Any write that changed the layout graph's roster (a frame or region created, freed, re-owned or re-pointed) made the next resolve derive the whole graph: about 1.35 ms for one `CreateTexture` or `CreateFrame` on the stock UI, against 0.05 ms for a no-op.
A write now names the nodes it touched, and the resolve re-derives just those in the cached graph: a joining node is added with its edges, a leaving one is removed and its dependents seeded, a staying one has its edges and inputs re-synced. The full derivation stays the fallback. A region's owner and an art change are now layout inputs that name their region. Verify now also compares the graph itself.
A randomized test checks rects and graph against a scratch solve after every batch of creates, frees, re-owns and re-anchors. `CreateTexture` now costs 0.14 ms, `CreateFrame` 0.15 ms, a backdrop rebuild 0.22 ms.
Fixes#903
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GameTooltip's lifecycle took shortcuts the reference does not: a 0.5 s linear fade armed once, a fade cancelled by new lines, cursor-anchored plates that faded instead of hiding, a SetOwner that never hid first, the last anchor kept after a hide, and a health bar shown and hidden without its handlers.
In the reference, FadeOut re-arms a 2 s countdown that holds full alpha for half and then ramps, or hides at once at the cursor anchor. Only the owner core, the show and FadeOut write the fade. SetOwner runs the tooltip's own hide first. The un-own leaves ANCHOR_NONE. The clear hides the bar through its own transition.
Every builder, verb and world or minimap seat now goes through these functions; an uncached item hides and un-owns as in the reference.
Tests pin each of these.
Fixes#889Fixes#134Fixes#896Fixes#893Fixes#895Fixes#843
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The tooltip fade, a model's animation clock and its OnAnimFinished ran after the whole OnUpdate walk, and OnUpdateModel fired in creation order. A tooltip hidden through its parent also kept its owner and fade.
In the reference, the walk makes one call per frame: its OnUpdate, then its kind's engine update (a tooltip's fade step, a model pane's clock and completion test). OnUpdateModel fires later, from the paint, in draw order. A tooltip's own hide and show run from any caller, its parent's cascade included: the hide un-owns and clears it; the show needs an owner.
The walk now runs a per-kind update in each frame's slot, OnUpdateModel fires per stratum in draw order after the hover re-pick, and show and hide dispatch to a kind's override. CharacterModelBase's override is left for #890.
Tests pin each order and a parent's hide and show.
Fixes#861Fixes#863Fixes#864
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Script errors reached the Lua error handler as mlua's display text: a `runtime error: ` word and a `stack traceback:` block, which overflow the stock `ScriptErrors` dialog. And a raise in a `CreateFrame` template's OnLoad was logged as a warning and never reached the handler at all.
In the reference, handlers run under `lua_pcall` with the error function `0x703b40`, which calls the `seterrorhandler` function with the bare message. `CreateFrame` builds its template through the XML instantiator `0x6ee280`, whose OnLoad runs under that same function (`0x704d50`).
Now every caught error records two texts: the decorated one for benilla's log and the bare message for the handler. A load report pairs the two per row, and `CreateFrame` records its template's errors as script errors.
Tests: the handler hears bare messages from handlers, and a template OnLoad raise reaches it.
Fixes#697Fixes#859
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Show and hide propagated a whole subtree first, then fired from a snapshot, parent first: a parent's `OnShow` ran before its children's, and a child hidden by an earlier handler still got its `OnShow`. A frame visible at the end of its load got no `OnShow` at all.
In 1.12.1 show `0x76ae10` and hide `0x76ad50` mark the frame, walk the live child list re-checking each child's gate, and fire the frame's own handler last (`0x76aef5`, `0x76adfd`). The post-load hook fires `OnShow` after `OnLoad` for a visible frame (`0x76a3b3`), from XML and `CreateFrame` alike.
Now handlers run inside a live post-order walk, and the post-load `OnShow` fires. Tests cover the order, a sibling hidden mid-walk, and one or two `OnShow`s after a load; they fail without the fix. Four existing tests' expected orders flip to the reference's.
Fixes#350Fixes#59🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The host's state leaves `_G`: the eleven `__benilla_*` globals, which any addon could read and overwrite, move into the host's model (and, for the error handler, the Lua registry, as the reference keeps it), and the getters over them are Rust bindings answering the reference's shapes.
