b7a8e1c406ff2940a8d693898c1bcb6a415ffa2f
1670
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
  gromandClaude Opus 5
|
b7a8e1c406 |
ui: a Slider answers SetEnabled, because OpenMap's options tab called it unguarded
Rendering a panel is not the same as it working. OpenMap's options menu opens 500x500 with 104 frames and 49 placed labels, and its Look tab died on click: Map/Panel.lua:1084, attempt to call method SetEnabled (a nil value) -- so a third of its options was unreachable. SetEnabled is already composed here for Button, CheckButton, EditBox and LootButton, and was deliberately withheld from a Slider, because addons duck-type on the enabled family to tell a control from a bar. What they duck-type on is the TRIO (if widget.IsEnabled then), that is what tests/slider.rs pins, and the trio is still absent; SetEnabled is beyond 1.12 by construction, so a Slider answering it invents no 1.12 behaviour. OpenMapClassic calls it unguarded on every options row with a condition, which is what settles it. On a Slider it writes the mouse -- input refused, art not greyed, which 1.12 has no slot for; the caller dims it itself on the next line. The argument goes through GetBoolOrDefault, not Lua truthiness, because DoiteAuras writes SetEnabled(ok and 1 or 0) and 0 is truthy in Lua. Also records a count that was wrong until checked: grep says 39 calls across five addons, but AdvancedTradeSkillWindow2 and SoftSpoken define their own SetEnabled and SoftSpoken's slider row is fully guarded. The real unguarded widget calls are OpenMap's three and DoiteAuras' two -- and SoftSpoken's blank settings page is NOT this, and is still open. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
2281179874 |
capture: a setup slot, for the moment before the scene ages
$WOW_CAPTURE_SETUP runs a chunk at the first aging frame -- early enough to change a setting or create a frame and still have the scene settle around it. $WOW_CAPTURE_EVAL keeps the shutter, where a read and the pixels it explains come from the same frame; anything created there is a frame too late to be in the image. The deep-zoom map shot needed it. OpenMapClassic resolves minimap tiles by FileDataID by default and this client has none, so its own status reports "loaded YES=0 NO=8"; it already carries a path mode for exactly that, asking for textures\Minimap\<folder>\map<col>_<row>, of which this install ships 10,884. Switching to it raised, because S.ResetMinimapTiles reads lodPieces/lodPool sixteen lines before they are declared and so bound them to nil globals -- an upstream bug, and the only two callers are the two switches that reach the mode. Port edits mirrored to Addon_Ports/openmap_port/: the declaration moved above its use, and the tile mode defaulted to the one that can work here. Captured at zoom 40: the minimap's terrain draws through the map. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
62c019e744 |
blp: a short colour map is a sliver tile, not a corrupt file
Seven of the client's own Interface\WorldMap textures never decoded. They are the last column or row of an exploration overlay -- a few pixels wide, a handful of colours -- so they store a SHORT BGRA palette and end: RuinsOfThaurissan4.blp is 404 bytes in all, LakeEverstill6.blp is 276. The parser sliced the nominal 256 entries and refused anything shorter. A refused decode is a quad that draws nothing, and until this week a texture that drew nothing also said nothing, so the symptom was a hole in a zone's revealed art -- indistinguishable from not having explored it. It surfaced only once the stitched world's geometry was right and the render reached those tiles. The palette is now as much of it as the file has, and an index past the end is transparent black, which is what an unused index means. every_world_map_texture_in_the_chain_decodes walks the folder: 3553 textures, 7 failures before, 0 after, and it fails on the old decoder. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
6c01be926d |
map: a map's rect on another map describes art, not a click box
C_Map.GetMapRectOnMap answered, for a continent on the world sheet, the world-level CLICK box -- continent_sheet_rect over the WorldMapContinent tile bounds, which trace the landmass. That is the reference's own 0x4a5d00 and right for what 1.12 uses it for; the two boxes are disjoint so a world-level click can tell the continents apart. It is the wrong answer for art. A continent's map is a 1002x668 sheet with the land inside it and ocean around, so its art rect is 792x528 px on the world where its click box is 300x564. Every 1.12 WorldMapArea rect is 3:2, because every sheet is, and the click box is 0.355 of that. A zone's rect on its continent was already a yard un-lerp and correct, so the error entered once and multiplied through every zone: Dun Morogh landed 42x79 where its art is 1002x668. A stitched world built on that draws each zone a third as wide as it should be, which is why the canvas was grey between the zoom steps. map_proj::continent_art_rect carries the continent's own WorldMapArea rect through world_uv, the projection a position on it already takes. The view keeps both rects; the Era map graph publishes the art one. Measured after: EK, Kalimdor, Dun Morogh and Elwynn all 1.500. Rendered: the zone view stitches its neighbours in, zoom 3 is the continent with its zones, zoom 1 is the world sheet. The stock 1.12 map is unchanged. Three instruments came with it and are kept: $WOW_CAPTURE_EVAL, a chunk run in a capture's VM at the shutter; $WOW_CAPTURE_EXPLORED, so a stitched map's fog can be ruled out in one run; and a named warning for a UI texture that is not in the chain, which is what refuted the other hypothesis -- a grep said the addon's art was missing, a real render asked for zero missing files. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
3e8c995e27 |
map: clamp the canvas onto whatever zoom ladder is in force
CreateZoomLevels is the extension point a canvas addon replaces wholesale, and OnLoad seeds a scale (1) before any addon has run. OpenMapClassic's ladder is 27 steps from 0.0125 to 0.899 -- its whole range below this file's own 1, because its stitched world is reached by zooming out past a single zone -- so the canvas sat above its own ceiling, and ZoomIn, which walks the ladder by value, found no step at all above it. ClampZoomToLevels is called from GetCanvasScale rather than from CreateZoomLevels, because CreateZoomLevels is exactly the function the addon will not have. Also records where OpenMapClassic stands, measured: 808 textures in the canvas subtree, 166 placed, spanning ~4600x3270 units of a 1002x668 canvas, with a ladder floor implying a world ~80000 units wide. The Era map tree under it answers correctly; GetMapArtLayers answering one constant for every map is the thread to pull next. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
07dd5ba7e1 |
capture: a shot taken of a player's config must not write it back
`$WOW_CAPTURE_SAVED` points the saved-variable READS at a real player's configuration so a capture reproduces their interface. The writes went to the same place, so taking the shot overwrote the configuration it was taken to study -- with a fixture character's state, since a capture logs in as nobody. Three of grom's files came back changed from one run, his pfUI chat scrollback among them. The shutdown tail returns before all four writes under $WOW_CAPTURE. Verified by md5 across a run: zero files touched, zero "wrote" lines. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
abb4dc1dd3 |
ui: a Slider's thumb had no rect, so no scrollbar could be skinned or seen
"Half of the info on the right is out of the frame" was a scroll frame with 699 of content in a 451 window, a scroll range of 228, and a scrollbar with nothing on it. The thumb's rect was only ever computed inside the draw pass and never published: the region sweep skips anchor-less regions and a thumb is anchor-less by construction, because its slider places it. That is invisible until something anchors to the thumb, and pfUI does -- SkinScrollbar clears the stock knob and hangs its own off the thumb's two corners, so every pfUI-skinned scrollbar in the interface was a bare track, which is a scroll frame that silently cannot be scrolled. The thumb is on the layout roster now, survives the unanchored prune, and carries its slider's (vertical, fraction) in its node hash -- without the last one a named touch moved no hash and seeded nothing, which WOW_LAYOUT_VERIFY caught as an incremental pass disagreeing with a full derivation. The draw pass keeps computing its own rect: extract takes &self and cannot resolve, so a SetValue with no resolve behind it still has to draw where the value now is. map: the window has two layouts and the canvas only knew the maximized one Pressing maximize left the art a 1002x668 patch in a full-screen window and minimize could not put it back, because the container was anchored once at build to WorldMapPositioningGuide TOP -502,-69 -- 1.12's maximized anchor against a fixed 1024x768 guide. Read as insets instead (15/4/33/20 windowed, 10/12/69/31 maximized) the same numbers reproduce 1.12 at 1.12's sizes and fit any other window, which is also the left-edge bleed: -502 against a guide stretched to pfUI's 1017-wide window starts the art 6.5 px in, not 10. The zoom ladder's floor follows: it was the constant 1, true only while the container was the art's own size, and is the measured fit now. A view outside a rebuilt ladder is pulled onto its nearest end, since ZoomIn and ZoomOut walk it by value -- OpenMapClassic's whole ladder sits below ours. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
41512a04ef |
map: 1002x668 is a contract, and the canvas counted the windowed scale twice
Every 1.12 map addon places its art in absolute pixels against WorldMapDetailFrame's 1002x668 — pfUI's reveal does it one SetPoint per tile — so that size is a contract, not a size. 1.12's windowed mode puts WORLDMAP_WINDOWED_SCALE (0.7) on the frame and on WorldMapButton; under the canvas the container already carries the window, so that scale was a second authority on the same rect and, with the frame SetAllPoints to the canvas, came back as unit size: 1002 / 0.7 = 1431. Everything absolute inside it landed at 0.700 of where it belonged, which is grom's screenshot drawing Coldridge Valley twice, and the mouse surface kept its XML 1002x668 beside 1431 units of art, so clicks read 70% of the map. SeatOnCanvas makes the canvas the only authority: both frames to scale 1, the frame filling the canvas, the button given the art's rect — explicitly sized, because pfUI reads WorldMapButton:GetSize() to size the whole map window. Re-run after Minimize/Maximize and on every canvas resize. Rescaling the children this file knows the names of could never have fixed it; the children it does not know the names of are the point. saved-variables.lua now goes through saved_home, so a capture with $WOW_CAPTURE_SAVED opens the player's windowed map — the only mode the bug exists in, and the reason neither harness had ever seen it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
d11b9ac3ae |
map: the art was scaled by the canvas while living in the detail frame
Grom's report was "textures are not where they should be", and still wrong with OpenMapClassic switched off — which is what made it ours rather than the addon's. His own session log had both states 1.4 s apart, same map, same frame: detail 1431x954 with tiles 366x366 (correct, 1431/1002 = 1.428), then the same frame with tiles back at the XML's raw 256x256. `ScaleDetailArt` was being handed the CANVAS's size by `OnCanvasSizeChanged`, on the assumption that the detail frame is `SetAllPoints` to the canvas. 1.12's own windowed mode breaks that: `WorldMapFrame_Minimize` puts `WORLDMAP_WINDOWED_SCALE` (0.7) on `WorldMapDetailFrame` and `WorldMapButton` and insets the frame to TOPLEFT 15,-33, so the same screen area is a different number of units in each — his trace reads `detail 1431x954 scale 0.56` inside `WorldMapCanvas 1002x668 scale 0.80`. Scaled by the canvas's number every tile came out 1.43x too small: the art covered 1024x768 of a 1431x954 frame, the rest bare parchment, and the exploration overlays scaled to match landed nowhere near their terrain. The detail frame's own size is the authority, since the tiles are its children. `GetLeft()` first to settle the layout, because that size is derived and an unforced read answers the size from before the last change; the caller's w,h stay as the fallback for what cannot be measured. Reproduced and controlled in the harness with the windowed scale applied as the stock code applies it: tile 362 where 362 is wanted, against the control's 256 — the number his log recorded. Two earlier attempts are written up in docs/ADDON_COMPAT.md as dead ends, because both looked right: caching the canvas size prefers the wrong quantity, and converting it by the effective-scale ratio gives 366 against a true 362, since Minimize also insets the frame. `$WOW_CAPTURE_SAVED` now reaches the enable-state file too, so a capture reproduces which addons the player has turned off — without which "I turned it off and it is still wrong" cannot be answered offline at all. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
5e2f1f3ac8 |
ui: SetClipsChildren, the verb an addon asked for by name and was refused in silence
Era's map canvas is built on one verb this client did not have. Its `ScrollContainer` is a plain Frame with `SetClipsChildren(true)`, and everything below it is cropped to that rect. `assets/ui/MapCanvas.lua` substituted a real ScrollFrame, because this client crops a ScrollFrame's scroll child and every descendant of it — which covers the canvas and nothing parented to the container beside it. On Era that distinction does not exist, so an addon has no reason to respect it. OpenMapClassic parents its pin layer exactly there and asks for the verb by name, behind an `if` (`Map/Stitch.lua:1070`), so the absence was not a raise. Measured on grom's own config with the map open: the container sits at L=1151 R=2153 and its seventeen direct children span L=-765 R=2153 — about 1900 units off the left edge, cropped by nothing. Implemented as a frame flag with `SetClipsChildren`/`DoesClipChildren` and the `clipChildren` XML attribute; `effective_clip` now intersects the rect of every ancestor that sets it alongside the ScrollFrame rule, so the two nest and compose. An ancestor rather than the frame itself: the verb clips a frame's children, and stock art overhangs its own frame on purpose. `MapCanvas.lua` sets it on the container too, so a layer parked there is cropped whether or not its addon thought to ask. Not claimed as the whole of grom's map report: the screenshot behind it has not been reproduced offline, because a capture's view is not the view his session was in. This is the part that was provably drawing uncropped. Also here, both from the same play-test: - `C_Spell.CancelSpellByID` answered with `SpellStopCasting` only. pfUI's buff bar cancels with it on a right click where 1.12's calls `CancelPlayerBuff`, so with nothing in flight the click did nothing and said nothing — grom had never been able to right-click a buff off on this client. It now finds the id among the player's own auras and queues it on the same CMSG_CANCEL_AURA path, behind the same `cancel_authorized` gate, and falls through to the cast only when there is no such aura. - `$WOW_CAPTURE_SAVED=<Realm>-<Character>` points a capture at a player's own saved variables. A capture is hermetic by design, but with addons and no saved variables it renders an interface nobody is configured to be looking at — which is why the world map came out correct in a capture while being wrong on his screen. It doubles as the identity because a capture never logs in and `pfUI_config` is SavedVariablesPerCharacter. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
afc5379911 |
auras: the player's first buff of a session was swallowed as a first-time diff
DoiteAuras reads the whole aura block once at PLAYER_ENTERING_WORLD and from then on keeps its
table purely from nampower's four self-aura events (`Modules/DoitePlayerAuras.lua:403-490`;
`UpdateAuras` is called nowhere else). The arrival loop here seeds a unit's memory and fires
nothing, which is right for a unit streamed in mid-aura — but it only ever seeded the harmful
half, while the comment above it claimed it seeded the player's helpful half too ("without this
its first real edge would be swallowed as a first-time diff, which is one lost BUFF_ADDED_SELF
per login"). The guid was destructured away as `_guid` and `own_buffs_seeded` was set nowhere but
in the edge loop.
