybox id, every one of them DeathClouds.mdx, reached from the death profile (slot 4) of all 374 Light rows and gated on the ghost bit, with a hardcoded weight of 1.0 that skips the WMO skybox outright; it resolves off the same (map, camera position) seed the atmosphere uses, so the ghost sky and the death-profile fog it hangs over switch on the same frame — and the lane sheds its `wmo_sky` name for `skybox`, no longer being WMO-only - a corpse cannot steer: the reference's two movement-input predicates share a precondition benilla never modelled — `UNIT_FIELD_HEALTH > 0` — so at zero health the keyboard turn and the right-drag close together, exactly as they do inside an Ice Block, while the ghost (health 1) keeps moving; the wow-re §5 round then corrected three byte trails (the mouse turn has its own eight-conjunct predicate, the supposed on-taxi test is the Knockdown animation lockout — set on exactly one of 208 AnimationData rows — and the "charm bit" is the far-sight latch) and added three verified health gates we lacked: the jump IS health-gated, sit and stand are refused while the body reads dead (a feigner too), and a follow tears down only when BOTH predicates drop — we had that wrong in both directions - the walk-gait cross-check folds back confirming `min(walk, run)` and the speed-only animation channel (by a 2,304-function closure census), and finds the trap we lacked: every incoming spline re-authors the gait, and the spline's RUNMODE flag is an INVERSE — a spline marked as a run clears walk mode, one not so marked SETS it — so a charge, taxi, knockback or flee hand-off leaves the body in whatever gait its path was authored with, applied once on the spline's start edge; the post-worldport zero-walk fallback is retracted on a re-run (the ctor's zero is a transient thirty instructions wide, never observable), leaving the plain minimum the bytes compute — the better failure mode too, since a fallback makes a walker silently RUN if a walk speed ever goes missing, where the transcription makes them visibly stop Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 20:18:16 +07:00
SamandClaude Fable 5	a0d2fc383b	the world map blips are the reference's, and particles go untinted ... - the world map speaks the reference's names: WorldMapTooltip is declared (six corpus addons index it; pfUI dies on it at file scope), the blackout frame takes the reference's spelling BlackoutWorld, and the four hand-written party blips become the reference's own unit-blip templates — all 44 blips as instances, each with its published $parentIcon, the name pfUI recolours group blips by class through; the 40 raid ones are declared but not positioned yet, named in the OUT list rather than faked - the templates addons build their own buttons from are published: ItemButtonTemplate, ContainerFrameItemButtonTemplate and the bag-slot chain — most of the handler scripts were live the whole time (this client executes the reference's own container and paper-doll code off the player's patch chain), so what was missing was mostly declarations, not behaviour — the bank arm stays out and is named as the omission it is; OptionsFrameSliders becomes real, with the three Graphics rows reading it and re-reading on show, so an addon's write to a slider's bounds actually moves the slider; and the template census that had ranked these at "zero demand" learns to see a local bound to a literal — its table went from 3 rows to 7 - the bindings answer in 1.12's own shapes: Button:IsEnabled() answers the NUMBER 1 or 0 (0 is truthy in Lua, so the reference's own == 0 tests got false forever, and three of our files had hand-worked around the divergence rather than fixed it); the constructors' string arguments are four shapes, not one rule — pfUI passes a Button where a template name goes and the reference constructs the FontString and ignores it; SetInventoryItem returns THREE values, the reference's own paper doll destructuring repairCost; UnitIsVisible is object presence and nothing else (57 bytes, one branch, no distance — an out-of-range party member is exists-yes, visible-nil); the tapped pair reads the two dynamic-flags bits, and its repaint wire fires UNIT_FACTION on the tap bit — the half a passing predicate test never sees; UnitIsPartyLeader is two ORed legs covering disjoint failures (a held stranger leading their own party, and an out-of-range member with no object to read a flag from); and the four nameplate verbs take no argument and return zero values - the ghost's wisps compose onto its body instead of over it — the four emitters drew at full strength over a body at alpha 0.5, and now inherit the instance alpha down the attachment chain the way the reference recomposes it per frame; the residual we had parked ("the wisps should inherit the tint too") is REFUTED at the bytes: a particle batch commits hardcoded white RGB, and even that never reaches the fragment — particles are never tinted, only the alpha crosses, applied to the alpha byte alone — so the shipped shader split was already right and no code changes Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 20:15:16 +07:00