- `GetTime` returns one number (`0x515ea0`), and `GetGameTime` returns hour then minute (`0x515ee0`). The tick advances the model clock, and every cooldown and buff reader reads it.
- `GetZoneText`, `GetRealZoneText`, `GetSubZoneText` and `GetMinimapZoneText` always answer a string, `""` when unset (`0x48a0a0`-`0x48a100`). `GetZonePVPInfo` answers three values (`0x48d540`).
- `seterrorhandler`/`geterrorhandler` work over a registry slot that starts empty, as the reference's does (`0x7039e0`, `0x702950`, `0x702900`). A non-function raises `Usage: seterrorhandler(errfunc)`. The engine dispatches caught errors through the slot, not through the global (`0x703b40`). The stock `BasicControls.xml` installs `_ERRORMESSAGE` before any addon, so play is unchanged.
- The engine's surface allowlist now holds only `_G`. `scripts/api-coverage.sh` no longer excuses `__` names.
Verified:
- New tests show each of the eleven names absent from `_G` and that writing it changes nothing. On main, `__benilla_now = 0` rewound `GetTime` and every buff and cooldown.
- The getter shapes and the error-handler slot are pinned. The surface gate fails when a name is re-added.
- `scripts/gates.sh` and `scripts/smoke.sh` are green.
- A live run read the clock, zone texts, PvP info and game time, then showed that writing the old names changes nothing.
Built on macOS; no platform seam touched.
Fixes#576🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
The Lua VM keeps no surface the 1.12.1 client lacks.
- **Removed:** the nine Era-only globals: `UnitAura`, `CancelUnitBuff`, `GetCursorInfo`, `GetInventoryItemID`, `GetPlayerFacing`, `GetTradePartnerName`, `IsGossipOptionCoded`, `UnitIsAFK`, `UnitIsDND`.
- **Hooks off Lua:** `GameTooltip:BenillaSetItemById` and `:BenillaGetTooltipOwner` are replaced by Rust-side calls. The uncalled `Benilla*` hooks are deleted.
- **Frame strata:** `SetFrameStrata` knows the reference's eight strata and raises its message for any other (`0x774450`).
- **`SetDrawLayer`** reads the layer name alone (`0x79a780`).
- **`Set<State>Texture`** reads a number as a path (`0x781b23`); there is no colour form.
- **`PlaySound`** names a kit and returns nothing (`0x4586d0`). `PlaySoundFile` returns one value (`0x458780`); benilla answers 1, a recorded deviation, since it plays after the call.
- **`OnClick`** sets only `arg1` (`0x779540`).
- **Quest getters:** `GetQuestItemInfo` and the quest-log reward and choice getters answer the reference's five values and its miss behaviour (`0x501f80`, `0x4e0b00`, `0x4e0cf0`).
- **`PickupContainerItem`** returns nothing (`0x4f9b30`).
- **Surface gates:** the bare VM's globals and every widget class's methods are held to `reference/1.12-globals.tsv` with one commented allowlist. A new gate on the full production load (core plus layer) allows beyond that only the layer's `Benilla`-prefixed globals.
Verified:
- A test per removal and per corrected form; a scripted undo of each fails its test.
- Each gate fails on a re-added name.
- `scripts/gates.sh` and `scripts/smoke.sh` are green.
- A live run showed the same warnings and 0 script errors before and after.
- In the 219-addon corpus, AckisRecipeList's bundled AceGUI-3.0 and Roid-Macros' Grid extension now behave as they do on 1.12.1.
Built on macOS; no platform seam touched.