So the player's first buff edge of every session reported nothing. That does not heal: a later
cast of a buff that is still running refreshes its slot with the same spell id, which is not an
edge this watch sees at all, so nothing reports the aura again until it drops completely. A
warrior's Battle Shout, re-cast while it still has a minute to run, never drops — grom's rule for
it read "missing" with the buff plainly up, all session.
Every other link was measured clean first: `GetUnitField("player", "aura")` answers a table, the
arg vector matches what the handler reads (arg3 spell id, arg4 stacks, arg6 raw slot, arg7 state),
and all seven ranks of Battle Shout resolve through both `GetSpellRecField(id, "name")` and
`GetSpellNameAndRankForId`.
The pin asserts both halves — the first buff is an add, and a refresh is not an edge — and its
control is worth noting: with the seed removed the first buff reports `[]`, not a wrong value,
which is why no session log ever named this.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
5ed9dded4c |
map: the Continent and Zone pickers were blank on the first map of a session
The realm's `WorldMapFrame.xml` fills them from exactly one place: the Zone dropdown's `<OnEvent>` on `WORLD_MAP_UPDATE` (:321), which re-runs both pickers' initialisers and then the two `WorldMap_Update*DropDownText` functions. That dropdown registers the event in its own `<OnShow>` (:318 — the realm commented the `OnLoad` out and moved it there), while `WorldMapFrame`'s own `<OnShow>` fires it through `SetMapToCurrentZone()` (:711), and a parent's OnShow runs before its children's. So the inline event landed in a gap where its only listener did not exist yet. Measured through the addon harness: the Zone dropdown answers `IsEventRegistered` nil before the open and 1 immediately after, and the pickers come up blank on the first map and correct on the second — which is what grom's recording shows, blank over Dun Morogh and filled over Elwynn later in the same clip. `SetMapToCurrentZone` now queues a second `WORLD_MAP_UPDATE` on the deferred lane as well as firing the inline one. The inline fire stays: it is what repaints the detail tiles inside the OnShow rather than a frame later (the world-map flash). Only this verb doubles up, and only this verb needs to — `SetMapZoom` is a click on a map that is already open. The first attempt at this was reverted and the reason is in the test's own comment: its pin opened the real map on the shipped interface and passed with *and* without the fix, because the stock ordering happens to be fine and the symptom needs the addons. The pin that replaced it asserts the engine half directly — a handler that registers after `SetMapToCurrentZone` returns still receives the event — and answers 0 without the change and 1 with it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
ae929e8a81 |
map: the art trace reads the tiles the repaint put up, not the ones it replaced
The line fired in the frame the selection changed, and the repaint that fills the tiles is Lua's own off a queued `WORLD_MAP_UPDATE` — so it read the map that had just left. That is how "the map shows Dun Morogh and the tiles are Azeroth's" got into a log with nothing wrong behind it. It now waits three frames, restarting the wait if the selection moves again. docs/ADDON_COMPAT.md records this round: the cursor position no verb ever wrote, the per-kind nameplate CVars a nil answer greys out, the capture lever above with its two limits measured rather than assumed, and one open item — the map's Continent and Zone pickers, blank on the first open with the addons loaded and fine without them, with the candidate engine fix rejected because the test for it passed either way. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
ba11fc3315 |
capture: WOW_CAPTURE_ADDONS=1, to put an addon's own drawing in front of the renderer
A map report could be answered from neither harness: `addon_harness` runs the VM with no renderer, so it can say what a region is and never what it looks like, and the capture harness renders for real but is hermetic — `local_state::home` answers None under $WOW_CAPTURE, so no config, no saved variables and no addons, which is what makes a baseline comparable between machines. An addon that draws its own art was therefore unseeable without a server, and the only evidence was a player's screen recording. `WOW_CAPTURE_ADDONS=1` opts the player's addon folder back in for one run — `ui_script::addons::root`, the single narrow exception, with saved variables still out, so an addon runs on its shipped defaults. The world-map fixture will also ask a stitched-map addon for a view through its published API, retrying until the map is stitched (`WOW_CAPTURE_MAP_ZOOM`, `WOW_CAPTURE_MAP_AT`). `scripts/visual.sh` never sets any of it, so no baseline sees an addon. The first shot taken this way answered the open question: with OpenMapClassic loaded this client draws the stitched Dun Morogh map correctly, art, neighbours, labels, lore panel and player dot — so the map report is not the art pipeline. Also corrects a stale comment: the shot has never been Elwynn; continent 2, zone 10 is Dun Morogh. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
d39d3ec5cf |
cvars: the later clients' per-kind nameplate switches, so an addon's rows are editable
Cinematic's Nameplates page asks `GetCVar(name) ~= nil` before it will offer a row (`Effects.lua`'s `HasPlateSettings`). A nil answer is not "off" to an addon written for a later client — it is "this client has no such setting" — and the whole row goes grey, which is why grom could not touch Friendly NPCs. The two rows that worked were the two that happen to name `nameplateShowEnemies`/`nameplateShowFriends`, the only nameplate CVars this client had, because 1.12 has none at all. Seventeen rows registered. Sixteen are live: friendly players, friendly NPCs, hostile players, and each side's pets/guardians/totems/minions — eight names on the friendly side and four on the hostile one, each side driving one field, because a 1.12 object update carries an owner and not a kind of owner and an addon writes the whole set together. `nameplateMaxDistance` joins `NameplateRange` and `NP_NameplateDistance` as a third name for one reach. All default "1", the least-restrictive value, for the reason `NameplateRange` gives: the two masters already decide whether plates show at all, and a refinement shipped off would take away plates a player has today. Paired against `VPlateMode::default()` in the defaults test, the pairing the derived-Default trap forced on `occlusion`. Five are registered and inert, each saying so on its own row and in the orphan test's list: `nameplateShowAll` (1.12 has no in-combat-only mode), `nameplateShowSelf` (no plate over the player), `nameplateShowOnlyNameForFriendlyPlayerUnits` and `nameplateForceShowUnitName` (no such plate mode here), and `nameplateShowEnemyMinus` (the plate path has no creature rank). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
36f504820d |
ui: GetCursorPosition answered 0,0, so nothing could drag or size a frame
`Model::cursor_pos` is read by the verb in two places and was written by nothing in the workspace: declared, initialised to the origin, never assigned. So `GetCursorPosition()` answered `0, 0` for every call of every session. That kills the universal drag-and-size idiom — stamp the cursor on `OnMouseDown`, subtract it each `OnUpdate`, size by the difference. Every delta came out zero, the frame was re-anchored to its own current size sixty times a second while the player dragged, and the addon's own "that was a click, not a drag" guard then declined to save. Nothing raises and nothing logs. grom reported it twice as OpenMapClassic's map window refusing to resize; the round before this one painted its invisible grips and left this underneath. Fed in `ui_script/input.rs` beside `set_mouse_buttons`, for the reason that one states: the pointer feed below is skipped for a cursor off the window and for a UI hidden by TOGGLEUI, and the reference answers through both. UI units, y-up from the bottom-left, undivided — UIParent is scale 1 here, so an addon's `x / frame:GetEffectiveScale()` lands in the frame's own units as on the reference. The test pins the whole idiom rather than the getter, because the getter answering *something* was never the question. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
204184afb2 |
ui: the map's own art line, in every session's log
A play report about how the map looks arrives without the diagnostic that answers it, and the only way to ask for the line again is to wait for the player to play again. It cost a round trip on 2026-11-02, so the trace is unconditional now: one line per world-map selection change, with the frame the detail is drawn into added beside the detail frame's own rect, since a canvas zoom and an oversized tile change the picture the same way and only the two together say which one did it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
7f97777597 |
ui: the stock map art has to follow the detail frame it is drawn into
`WOW_MAP_TRACE=1` named it on grom's first run, in one line:
detail 1431x954 @scale 0.70 | 1 256x256 @(10,1122) | 2 256x256 @(266,1122) | ...
1.12's `WorldMapDetailFrame` is always 1002x668, so everything drawn into it is placed in absolute
pixels against that: twelve 256x256 tiles anchor-chained from its TOPLEFT, each exploration overlay
at the offset `WorldMapOverlay.dbc` gives it. The canvas breaks that the moment it sizes the frame
to anything else, and it does — `OnCanvasSizeChanged` falls back to the container's own rect
whenever it has no art layer to read, which is every moment the canvas has no `mapID`, including
the world-view blip the same log shows on every map open.
So: a 1024x768 sheet of art pinned to the top-left of a 1431x954 frame, covering 72% of the width
and 80% of the height, with the rest of the map drawing nothing. That is the grey he reported,
with the art magnified relative to what was left of the frame.