SamandClaude Fable 5	961d7e46a0	the select-screen ghost is four mechanisms, and a click raises windows ... - the select screen's ghost is four mechanisms off one roster-row bit, and the rest of them land: the FFX pass swap — the reference builds TWO pass pairs, world and glue, and a ghost row installs the glue DEATH pass over the whole diorama while the character list and buttons composite clean above it; the ghost-gated sun — the scene's light gather is wiped and one hardcoded overhead directional injected, its colours turning out to be Light.dbc's death preset (diffuse (94,153,198), ambient (26,56,85)) rather than the live capture we had parked them as, with the sign settled at the bytes (an authored light already carries one negation, so reusing that convention lights the ghost from underneath); and three glow weights corrected — the glue screens pin 0.30 login / 0.40 living select / 0.15 ghost where we rendered a flat 0.5 everywhere, so login and create get slightly less glow too, a fidelity correction to a value we had never verified - a press brings its window to the front: the mouse-DOWN handler raises the capture-or-hover frame unguarded, the toplevel gate living inside the worker — which is what makes a click on a child raise the window that owns it — and the release never raises; behind it, a frame-by-frame diff against the reference put back what the transcription dropped: toplevel on 47 frames (without it every raise trigger was a silent no-op across the whole interface), the mouse on 16 window bodies (a click on the bank's or merchant's own background was turning the camera or dropping your target), and id= on 81 frames — the slot index a 1.12 addon reads, where GetID() answered 0; a new gate reads our flags off the LOADED engine against the reference's, both directions with a reason each, so the next drift can't hide behind renamed templates - the jump stops being frame-rate dependent: the fall step becomes the reference's closed form (the mean of the endpoint velocities is the exact displacement rate, and both our movers had the error in opposite directions), taking the apex from 1.668 yd at 144 fps and 1.844 at 20 to a flat 1.640 against the analytic 1.640 — a live run apexed at 1.6410; and the knockback's freeze is its PLANTED direction bit, not a velocity test — air control opens exactly while the direction nibble is clear, the launch plants forward whatever its speed, and the bit now rides the wire, which we had never sent at all - the walk-away leash belongs to the OBJECT: the reference's watchdog overwrites the opener's stored range on the player typemask bit alone with (10+4)², so a player-shared quest panel is leashed at 14 yards — not at the 5.56-yard NPC reach, and not never, as two earlier readings had it — while a census of all 14 arming sites confirms every other NPC window already closed at the right distance; the walk-away also sends strictly LESS than the button: an NPC panel closes silently while a share still owes its decline Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 20:14:48 +07:00
SamandClaude Fable 5	1052393487	the HUD becomes UIParent's child, and cinematics fade through black ... - the transcription had dropped parent= on 72 declarations the reference carries (37 to UIParent, 28 to the action bar's art frame, 7 more), so the HUD was a field of top-level frames and UIParent:Hide() reached almost nothing — which is why a cinematic had to hide the UI at the draw level and then special-case the pointer; both workarounds are deleted and the HUD hides through the reference's own route; restoring only the UIParent half would have been the dangerous part (draw order is one flat sort, so a partial restore lifts frames over the neighbours they carry), and manifest order turns out to be load-bearing — a parent name that doesn't exist yet is silently ignored, so UIParent.xml moves to the reference's own slot with a test that fails if anything declares a child first; the chat still drew over a fly-by after all that, because the reference parents the chat windows on their TEMPLATE, invisible to a name-for-name diff when both sides rename — two declarations fix it, and the sweep that replaces the method loads the whole shipped manifest, fires CINEMATIC_START and asserts the survivors are exactly three: 91 frames visible before, 3 after, and the count comes back on stop - every cinematic boundary cuts through the reference's 0.25 s fade to black, and the next step hangs on the fade's COMPLETION, not a delay — the shot boundary is exactly where the reference does a blocking terrain load, so a delay-driven cut shows through a hitch and a completion-driven one cannot; and a fly-by now ends into black instead of a ~0.3 s loading cover: the reference raises no cover on that path, it holds black while the ground under the player loads back, so ours holds black while the stream catches up — budgeted, never an unbounded wait - the death arc keeps its promises across a relog: the body's own state machine now dies with the session — re-entering the world as a ghost used to match no arm of the world-edge machine, so