Fixes#171🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- Spell tooltips for pet spells, quest reward spells, Beast Training, enchant links and pet or target-of-target debuffs show whole on the first hover
- The spell-tooltip tests' doc comments say the one thing the test name does not, in a line
- The spell feed builds each view once per frame, the pet-bar Firebolt test slot is one helper, and the wait's comment reads in one pass
- A spell tooltip completed by its view fires no second OnTooltipCleared: one setter call, one event, as in 1.12.1
- The spell feed reads every aura and the tracking spell from the VM alone, dropping the three app-side aura reads that duplicated it
- The spell feed reads the auras from the VM's player list and every guid's list, the aura model main moved to
- A spell tooltip waiting on its view re-renders only while it still shows what the miss drew, so a line Lua rewrote in place, by AppendText or a cell's SetTextColor, is kept
- The spell tooltip's one-frame re-render is stated as a known difference from the reference, which builds every spell tooltip at the call
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-Authored-By: Sam <samuv.eth@gmail.com>
`"mouseover"` now names nobody once the mouse leaves a unit, as in 1.12.1, where the publisher `0x492890` zeroes the mouseover guid pair on teardown (`0x4928e8`, `0x4928f2`) and a corpse or GameObject hover leaves it naming no unit. Before, `UnitExists("mouseover")`, `UnitName("mouseover")` and `UnitIsUnit("mouseover", …)` kept answering for the last unit hovered.
- The world tooltip's health bar follows its subject's guid, as the builder `0x529fe0` registers its health watcher on the guid (`0x52a0a8`), so clearing the token leaves the fading plate and its bar as they were; the clear fires no `UPDATE_MOUSEOVER_UNIT`, whose one fire site is the unit arm (`0x4929b3`).
- One definition of the hovered unit (none when a nearer GameObject wins the pick) serves the tooltip, the unit tokens and the aura feed, so they clear together.
Fixes#500
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Part of #41: the comments of `benilla-ui` (the Lua UI engine, its FrameXML loader and its widgets) are rewritten to the rule in `docs/METHOD.md`, which is what the code does and the reference fact behind it, in a line. 43,140 comment lines become 15,161.
- **The mechanical strip.** The citations that were only a private record number go first, and each edit is shown to be a pure removal.
- **Wrong facts corrected.** The rewrite corrects the comments that stated wrong facts. Among them:
- `0x64a480` is the case-sensitive compare (a byte `repz cmpsb`), and `0x64a4c0` the case-insensitive one.
- A frame's id sits at `+0xb0`, not `+0xb4`. `EnableKeyboard`/`IsKeyboardEnabled` are `0x776ec0`/`0x776f90`.
- A region's unpinned edge has no fallback to its owner's rect: the owner supplies only the scale.
- `$parent` expands to the nearest named ancestor, or "Top" when none is named (`0x76c5b0`), not simply to the parent.
- A StatusBar fills from its internal normalised fraction, not from `GetValue()`.
- The addresses of `StartDuel` and `StartDuelUnit` were swapped.
- Many "untraced" notes now give the traced reference fact: the drag threshold, the per-grip sizing edges, the empty-`autoFocus` skip, and `SetTexCoordModifiesRect`'s default.
- Stock FrameXML line citations are checked against the stock 1.12 files and corrected. `ref-` file names become the stock file names.
- **Deviation labels.** "Deviation:" now marks only a choice with a reason. A surface that exists only for the retired Era addon target says so plainly: it is not 1.12.
- **Verification.** Every file strips to identical code, and `cargo fmt --check` is clean. Nine second readers compared every file old against new, restored the facts the rewrite had dropped, and checked each new citation against the code, the stock files, vmangos and the reference.
Behaviour differences the rewrite surfaced are filed as their own issues (#131–#142), and more are being validated. The comments here state what the code does. String literals are untouched; they are #92.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Comments that pointed into the maintainer's private reverse-engineering notes (a note by name, a
path, a section or finding id, the analysis's own name for a client function, a process word) now
carry the fact itself, cited by the reference client's address, by vmangos file and line, or as
"the reference". Nobody outside could follow those pointers.
Comment text only, plus the test-assert and log strings that carried the same pointers. No claim
changes: a comment the notes have since corrected keeps its wording here, and correcting it is
separate work.
The two reference generators, `scripts/gen-reference-globals.py` and
`scripts/gen-reference-verb-events.py`, named the private notes' layout to find their inputs. They
now take the input files as arguments (`--capture`, `--framexml`, `--names`, `--extents`,
`--disasm`). Run on the same inputs, both write the committed tables row for row; only the two
header lines that named the old environment variable change.