The art is now rescaled to whatever size the frame has. The tiles need only their size set, because
the XML chains them TOPRIGHT-to-TOPLEFT and BOTTOMLEFT-to-TOPLEFT. The overlays carry absolute
offsets, so they are captured from what the stock repaint last wrote — hooked on
`WorldMapFrame_Update`, the only moment those base values are readable — and scaled from that base,
so a second pass cannot compound. Measured: a 1431x954 frame now holds a 1462x1097 grid, the same
1024/1002 and 768/668 proportion the reference frames.
Two smaller faults fell out and are fixed with it: the container resizing was not treated as a
canvas size change at all (the canvas was re-sized only on a map change), and a derived resize
fires no `OnSizeChanged` here, so the rescale takes the size its caller just set rather than
reading the frame back, which would answer the old one.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
e62327424f |
app: WOW_MAP_TRACE, because no instrument here can see the map
A player's report about how the map *looks* is the one question this side cannot answer. The offline addon harness seats no renderer, the capture harness seats no UI or addons, and a scripted login would kick the player off their own account — so a full audit (art present and correct, catalog right, canvas geometry right, overlay tiling arithmetic right, addon setup complete) still could not say why grom's Dun Morogh map came up magnified, clipped and grey. Every number that decides the picture is readable from the VM. `WOW_MAP_TRACE=1` prints them once per map change: the detail frame's rect and effective scale, then per tile its rect, the texture path it holds, and the decoded size of that file. The file size is measured through a scratch texture with no size of its own, because the tile declares 256x256 in `WorldMapFrame.xml` and would answer that whatever the file holds — and this install's map art is a 512 upscale pack, which is exactly the disagreement worth seeing. Off unless the variable is set: it runs a Lua chunk per map change. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
c93a0be93f |
ui: a nameplate now carries the unit token its own event names
`plate.namePlateUnitToken` was set on nothing. It is how every later-era plate addon gets from a frame back to its unit — `plate.namePlateUnitToken or (plate.UnitFrame and plate.UnitFrame.unit)`, six sites in `Cinematic/Effects.lua` and `PlateParts.lua` alone — so an addon walking `C_NamePlate.GetNamePlates()` had a frame's rect and nothing about its unit, and could not tell a friendly plate from a hostile one. The measured symptom was grom's friendly nameplates switching themselves on the moment CineMode started, on a profile whose every friendly-plate box is unticked. That CVar write is the addon's and it is deliberate: outside instances Cinematic switches every plate kind on so visibility is decided by *fading* instead, because fades can change in a fight and plate settings cannot. Having switched them all on, it then faded none of them, because it could not identify one. The token already existed — `script/nameplate.rs` has had the slot-bound `nameplate<N>` tokens and the three plate events since the pool was built. All that was missing was publishing the slot's token on the frame: set before `NAME_PLATE_UNIT_ADDED` and cleared after `NAME_PLATE_UNIT_REMOVED`, so a handler for either event reads the field that event is about, and a pooled frame reused for the next unit cannot answer the departed one's token. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
6e8c4488e1 |
ui: the override-binding family exists and says what it does not do
`ClearOverrideBindings` was absent outright, so `Cinematic/Camera.lua:470` raised in grom's chat on every camera shot that ended — and a raise mid-handler costs the rest of the handler, not just the line. These are 2.0's secure-binding layer: a second, unsaved binding table a frame owns, outranking the saved one while that frame wants it and dropped whole by `ClearOverrideBindings(owner)`. This client models the saved table only (`script/keybind.rs`), so: - the five setters are no-ops that **warn once each** through the script warning channel. Silence was the alternative and it is the worse one — a setter that quietly binds nothing is the same class of fault as a hook that quietly arms nothing, and the caller cannot find out. - `ClearOverrideBindings` is a plain no-op and does not warn. Dropping a table that was never filled *is* nothing, and warning there would fire on every `ClearRecast()` while the player does nothing unusual. The module header lists what a real layer needs — an override table keyed by owner in `Model`, a generation bump for `bindings.rs`'s `sync_dispatch`, an override arm in `BindingDispatch::map` that wins over the saved chord, and a `Bound::Click` — together with the `SecureHandler*` half in `secure.rs`, which is the same table seen from inside a snippet and no-op for the same reason. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
bb2e40d1dd |
ui: GetDebugName, the name an unnamed frame still answers
`Cinematic/Frames.lua:3734` is `entry.name or entry.frame:GetDebugName()` — an addon that keeps a list of frames reaches for this precisely because some of them have no name, which is the one case `GetName` cannot serve. With the method absent, `/cine debug frame <name>` raised on the first unnamed row in the list instead of printing the row the player asked for. On all three method tables, since the era puts it on ScriptObject. A named frame answers its name, so a feature test on it learns nothing a 1.12 client would answer differently. The unnamed leg deviates in its fallback only: it walks the parent's own fields for the key the child is stored under (frames here are plain tables, so `parent.icon = tex` is visible) and answers `Parent.<unnamed>` when there is none, where the era answers an index among the parent's children — an index here would be the order of a list this client does not keep, and a wrong number reads worse than an honest absence. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
09aca72c46 |
ui: hooksecurefunc over a widget method has to read through the metatable
The table form read the field with `raw_get`, on the reasoning that a field
reached only through a metatable is not that table's to replace. A widget
carries none of its own methods — they live on the shared class table its
metatable's `__index` names (`object.rs`'s `frame_meta_for`) — so
`hooksecurefunc(frame, "SetAlpha", cb)` found nothing there and armed nothing,
silently. Measured: **0 fires across 2 `SetAlpha` calls**, while the plain-table
and global forms fired correctly.
That is not a corner of the form, it is what the form is for. 29 files in grom's
corpus hook a widget method, and the one that shows: `Cinematic/Frames.lua:459`
hooks each pfUI frame's own `SetAlpha` so that Cinematic can **follow** pfUI's
out-of-combat fade rather than fight it ("their frames have SetAlpha followed
straight away by ours, taking what it set as the new base"). With the hook
unarmed the two wrote a different alpha to the same frame on alternating frames
— which is grom's "still flashing my unitframe when cinematic is on".
The reference reads `table[name]` plainly and assigns the wrapper back, which
*shadows* the class method for that one object and leaves its siblings alone.
This now does the same: read through `__index`, write raw onto the host, which
is the same thing a plain assignment would do since frame metatables carry no
`__newindex`. A test pins both halves — the chain fires on the hooked frame, and
a sibling of the same kind is untouched, with nothing written to the class table.
Corpus roster byte-identical (87 ok / 1 fail) with all 29 files' hooks armed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
aa2ee515aa |
ui: MOD multiplies, and a gradient is not an alternative to a fill
Cinematic's colour tint painted the whole screen opaque white for as long as its
CineMode ran: `srgb(255,255,255)` across the world with the UI still drawn on top
of it, 16.8 s of it in grom's 2026-11-02 recording, starting the instant he picked
a profile. Three faults on the one path from the addon's call to the pixel, each
of them enough on its own.
`SetBlendMode("MOD")` parsed, answered `GetBlendMode` and then drew as straight
alpha. One pipeline served BLEND and ADD — the shader emitting `(rgb·a, a)` or
`(rgb·a, 0)` into a single `(One, OneMinusSrcAlpha)` state — because no scalar
source alpha can stand in for a per-channel destination factor, and the mode had
no caller worth a second pipeline. It had one: nine of the addon's ten tint
presets are MOD (`Tint.lua:21`), and a MOD colour is built to be multiplied —
`1 − (1 − c)·strength` at **alpha 1**, white at strength 0 — so drawn with alpha
instead it is an opaque near-white quad over UIParent. MOD now keys its own
pipeline, `(Dst, Zero)` with the destination's own alpha kept, off the same
`0x70c190` table the world's Mod batches already read; the shader returns the
multiplier un-premultiplied, since neither alpha appears in that formula, and runs
no longer merge across the flag.
The extract read `fill.or_else(|| gradient.map(midpoint))`, so a region that set
both drew the fill and dropped the gradient. They are not alternatives:
`<Gradient>` writes the four vertex colours (`+0xb8`, `0x77304d`), which
`loader/regions.rs` already reads as "unlike `<Color>` it tints the art". The tint
is exactly that shape — `SetColorTexture(1, 1, 1, 1)` once, re-gradiented a few
times a second — and so is its vignette, whose four edges were drawing as opaque
white bars.
And `SetGradient` took only 1.12's six loose numbers. This client answers
`CreateColor`, so an addon that feature-detects takes retail's colour-object form
and never tries the numbers; the two objects coerced to six zeros through
`as_f32`, the call **succeeded**, and the `pcall` that was the feature test
(`Tint.lua:546`) reported it understood. Both spellings now read a colour object
when the first stop is a table.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
33d0868009 |
ui: SetEnabled, the one setter retail put in place of Enable/Disable
A later-era addon writes `SetEnabled` **unguarded on its own controls**, even where it guards on someone else's. `Cinematic/Options.lua:4176` is a tick's own `Refresh`, greying it when a rule owns the row; `:305` twelve pages up is the same file writing `control.SetEnabled and …` about a control it did not make. The raise at 4176 unwound `RefreshOptions` → `SettingsChanged` → `SetProfile` → `Welcome.lua:43`, so picking a profile from the addon's welcome window applied the profile and then abandoned the panel refresh, leaving the traceback in chat and the window still reading "Nothing runs until you pick one." That is the chat error grom recorded on 2026-11-02. Composed from the stock `Enable`/`Disable` pair (`0x780160`, `0x7801b0`) on Button's table, which CheckButton and LootButton reach down the same chain the pair travels. EditBox takes it too, written out rather than composed because the reference gives a box no enabled state at all: it writes the mouse and keyboard the frame already owns and drops focus, so input is refused and the text is NOT greyed, which 1.12 has no slot for — `Options.lua:1629` is the call that needs it. A Slider still answers neither this nor the trio: addons duck-type on them to tell a control from a bar, which `tests/slider.rs` pins. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
70e4278c79 |
ui: the map's hovered-zone label is chrome, not map content
The canvas reparents 1.12's `WorldMapButton` into the zooming child so that every 1.12 addon's icons pan and scale with the map. `WorldMapFrameAreaFrame` hangs off that button in the client's own XML — it covers the art, so that is where the label that names what the cursor is over was hung — and it went along with it. Its label is a 102 px THICKOUTLINE FontString. Inside the canvas it scales with the zoom, so at zoom 2.5 every letter is a 261 px glyph, and one zone name's worth of those is most of the 2048² glyph sheet. Measured in grom's 2026-11-02 session: one `ui_text: a 261.10907 logical-px request … past the 256 ceiling`, then 20 `glyph sheet full … resetting at the frame boundary`, one per hover, each of them a frame in which every string on screen is missing — the map "flashing the interface". Frame cost with the map open went 11 ms → 16-20 ms over the same window. So it goes back on the window, anchored to the container's top, which is where the art's top edge is now; at rest that is the pixel it was already at. Measured through the harness: the biggest text under `WorldMapFrame` was 102 px at zoom 1 and 1632 px at zoom 16, and is now 102 px at every zoom. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
c6b19c7926 |
lua: a size this library derived is a cache, and a # append invalidates it
This is what froze his client. The bound named `Cinematic/Camera.lua:2597`:
while turnStarts[1] and now - turnStarts[1] > WINDOW do
table.remove(turnStarts, 1)
end
`remove` drains the table and remembers a size of 0 — faithfully, that is
`luaL_setn`. The addon then appends with `turnStarts[#turnStarts + 1] = now`,
which is a raw set this layer never sees, so slot 1 is full while the
remembered size is still 0. `remove` reads 0, takes its "nothing to remove"
path, changes nothing, and the caller spins forever. 15 s of Lua in every
frame, for the whole session, in a loop that in the reference cannot fail.