the corpse query never went out and "Resurrect Now" stayed down until the next death; and a corpse wears its owner's guild crest — the corpse descriptor carries the guild field and resolves through the same crest cache a living tabard does, where ours dressed the body crestless - the char-select ghost gets its body kit: the screen resolves the state kit of the very "Ghost" aura a released ghost rides in the world — tint (140,185,253), alpha 0.5, and the Ghost_state wisp model at the body's attach point, pinned by a golden that decodes the row out of the shipped DBC; the airborne model is corrected before its fix lands — the address six comments called the gravity integrator is really the fall-velocity query, and a knockback is unsteerable because its launch PLANTS the forward bit, not because the arc is airborne (our guard was a velocity proxy, so a near-zero-speed knockback steered from frame 2); and the hover inspect card now decodes a picked part's mesh tag — alpha, rig slot, and both readings of the lighting payload side by side — turning a rebuild-and-two-probes hunt into a hover Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 20:13:31 +07:00
SamandClaude Fable 5	1d65531ef8	knockback is a jump the server aims, and tabards wear their guild crest ... - knockback is a jump whose seed came off the wire, not a server spline: SMSG_MOVE_KNOCK_BACK's four floats are exactly the MovementInfo jump tail in a different order (speedZ down-positive — a rising launch carries a negative value, the sign vmangos's own source marvels at), flown at the same take-off site as the keyboard jump and acked with the post-launch pose, which the server matches float-for-float within 0.01 before relaying to anyone; it is the one take-off that seeds BOTH axes, absolute in world XY, so it skips the arm that would hand the momentum back to a held key and reduce the whole knockback to a hop; root refuses it silently, a seated player stands up first, no MSG_MOVE_JUMP goes out — that packet is the keyboard's, and vmangos logs a jump at knockback speed as an overspeed cheat — and MSG_MOVE_KNOCK_BACK turns out not to be dead: an observed knockback replays as the arc it is - your guild crest reaches your tabard: the three GuildEmblems layers — background, border, emblem, on both torso tiles — belong to no equipment slot, so the body composite never wrote those cells and every guild wore the Guild Tabard's own default art; the install is gated on a display flag only seven of 29,604 rows set, of which exactly one is worn by any item — which is why every battleground and Argent Dawn tabard keeps its own art; the five wire indices are signed, because an undesigned guild carries −1 in all five and the client refuses to paint on the first −1 (unguarded, the commonest guild state on a young realm would have worn a blank tabard); the emblem keys the skin cache, so two guilds in the same tabard display cannot lend each other crests; and a guild tabard's background hides the bra where a plain tabard does not — it is a cutout, 79.5% opaque off the shipped art, and what shows at the sides is bare skin - a fly-by is framed by the world camera, not the M2 camera record's fov: a census of every reader shows the record's fov and clip planes reach only the portrait and Model-frame paths — nothing on the cinematic path reads any of the three — so fifteen of the sixteen shipped shots were rendering at HALF their intended vertical FOV, visibly over-zoomed on every race intro but the Undead; the fix is a deletion, and it dissolves a standing puzzle: the authored 0.22 and 27.8-yard "clips", which would render the dwarf intro's mountains as empty sky, are not clips at all on this path — two floats nobody consumes look exactly like that - 104 of Onyxia's lava traps stood on an interior floor classified as outdoors, taking the day/night sun inside a windowless lair: a GameObject's light node anchors at its model's AUTHORED bounding-box centre, and the trap's box reaches 98.42 yd past its bind pose — the anchor sits ~40 yd under the chamber floor, where a downward ray finds nothing above it to claim; the reference recovers with a 1000-yd upward retry at the claim itself, which benilla had only at the colour sample, so the verdict was decided before any colour was read; all 104 now classify interior, beside the 26 rocks and 119 eggs that always did — and the drowned-corpse pose came back from its cross-check confirmed at the bytes (a strict liquid-depth test at exactly two-thirds of a yard, Dead or Drowned), with the two reference-side notes it contradicted corrected over there Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 19:50:50 +07:00