Checked per file:
- the code is identical: each file with its comments stripped compares equal before and after; a
changed string literal is compared under a mask that proves only literal text moved and every
literal kept its line count;
- no client address the old text cited was dropped;
- a second reviewer checked every added address against its derivation;
- a detector for the notes' names, paths, ids, function names and process words finds none left;
- `scripts/gates.sh` is green.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- a creature walking a path no longer cuts its first corner: the path's first waypoint was dropped at decode (2342)
- two characters whose names differ only in accented or non-Latin letters no longer share macros, bindings and SavedVariables; a file that failed to load is not overwritten at logout (2342)
- custom chat channels survive a logout, a collapsed quest header keeps its quests watched, and a trackpad swipe scrolls as far as it adds up to (2342)
- a hearth off a zeppelin no longer lands back on its deck, and an opened lockbox with loot left in it unlocks (2342)
- every snapshot runs on Vulkan, in a container, before it ships (2344)
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
- the shift-hover item compare is gone: 1.12.1 compares only from the vendor and auction rows, and the plate those seat picks its side so it never slides back over its own tooltip (2210, 2207)
- UnitIsPlusMob is implemented, off the flag bit and not the creature rank, so an addon that calls it no longer raises (2209)
- a top-level frame named $parentX with a parent= attribute is published under that parent's name, so ClassIcons loads (2208)
- a parent= that resolves to nothing leaves the frame parentless and logs it, as the reference does, instead of falling back to the enclosing frame (2213)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- loot rows show their Quickslot border, depress on click and light up under the cursor: the stock `<LootButton>` is parsed by the Button loader, as the reference's vtable has it (2195)
- holding W keeps you running through Enter to chat and M to close the map: a UI frame eating a key never releases a held movement binding (2196)
- food, drink, bandages and mounts grey out on the action bar in combat: an ITEM slot runs its on-use spell's usable walk, and its cooldown gate is the item-keyed on-hold record (2200, 2201)
- re-clicking the character you already have selected keeps its facing instead of snapping it square (2194)
- the shift-compare plate seats beside the tooltip and is owned by the engine, so a stale plate never rides over the next hover (2202)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- Questie's arrow points at the turn-in, not 100 yards off: the map selection sits on your zone from the moment you enter the world, as the reference's does, so Astrolabe reads the right sheet (2192)
- an area the map has no row for sends the selection to your continent, not the world view; the reference fails its zone lookup and its continent lookup separately
- a tooltip shown with no lines hides itself instead of drawing an empty plate at the last hover's size, as Questie's tracker did on some quests; each fresh hover restores full alpha, so a plate an addon dimmed does not stay dim
- Questie's minimap button turns the quest icons back on: a nil handed to a colour setter stores zero and carries on, as the reference does, where benilla raised and jammed Questie's event queue from login (1973)
- MonkeyDev loads without three red dialogs at login: an unknown frame type in an addon's XML is logged and skipped, as the reference does; only a script creating a frame of a bad type still raises (2191)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- addons load in the order NTFS lists a folder, case-insensitive with an underscore before the letters, the order the vanilla corpus was written against and the one that decides which addon provides a shared library (2178)
- GetNumAddOns and the index form of every AddOn verb walk a title-sorted list that hides the Blizzard addons the server says to hide, and the character-select AddOns list is in title order (2175)
- SetPoint and SetAllPoints raise the reference's own errors for a frame name that does not exist or an omitted relative frame instead of quietly anchoring to the parent, and an explicit nil means the screen root (2176)
- a tooltip owned with ANCHOR_CURSOR follows the mouse, which pfUI's tooltip, xpbar and chat modules, pfQuest's browser and TipBuddy ask for; an omitted or unknown anchor is ANCHOR_LEFT (2176)
- a $parent name handed to CreateFrame, CreateTexture or CreateFontString expands as it does in XML, so FonzAppraiser's thirty dropdowns and pfQuest's search icon get the names they asked for (2176)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- the last app-side lines become GlobalStrings keys — tooltip cells, party, petition, duel, quest, PvP and rest lines read the install's own wording, so a non-enUS chain shows its own text (2045, 2080)
- a locked chest says Requires Pick Lock, ammo and thrown weapons drop their invented slot word, a met set bonus reads Set: without the count, and a 100 s cooldown reads 1.67 min as the template prints it