It cannot fail there because 5.0 has no `#` operator: every append goes through
`insert` or `getn`, so a remembered size is always the truth. The collision is
a later-era addon's `#` meeting 5.0's bookkeeping, and it also loses data the
quiet way round — `insert` after a `#` append writes *over* the appended slot
instead of after it, which is a test here now too.
So the two kinds of size are kept apart. One this library **derived** from an
`insert` or a `remove` is a cache: if `t[n + 1]` is not nil the table has grown
since, and the count wins. One an addon **declared** with `table.setn` is
honoured exactly as before, because `setn(t, 0)` as a logical wipe over full
slots is the entire point of it — AceLibrary and Mik's capture tables both turn
on that, and both still pass. An explicit numeric `t.n` field is a declaration
too, and is untouched.
Measured: the corpus roster is byte-identical (87 ok / 1 fail, the failure the
pre-existing AdvancedTradeSkillWindow2), and the drain loop that spun 44 times
in his log now ends in one removal a round.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
d1edd74cce |
ui: a runaway frame writes its own line, and stops costing the whole bound twice
Two of his sessions froze this morning with nothing in the log. The journal says what happened: every frame 15,0xx ms, `main_ms` 14,970 of it, GPU 3 ms, the main thread `state R` with 64 s of CPU in 68 s of wall. That dead-constant 15.02 s is `RUNAWAY_FRAME_SECS` — a Lua loop that never returns, re-entered the next frame and the next. The bound did its job and said nothing, because its raise unwinds to whatever `pcall` caught the call. On this corpus that is always an addon's own guard, so the trip became nobody's script error: `session errors: 0`, no log line, no file, no line number. `WOW_UI_RAISES` would have caught it and is `eprintln!` only, which is no use to someone who is in the game. So the hook writes its own report before raising, and the host logs it at `error!` beside the ordinary error drain. Each kept report carries a sixteen frame traceback, because the innermost line is usually inside a library and the handler that entered the loop is the name worth having. And the bound tightens to a second once it has tripped. The first one has to be far beyond any legitimate frame, which makes it far beyond playable: a loop is re-entered every frame, so the player gets one frame every fifteen seconds — indistinguishable from the freeze the bound exists to prevent, and that is exactly what he got. One trip has already proved the call non-terminating; the ones after it need only enough rope to say so. A second is still some 125M VM instructions, two orders past the heaviest handler ever measured here. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
75ba592fba |
ui: the Era map canvas over 1.12's own world map, and the Line region
What 1.12 withholds here is a framework, not a verb. Its world map is a fixed 1002x668 frame showing one zone — twelve tiles, a hit button, POI buttons, two dropdowns — with no canvas, no pan, no zoom, no pins and no notion of where one map sits on another. Every map addon written after 4.0 drives the canvas framework instead, and there is no stock file to migrate from because there is no stock feature. The constraint that shaped every decision: `WorldMapFrame` already has eight 1.12 consumers on this install, all anchoring icons to `WorldMapDetailFrame` and `WorldMapButton`. So nothing is replaced. The canvas is added, and those two frames are reparented INTO it — at rest the rect they had, at scale 1 with no pan, so the stock window is pixel-identical, and the payoff is that the 1.12 art and every icon on it now pan and scale WITH the canvas for free. Clipping comes from a real ScrollFrame: this client clips a scroll child and every descendant of it from one source, in the draw pass and the hit-test, so drawing and hitting agree. The base-art layer a canvas addon hides is a proxy that switches draw layers rather than the frame, because hiding the frame would take the POI buttons and every addon's icons with it. Three regressions, each found by measurement, each pinned: the anchor cycle (container to detail frame to canvas to container — a cycle resolves to nothing, and the symptom was `GetCenter()` nil inside the stock hover pass); the base scale (a minScale below 1 opens the map zoomed out of its own window); and `dataProviders` being a set rather than an array, because callers iterate it with `pairs` and bind the key — an array handed that loop the number 1. `CreateLine` is a Texture carrying two endpoints, drawn through the explicit-corner quad path the cooldown wedge already uses. Nothing stands in for it: a line between two moving points is a rotated quad whose angle and length both change. `C_Texture.GetAtlasInfo` answers nil for every name, and the nil is the answer — but the getter's presence is what makes the common atlas guard take its fallback branch instead of raising. The harness grew a second surface, and the finding is worth keeping: the full screen map hides `UIParent`, so opening it in the ordinary render probe dropped the corpus's draw column from 37/86 to 18/86 — nineteen unrelated addons. As its own pass: no addon lost a quad and nineteen gained, pfQuest 374 -> 911, pfUI 1063 -> 2259. OpenMapClassic now reads 1/1 on all five columns, stitching 74 of the realm's own zones. Corpus roster unchanged at 85 ok / 1 fail; the 12 data-half reds byte-identical to HEAD's. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
b49856c5cb |
ui: texture masks, the later-era texture dials, and the pools a canvas map needs
The mask pair is the one that matters and nothing can stand in for it. 1.12 has no texture masking at all — its single masked thing is the minimap disc, cut engine-side — and no idiom reaches the same end, because the point is that a rectangular texture draws as a non-rectangular shape. A seamless world map is that operation and nothing else: a zone's art is a rectangle of parchment and the map wants it cut to the zone's coastline. It was affordable because the render pass already had the slot. `UiQuad::mask` is a screen-anchored alpha mask built for the minimap's fixed circle with tiles panning under it, which is exactly `AddMaskTexture`'s semantics — the mask is another region with its own anchors, and the subject keeps the fragments its alpha covers at that place. So this is plumbing, not a new material. One mask per quad, said rather than hidden: widening it is a second texture binding and a second uniform on every UI quad in the client. The list is kept whole, because the order decides the winner — first added, which for a stitched map is the land mask. Two traps are pinned by tests: a mask region is skipped at extract rather than marked hidden, since its rect is what it covers; and a mask whose art is missing drops the mask, not the quad, because zero coverage would blank the art it was meant to shape. `Texture:SetRotation` was nearly free — the renderer has carried the angle since the minimap arrow, so this is the Lua setter for it. `SetDesaturation` is the fractional form of 1.12's own flag, rounded onto the one flag this renderer has, with the getter answering the caller's own number. The tile pair is applied: the host computes the repeat count once the art has loaded. `SetPropagateKeyboardInput` is stored, not applied: 1.12 already always propagates, so there is no swallow mode for false to select. `EditBox:HasFocus` publishes a fact this client already tracks and never exposed. Corpus: 85 ok / 1 fail, roster and digest unchanged. The 12 data-half reds are byte-identical to HEAD's. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
87f93fea46 |
compat: the uiMapID graph over the realm's own WorldMapArea, and C_Map's other half
1.12 addresses a map by a (continent, zone) pair of dropdown indices — a position in two lists, which moves when a zone is inserted — and publishes no map's rectangle on any other map at all, because its window draws one zone at a time into a fixed 1002x668 frame and never needs to know where that zone sits. A seamless, pannable map is exactly that missing relation. `script/uimap.rs` builds the tree as a pure function of the catalog `ui_world_map.rs` already pushes, rebuilt inside both setters so the graph and the catalog cannot disagree. No new DBC read, so nothing here can contradict what the stock map window itself draws. It is built from OctoWoW's own rows, which is the point: 175 against stock 1.12's 51, with 124 maps added and 49 shipped rects re-tuned. Era's uiMapIDs are reused where the map is the same map, joined on the WorldMapArea art folder rather than the zone name, because the folder is the column a realm has no reason to touch — this realm spells Azshara's `Aszhara` and Hillsbrad's `Hilsbrad` exactly as 1.12 does. A row Era has no counterpart for is minted from its own AreaTable id. One realm quirk is handled with its measurement: a third `areaID == 0` row whose art folder is `World` is the sheet's own row, which the reference's catalog walk files beside the two continents — `SetMapZoom(3, 0)` then `GetMapInfo()` answers `World`. Left in, it is a phantom continent carrying a second copy of all 39 Eastern Kingdoms zones. Art is answered as paths, so the chain serves the upscaled pack; the layer geometry is the logical 1002x668 / 256x256, because normalized fractions are invariant under a uniform upscale. The harness now seats the player a world position and the per-frame map feed, because every verb that answers *where* the player is reads the position, not the zone name. Corpus: 85 ok / 1 fail, roster and digest unchanged. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
0886952611 |
ui: a numeric texture argument is a file id, not a colour channel
With this and the font-object rename below, **OpenMapClassic loads clean, survives a session
start and survives a UI probe** — from dying on its first file, inside Ace3, before a line of
its own code ran.
Later clients gave every shipped file a numeric id and addons pass those instead of paths. Both
of our texture setters read a number as the **colour** arm, and they failed differently, which
is why this is worth a module of its own:
- `region/paint.rs`'s `SetTexture` needs *two* numbers for a colour, so a lone
`SetTexture(130763)` fell through to the catch-all and drew **nothing**.
- `button.rs`'s slot setters take one number as a colour with green and blue defaulting to 0,
so `SetNormalTexture(130763)` painted a **solid block** where the button art belongs. That is
the arm Ace3 calls, and it failed silently: nothing raised, the art was simply wrong, which is
the worst way for this to go.
**The rule**: a lone numeric argument of 2 or more is a file id. A colour channel is 0..1 and
every id a client ever minted is six figures, so the two cannot collide and `SetTexture(0.5)`
keeps answering exactly as 1.12 does.
**The table is transcription, not invention.** Every numeric literal passed to a texture setter
anywhere in the installed corpus *and* in both `Addon_Ports` addons — 21, all from Ace3/AceGUI
— is annotated by its own author with the path it means:
button:SetNormalTexture(130763) -- "Interface\Buttons\UI-DialogBox-Button-Up"
So the ids stay the real ones rather than something of ours, and the source is the call site. An
id with no row draws nothing and records a warning naming it, because a wrong texture is harder
to notice than a missing one — and that warning is what says which id to add. `GetFileIDFromPath`
answers only what the table knows and nil otherwise: minting a synthetic id would hand back a
number no other client agrees with.
**`SetNormalFontObject`** is the later-era spelling of 1.12's `SetTextFontObject` (`0x780400`)
— the one name this family left behind, since `SetHighlightFontObject` (`0x780700`) and
`SetDisabledFontObject` (`0x780580`) are 1.12's own. It hands out the same function value, so
the spellings cannot drift. Ace3 calls it on every button it builds
(`AceConfigDialog-3.0.lua:586`), one line above the `SetNormalTexture(130763)` above.
What openmap still cannot do is **draw**: it does not build a map, it re-skins Era's canvas
`WorldMapFrame` (`Map/WorldMap.lua:823`), and that canvas does not exist here. That is the map
stack, not this.