SamandClaude Fable 5	d4690cc67c	a cinematic holds your body, and all seven chat windows exist ... - a cinematic holds the body while the camera flies: the streaming focus follows the eye — a Tauren's intro opens 1741 yards from the body — so the body's own ground unloaded under it and it fell (z=59.4 to −62 in four seconds, the server yanking it back); it now takes the same hold a teleport takes, released when the focus comes home to a resident tile, and the loading screen stops covering the fly-by, because "the player's ground is missing" is not a load while the eye is deliberately elsewhere; the letterbox measurement prints as an ordinary info line, and scripts/cine.sh plays any one cinematic and prints what the client did to the world while the camera was away - a cinematic cuts the zone music: the reference asserts exactly what /console EnableMusic 0 asserts, through the same setter — a cut, not a fade, the music pump dead for the duration, resume at +3.000 s — and benilla asserts the runtime flag, never the CVar, or quitting during a 102-second intro would save your music off for good; ambience is deliberately untouched (it has the analogous flag and the cinematic never writes it); the narration has no category at all — its flags route it around every volume slider but not around the Enable Sound Effects gate, so unchecking that silences it — and the listener follows the cinematic camera for the duration - addons can walk a UI they did not build: GetChildren/GetNumChildren/GetRegions/GetNumRegions land — the basis of the whole reskin genre, and the wall four of the twenty most-installed 1.12 addons died on: pfUI drew nothing before and now paints 495 quads, ShaguDPS 71 → 566, pfQuest 3 → 498 — with the reference's own quirks: the title region is NOT in the list (its creation path never reaches the region linker), hidden regions are returned and counted, oldest first; the Model widget gets its Lua surface (SetModel, SetSequence and siblings, read out of the shipped image, where SetRotation and SetFacing are two names for one slot); and the Minimap carries the nine engine-created Model children no XML declares, built before the frame descent, the ninth being the player arrow Questie indexes as ({Minimap:GetChildren()})[9] - all seven chat windows come from one template, tabs and scroll buttons included — whether a window is shown has nothing to do with whether its globals resolve, and pfUI reads every piece of that furniture for all seven, unguarded; isDocked only on windows 1 and 2 (the chat-cache's own DOCKED byte), SELECTED_CHAT_FRAME seated beside DEFAULT_CHAT_FRAME (the window you last picked vs the one FrameXML prints to), and the four per-window fields ChatFrame_OnLoad seats; two more one-name walls fell — a Button's label answers to its second spelling NormalText, which KLHThreatMeter's column headers die without, and ABANDON_QUEST_WITH_ITEMS is declared because a dialog table is read at load whatever the client later shows, leaving Questie with zero errors of ours — and the addon harness gains --probe/--eval, one addon's VM answering a question in ~5 s instead of a 90 s corpus run Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 19:50:08 +07:00
SamandClaude Fable 5	e9e38bf6ec	your corpse is your own dead body, and particles take the room's light ... - corpses render: a corpse create spawned a real entity, latched its guid for the reclaim send, and drew nothing — one missing EntityKind variant folded it into the catch-all and four downstream lanes then failed by default, not even the debug cube fired; the reference forks once, on the bones flag: a fresh body builds a character composite from the corpse's OWN appearance bytes and 19 item slots through the same compositor a living player is dressed by (its own hide-helm/hide-cloak bits honoured, the ranged slot never dresses, and weapons never dress because the client looks a packed display word up as a guid — we reproduce the outcome, not the dead lookup), while a bone pile takes its race and sex's DeathSkeleton, sixteen of which ship; the pose is one settled arm — Dead, or Drowned when the body lies more than 0.66666669 yd under a liquid surface — and a corpse is not a unit, so clicking one no longer sends a selection for a guid no unit owns - a lit particle takes its MODEL's committed light node, not the sky's: indoors the reference lights an object's particles from the object's own node — the footprint's baked floor colours on the fixed interior axis plus the room's placed lights — never the day/night sun; a particle quad carries exactly one normal, world up, one constant per draw, so the whole node collapses to three floats, published off the same fold the model's mesh renders from so a mesh and its own particles can never disagree about the room they stand in; in Onyxia's Lair the 78 lit falling-rock emitters on the 26 anchors that classify interior now take the room's light - and that lit lane is FIXED FUNCTION, not the mesh's soft SH curve: the shipped client null-binds both shader slots for a particle batch and re-zeroes the shader mask at the head of every batch, so a model and its own particles share the same committed light words and legitimately differ in the curve applied to them — the mesh's wrapping lobe against the hardware's max(N·L, 0) is 0.900 vs 0.869 on the key axis, up to the three nearest room lights arrive diffuse-only under GL's own 1/(0.7d + 0.03d²) falloff (a light at or below the emitter cannot reach it at all), and the clamp sits on the product, never the multiplier, because it is GL that clamps