- the level-up block and the free-professions line printed twice in chat and print once; a declined invite, a friend coming online, a duel request and a failed quest play the sound 1.12 gives them
- Duel cancelled. is the yellow line at the top of the screen, and a trainer refusal or a failed quest with no cached template shows nothing, as on 1.12
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- 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>
- the bank window is the reference's own `BankFrame.xml` run off the player's chain — our 890-line copy is deleted whole; putting bags into the bank's bag slots works (the two verbs its buttons are wired to now exist), the purse is a real `SmallMoneyFrameTemplate` splitting 987654321 copper into three coin buttons instead of one string, `BANKFRAME_OPENED` loses the argument the real client never fires, and the bank stops auto-opening your bags while the vendor and mailbox keep doing it; the empty-slot click still plays the bag sound, because the reference's own line plays it on every click the cursor did not consume — named as faithful rather than quietly fixed
- underneath it, the bank's inventory band is a VIEW over the container feed, not a second copy — inventory ids 40..69 had answered nil because the doll array stops at 24, so the stock file swapped in would have drawn an empty bank; the six bank bag slots become real inventory slots (64..=69) fed off the player descriptor, the three bag verbs are carved from the binary (`PutItemInBackpack` IS `PutItemInBag(0xFF)` through a ten-byte thunk, the occupied-by-an-ordinary-item leg auto-stores with no destination slot at all, and the return is the Lua number 1 or nil — the nil being the whole reason a plain left-click opens a bag), and the swap caught `BankButtonIDToInvSlotID`'s bag arm six slots wrong: it took a bag number 1..6 where the reference passes the button's own id, which is the container id
- far sight freezes your own body while the view is out — the fifth conjunct of the movement predicates' shared precondition, joining the health term so the walk, the turn and the follow teardown go down together; both halves are load-bearing: the `IsActivePlayer` half is what keeps a Mind Control victim walking (possession rides the same PLAYER_FARSIGHT field), and the latch half is the RESOLVED engage rather than the raw field — and the camera is untouched, you still look around freely, which is the whole point of the spell
- the reset offer no longer outlives the character that earned it — log out after walking out of the Deadmines and your next character was offered a reset the first one had earned — and the one lockout leg no probe can ever reach (vmangos writes a last-instance map only from raid binds, which the client's handler discards as not-a-party-dungeon) is now covered end to end through the real system, its module doc saying plainly: tested and never observed
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the bag windows are the reference's own `ContainerFrame.xml` running off the player's patch chain — decision 1751 makes the stock 1.12 UI off the install the end state and `assets/ui` scaffolding that retires file-by-file (86 files against the 99-file stock corpus, 72% of the 1128-function global surface already built) — via one ordered manifest, `benilla.toc`, where a bare name is a file we ship and an `Interface\FrameXML\` path is the player's own file; the windows on screen are `ContainerFrame1..12`, the bank's six copy-pasted popouts fold into them at ids 5..10, net −750 lines, and the swap forced real engine verbs into being rather than stubs (`SetBagPortaitTexture` with its id-0 no-texture arm, `GameTooltip:AppendText`, `SetButtonPulse`, and a `<Model>` whose file is the cooldown indicator materializing as the native sweep, which three corpus addons inherit)
- the 1.12 walk toggle lands: `TOGGLERUN` latches one bit with `ToggleRun`'s own silent refusal chain (dead, rooted, on a live server spline), and the walk animation was never missing at all — the flag reaches it through exactly one channel, the speed (`0x5fd100` picks Walk over Run on `speed > 2 × walkSpeed`); pinning it caught our remote extrapolator with the two min rungs backwards — `min(walk, run)` runs BEFORE the backward min, so a walking backpedaller moves at 2.5 yd/s where we would have played them the Run clip at 4.5 — and one shared `current_speed` now keeps the local controller and the extrapolator from ever disagreeing about it
- the instance and raid lockout messages exist — six packets that were dropped as unknown opcodes now land, four of them as the CHAT_MSG_SYSTEM lines the client composes itself, which is why no Lua handler ever surfaced them — plus `IsInInstance`, `ResetInstances`, and the unit popup's "Reset all instances" row behind the reference's own predicate; where the binary contradicts vmangos: the ownership flag means "any instance bind" but vmangos only ever names raid binds, so the row was dead for every character without a raid lockout — term 1 is now also satisfied by our own eyes, walking out of a party dungeon the server itself bound us to