Corpus: 100 addons, 85 ok / 1 fail, roster identical. benilla-ui 2085 passed. Chain green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
4d19090259 |
compat: the cast-info pair as bare globals, and IsMouseButtonDown over a real feed
Cinematic's last two session errors. With these it loads clean, survives a session start, survives a UI probe and **draws** — all four of the harness's runtime columns. **`UnitCastingInfo`/`UnitChannelInfo` as bare globals** is an export, not a feature: the namespaced pair was already built with the whole contract in `cast_info` (the two return shapes, the millisecond unit, every position this client leaves nil and why), and these hand out the **same function values** the `C_Spell` table holds, so the two spellings cannot drift. Six calls of each in `Cinematic/Camera.lua` and its siblings, unguarded. `CastingInfo`/ `ChannelInfo` stay namespace-only: nothing on this install calls them bare, and pfUI reaches the one it wants through the namespace (`libdebuff.lua:887`). **`IsMouseButtonDown` needed a feed before it could be built honestly, and the reason it was declined is why.** `compat/classic_api/input.rs` had said no on the grounds that the only press state here is `Model::mouse_down_on`, which the pointer layer fills *only* for a press that landed on a frame — and `Cinematic/Camera.lua:612` asks in order to spot a right-drag **on the world**, which is exactly what that map cannot see. Answering from it would have been wrong for the one caller. So `Model::mouse_buttons` is new: `[Left, Right, Middle, Button4, Button5]`, pushed each frame from `ButtonInput<MouseButton>` beside the modifiers. Placed there deliberately rather than with the pointer feed below it, because that feed is skipped for a cursor off the window or a hidden UI, and "is the right button still held" has to stay true through both — which is the camera's question. Both spellings (`RightButton` and `Button2`), no argument means any button as upstream's does, and a button no device has answers not-down rather than raising. The reference capture of the real client's `_G` has no mouse-button getter at all, and `ClassicAPI.dll` carries the name, so it is compat's rather than a 1.12 gap. Corpus: 100 addons, 85 ok / 1 fail, roster identical. benilla-ui 2082 passed. Chain green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
482ad0bc5c |
sound: the later-era Sound_ CVar names, which three installed addons already wanted
`GetCVar("Sound_MusicVolume")` answered nil, so `Cinematic/Core.lua:3621` raised inside a
`format` and took the rest of its `PLAYER_LOGIN` handler — the music restore — with it. 2.x
renamed the sound CVars with a `Sound_` prefix and 1.12's names stayed what they are; Cinematic
is a multi-flavour addon and uses the modern spelling at **all 47** of its music sites, which
is the whole of its music-fade feature.
**This is a fix for the install, not a concession to the port.** Three addons already here ask
with the modern spelling and have been reading nil: DoiteAuras' editor and Tmog
(`Sound_MasterVolume`, `Sound_DialogVolume`, `Sound_EnableDialog`), SoftSpoken
(`Sound_EnableAllSound`, `Sound_SFXVolume`, `Sound_EnableSFX`).
Ten rows, each on the footing the `BackgroundSound` pair in this file already set: two names,
one knob, dispatched together in `sound/mod.rs`'s observer so the spellings cannot disagree
about a setting. Each default is its 1.12 sibling's, since they are the same setting twice.
Two honest compromises, labelled where they live: 1.12 has no dialogue channel, so
`Sound_DialogVolume` rides the SFX volume and `Sound_EnableDialog` the error-speech toggle; and
it has no SFX-only enable, so `Sound_EnableSFX` rides `MasterSoundEffects`, the only sound
enable it has. Riding what the mixer actually has beats inventing tracks it has not got.
**`GetCVarBool`** is absent from 1.12 on this repo's own evidence —
`reference/1.12-globals.tsv` is a capture of the real client's `_G` and carries `GetCVar` and
`GetCVarDefault` and nothing else of the family — and present in `ClassicAPI.dll`'s string
pool, so it belongs to the client this build announces itself as. It answers a real boolean, as
the name says, and **nil for a name with no row**, so "there is no such CVar" stays
distinguishable from "it is off". `Cinematic/Camera.lua:493` reads `ActionButtonUseKeyDown`
through it.
Not built, and deliberately: `IsMouseButtonDown`. `compat/classic_api/input.rs` already
declined it with the right reason — the only press state here is `Model::mouse_down_on`, which
the pointer layer fills *only* for a press that landed on a frame. `Cinematic/Camera.lua:612`
asks in order to spot a right-drag **on the world**, which is exactly what that map cannot see,
so answering from it would be wrong for the one caller. It wants a real mouse-button feed.
benilla-ui 2081 passed. Full chain green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
139965f415 |
ui: PreClick/PostClick, and a comma-separated template list
Two of the `Addon_Ports` pair's load blockers, both measured with `addon_harness` either side. **`PreClick`/`PostClick`** follow the convention `SCRIPT_KINDS` already set for names that are not 1.12's — `OnHyperlinkEnter`/`Leave` sit there with the same justification: a later client added them, addons register them at load, and refusing the name raises and takes the rest of the file with it. `Cinematic/MenuWheel.lua`'s wheel slots are the callers here. They arrived with the secure template system so a button could act either side of the protected part; nothing protected sits between them here, so the order is all they are. `button::click_button` fires each of the three **separately**, so one that raises is recorded and the next still runs: the pair exists to bracket the click, and a broken bracket must not eat it. That is the same reasoning the attribute-click dispatch below it already carries. **A comma-separated template list**, which 1.12 does not take: its `inherits` is one name, and `object.rs`'s lookup is a transcription of that (`0x6ee6f0` at `0x7061dd`, no splitter). Era takes a list and applies them in order, and that is what an addon written for Era passes — `Cinematic/MenuWheel.lua:312` asks for `"SecureHandlerShowHideTemplate, SecureHandlerAttributeTemplate"`. **The evidence is weaker than `compat/`'s other relaxations and the code says so.** The one-argument `SetPoint` rests on the real `ClassicAPI.dll` writing that form itself; this rests only on Era doing it and the addon targeting Era — nothing in the DLL's own Lua passes a list, and **no addon in the installed corpus does either**, which is also why it can break nothing. Safe by construction: the split is attempted only **after** the verbatim name misses, so every existing call takes exactly the path it did before and a template whose own name contained a comma would still win. One unknown name anywhere in the list and the whole call misses, reporting the literal **unsplit** and leaving no frame behind — so a typo reads as a typo rather than as a half-built frame. `a_comma_list_is_one_name_and_a_single_name_still_chains` was a deliberate, traced pin on the old behaviour, so it is rewritten rather than deleted. It now pins the shape of the relaxation: the verbatim name tried first, both templates applied in the order written with `$parent` resolving against the instance, a list with one unknown name still missing unsplit and leaving nothing behind, and a plain unknown name unchanged. Its doc states the weaker footing, which is why the miss path is pinned that closely. benilla-ui 2081 passed. Full chain green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
af823dd708 |
compat: the secure-handler family and the animation surface
Both modules land together because `classic_api/mod.rs` and `widget_surface.rs` carry rows for
each: split, the first commit named a module the second added, and would not have compiled.
## The secure-handler family, because 1.12 has nothing to be secure from
`frame.rs` declined it with a stated reason — "modern restricted execution, which 1.12 has no
equivalent of" — and that is right about the *sandbox* and is the whole of why it is now built.
A secure handler exists so an untainted snippet may call **protected** functions; 1.12 has
neither taint nor protected functions, so the restriction has no counterpart here, while the
plumbing underneath (a named frame store, a snippet on a child's handler) is ordinary
behaviour the snippets need. The body runs as plain Lua with the names the secure environment
would bind.
**The per-header shared table is the part worth getting right.** `MenuWheel.lua:484` writes a
bare `held = false` inside a snippet, and in the real secure environment that lands in a table
per header, not `_G`. `setfenv` onto one table per frame, `__index` to `_G` so the rest of the
API resolves; the test asserts `held` is **not** a global afterwards. My first test asserted
through `_G` and failed — the implementation was right and the test measured the wrong thing.
**`SecureActionButtonTemplate` is not empty**: reading your own attributes on click is plain
dispatch. `type="click"` clicks the frame in `clickbutton` (how the wheels reach the game's
micro buttons), and the lookup follows upstream's order (`type-<button>`, `*type*`, `type`).
**That exposed a real bug of ours**: `SetAttribute` silently dropped a frame.
`attribute_value` kept bool/int/number/string and sent everything else to `ScriptValue::Nil`,
so `SetAttribute("clickbutton", SomeFrame)` stored nothing — and upstream's `clickbutton` is
routinely a frame, which `Cinematic/MenuWheel.lua` relies on. A widget table is now kept as
`ScriptValue::Frame(id)`, the variant `NAME_PLATE_CREATED`'s arg1 already uses, which
round-trips through `object::frame_wrapper`. A non-widget table is still nil, documented.
**Not built, named rather than left to be discovered**: the binding half (`ClearBindings`,
`SetBindingClick`, …) are no-ops — no override-binding layer here to write into — so a wheel
opened with the mouse works and hold-a-key-and-release does not; and the `_onshow`/`_onhide`
attribute snippets are not run, since `SetAttribute` is a plain store with nothing to drive
them.
Templates in `assets/ui/CompatTemplates.xml`, listed on that wall with the reason: no stock
file to migrate from, and XML rather than `compat/` only because `inherits=` is loader-resolved.
A trap caught before shipping: `BackdropTemplateMixin.GetBackdropColor` forwarding to
`self:GetBackdropColor()` calls **itself** once `Mixin()` copies it onto the frame — the stock
getters answer, so there are no getter rows, and a test mixes in and reads back.
## `CreateAnimationGroup`: the surface, with nothing animated
1.12 has no animation system and this UI pass has no interpolation to hang one on. The cost is
the **absent constructor**: `Cinematic/Options.lua:7106` builds a rotating spinner and the nil
method aborted the whole options panel. Measured first — that is the addon's only animation in
55,661 lines, so this is a surface for a decoration.
**A non-looping group finishes the moment it plays**, firing `OnFinished`, because nothing
moves and that is the reading that keeps an addon's flow right; one that chains its next step
off `OnFinished` would otherwise wait forever. A looping group plays until stopped. Stated in
the header, since the name implies motion: a spinner does not spin, a fade appears at its
destination, a sequence lands on its last frame. Setters and getters round-trip.
Corpus: all 100 of grom's addons surveyed before and after — 85 ok / 1 fail, rosters
identical. benilla-ui 2081 passed. Full chain green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
64f07d9864 |
compat: SetPoint("POINT") with nothing after it, because ClassicAPI writes it
**A deviation, and the reasoning matters more than the code.** 1.12 raises `Usage:` on a
one-argument `SetPoint`, and that is traced rather than assumed: the `lua_type(L,3)` tag gate
at `0x7a25e6`..`0x7a2620` accepts four tags and an absent argument is none of them, which
this file's own header has said all along. Later clients read it as the default parent at the
same point with no offset, and that is what every Era-era addon writes —
`Cinematic/CameraWheel.lua:243` is `face:SetPoint("CENTER")`.
Allowed because **the real `ClassicAPI.dll` writes that form itself**: its embedded Lua
carries `f:SetPoint("CENTER");` and `tx:SetPoint(q[4]);`, so on the client this build already
announces itself as (`CLASSIC_API_VERSION` is set) the one-argument form has to work, or
ClassicAPI's own windows would not build. Claiming that client and withholding this is the
inconsistent position `stdlib_backport.rs` settled for the stdlib half.
It costs nothing in the ladder because of how it is spelled: the leg resolves to exactly what
the reference's own **three-argument** shorthand does (`RelTarget::DefaultParent`,
`0x7a270e`) — default parent, `rel_point` defaulted to `point`, offsets falling through to
(0, 0). No anchor shape is added; one spelling is mapped onto another the reference answers.
**It lives in `parse_set_point` and not in `compat/`, and that is load-bearing.** My first
attempt wrapped `SetPoint` per method table and `tests::region_map` caught it: the reference
registers it once in the Region method table, so every widget type runs the same function
object, and three wrapped copies broke an identity `region_map.rs` maintains on purpose (an
addon can pull `WorldFrame.GetHeight` off and apply it to a Texture). The header now says so,
so the wrapper route is not tried again.
Two tests pinned the old behaviour and are rewritten rather than deleted:
- The trailing-nil test gets **stronger**. It existed because `lua_type` tells `LUA_TNIL`
(the screen root, `0x7a2710`) from `LUA_TNONE`, and the ladder must see the count Lua saw.