the lit vertex colour - the cinematic letterbox has no aspect test in it: CinematicFrame.lua's if guards a recompute, not whether there are bars — below 4:3 the reference's bar textures simply stand at their declared 1024x128 on the client's native 1024x768 sheet, which is height/6, the same formula evaluated at exactly 4:3 — so one expression covers every window shape, min((h − min(w/2, h))/2, h/6): a 2:1 picture when wide, bars at 4:3 too; and a shown keyboard frame swallows EVERY key, not the ten names the host used to feed it — you could walk around underneath your own intro; the proof is the reference's own handler, which has to RunBinding("SCREENSHOT") by hand to get one key back; ESC still skips, and a W already held when the shot starts keeps running, which is faithful too Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-31 19:49:15 +07:00
SamandClaude Fable 5	d7fb26713c	cinematics play, and a lit particle stops brightening as you turn ... - cinematics play: `SMSG_TRIGGER_CINEMATIC` resolves through `CinematicSequences.dbc` → `CinematicCamera.dbc` → a `Cameras\*.m2` whose camera record is an eye path, a look-at path and a roll — cubic tracks with the full step / linear / Bézier / Hermite dispatch the bone loops collapse to step-or-linear (every shipped fly-by is Bézier), planted at the row's origin and facing (`Rz(+facing)`, checked against vmangos's own waypoint table: 8.5 yd mean on the human path, 154 with the other sign); FOV is per shot (the Undead intro authors 90°, the rest 45°), roll is authored around whole turns, the flight ends at the sequence band's end, a trigger arriving before the loading cover lifts is latched and started when it does, ESC skips through `CinematicFrame`'s own `OnKeyDown` (`StopCinematic` has zero native callers), and every path acks — natural end, ESC, or an unresolvable trigger immediately, as before; `.debug play cinematic <id>` replays any of them - a cinematic hides the HUD — `UIParent:Hide()` reaches nothing here because the HUD frames are not its children, so the ALT-Z draw-level hide is used and the letterbox moves out of FrameXML onto Bevy UI nodes where that hide cannot reach it, recomputed per frame (bars only wider than 4:3, the picture cropped to 2:1) — and the restore only un-hides a UI this module hid - a lit particle's normal is world up, one constant per draw, not camera-facing: Onyxia's always-on floor haze read about four times brighter from a level camera than from above because the shader dotted a to-eye normal against a fixed light; the reference uploads row 2 of the view matrix — world +Z carried into eye space — on every vertex and carries the light into the same frame, so the rotation cancels; 400 lit emitters (the waterfalls, chimney and campfire smoke, Westfall dust, Dun Morogh snow) stop brightening and dimming as you turn - the `play` profile drops ThinLTO — rustc #86049 was intermittently poisoning the incremental link with `_anon.<hash>.llvm.<id>` symbols promoted by an earlier compile (48 unresolved across 7 codegen units; 1065 such symbols → 0 after) — and `benilla-formats` stops taking `image` with default features, which had been compiling a complete AV1 encoder into a crate that writes two PNGs: 499 → 466 workspace crates Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-29 17:24:50 +07:00
SamandClaude Fable 5	1fb4d75acd	effects ride the lift with you, and a long frame keeps you on the deck ... - a rider's effects are stored in the deck's frame: standing still on the Thunder Bluff lift, a weapon enchant's streak hung four yards below the blade and through the floor, because its edges were laid down in world space while the car climbed out from under them — the reference folds the transport matrix (translate · Rz(facing), NULL on the ground) into the emitter at birth and back out at draw, so committed content rides the deck and only a boundary re-expresses it; a `RideFrame` on the rider's model, inherited down the parent-model chain the way the reference copies the transport pointer to every child each frame, one fold at birth, one at draw, one at the boundary — ribbons take it unconditionally, world-mode particle clouds take it, model-mode clouds never see it; measured on the car: streak 4.2 yd → 0.00 over a 49 yd climb, while the same rider thrown off the top streaks 0.9 → 24.8 yd on the way down - switching away from the window while riding a lift dropped you through it: a collider moved in `Update` is invisible to `Update`'s own spatial queries until the next physics step, so on a 1 s frame the deck descended 7 yd with its broad-phase box left behind, the ground probe found nothing under the feet, and the second long frame's clamped 250 ms of gravity put them under the deck; avian's own `update_moved_collider_aabbs` now runs inside the transport chain right after the deck moves (0.03 ms p50, flat across scene density), and `WOW_STALL` injects long frames on demand, since a probe window is never throttled by the OS - `WOW_TRAIL_CENSUS` prints each live ribbon trail's world-space spread beside its spread on the deck — the pair that separates "not riding" from "riding but not folded"; it first matched zero trails in every scene by requiring a component trails deliberately never carry, recorded rather than quietly fixed, because an instrument answering a confident zero cost an afternoon Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-29 17:17:25 +07:00
SamandClaude Fable 5	80551cab5e	the cave hides the lake, the lifts come back, the bolt gets its width ... - from inside a building the open-world water outside draws only through a doorway you can see, as terrain and trees already did — the Caverns of Time lake is gone from the cavern's ceiling: there was never a second visibility lane for open-world liquid (the old deferral said "it already has an authority"; a sweep of every visibility writer found none, so Bevy's frustum test was the lake's only gate), an ADT liquid layer is tagged exterior scene at spawn and the existing window cull owns it, one 33 yd chunk layer at a time — which the reference's bytes then confirmed is its unit too (`CChunkLiquid`, reachable only from the per-window walk, a sealed room's supply discarded not deferred); a building's own pool also rides the Map Objects toggle now, so switching buildings off no longer leaves the canals hanging in the air - the Thunder Bluff elevators and every other Kalimdor lift were invisible for the whole session whenever you logged in on that continent: the lift arm snapshotted the current map one stage before the login frame had written it, so every lift compared itself against map 0 and hid as off-map — a lift carries no map at all now (it is only ever streamed to players standing on its own map), the off-map term stays for the boats it was written for, and `WOW_LIFT_CENSUS` prints each transport's arm state, cycle and visibility (13 transports, 11 hidden → 0) - the lightning across Razorgore's room and Chain Lightning's mob-to-mob hop drew as hairlines: the chain renderer took each hop's cross-section as the world-horizontal perpendicular, where the reference takes `(−d.y, d.x, 0)` in eye space — the ribbon is view-plane aligned and keeps its authored 2 × halfWidth yards at every angle, so a long level beam watched from its own plane no longer thins to a line (the "a hop seen end-on genuinely collapses" note was wrong and is gone) - a display that names a model builds even when it draws no batch — Naxxramas' Unholy Axe / Swords / Staff are an invisible stalker holding a Monster weapon, and the held weapon had nowhere to hang; the same gate was keeping 43 emitter-only GameObjects dark (house and blacksmith smoke, Ironforge's lava steam, Stratholme's embers, every firework), and a name sits at the authored PlayerName attachment instead of at the feet; and a vendor's foot-shuffle on turning to you now blends the whole way in and runs to its own clip window — a 500 ms step on 1100 of 1130 shipped shuffle records — where the latch had tested the gap left against a band sixty times too wide, and nothing ever ended the step Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-29 17:15:08 +07:00
SamandClaude Fable 5	1a1d7e50dc	the world redraws on resize, and terrain takes its DXT blocks ... - the world froze on every resize or graphics change: the backdrop was resized by writing a fresh image through the existing handle, and a mutated `Image` is a whole new GPU texture while a material's bind group is captured once and never re-captured — so the camera drew into a texture the UI never sampled, and the last frame stretched over the window forever; a resized world target is now a NEW asset (publish, retire the old one, and the UI pass forgets materials for removed images so a window drag does not leak a 46 MB target per frame), and a resized capture now matches the native render bit for bit - terrain tile layer arrays take the DXT blocks when the tile's whole layer set allows it — one array has one format, so eligibility is the tile's, not a texture's — and the install's own census says narrow is enough: 87 % of Azeroth's tiles and 92 % of Kalimdor's pass, layer bytes 1.57 GB → 0.63 and 2.23 → 0.76; the decode-tax comment is corrected on the way past — decoding costs 8x DXT1's bytes and 4x DXT3/DXT5's, not 4x and 2x - Environment Detail is the reference's slider, not a dropdown — row 27 of 1.12's own slider table, with the readout keeping the stop names because a bare "1" means nothing to a player — and the shipped ground cover comes off High to Medium: the x3 clutter density had been matched to a configured install, and Medium is the nearest stop no sparser than what a fresh install draws on capable hardware - Display Mode is modern Classic's two-entry dropdown — Windowed (Fullscreen) / Windowed, a boolean at Blizzard's own source, where the three-mode list is pre-8.0.1 folklore — riding `gxWindow` so saved configs keep meaning, with the label-against-value polarity pinned by the row's first test; the Multisampling row comes off the Graphics page for now (the CVar, the device clamp and the bindings all stay): with alpha-to-coverage set nowhere, MSAA touches polygon silhouettes only, and render scale on the same page measures four times the visual change Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:16:01 +07:00
SamandClaude Fable 5	a48441bca8	trilinear ships on, and every line of text rounds once ... - the texture filter is the reference's policy — two globals, one process-wide mode, no per-texture override — registered as real CVars every sampler reads: the hardcoded trilinear + aniso 8 at five sampler sites had been read off an apitrace of an install whose `Config.wtf` set `anisotropic "16"`, a configured value taken for a default; and `trilinear` ships ON, because the registrar's `"0"` is not what a fresh install renders at — first-launch hardware detection overwrites sixteen video CVars from its hardware table, both rows a real GPU can reach set trilinear, and the reference's own `gx.log` on this machine names that row — while `anisotropic` is not among the sixteen, so anisotropy stays off - the MSAA device clamp ran before the config was read, so a saved `gxMultisample = 8` carried from a machine that offers it still killed the render thread on one that stops at 4; the ceiling is now a property of the write itself, covering the config file, `SetCVar`, the dropdown and any writer added later — and an empty device list is no opinion, so headless runs pass through - one line, one rounding: at a fractional display scale the per-glyph snap's f32 round-trip moved half-pixel ties by an ulp, so caps and ascenders rounded up while x-height letters stayed — the letters left the baseline, the thing the reporter photographed; the line's two anchors now round once and each glyph adds integer device-pixel offsets, so a line cannot be anything but flat at any scale, proven over 15,680 layouts — and `$WOW_DPI`'s size guard, which compared a physical request against the logical size and refused every non-1x run, now reads the unit it is denominated in - a test that panics while holding `ENV_LOCK` no longer poisons the tests behind it — one missing knob resource had reported as twelve red tests, eleven naming the mutex and none naming the bug; there is nothing for the poison to protect, since the guard's `Drop` restores the environment during the panicking test's own unwind Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:15:45 +07:00
SamandClaude Fable 5	72ca69ef5d	the world shrinks without the UI, and a party member keeps their face ... - `renderScale`: the world renders into its off-screen target at a fraction — or a multiple — of the window while text, icons and frame art stay native; the logical viewport comes out arithmetically unchanged, so picking cannot see the scale at all, and every hand-written sampler takes the `log2(scale)` mip bias below 100 % and none above it, where the sharper mip is what makes supersampling work — measured here at 26x the world pixels for +0.02 ms, and reachable as a Render Scale row on the Graphics page (50–200 %), staged to Apply so a dragged slider rebuilds the world target once instead of thirty times a sweep - a party member who walks out of range keeps their numbers and their face (B334): the client asks — `CMSG_REQUEST_PARTY_MEMBER_STATS` goes out on the despawn edge and for a roster member gained unseen, the same two edges the reference sends from — the record is snapshotted whole from the live descriptor at the moment the object deactivates, a never-seen member seats at the reference's 1/1 placeholder rather than an empty bar, and the raw-to-display divide happens at the read, so an out-of-range warrior stops reading ten times the rage; the portrait binds a guid-keyed bake cache built as a handover — the booth's own target image moves into the cache as its member streams out, so the face freezes with no copy and no flicker — and the self-review adds that a changed occupant is a re-bake and the record's dead/ghost bits are read - a slider row's number sits on its label's line (B232): the readout hung off the control frame and inherited a +3 art nudge that exists to seat the groove texture — four pixels off on an eleven-pixel cap height at the reporter's geometry — so the nudge moves onto the two frames that draw the art, across eight rows on three pages, and the bar itself does not move a pixel; `$WOW_DPI` renders on a player's pixel grid, so a 1x reporter's screenshot and a capture from this 2x machine compare pixel-for-pixel Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:15:06 +07:00
SamandClaude Fable 5	6b2a75e8a0	instruments for motion at the glass, and a macro hovers to its name ... - the jitter meter measures the palette and the pixels, not just the pose: `pal` probes the rows the vertex stage actually blends (the old bone columns read a matrix the GPU never sees), `SCREEN` carries them to viewport pixels by the shader's own arithmetic, `camrot` watches the camera's rotation where only its translation had been vouching for "bit-stable", and `WOW_JITTER_RAW` dumps one bone's series unaggregated — rotations too, since a translation can be smooth while its frames are not - `WOW_FIXED_DT=<ms>` pins the animation advance so a screenshot burst measures the render rather than its own stall — arming on the wall clock, because during load the app runs ~385 updates a second and a pinned clock reached "90 seconds" in 14 s, photographing an empty world - `benilla-visual flow` estimates the whole window's motion by least squares over every pixel (noise floor 0.0000 px on static regions, and its brightness-constancy limit is measured and stated: invalid on a lit deforming body), `edge` reads a silhouette's sub-pixel crossing renormalised against the lighting with an honest ±0.5 px floor, `WOW_TWIST_GAP` pins a yaw gap so the body counter-twist is exercised at all — every earlier "removing it changed nothing" had measured a pass that never ran — and `WOW_DITHER` adds the dither the frame's 8-bit output has never had, off by default because the reference's framebuffer was undithered - hovering a MACRO action slot shows one white line, its name — the reference's macro leg renders exactly that and no more, in its "normal" white, where ours was a `_ => Ok(())` from before macros shipped; and `WOW_MSAA=8` is refused with a loud warning instead of panicking the render thread, through one parser both world-camera spawners now share Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:14:35 +07:00
SamandClaude Fable 5	bf7201f4d3	action-bar hover stops re-deriving, and multisampling asks the GPU ... - hovering the action bars re-derived the whole layout graph and hovering bags did not: `GameTooltip:SetOwner(owner, "ANCHOR_NONE")` dropped the tooltip's anchors through the conservative whole-graph derive, and with `UberTooltips` at its shipped default every action-bar button routes through that arm while a bag slot anchors `ANCHOR_RIGHT` and never does — the arm now retargets to the empty target set, and derive frames fall from 15.9 % of hovered frames to 0.2 % - `gxMultisample` clamps to what the GPU actually offers: Apple GPUs stop at 4x, so a typed 8 was a wgpu validation panic on frame one — the clamp intersects the supported counts of every format the client multisamples into and steps the request down the way the reference's device-init retry loop does, with a warning that says so - the Multisampling dropdown lands on the Graphics page with the device's own stops, built from the same enumeration that clamps a typed value on the reference's three bindings (`GetMultisampleFormats`, `GetCurrentMultisampleFormat`, `SetMultisampleFormat`) — a fixed Off/2x/4x/8x ladder would put a crash on the menu of every GPU that stops at 4 - capture hardening: a run whose `$WOW_WIN` the display clamped is refused in ~2 s instead of producing a wrong-sized shot to diff, a capture never reads the OS pointer (a cursor resting over the window arms a hover and a tooltip that a cursor an inch to the left does not), and the stall sampler writes into `benilla-config` instead of a hand-built `~/Library/Logs` path Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:13:27 +07:00
SamandClaude Fable 5	17a47acd71	borderless fullscreen ships, and the 4x MSAA nobody chose is off ... - the display mode is borderless fullscreen by default, and there is no exclusive mode at all — winit ignores exclusive on Wayland, X11's path cannot restore the desktop after a crash, macOS has none, and WoW itself dropped it in 8.0.1; 1.12's `gxWindow` row keeps its name, polarity and default with "0" redefined as borderless, read before the window exists so it is born in the player's mode (a window that does not fill gamescope's nested output is a documented input break, and the recorded workaround is "set the game to fullscreen"), `gxResolution` remembers the windowed size, and the Graphics page grows 1.12's own Windowed Mode checkbox - the UI camera says `Msaa::Off` out loud, because silence there meant Bevy's default 4×: the world arrives on this camera as an already-resolved quad and everything it draws itself is axis-aligned rects, so four samples bought a 4× colour target, a 4× depth texture nothing ever writes, 4× the fill on every quad and a resolve — for a measured difference of MAE ≤ 0.024, all of it quad borders now drawn at their authored crispness, which is also the reference's hard-edged 8-bit direction - `gxMultisample` is a real player setting, and it ships off — the reference's own default: the reference registers no literal, it formats field 21 of the `VideoHardware.dbc` row the GPU matched, a 193-row table that holds nothing above 2 anywhere and whose every 2004-era id misses a modern card, so the fallback rows answer and all of them hold 1; the setting persists, all three world-lane cameras read it on the reference's [1,16] scale rounded down the way its own retry loop steps, and `$WOW_MSAA` survives session-only - two instruments for the tooltip hunt: `rebuild_ui_mesh` — which sorts, clips and re-splits every quad in the interface on any dirty frame — gets a meter carrying total/sort/split/write beside the script phases whose flat reading had stood as an alibi, and `WOW_PROBE_HOVER` sweeps the real cursor across named frames through the real pointer path, pressing nothing: the hover recorder records what a hand does, and this supplies the hand Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-08-27 15:12:37 +07:00