- opening the world map showed 3D world at the sheet's margins: `BlackoutWorld` was parented to the very `UIParent` a fullscreen panel hides, and three more frames had crossed the same boundary — the shared dropdown lists (so the map's continent and zone pickers answered a click with nothing at all), the master-looter menu, and the screenshot confirmation, which could never appear during a cinematic; `frame_flag_gate` now checks the parent seat against the reference both ways a frame acquires one (the attribute, template-inherited included, and `<Frames>` nesting), and its first harvest took 30 divergences down to three accepted ones
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the logon address is the reference's own `realmList` CVar — name, `host[:port]` form and the "Address of realm list server" prompt read straight out of `WoW.exe`'s string table — persisted in `benilla-config/config.toml` and edited on the login screen: a Realmlist button in the slot `AccountLogin.xml` already reserves under the Quit stack, the current address in grey beneath it, opening the reference's own `hasEditBox` glue dialog; before this the address was `$WOW_HOST` read once at process start with no way to change it; one text field and no saved list (every client at this scale ships one field), a pasted `SET realmlist "…"` line is accepted, and the default is `localhost` because the retail host has not resolved since 2019
- the README's "Not built yet" becomes "Where it's going": what done means, no dates, and one list in probable order — the remaining basics (battlegrounds, guild charters, the stable master, meeting stones, GM tickets, master loot, cinematics), addons / options / performance as ongoing, the no-brainer VanillaFixes- and SuperWoW-class fixes, playable downloads with Linux first; not planned: other expansions, Warden
- a UI region's span comes from its content and the owner-edge fallback is deleted: the reference's resolver core has no parent pointer at all, a FontString floors at one unit on both axes, a texture takes its texel span, and a region resolves from its own anchors and span or not at all — a live census over 8772 rects: 0 stopped resolving, 923 FontStrings shrank (842 to exactly 1×1 — empty labels that had been the whole screen, or an inverted rect), 0 textures moved; and `Region:GetWidth`/`GetHeight` are the same virtual getters the resolver calls (authored value wins per axis, natural width, wrapped height, texel span for art), so a script reads the number its rect was built from
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- a tooltip change costs ~1.3 ms — five whole-interface recomputes, each working as designed: the quad re-convert at 474 µs, the Lua tick at 287, the layout resolve at 200, bevy's `Assets<Mesh>` tax at ~270 and the mesh rebuild at 146 — pinned with new hover probes (a jitter dial, a `WOW_PROBE_HOVER_DUTY` duty cycle so run drift cancels, an `AssetChanged` tax meter), and the layout preamble's two whole-roster scans become lists maintained at the writer (52–83 µs → 1–11)
- the splice compared this frame's entries against last frame's by index, so one inserted highlight shifted the 160 entries behind it and read as "everything changed"; it now aligns on `z` — an identity, not a position — and a `Text` entry that writes only quads became spliceable, so a tooltip change re-converts 22 entries instead of 520 (`convert` +439 µs → +107), with `WOW_UI_SPLICE_VERIFY` running the full conversion behind every spliced frame: 0 mismatches over 1,078
- a spell-less macro is a usable action button, not the 0.4 grey — the reference's usable compute has a dedicated leg answering usable for the bound-spell zero, cross-checked at the bytes in the reference's disassembly — and the button's Name line is written unconditionally from `GetActionText` as the reference does, so the macro's name sits under its icon and a `/cast` of a spell the character does not know now greys, which is also 1.12's behaviour (B340)
- a capture bounds its run in wall-clock seconds, because a frame cap is no cap at the ~1 fps a starved `nextDrawable` leaves — 1800 frames is thirty minutes there — and the teardown backstop runs on every run as a liveness guarantee, carrying the exit code the run asked for; the Multisampling row reads Off / 2x / 4x, since "1x" contradicts the no-multisampling byte both of the reference's enumerators normalise to and the bit-depth prefix was identical on every row
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the glue scene widens only as far as its art goes: none of the seven login dioramas is drawn wider than about 3:2 — measured per texel by `benilla-extract glueextent`, which rasterises every batch into the camera's own tan-space, painting where the reference's alpha-key test passes — so the frame zooms down to the reference's own 16:9 opening (never tighter) and pillarboxes past the aspect the art cannot fill, instead of showing the clear colour past the backdrop's slanted edges; 4:3 stays byte-identical
- the character screen opens on whoever you last logged in as: the reference's `lastCharacterIndex` CVar, written at Enter World and only there — clicks and arrow keys never touch it, which is what makes the value mean "last logged in" rather than "last clicked" — restored on every character-list rebuild and clamped into range
- spam-hovering tooltips stops costing 2 ms: two write sites gave up naming their node and re-derived the whole layout graph (1,812 µs against the ledger's 44, on 919 of 1,262 resolves) — the wrap pin now touches the one region it wrote, and re-pointing a tooltip at the button under the cursor retargets the node's edges in place instead of throwing the cached graph away