Both forms resolve now, so it resolves a real layout and asserts they land in *different*
places: the explicit nil at the screen's corner, the absent one on its parent.
- One genuine ordering change, asserted: with no third argument there is no tag to gate, so
`SetPoint("MIDDLE")` reports "Unknown region point" where it reported `Usage:`. The
tag-gate-runs-first claim is kept for a *bad* tag, `SetPoint("MIDDLE", true)`.
Movement: Cinematic's load `CameraWheel.lua:243` → `MenuWheel.lua:312` (the secure-template
family, which `compat/classic_api/frame.rs` declines with its own reason).
benilla-ui 2077 passed, 0 failed. Full chain green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
164eb7e624 |
ui: the ClassicAPI-era templates addons inherit by name
`CreateFrame(): Couldn't find inherited node "DialogBorderOpaqueTemplate"` is where OpenMapClassic stopped — on its first file, inside Ace3, before a line of its own code ran. It is not that addon's problem: `BackdropTemplate` is named **unguarded** by Ace3 itself (`AceGUIContainer-Frame.lua:179` and `:206`, `AceGUIContainer-DropDownGroup.lua:128`), so every Ace3 addon on this install hits it. `assets/ui` does not grow, so this file carries its reason on that wall's own list: there is no stock 1.12 file to migrate from, because 1.12 has neither template. They are the same kind of surface as `script/compat/`'s verbs and live in XML only because `inherits=` is resolved by the loader and cannot be answered from Lua. **`BackdropTemplate` is correctly empty.** 9.0 moved the backdrop off the C side and invented `BackdropTemplateMixin` to put it back; 1.12 still has `SetBackdrop`/`GetBackdrop`/ `SetBackdropColor`/`SetBackdropBorderColor` as real frame methods, so a frame inheriting the template already answers all four. The mixin table exists for the `BackdropTemplateMixin and "BackdropTemplate" or nil` idiom and for `Mixin()`, its rows forwarding to the stock methods. **No getter rows**: a forwarding `GetBackdropColor` would call itself once `Mixin()` had written it onto the frame, and the stock one already answers. The dialog border is one `<Backdrop>` over this client's own dialog art — the shape every stock 1.12 dialog draws, and the same block `GameMenuFrame.xml` uses — rather than a transcription of Era's nine `SharedUIPanelTemplates.xml` textures. The opaque variant differs only in background alpha, which is the one thing the two names disagree about. `enableMouse` stays off: it is a decoration over a frame that handles its own input, and a border that ate clicks would make an Ace options window unusable. Movement: openmap `AceConfigDialog:573` → `:586`, now wanting `SetNormalFontObject`. Full chain green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
e731a77a42 |
compat: RegisterUnitEvent, the fixed-strata pair and SetWordWrap
The next three blockers of the `Addon_Ports` pair, each measured with `addon_harness` before and after rather than picked off a list. **`RegisterUnitEvent(event, unit1 [, unit2])`** — Cinematic's load blocker, and `docs/ADDON_COMPAT.md` already had it on the "plausible and untraced" line. The registration is the stock one: same `event_to_frames` list, same registration order, and `Model::frame_unit_events` holds the units per frame per event. `tick.rs`'s walk skips a frame whose row for this event does not name `arg1`; the map is empty on an install with no such addon and the walk tests `is_empty()` first, so the common path costs one load. All three unregister paths drop the row, including the one that is easy to miss: a plain `RegisterEvent` after a unit register **widens** the registration back, as upstream's does — the frame keeps its place and loses the filter. A row surviving an unregister/re-register cycle would swallow that event forever, so the test walks the whole cycle. `Cinematic/Frames.lua:2611` registers `UNIT_AURA` for `"player"` at file scope, so the nil method stopped the file dead. **`SetFixedFrameStrata`/`SetFixedFrameLevel`** are no-ops, and that is the honest answer rather than a stub: 9.0 added them to tell the *frame manager* to stop re-deciding a frame's strata and level, and this client has no such manager — both are whatever the setters last wrote and nothing else rewrites them, so the guarantee the caller asks for already holds. Ace3 calls both unguarded (`AceConfigDialog-3.0.lua:575-576`). **`SetWordWrap`/`CanWordWrap`** is 3.x's whether-to-wrap switch; 1.12's `SetNonSpaceWrap` is a different question (where a wrap may fall, not whether one happens). **Stored, not applied**, on the footing `SetSpacing` and `non_space_wrap` already set in `types.rs`: the text layout wraps to the declared width unconditionally and has no one-line-and-clip mode for the flag to drive, so honouring it would mean inventing a layout mode. Kept because the pair must round-trip and because the missing *setter* is what costs — `Cinematic/Options.lua:4130` calls it on a label and aborted the options panel there. All four are registered in `widget_surface.rs`'s `ALLOWED_BEYOND` with their call sites. The first `RegisterUnitEvent` row said "Frame" and the wall rejected it: event registration lives on the one shared frame method table, which is where 1.12 keeps `RegisterEvent` too, so the two are registered alike on all twenty frame kinds or neither is. Movement: Cinematic's load `Frames.lua:2611` → `CameraWheel.lua:243`, session `Options.lua:4130` → clear; openmap `AceConfigDialog:573` → `:575`. benilla-ui 2077 passed, 0 failed. Full chain green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
79e2494fc6 |
compat: math.huge and SetColorTexture, the two load blockers of the ported pair
Measured through `addon_harness` before anything was written, which is the only reason this
is two small changes rather than a guess at a list. Both `Addon_Ports` addons failed to load
at all; these are their first two stops.
**`math.huge` was nil, deliberately.** `stdlib_backport.rs` restores the 5.1 stdlib members
ClassicAPI backports, scoped to what the corpus calls, and `math.huge` was on the stated
"nothing calls them" list. `Cinematic/Core.lua:1103` is `local lastMovingOff = -math.huge`
at **file scope**, so the nil stopped the addon's main file dead ("attempt to perform
arithmetic on field `?`") and nothing after it ran — one site, but a load-time one, which is
the bar that table sets. The real `ClassicAPI.dll` carries `huge` in its own stdlib name
table beside `fmod`, `modf`, `gmatch`, `match`, `select`, `maxn`, `cosh` and `reverse`, so
this is what the client we already claim to be actually has.
Two tests pinned it absent and are updated rather than deleted: `lua50.rs`'s keeps asserting
this *layer* removes it (the registry stash is the proof), and `stdlib_backport.rs`'s keeps
the narrowness assertion for the four that stay gone. The new assertion is `-math.huge`,
because file-scope arithmetic is the shape that failed.
**`Texture:SetColorTexture`** is the modern spelling of `SetTexture`'s colour arm, which
1.12 already has (`0x770360`'s 8×8 solid surface, and `region/paint.rs` defaults the alpha
to 1 as it does). Retail split the names when `SetTexture` stopped taking a colour, so every
Era-era addon writes the new one. `Cinematic/Options.lua:701` paints its list border with it
and died at session start; OpenMapClassic its map chrome. Four lines of the same composition
chunk `SetShown` and the geometry pair use, and an `ALLOWED_BEYOND` row with both call sites
— `docs/ADDON_COMPAT.md` had it on the same "plausible and untraced" line as
`RegisterUnitEvent`, and this is that line's follow-through.
Movement, same harness both ways:
- Cinematic load: `Core.lua:1103` → `Frames.lua:2611` (~10 files further; wants
`RegisterUnitEvent` next). Session: `Options.lua:701` → `Options.lua:4130` (3400 lines on,
wants `SetWordWrap`).
- OpenMapClassic: unchanged at `DialogBorderOpaqueTemplate`, which neither fix touches.
benilla-ui 2076 passed, 0 failed. Full chain green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
2a65266837 |
assets: the shader asset type exists for an app with no render plugins
Every end-to-end asset-pipeline test — the M2, WMO, BLP and ADT loaders, the minimap's `load_with_settings` path and the Ironforge group grid — panicked with "Cannot allocate an Asset Handle of type 'bevy_shader::shader::Shader' because the asset type has not been initialized". `register_shaders`'s `load_shader_library!` calls allocate a `Handle<Shader>`, and on Bevy 0.18 it is the render plugins that initialise that type; a headless app has none, so all six went red at once on the upgrade. They are the only tests that exercise the loaders every zone load depends on, and the gate could not see it: each one is data-gated, so it skips without the install, and the no-install gate is the one that runs everywhere. Found by running the workspace suite with the install mounted — 44 red there against 0 on the gate. Initialised here when nothing else has, and **guarded**: `init_asset` *inserts* `Assets<Shader>`, so calling it unconditionally in the real app would drop every shader the render plugins had already loaded. benilla-assets: 109 passed, 0 failed (was 103 and 6). Full chain green. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
e646b5a678 |
compat: the usable pair tells the truth — 1/0 from nampower's verb, the walk behind C_Spell's
Two defects in one pair of verbs, both found by reading what grom's addons do with what we answer rather than from a report. A DLL whose source is not to hand still has its contract written down: in the addons written against it. **`IsSpellUsable(id|name)` answers two numbers.** Ours gave 1/nil for the first — transcribed from `IsUsableAction`'s own convention — while already giving 0/1 for the second. Four readings in the corpus say otherwise, and `AlternativeClassicActionBars/Button.lua:1907` is the clearest: `elseif usable and usable ~= 0 then` is not a line anyone writes against a function that answers 1/nil. `SuperCleveRoidMacros/Core.lua:1035` labels its own branch "IsSpellUsable returned 0 = wrong stance/form, not enough power type", reading the value straight out of `pcall(IsSpellUsable, …)`. And the one with teeth: DoiteAuras' `_SafeSpellUsable` keeps a verdict only `if ok and u ~= nil` (`DoiteConditions.lua:970`), so a nil sent it down a fallback chain ending at `:1022`'s "treat as usable so icon doesn't die forever" — **every** Usable-mode aura lit on cooldown alone, stance, rage, range and reagents ignored. Answering the wrong shape and not answering at all land in the same place. 0 is truthy in Lua, so that was checked before changing it: no caller tests the value for truth unguarded, and the rest compare `== 1`. The DLL's string pool publishes the signature and the name-lookup error but not the return shape, and the five mirrors of the fork's `SCRIPTS.md` I could reach are each behind a Gitea login or 404 — the corpus was the better source. A name it cannot resolve raises there, which is why every caller wraps the call in `pcall`; ours answers `nil, nil`, and all four callers treat those identically. **`C_Spell.IsSpellUsable` was computing the wrong thing and raising on its caller's argument.** It was approximated as "known and not on cooldown" — the one reason upstream does *not* count (its own list is "hasn't learned it, lacks mana or reagents, reactive conditions haven't been met") and none of the ones it does. It now reads the same per-spell map the global does, the reference's usable walk (`0x6e3d60`), which agrees about cooldowns: "a running cooldown never greys here" (`spell/usable.rs` leg 11), leaving `GetSpellCooldown` the question about cooldowns. That makes it the honest answer for a reactive proc, which is exactly what its only caller asks: Overpower greys on the caster's combo-point byte (leg 5, `PLAYER_FIELD_BYTES` byte 1, the byte the server sets when the proc banks), so the walk answers it inside the proc window and nowhere else, while "off cooldown" showed it nearly always. pfUI's reactive indicator (`modules/nampower.lua:130`) passes a **name** against a `u32` signature, which mlua refuses for a non-numeric string — a raise once per frame while the indicator was switched on. `spell_ident` is shared now, so a name resolves through the book as the global's does. Both returns there stay booleans, unlike the global's: the caller hands the first straight to `SetShown`, where a numeric 0 would be truthy and show every icon forever. Two conventions, each its own caller's, written down where they live. Tests: the global's pair is asserted as `1|0` and `0|1`, with the callers' own accept test (`pcall` + `u ~= nil`) quoted as an assertion; `C_Spell`'s is asserted over Overpower outside and inside the proc window with nothing about cooldowns changing, by id and by name, plus the not-in-book case answering false rather than raising. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
315ebe9138 |
ui: an <IMG> with an empty src names no file
`src=""` was stored verbatim, so the region carried `texture: Some("")`, the resolve asked
the size probe for a nameless file and the renderer searched the patch chain for one and
logged `texture miss: ''` — a defect-shaped warning that can name neither a file nor an
asker. `""` clears on every other texture setter here, as the reference's path arm treats
it like nil (`0x770200`), and `a9ed091eb` closed the three that had stored one; this was
the last route in the UI that still could.