- two corrections off the project's first GPU-bound report (a Steam Deck at 94% GPU busy on 8.4 W of a 15 W budget, the memory-stall signature): `farclip` ships the reference's registered 350 instead of the clamp's 777 — the value also sizes terrain residency, so 777 streamed, held and drew 4.8× the area — and BLP textures upload WoW's own DXT blocks verbatim the way the reference does, instead of decoding to RGBA8 at 8× the bytes of DXT1; the terrain slider still reaches 777, and the lanes that read texels on the CPU (terrain arrays, the character-skin compositor) stay decoded and say why
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- a text write names its own label, so the per-frame measure sweep drains a short ledger instead of hashing every font string twice — ~0.3 ms a frame at the Stormwind gates, gone
- the idle interface tick stops re-writing settled state: buff durations and the weapon-enchant frame write only on change, and the fade and unit-popup drivers park until woken
- the interface extract skips a hidden region before cloning its row, not after
- WOW_UPLOAD_BUDGET=<MB> brackets the arrival burst's upload pacing — a lever for measuring world entry, not a fix yet
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the interface's layout graph is maintained as anchors change instead of rediscovered from scratch: 1.48 ms per mutated frame becomes 34 µs, and end to end a running cast now costs +0.14 ms where it cost +4.33
- the autocast shine draws its authored population — 1200 particles through the star's own projection, spinning as the model says
- the shine is clipped to its own widget: the model frame installs its rect as the viewport, so nothing it draws leaves the button
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the social window's third tab is live — the roster in both its views, the member card with its promote/demote/remove verbs, the rank editor's thirteen permissions, the guild information board, and the invite, leave, kick and note dialogs behind them
- setting your hearthstone at an innkeeper now asks before it takes and says so afterwards, and the innkeeper's gossip lines get their own icons back
- unavailable talents grey out, and every greyed icon in the client now goes through the same luminance fold the original used
- the buff tooltip's remaining-time line reads gold with the client's own hour/minute ladder, the macro editor takes a shift-click to paste a spell, trainer rows repaint after a purchase, and a long book page stops flickering its ellipsis
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
- the one shared region table splits into the client's own maps: a Texture no longer answers SetText, a FontString no longer answers SetTexture, and feature-detecting addons see the true surface
- CreateTitleRegion lands whole — object and drag — so an options window that is its own drag handle actually moves
- font-object names resolve whatever their case, and the CursorUpdate/CursorOnUpdate pair arrives for the container-item templates
- GetStringHeight is pruned (1.12 has no such method) and SetPortraitToTexture moves to the global it really is
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the justify verbs had been written three times and drifted three ways; one implementation now serves the region, font-object and message-frame tables, and a cleared axis paints centre, as the client's draw path really does
- MessageFrame and ScrollingMessageFrame gain the ten-verb font block they were missing entirely, so font calls on message frames stop raising
- EnableKeyboard and IsKeyboardEnabled land as real Frame methods — eight addons were dying on their absence
- the harness's METHODS ranking states its own limits instead of reading as a build queue
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- benilla loads real 1.12 addons from Interface/AddOns — TOC, XML and Lua through one loader, with per-addon saved variables, Bindings.xml and the lifecycle events addons expect
- both AddOns screens are in: the character-select list with per-addon enable state, and the ESC menu's in-game list
- the API surface is measured, not guessed — a corpus harness runs hundreds of real 1.12 addons and ranks exactly which globals, templates and verbs they demand
- the Lua dialect is the client's own, read off the binary — 5.0 semantics with the 1.12 additions, and UI source files are bytes, not UTF-8
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- the whole loop plays: the line arcs rod tip to bobber, the bite splashes with its own sound, and clicking the bobber reels in
- the verdicts read like 1.12: yellow "your fish got away" lines, the reel-in sound, the fishing loot window
- "You feel rested." prints on the transition, rest states come from the client's own Exhaustion table, and level 60 trades the XP strip for the max-level rail
- the Skills pane lists what the real client lists: the hidden rows, the sort, gray numberless single-rank lines
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- The keyring is built. Your first key makes the keyring button appear by
the bags, opening the window the original keeps its dungeon keys in — the
server had been sending the keys all along; nothing had ever shown them.
- Keys open their doors now. The client used to cast the open as a bare
spell, which the server rightly refuses; it now sends the item-use the
original sends, carrying which key from which slot. The tooltip plays
along: "Locked" turns the soft green when something you carry opens it,
and a key stops inventing a "Use: Opening" line it never had.
- A building can own the sky: a WMO can name its own painted skybox, and
Stratholme's burning city now draws its near-black-to-red dome instead of
the zone's khaki gradient. The exact rule for when it applies settles in
the next snapshot.
- Signpost tooltips follow the cursor again — the tooltip anchor forks per
object class, corner for doors, cursor for signs — and the Blasted Lands
lightning stops striking the same spot for ever: a placed doodad re-rolls
its animation variation every play-window, like the original.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- The UI stopped paying to discover that nothing changed. Layout used to hash
~1,900 frames' worth of inputs every frame just to decide "quiet"; it now
tracks a mutation epoch and answers with one integer compare, and a settled
frame reuses last frame's rendered quads instead of re-rasterizing every
glyph. ~0.9 ms off every idle frame — with a verify mode, forced on across
all 489 UI tests, that re-runs every skipped resolve in full and asserts it
would have produced identical results.
- Across this run of snapshots, the one number everything is graded by: CPU
work per frame at the crowded dungeon spot went from ~29 ms to ~16.5 ms.
- A lighting question got the full treatment and an honest answer. Units
briefly took a different day/night lighting law; it read wrong on screen the
same evening, and a build-by-build hunt a full week back confirmed the
world's look never regressed in the first place — the change is reverted,
nothing about the look changes in this snapshot, and the question stays
parked until the one remaining unknown in the reference client is pinned.
- The character pane's model booth grew a headless eye — a dump of exactly
what it draws, no live session needed — built for that hunt, kept for the
next one.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Particles, ribbon trails, blob shadows, the selection ring, ground decals,
water foam and the weather all push their quads into one shared vertex
stream now — one buffer, batched draws, indices built per frame in draw
order. ~1,100 emitters at the crowded dungeon spot stopped being ~1,100 mesh
assets the engine reallocated every frame, no material asset is ever mutated
for an effect again, and adjacent draws merge like the original's own batches
(its measured shape: 793 quads in 23 draws). Vertices went camera-relative on
the way — the fix for streaks and flakes vanishing in far zones, where the
raw coordinates get big.
- Hovering an item in your bags cost +14 CPU ms for as long as the tooltip sat
open. The tooltip's refresh loop is faithful 1.12 behaviour (the bag button
re-runs its hover every frame, unthrottled, so a late-arriving item fills in
without a re-hover) — but each rebuild dropped the text measuring cache, so
every frame re-shaped every line. One line keeps the cache: +14 → +0.1 ms,
and the paperdoll, inspect and merchant tooltips stop paying the same tax as
spikes.
- "Which room is this creature in" walks one anchor per model instead of every
lit part — ~13,700 part visits a frame became ~1,700 anchor rows. The
frame-time win was honestly flat (the cost had already moved elsewhere); it
lands for the structure, and the old sweep's repair duties became
event-driven with it.
- The perf probe defends its own window now — always-on-top from the first
tick, occlusion counted on the measurement line — after two long overnight
runs were silently destroyed by the display going to sleep.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