Found by sweeping the rest rather than by a new sighting, and two of the leads this list
carried are now dead: the `CursorPayload` arm it called most likely is unreachable in a
live session (`extract/mod.rs:1020` guards the whole block with `if capture.is_some()`),
and `PortraitSource::File` cannot be empty — `temporary_portrait` always returns a `.blp`.
The M2 and WMO sides filter at parse, and `texture_url("")` would read `'.blp'` rather than
`''`, so the warning's own text says the UI sprite path is where it came from.
Whether this closes the one line in grom's log is unproven: it sat in `benilla-2.log` at
+20.717s, that run may predate `a9ed091eb`, and three measurement launches have since
rotated his logs out. The test is the next sighting: with the filter removed it reports
`texture: Some("")`, and in the fixture the nameless path also reserved the probe's 48 px
of flow under the image — in game the probe answers nothing for a miss, so only the chain
search and the warning were real.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
fb0d37e3c8 |
sound: a stream whose callback stopped is rebuilt, whatever the device layer said
cpal's ALSA loop exits on a `POLLERR` out of `alsa::poll()`, which reaches us as a `BackendSpecific` error — classified non-fatal there, because an xrun must not tear the stream down. So `device_died` was never set, nothing rebuilt, the ring filled, the render thread parked on a full ring and kira's renderer stopped advancing every sound in the mix. In grom's 2026-10-18 log that was four minutes of silence out of one `soggyplace-zone5.mp3`, cleared only by a worldport replacing the slot; the `3D sound(s) never played — the spatial-voice arena was full` count climbing to 267 in the same session is the proof that nothing at all was completing, music included. `StallWatch` reads the cycle count instead of the error string: a running stream that reports no cycle for 2 s is dead however it died, including with nothing raised at all, and `service` closes and reopens it exactly as it does a lost device. Reclassifying the error could not have worked — an xrun arrives in the same variant and the string is all we get. A window with no cycles is not a wall clock with no cycles. `service` runs once a frame on the main thread and a 4 s load hitch is one window the callback ran right through, so the span is taken between two `Instant`s and any window that saw a cycle resets it, however long that window was. That is the same trap as `510dda3d3`, one layer down. Measured while verifying it, and the reason for the other two pieces here: the rebuild held the main thread 4.5 s with the audio server down. Attributed with `WOW_SYSTEM_TIMING=1 WOW_STALL_MS=400` to `sound::poll_mix_health`, then split — close 5 ms, finding the default 2 ms, `Stream::open` 4521 ms. Opening a device is a blocking call into the host's audio stack. `SLOW_OPEN` now says so in the log, and the reopen wait backs off to 8 s while opens are slow or refused, so an outage costs one blocked frame per attempt instead of one a second. It did not happen in grom's own failure, where the server stayed up and the reopen would be instant. Taking the open off the main thread needs a device thread owning every open and close; not done here. Verified: four offline tests — a stream calling back at 3 ms frames, the 4.1 s login hitch arriving as one window of 380 cycles, a real stop named once on the limit, and the backoff's escalation — plus two live runs on Linux/PipeWire, restarting pipewire under a running client at the glue screen. Both read the same: position frozen, two genuine starve warns, `output stream dropped — the stream stopped calling back; reopening`, a reopen in the same frame, music advancing again. Before this, that was permanent. Built and run on Linux only; the device layer's own files are untouched apart from the comment that pointed the wrong way. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
510dda3d3c |
sound: the glue-music crackle never happened — a meter that summed frame deltas
`StreamWatch` asked "did this stream play as much audio as wall time passed?" by summing each frame's `Time<Real>` delta into `expected` and comparing it against the stream's own advanced position. But a Bevy clock delta describes the interval *ending* where the frame began, while `window_start` is stamped with `Instant::now()` part-way *through* a frame. So a window opened inside the 4 s world-entry frame was handed that entire 4 s by the next frame, as audible time that should have played — while the position, which tracks the speaker honestly, had advanced only by the window's real span. The difference was logged as injected silence. The tell was in the log all along: `position advanced` equalled the window `span` to the centisecond. When those two agree the audio was perfect, whatever `counted` says. All 101 "glue music stream starved" lines in every log to date are this artifact, plus 104 of 395 zone-music ones. Replaying the new formula over those logs silences all 205 and keeps all 291 real events — streams genuinely frozen with `position advanced 0.00 s`. Fixed by taking both ends of the window off one clock (`now - window_start`) and deleting the `dt` parameter outright, so the counted interval and the reported span are the same number by construction and no caller can supply an inconsistent delta. The old test `stream_watch_spans_the_window_it_counted` asserted exactly this invariant and passed, because its fixture fed a delta that agreed with its own wall clock — a fixture that supplies both sides of a relationship cannot test it. Replaced with the regression that matters: a stream which keeps up across a 4.1 s main-thread stall must not be reported, and one genuinely frozen through the same stall must still be named, or the fix has just deleted the meter. The warn now reports what it measured; "this is what a crackle sounds like" was read as evidence of a crackle and written into the defect backlog as one. `examples/stream_probe.rs` decodes a track through the chain exactly as kira's decoder does and compares the container's declared frame count against frames actually decoded — it killed the first hypothesis for the real dropout: 43 of 43 zone tracks are clean. The real dropout is still open and is not a crackle: one `alsa::poll()` POLLERR, which `output/cpal.rs` classifies as non-fatal and reports as "stream kept", and no sound advances again for the rest of the session. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
14d5dc36e9 |
formats: benilla-formats is green — the world-data goldens, and two tests that
were measuring nothing
The last 11. The realm reships `World\Maps\`, `HumanMale.m2` and
`TaxiPathNode.dbc`, so these tests had lost the *examples* they picked without
losing the properties they were for. Two had quietly stopped testing anything.
**A guard that was a proxy, not a fact.** The lava seam test skipped a vertex
whose UV was `[0.0, 0.0]`, standing in for "this vertex is dry". The realm's
lava leaves non-zero texcoords on undrawn verts, so 121 meaningless pairs were
entering the compare and one of them failed. Gating on the mesh's own `wet` map
instead leaves 277 pairs — the verts both neighbouring blocks actually draw —
and every one of them agrees. The invariant was never in doubt; the guard was.
**A keyless bone read as an empty band.** `empty_band_poses_the_bone` asserted
HumanMale's waist (bone 27) carries exactly one clamp key in Run. In the realm's
model bone 27 has no rotation or translation keys in *any* of its 156 sequences,
so there was nothing to clamp and the assertion was reading a keyless bone. The
subject is the five finger bones 100..=104 now: each appears in every one of the
156 sequences with exactly one clamp key at t = 0, which is the rule — an empty
band poses the bone and never drops it.
**The skeleton was renumbered, and the poses came with it.** 1.12's finger bone
102 is unposed at the HandsClosed frame; bone 47 carries the curl, within 0.004
of 1.12's `[0.2536, -0.016, 0.0577, 0.9654]`. Same for the eyelid: bone 75 ->
bone 24, still scale-only, still a 6633 ms loop that opens at t=0 and shuts
mid-loop. That the *pose* survived is what says renumbering rather than
re-animation, so the window arithmetic is still what is under test. (The model
does have HandsClosed throughout — `parse_m2_animations` not surfacing it is
exactly why the raw reader exists.) Ten new global-sequence bones came with it,
all parks: six 33 ms two-key constant scales and four rotation channels with
empty tables, recorded so a real loop appearing among them reads as a change.
**Tiles chosen instead of inherited.** `examples/liquid_scan.rs` walks a map and
reports which tiles still have each liquid property, so a re-recorded test names
a tile that was picked:
- the river mouth moved to `Azeroth_33_38`, three Still-over-Ocean chunks, all
disjoint — 1.12's 33_33 has no two-block chunk left. Four tiles *do* overlap
and are kept out deliberately: Kalimdor 26_54, 41_29, 41_30 and the realm's
own Azeroth 36_51, where 1.12 had only 26_54.
- the ocean depth divisor moved to `Azeroth_29_54`: 13210 depth bytes inside the
ramp and not one pinned at 255, so a wrong divisor cannot hide behind
saturation. 1.12's 41_41 carries no ocean at all now.
**The impassable band was thinned.** `Azeroth_33_44` authors 17 impassable
chunks where 1.12's copy had 48, and the chunk 1.46 yd east of the reported
`.go xyz -6601.98 -531.87 335.60` is among the ones it dropped — so the data no
longer walls the reported spot. The `Some(true)` leg would have stopped being
tested, so it moved to a column inside one of the 17 the realm does flag.
**The transport goldens no longer describe this client.** `TaxiPathNode.dbc` is
re-authored: 11219 rows against 9582, and 7975 rows kept their id while changing
path, index *and* coordinates. Path 436 is gone; the eight survivors miss
vmangos's 1.12 periods by −290815 to +271184 ms, and no transport path crosses
continents any longer. Closing that needs the realm's own transport periods,
which only live observation gives — so the calibration is a report now, with the
deltas printed on the same variant grid a future observation can be compared on.
What is still assertable is asserted: `client_period_ms` and the timetable's own
`period_ms` are one number on all eight. The two map-invariant tests never
needed a server period at all, so they run on taxi path 72 — 37 nodes over maps
{0, 1}, the only shipped path that still spans both — and additionally assert
that every transport golden is single-map, so none of them can silently stand in
for it again.
**benilla-formats: 445 passed, 0 failed.** `cargo fmt --all` and
`clippy -p benilla-formats --all-targets` clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn
|
||
|
gromandClaude Opus 5
|
c0685f0273 |
formats: the realm's tenth race, and what it switched on
Four goldens, and three of them turn out to be the same fact: `ChrRaces.dbc` has a tenth row — id 10, **High Elf**, on the Blood Elf art (`ClientFileString` "BloodElf"), Alliance-side. - **Every talent tab opened to it.** All 27 `TalentTab` rows went raceMask 511 -> 1023, so race 10 takes every class's tree rather than being gated out of any; the test asserted the opposite, that "race bit 10 is outside 511". `Talent.dbc` itself was rebuilt too — 476 rows, 317 ids added, 273 removed, 159 changed — so the trees are the realm's, not 1.12's. - **Dwarf Mages are real here, and the gate was already written for it.** `chain_is_stock` calls a chain stock when no playable race exceeds 8, so this install read as stock until race 10 arrived and the `UNUSED_COMBOS` Dwarf-Mage skip kept applying; the test's "patched chain or not" passed by accident. The skip only ever modelled a compare in the *stock* executable at `0x4706b0`, and this install's own `WoW.exe` has none — the whole builder holds no immediate compare against 3 or 8 and gates on a `0xFE` sentinel (`cmpb $0xfe,0x1(%ecx)` at `0x4706e8`). Re-verified against the binary rather than trusted from the comment. `CharBaseInfo` grew 103 -> 139 bytes, which at 2 bytes a row is 18 new race/class pairs. - **`CharHairGeosets` was re-authored, and took a test's only coverage with it.** 299 rows against 148, 136 of the survivors rewritten, and 1.12's one duplicated key — the goblin male's variation 0, listed four times as geosets 1, 2, 1, 2 — is gone with none replacing it. That key was the *only* thing exercising the loader's first-row-wins rule, so re-recording the count alone would have quietly dropped it: `or_insert` could have become `insert` and nothing would have noticed. The hair decode is now a `hair_geosets_from_bytes` seam and a synthetic DBC replays 1.12's goblin block to pin the rule, plus the reverse order to prove the later row is discarded and not merged. The real-data test asserts the absence instead. Race 9 is where the re-authoring shows: the goblin has a full hair set for both sexes now, 14 male variations and 17 female, and the male's variation 0 is `GeosetID` 0 where 1.12's winning row said 1 — both land on scalp geoset 1, since `visible_geosets` takes `max(1, GeosetID)`. Also: - **WorldStateUI** 20 -> 34 rows, and the realm rewrote the text and tooltip on every one of 1.12's 20. Only 33 ids are distinct — **168 ships twice**, both map 0 / area 44, one reading "Banners Controlled: %3809w" and one "Tower Score : %3811w / %3813w". Safe, and checked rather than assumed: the catalog exposes no id lookup at all, only `rows()`, and the app filters by (map, area), so both readouts reach the frame — which is the only thing two rows under one id can have been meant to do. Four zones narrow by area where 1.12 had two, the new three being outdoor PvP: Redridge Mountains, Hillsbrad Foothills and Silithus. Live `StateVariable`s run to 3810 where 1.12's three all sat under 3000. - **WMOAreaTable** 21115 -> 26653 rows and 21105 -> 26570 distinct (wmo, nameset, group) keys, so the repeats went 10 -> 83; 5633 ids added, 95 removed, 4080 changed. benilla-formats: 16 -> 11 failures (and one test more than before). `cargo fmt` and `clippy -p benilla-formats --all-targets` clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
d2414d9f50 |
stable: the realm sells four pet slots, and our window drew two
Found by the `StableSlotPrices.dbc` golden: the table has four rows where 1.12 had two, and `NUM_STABLE_SLOTS` was 3 (the current pet plus two). Three independent sources in the install agree on four, so this is not a guess: - `StableSlotPrices.dbc` ships 4 rows — 500c, 5g, 250g, and a fourth at one copper — against 1.12's two. - the realm's own `PetStable.lua` (patch-4) declares `NUM_PET_STABLE_SLOTS = 4`. - the realm's `WoW.exe` caps the purchase at four, and `0x4cb0bf` reads straight across: `cmp $0x4,%eax; je refuse` on the bought count, then `inc %eax` and `cmp 0xc0d728,%eax; jg refuse` against the price table's max id — which is exactly `num_stable_slots != NUM_STABLE_SLOTS - 1 && next_slot_cost != 0`. So the window drew two slots the server would happily sell, and `GetStablePetInfo(3)`/`(4)` answered nothing: the last two buttons were dead. `NUM_STABLE_SLOTS` is 5 now, the gate's comment cites the realm's binary rather than 1.12's 2, and the test fixture helper takes however many leading slots a case cares about so the existing three-slot fixtures still read. Also five goldens re-recorded, two of them findings: - **The spam filter is one line now.** patch-4 ships a 93-byte `SpamMessages.dbc` of one row where `patch.MPQ`'s 2161 bytes held 28, and the chain takes the whole winning file rather than merging ids — so Blizzard's 28 gold-seller patterns are gone and what is left matches a rival server's address, whitespace-tolerant between every letter. - **`md5translate.trs` stopped hashing.** 9794 rows against 8401, and an Azeroth tile now resolves to the readable path `Azeroth\map32_48.blp` rather than `ea283abc0bf9637c3fad5e840a65b38b.blp`. The realm ships the tile at that path in patch-4; 1.12's hashed copy is still in `texture.MPQ` at the same 33940 bytes, so only the table moved and the lookup still has one job. - ItemSet 172 -> 363 rows, and Defias Leather (161) gained a sixth item, 81007, one of the realm's own. Every one of 1.12's 172 sets also gained a name in five more locales — the same five-locale fill the realm gave `EmotesText`. - NPCSounds 156 -> 247, purely additive; the one pre-existing row patch-4 touched is 90, which gained two kits. benilla-formats: 21 -> 16 failures. `cargo fmt` and `clippy -p benilla-formats -p benilla-ui -p benilla-app --all-targets` clean. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
e61976bc0e |
formats: the world-map goldens, and a nondeterministic continent lookup
The four map tables against the realm's own files. One real bug, and three findings that are the realm's data rather than ours. **`WorldMapArea::continent` was a coin flip.** It `find`s the `area_id == 0` row for a map id, and 5875 had exactly one per map so order never mattered. The realm ships two for map 0 — 14 "Azeroth" and its own 694 "World" — and the lookup iterated the `by_id` `HashMap`, whose order is seeded per process. So `continent(0)` answered Azeroth or World depending on the run. Now it walks file order, which is the reference's own walk and picks 1.12's row; the test asserts 14 and passes eight runs in a row where it used to flip. - **WorldMapArea** 51 -> 132 rows loaded, of 175 in the file. The 43 dropped are dropped on purpose: `AreaName` empty, `AreaID` 0, the whole rect zero, one per instance map (33 Shadowfang through 533 Naxxramas, then the realm's 800s). An instance has no continent art — but `AreaID == 0` is exactly what `ui_world_map.rs:146` reads as "this row is a continent", so keeping them would put 43 unnamed continents in `GetMapContinents()`. `str_at` folds an empty string to `None` and the row falls out; the test now pins that as deliberate. The realm's third *named* continent is 694 "World", a second sheet over map 0 whose art ships (patch-M) but whose four rect columns are all zero — the only named row like that. `map_proj::zone_uv` already refuses a zero span and answers 0.0 instead of dividing, so that sheet puts a blip in the corner rather than NaN. - **WorldMapContinent** 2 -> 3 rows. The new row's `MapID` is 42, which the same patch *deleted* from `Map.dbc` along with the rest of 1.12's dev test maps (13 "Testing", 25 "Scott Test", 29 "CashTest", 37 "Azshara Crater", 42 "Collin's Test", 44 "<unused> Monastery"). It is inert rather than broken: the continent list comes from `WorldMapArea`, and this table is only ever reached by `wmc.get(map_id)` for maps 0 and 1, so nothing asks for continent 42. - **The zone bitmaps are repainted.** `Azeroth.zmp` and `Kalimdor.zmp` have the realm's own copies in patch-3 and patch-4, and every id they write is a **zone** where 1.12 wrote sub-areas and left the client to roll them up. Azeroth: 1874 labelled cells over 34 distinct ids, all zone-level, against 2630 over 334 of which 310 were sub-areas (Northshire Valley, Crystal Lake, Fargodeep Mine). After the one-hop rollup `ui_world_map.rs:193` applies, 1280 cells agree, 534 name a *different* zone, 816 lost a zone and 60 gained one — a redrawn map, not a mechanical flattening. Nothing of ours needed changing: the rollup is a no-op on an already-flat grid. The test asserts the structural claim behind the number (every labelled cell is a top-level zone), which 1.12's own file would fail, so a decode slip cannot hide inside a re-recorded count. - **WorldMapOverlay** 526 -> 707, and not an extension: 360 ids added, 179 removed, 347 survivors given a different texture and rect. None of this table's numbering is 1.12's, so only the joins are worth asserting. `dbc_diff` gained `--row=N`, which dumps a row from the winning copy *and* the baseline — the only way to read a row the realm deleted, which is how the six test maps above got their names. benilla-formats: 25 -> 21 failures, no regressions. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |
||
|
gromandClaude Opus 5
|
e58ddc10ab |
formats: re-record the sound and visual goldens, and a tool to say what moved
Six modules' data goldens against the realm's own files, plus the thing that made the last three honest rather than guessed: `examples/dbc_diff.rs`. The tool reads the copy of a DBC the chain *wins* and the copy it shadows and diffs them row by row. `Chain::copies` is ascending priority and several `patch-*` archives are mounted, so 1.12's own file is usually still underneath the realm's; the baseline is the highest copy in a stock archive (`dbc.MPQ`, `patch.MPQ`, `patch-2`) so the diff reads "what OctoWoW changed about 1.12". Schema-free — field 0 is the id, a row is its raw 4-byte slots — so an undecoded column still shows up as "id N differs at slot K". `--row=N` dumps a row with its text slots as text, which is how three ids below got names. Two findings, both the shape of a reference into a table that has no such row: - **The glow band.** The realm rewrote `ItemDisplayInfo.dbc` (patch-5) and left `ItemVisuals.dbc` as 1.12's out of base `patch.MPQ` — 34 rows, ids `1..=134`. 22 ids its new displays name run past that end, where a later expansion's table continues, so 61 items wear no glow. The reference drops them exactly as we do (`0x479700` skips a visual id past the max id), so this is the realm's data, not our decode. Pinned by id with the displays behind each. - **The silent kits.** `SoundEntries.dbc` grew 4623 -> 8805 rows and 4764 of its 16537 variation cells name a file the chain has not got: 236 directories hold no file at all (LesserManaFiend, ChineseDragon, ZoneMusic\BloodElf), and the ones that exist miss the TBC boss voices — Illidan, Akama, Lady Vashj, Kael'thas — plus 552 of `Sound\Doodad`. What makes that an observation about the data and not about our path join is *where* the misses fall: of the 2834 kits with a dead variation, 2832 have an id past 4623, the last row 5875 shipped. A faulty separator would miss evenly across old rows and new. Only 395 OgreFidget and 3792 WaterFallLarge predate the realm, and only 3792 is wholly silent, so nothing that used to make a sound has gone quiet. The rest are censuses re-recorded with their 5875 figures beside them. Every one is safe to re-record for a mechanical reason: `benilla-dbc` fails the load outright when a schema's expanded field count differs from the header's (`lib.rs:215`), so a test that *reaches* a count assertion has already proved its layout unchanged. - SoundEntries 4623 -> 8805 rows, EAXDef census (2072,2,2549) -> (6241,4,2560), type-17 NPC voices 275 -> 529 (still every one authored dry), variation cells 8961 -> 16537. Kit 8940's ten ASH_SPEAK misses stay, and now assert *why*: the asset ships and it is the authored leading `\` that cannot reach it. - VocalUISounds 533 -> 688 rows, live normal kits 785 -> 1085. - ItemSounds nonzero field-11 20513 -> 32728. - CreatureSoundData 406 -> 525 rows. A fifth row carries pet barks, and `dbc_diff` settles what it is: row 348 is 1.12's own and columns 27-29 are the only thing patch-4 changed about it, filled with the Imp's own kits — so an imp barks whichever of its three models is in play, `Creature\Imp\Imp.mdx` and TBC's `Impoutland` and `Moargminion`. Row 348 says nothing else at all. - SpellVisual 2165 -> 2313, SpellVisualKit 1772 -> 1969, SpellChainEffects 18 -> 21, chain CharProc census (68,48,20) -> (110,51,59). The three new beams are HarvestBeam, SpiritLinkBeam and Beam_ChainIron_NoAlpha, and unlike the sound rows the art ships with them; 23 zeroes every parameter the others animate. All three keep 1.12's segment length. - ItemVisuals: enchants carrying a glow 102 -> 109, and 108 of 109 resolve — still exactly the one -1, so every new glowing enchant names a row that exists. Only the display side dangles. benilla-formats: 38 -> 25 failures, no regressions. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn |