TOGTools-v0.15.0
What's new
Changelog
[v0.15.0] (2026-11-16) - The rest of the Leatrix Plus wish list
New Features
- Transports tab (Classic Era) (operator, 2026-11-13: "can we bring in the function from nauticusclassic into UT? showing the zepp/boats on the map and timers, with the titan panel integration?").
Modules/Transports/Transports.lua+GUI/TransportsTab.lua. Every zeppelin, boat and Deeprun Tram car is drawn where it is now on the world map (turned to its heading) and, in a zone it serves, on the minimap, through HereBeDragons-Pins; a schedule lists each route's next arrival or departure per dock; a LibDataBroker data source (UltimateToolsTransports, so Titan Panel picks it up) shows the followed route ("UC 1m 05s"), left-click for the next timed route, Shift-click for a departure alarm (one warningalarmLeadseconds before it leaves), right-click to open the tab. How it knows: each route's recorded path and exact round trip, plus one moment the ship was at a known point -- from riding it (faster than 17 yards per 0.8 s within 20 yards of a trigger point) or from another player over NauticusClassic's ownNauticSync/KWN4data-version-4 addon messages, which this module reads and writes byte for byte, so Ultimate Tools and NauticusClassic players share one pool of timings. Whose timing wins follows the same rules (fewer restarts, then fewer hand-offs, then newer by whole cycles), and a PC restart moves the stored cycle starts (T.ShiftForRestart, over 180 s of clock jump). Data: NauticusClassic (MIT, Copyright (c) 2020 psynct) -- the route facts are hand-written inTransportData.lua, the 1,903 recorded path points and four map icons are copied verbatim by the newtools/build-transports.py(desk tooltogtools-transports) with the licence beside them; no NauticusClassic code is used. HereBeDragons is the operator's SandPacker copy (Questie's fork,HereBeDragonsQuestie-2.0, operator: "i have a custom copy of herebedragons on sandpacker, we should probably use that"), vendored intolibs/HereBeDragonsby the same script. Loaded by the Era TOC only (isVanillaand notisForever);Tests/toc_spec.luarecords the six omissions. One OnUpdate drives it, no C_Timer. New specTests/transports_spec.lua. Not seen in game. - Range tab (operator, 2026-11-13: "i also want to bring RangeDisplay into UT, it's also installed on my PC").
Modules/Range/Range.lua+GUI/RangeTab.lua: a coloured yard readout ("8 - 15", "40 +") for target, focus (only where the client hasFocusFrame), pet and mouseover (beside the cursor), each with its own switch and an enemies-only tick; a text size, a lock (unlocked readouts show a green box to drag, positions saved per unit) and a reset. Bands: within 5 yards white, 20 teal, 30 green, a 35-yard band that is off by default, red from 40. The measuring is LibRangeCheck-3.0 (MIT, MINOR 38), the library RangeDisplay itself uses, embedded inlibs/LibRangeCheck-3.0by the newtools/build-rangecheck.py(desk tooltogtools-rangecheck) because its standalone release is load-on-demand with no Forever TOC; it is in all seven TOCs.GetRangeruns in a pcall and a secret result is skipped (Retail / Forever). Not built from RangeDisplay: warning sounds, custom colours and band yards, backgrounds, fonts, arena units. New specTests/range_spec.lua. Not seen in game. - The embedded LibRangeCheck-3.0 threw in a WoW Forever dungeon, through Plater (operator, 2026-11-14: 1323x "LibRangeCheck-3.0.lua:4081: attempted to index a table that cannot be indexed with secret keys", stack
Plater.lua:932 rangeChecker->CheckRange->UpdateTarget). The stack namesTOGTools/libs/LibRangeCheck-3.0, so through LibStub our embedded copy was the one every addon on that client used, Plater included. The library'sgetCachedRangekeys its cache onUnitGUID(unit)and swaps in the unit token only whenisMidnight(Mainline, interface 120000 and up) andissecretvalue(guid). Forever is not Mainline, but its client hasissecretvalueand returns secret GUIDs (FrameScriptDocumentation.lua:286in the Forever tree), so the secret GUID became the table key and the lookup threw. The check now feature-detectsissecretvaluerather than testingisMidnight. That is a local patch to upstream MINOR 38, applied bytools/build-rangecheck.py(PATCHES; the build refuses if the upstream line changes and the source does not already carry the fix), so a rebuild keeps it. The Range module's ownpcalldid not help here because Plater calls the library directly. Not fixed: another addon embedding an unpatched MINOR 38 that loads first would win the LibStub race and bring the bug back on Forever. New specTests/librangecheck_spec.lualoads the real library on Forever, Midnight and Era builds. With the old line put back, only the Forever case fails. The harness raised on the library'sGetInventorySlotInfo("HANDSSLOT"); WoWAPITesting now matches slot names ignoring case (inbox 768985d0), and the harness pin moved to48872defor it, so the spec has no stand-in. Not seen in game. - Automatic chat messages blocked on WoW Forever again, by a different route (operator, 2026-11-13: 8x
ADDON_ACTION_BLOCKED"UNKNOWN()" in one session; the client is confirmed by the stack being in_forever_'s BugGrabber saved variables). The stack ran LibGuildRosterOnGuildRosterUpdate->ReadRosterLevels-> GratzonMemberLevelChanged->fireGuildLevelUp->tryFire->sendToChannels->addon.Chat.Send->C_ChatInfo.SendChatMessage(Compat.lua:443), so the v0.13.0InChatMessagingLockdowncheck had answered false. Which restriction was in force is not known.addon.ChatinCompat.luanow closes every route found in the Forever API docs, rather than guessing one:MessagingLockedalso asksC_RestrictedActions.IsAddOnRestrictionActiveabout five restriction types (Encounter 1 to Chat 5,RestrictedActionsConstantsDocumentation.lua; Combat 0 is left out because it would silence every Smack line, which is spoken in combat, and Gratz already holds its lines in combat);Sendrefuses a message or target that is a secret value (SendChatMessagetakes secret arguments only from untainted code,SecretArguments = "AllowedWhenUntainted",ChatInfoDocumentation.lua:564-568); and a frame watchesADDON_ACTION_BLOCKED, which the client fires synchronously inside the blocked call, so a send it blocks anyway is seen at once. Then the operator saw it 9x more, through all of those checks (stack at the newpcallinCompat.lua, trigger a LibGuildRosterOnLevelComm). Two triggers, neither player input, and no restriction active: on Forever the client refuses automatic GUILD chat outright. That it wants a hardware event is inferred from this, not documented (the docs say onlyHasRestrictions); the 12.0 rule Compat already records for SAY / YELL / EMOTE now evidently covers GUILD on Forever too. So on Forever (gameVersion.isForever) every automatic line is held and sent on the player's next key press (addon.Chat.NeedsKeyPress,FlushKeyHeld), through a hidden frame that passes every key on to the game, armed out of combat like the GreenWall relay frame; on any other client the first blocked send switches that on for the session and says so once. Held lines wait through a key press while chat is locked, are dropped and named afterRELAY_MAX_AGE(2 minutes) or past 10 waiting, and a block even on the key press is said once with every waiting line named, not retried.Sendreturns true for a held line, so Gratz counts the fire as it does a combat-queued one. Only Gratz and Smack calladdon.Chat.Send; a Smack line sent inside the player's own key press (isHardware, a new fifth argument) goes out at once. NOT VERIFIED IN A CLIENT: that an insecureOnKeyDowncounts as the input Forever wants for a GUILD line. The call itself is wrapped inpcallso the in-progress flag always clears.C_RestrictedActionsadded to.luacheckrc,.luarc.jsonand.vscode/settings.json. Seven new specs inTests/gratz_spec.lua(the world now exposes Compat's second and third load-time frames asblockFrameandkeyFrame). Not seen in game. - The Forever chat fix, refined after Peer Review (thread 58a37622). The block watcher in
Compat.lualistened forADDON_ACTION_BLOCKEDonly;ADDON_ACTION_FORBIDDENis a separate SynchronousEvent with the same payload (RestrictedActionsDocumentation.lua, Forever tree), so a send the client refused as forbidden was counted as sent, bridged to GreenWall, and never switched on key-press mode. Both are watched now. The payload's first field, which the docs nameisTainted, is still read as the blamed addon's name (what the operator's stacks show); a refusal blamed on another addon's taint is missed, and that is accepted in the comment. Off Forever, the first block switched every chat type to the key press for the session (onekeyModeflag); it is now per chat type (keyMode[chanType],NeedsKeyPress(chanType)), so one refused EMOTE no longer makes GUILD lines wait. Forever keeps its blanket rule. And a guild line sent inside a key press (flushed from the key-press queue, or a Smack line withisHardware) no longer waits for a second key press before an older GreenWall relays it:bridgeThroughGreenWalltakesinKeyPressand flushes the relay at once. The two hold-for-key queues (the GreenWall relay's and the chat one) stay separate. Three new specs inTests/gratz_spec.lua. Not seen in game. - Quest DB records quest rewards (QuestDB contract, inbox thread 8c13395a: QuestDB's
GetRewardsanswered nil for all 775 harvested Forever quests because the walk record had no reward fields).Modules/QuestDB/QuestDB.luaappends three optional fields toquests[id]once they are seen: rewardItems and rewardChoices asitemId:countentries joined by\30(bareitemIdfor one,QuestDB.RewardEntry), and rewardMoney in copper (empty for none); a record without them is the original seven fields exactly. They are read when the game shows the quest's offer or hand-in page (QUEST_DETAIL,QUEST_COMPLETE), from the same calls the Forever Vanilla quest frame uses there (GetNumQuestRewards,GetNumQuestChoices,GetQuestItemInfo,GetQuestItemLink,GetRewardMoney; wow-ui-source-foreverBlizzard_UIPanels_Game/Vanilla/QuestInfo.lua:479-482, :557-558), because nothing in that UI reads a reward by quest id for a quest not in the log, so the walk itself cannot. Only with Developer Tools on, only on Forever, and only for ids the walk wants (LibQuestDB "uncovered"). Rewards are kept inquestDB.rewards[id]as well, so rewards seen before the walk reaches a quest are put on the record it writes, and Purge clears them. The item count is read withpcalldirectly: the module'scallhelper keeps only the first return value. Four new specs inTests/questdb_spec.lua. Not seen in game. - Quest DB records a quest's description, objective text and completion text (QuestDB contract, inbox thread de72ae36: Questbook showed "The full quest description is not available yet" for Forever's own quests, such as Friend of the Library, 78150). The client gives these only while the quest window shows the quest, through the calls Forever's own quest frame uses (
GetQuestText,GetObjectiveText,GetRewardText; wow-ui-source-foreverBlizzard_UIPanels_Game/Vanilla/QuestInfo.lua:162, :333, :349).QuestDB:_onQuestShown(event)now also keepsquestDB.texts[id] = "details\31objectiveText\31rewardText": the offer page (QUEST_DETAIL) fills the first two, the hand-in page (QUEST_COMPLETE) the third, each keeping what the other saved. They are appended to the record afterrewardMoney(QuestDB._extras, which writes three empty reward fields in front when no rewards were seen), only once some text was seen, so a record without them reads as before; a secret text is stored empty, and Purge clears them. Separators are stripped as for every other field. Same gates as rewards: Forever, Developer Tools on, a quest the window capture wants. That gate was the walk's own (_wants, LibQuestDB "uncovered"), and it was wrong for the window: once QuestDB ships a structure row for every quest the walk found,GetCoveragecalls them "shipped", so the reward and text capture never fired for the very quests it exists for (QuestDB thread e58a62d1: the operator opened Rare Books 91777 and Friend of the Library 78150 and nothing was saved).QuestDB:_wantsWindowalso takes a quest whose shipped row has an unknown required level, the "?" only the walk writes (GetQuestHeader(id).requiredLevel == nil, which reads only the shipped row and ignores Questie;LibQuestDB-1.0.lua:481-489, :1583-1605). The walk keeps_wants. One new spec. The three calls added to.luacheckrcand.luarc.json. Four new specs inTests/questdb_spec.lua. Not seen in game. - Quest DB records WoW Forever's NPCs (QuestDB contract, inbox thread 4144495b, for Questbook's trade-hub request; operator, 2026-11-14: "we need to add the faction trade representatives to our data so we can do all the things we need to with them in the addon"). QuestDB's Forever world data is the Classic dump, so Forever's own NPCs are in no offline source.
Modules/QuestDB/QuestDB.lua_onUnitSeenruns onPLAYER_TARGET_CHANGED(unittarget) andUPDATE_MOUSEOVER_UNIT(mouseover), both in the Forever tree (UnitDocumentation.lua:3974, :4917), only on Forever with Developer Tools on. It keepsquestDB.npcs[id] = "name\31subname\31unitFaction\31reactions\31positions": the id from aCreature-GUID (QuestDB.NPCFromGUID; players, pets and other kinds skipped), the subname from tooltip line 2 (C_TooltipInfo.GetUnit) unless that line starts withLEVEL,UnitFactionGroup(unit), reactions asplayerFaction:UnitReaction(unit, "player")per faction that has seen it (a reaction toward a faction the player is not in cannot be read, so each faction's comes from its own characters), and positions asmapID:x:yin map percent to one decimal, one per whole-percent cell (mapID::with no position, as inside an instance). The position is the player's (C_Map.GetBestMapForUnit/GetPlayerMapPosition), since the client has no call for an NPC's own position. A secret GUID or name records nothing.GetNPCs,NPCCountandClearNPCsadded; Purge leaves NPCs alone. The Quest DB tab shows the NPC count, gains a Clear NPCs button (with confirmation) and says where the two hubs are.UnitReactionandLEVELadded to.luacheckrcand.luarc.json. Seven new specs inTests/questdb_spec.lua. Seen in game 2026-11-14 on the Alliance side: the NPCs recorded include all seven Azeroth Commerce Authority profession vendors (256729-256735). That capture showed one of my defects: a position could be stored more than once, because the entry was rounded to one decimal while the duplicate check floored the raw value (40.98 stored as "41.0" in cell 41, checked as cell 40). The check now reads the cell from the stored text; one new spec, and one spec's same-cell sighting moved off the rounding edge. - Profession Capture records a vendor recipe's non-gold cost (ProfessionDB contract, inbox thread 69192294, for TOGProfessionMaster's Merchant's Favor recipe vendors on WoW Forever). A Favor-priced recipe was stored as price 0 with
extendedCost = trueand its real cost lost.ProfCapture.ReadCosts(i)inModules/ProfCapture/ProfCapture.luanow readsGetMerchantItemCostInfo(i)and eachGetMerchantItemCostItem(i, j)-> texture, value, link, currencyName (the signature the Forever merchant frame uses,Blizzard_UIPanels_Game/Vanilla/MerchantFrame.lua:380, :596) intor.costs = { { value, currencyName } or { value, link } }for every recipe with an extended cost; a secret part is left out.SCHEMAis now 2. It had never been saved anywhere, so the parser could not read it:stampnow writesschemaonto every trainer, vendor and known-recipes record, and a record without it is schema 1. Both calls added to.luacheckrcand.luarc.json. Three new specs inTests/profcapture_spec.lua. Not seen in game. - Quality of Life tab fixes from the operator's first in-game look (2026-11-14). Opening the window froze the game for about 10 seconds, and the tab's sub-tab group ran past the window's bottom edge. AceGUI lays a container out again on every
AddChild(AceGUI-3.0.lua:437-453), re-measuring every wrapping checkbox description, so filling a sub-tab one widget at a time cost a full re-layout per widget.Tab:DrawGroupinGUI/QoLTab.luanow pauses the scroll frame's layout while filling it and lays it out once. The innerTabGroupsets its own height to its content's after every layout (AceGUIContainer-TabGroup.lua:475-478). The first fix,SetAutoAdjustHeight(false), was wrong: under the window's Flow layout nothing else set that height, and the operator then saw the options squeezed into a sliver (2026-11-15: "the setting scroll size is super small now").Tab:Drawnow sets the window's tab container toFill, asGUI/LogsTab.lua:73does for its own sub-tabs, which pins the sub-tab group to the window, and the auto-height call is gone. Because that container is shared,MainWindow:DrawTabnow resets it toFlowbefore every tab draws: IdHitThat, Login Digest, Name Prefix, Range and Transports never set a layout, so after Logs (which has always setFill) they would have shown only their first widget. Where LibAceGUIWidgets hasLAGW-TabStrip(MINOR 40, feature-detected withAceGUI:GetWidgetVersion), the ten sub-tabs are now one scrolling row on that strip, over aSimpleGroup(Fill, auto-height off) that takes the rest of the window and holds the sub-tab's scroll list, instead of a TabGroup wrapping them onto two rows (operator, 2026-11-15: "why aren't you using the widgets library for all the elements"); the TabGroup stays for an older copy, since TOGTools requires only MINOR 34. Not seen in game. Neither has been seen in game; that these were the causes is inferred from the AceGUI code, not measured. Minimap button bag opened empty with MinimapButtonButton loaded: it gathers the same LibDBIcon buttons and replaces each button'sSetParentwith a no-op (MinimapButtonButton/Logic/Main.lua:87, :107), so the bag'sSetParentdid nothing.QoL.ButtonCollector()now makes the bag stand down (no buttons taken, the minimap's click left alone, its own button hidden) and says why once per session, naming the collector. It also stands down for Leatrix Plus's own Minimap button bag (Enhance minimap): that hides every LibDBIcon button and re-hides it on show unless its own frame is open (Leatrix_Plus.lua:10224-10225) and replaces the minimap'sOnMouseUp, so this bag would show nothing. Leatrix is detected by its publishedLeaPlusGlobalMinimapCombinedButtonFrame, or byLeaPlusDB.MinimapModderand.MinimapButtonBagboth "On" (Leatrix writesLeaPlusDBback only at logout,Leatrix_Plus.lua:20165, so it holds this session's settings). Leatrix's "Hide addon buttons" and SexyMap do not stand it down: both only fade or place the buttons (LibDBIcon:ShowOnEnter;SexyMap/Buttons_Classic.lua:407-438) and never take them; two new specs. SexyMap's default "hover" fades addon buttons to alpha 0 once the mouse leaves the minimap (Buttons_Classic.lua:308-329), solayoutBagnow sets each shown button to alpha 1, or a bag opened from its own button could open full of invisible buttons (inferred from SexyMap's code, not seen in game). Square minimap switched off stayed square until a/reload: it now setsTextures\MinimapMask, the engine's default circular mask per wowprogramming'sMinimap:SetMaskTexturepage (the Classic and Retail XML set no mask), then calls Forever'sMinimapCluster:UpdateMinimapConfigwhere it exists (Camelot/Minimap.lua:31-57), which applies Forever's own mask. The spec's stand-in for that call now shows the underlay only while the map rotates, as the real one does. One new spec, three changed and the spec's Forever stand-in corrected inTests/qol_spec.lua. Not seen in game. - Minimap button bag: its own movable button (Discord request, inbox thread 2df31e79: "an option/mode for this feature where said minimap buttons are collected and added to a moveable/lockable button that expands"). Two ticks under Minimap button bag (
extras.minimapButtonBaginGUI/QoLTab.lua): Open from a button (db.global.qol.bagButton) shows a 28-pixel bag button, below and left of the minimap until dragged, whose click opens the bag beside it, anchored away from the nearest screen edges (QoL.BagHandleAnchor); Lock the button (bagButtonLocked) stops the drag. The position is saved inbagButtonPos.QoL.UpdateBagHandleshows or hides it and runs from the feature's apply on both switches. Opened from the button, the bag stays open while the mouse is over the bag or the button, not the minimap (QoL.BagTick); the minimap right-click still opens it beside the minimap. RawCreateFrame("Button"): a free-standing screen button AceGUI has no widget for. Only LibDBIcon buttons are gathered, as before; MinimapButtonButton also collects buttons that do not use it, which this does not. Its art is the backpack button's ownInterface\Buttons\Button-Backpack-Up, which the Classic Era bag bar sets (Blizzard_MainMenuBarBagButtons/Classic/MainMenuBarBagButtons.xml:76);INV_Misc_Bag_08was the first choice, but only the Retail and Forever sources use it. Three new specs inTests/qol_spec.lua; the spec's fake frame gains the draggable-button calls. Not seen in game. - Minimap button bag: the player sets the button order (Discord request, inbox thread 7e957e1a: "re-order the addon buttons within the button collection"). The bag was always alphabetical;
QoL.BagOrder(names, saved)now puts the player's saved order (db.global.qol.bagOrder) first, for the names still installed, then every other button alphabetically, so a new addon's button joins the alphabetical tail. Under the feature, "Order in the bag" lists every LibDBIcon button (QoL.BagButtonNames, hidden ones included so their place survives) with Up and Down buttons (QoL.MoveBagButton, which saves the whole list and re-lays an open bag at once); the rows are re-labelled in place rather than redrawing the tab. Two new specs inTests/qol_spec.lua. Not seen in game. - Mute game sounds: a Guns group (Discord request, inbox thread 83341f87: "please add player gun shooting sounds to Mute Game Sounds").
tools/build-mute-sounds.pygains agunsgroup:sound/item/weapons/(gun/)?gun(fire|load)\d+.oggandsound/item/weapons/gun/projectile_bullet_impact\d+.ogg, 17 files, rebuilt intoModules/QoL/MuteSoundData.lua.MuteSoundFilemutes a file for every source, so the description says it mutes every gun heard, not only the player's. Which of these files Classic Era's Shoot / Auto Shot plays was not checked in game. Also fixed in the script: the desk tool passes"<regex> <limit>"as one argument, so any search given a limit matched nothing; a trailing number is now split off. - Transports and Range, fixed after Peer Review (thread d9413d43). SandPacker's Pins file registers the standard name
HereBeDragons-Pins-2.0(MINOR 15) bound to the Questie fork, and the standard world-map pin poolHereBeDragonsPinsTemplate. Other Era addons on this PC ship the standard name at MINOR 14 (NauticusClassic), 15 (NovaWorldBuffs, NovaRaidCompanion) and 17 (TomTom, GatherMate2, Guidelime, Tamed), so whichever loaded won: with a 17 present our positions came from the fork and our pins from standard HBD, with only NauticusClassic's 14 its pins ran on the fork, and two Pins copies sharing one pool name remove each other's world-map pins (RemoveAllPinsByTemplate).tools/build-transports.pynow renames both into the fork's namespace (HereBeDragonsQuestie-Pins-2.0,HereBeDragonsQuestiePinsTemplate;PINS_RENAMES, refusing if the expected counts are not found), the vendored file carries the rename, andTransports.luafetches the renamed library. The operator's choice of SandPacker's copy stands. The script's docstring no longer says the copy carries a BSD header: neither file has a licence header and SandPacker ships no licence file. Transports now stays off when NauticusClassic is loaded (T.NauticusLoaded: its loaded flag or its global), since both would draw an icon per ship and announce twice; the tab says so. Sharing sends toPARTYin a party andRAIDonly in a raid (IsInRaid), and acceptsPARTYas it is. The own-echo filter compares the sender throughAmbiguate, so a "Name-Realm" sender is still recognised (it was harmless anyway, as the comment now says). Map pins are moved only when the ship has: the minimap pin is no longer removed before each add (an add updates a placed icon in place), and the world-map pin, which an add re-acquires, is removed and added only on a change. Range applies its text size and lock look only when they change, not as aSetFontper readout per tick. Two new specs inTests/transports_spec.lua. Not seen in game. CHANGELOG.mdarchived under the release-body ceiling. With the two entries above it measured 125,098 characters, over GitHub's 125,000 release-body limit. The two oldest sections, v0.9.2 and v0.9.1, moved whole to the top ofCHANGELOG_ARCHIVE.md, ahead of v0.9.0; the trailing pointer now says "v0.1.0 to v0.9.2".CHANGELOG.mdarchived again. With the LibRangeCheck entry it measured 132,008 characters. v0.9.4 and v0.9.3 moved whole to the top ofCHANGELOG_ARCHIVE.md, leaving 114,853; the pointer now says "v0.1.0 to v0.9.4". Then again at 119,822: v0.9.6 and v0.9.5 moved, leaving 105,104; the pointer now says "v0.1.0 to v0.9.6". And at release prep, at 120,797: v0.10.0 moved; the pointer now says "v0.1.0 to v0.10.0".- The rest of the Leatrix Plus wish list, without waiting for more votes (operator, 2026-11-13: "lets go ahead and just do the rest of the things on the list, and not wait for community feedback"). The 58 numbers v0.14.0 did not build are built on the Quality of Life tab, in new feature files that register into the same registry:
Modules/QoL/Automation.lua,Social.lua,Chat.lua,Frames.luaandSystem.lua, each listed in all seven TOCs afterQoL.lua. New sub-tabs Automation, Social, Chat and Frames & Text join the five from v0.14.0.QoL.Registernow takes agroupfield, so a feature file places its own features (the spec still asserts every registered key names a real sub-tab). Shared helpers moved onto the module for those files:QoL.Listener(key, events, handler)(one event frame per feature, created on first switch-on, the handler running only while the feature is on),QoL.Say,QoL.IsSecret,QoL.CFG. A feature's extra settings can now come fromQoL.Extras[key], set by aGUI/QoL*Extras.luafile, whichGUI/QoLTab.luafalls back to.- Automation (#1, #3-#7),
Modules/QoL/Automation.lua. Holding Shift as the window opens skips each window-driven one. Nothing in the file waits on a timer: every game call runs from the event itself. Automate quests (#1):QUEST_DETAILaccepts (AcknowledgeAutoAcceptQuestfor an auto-accept quest; aQuestFlagsPVPquest is left alone because the game asks to confirm it,Classic/QuestFrame.lua:816-826);QUEST_PROGRESShands in a completable quest unlessGetQuestMoneyToGetis above 0;QUEST_COMPLETEtakes the reward only with no choice or one (QoL.PickQuestReward, the same rule as the game's own button,Classic/QuestFrame.lua:162-180); a gossip window opens a finished quest first, then the first non-grey available one (QoL.PickGossipQuestoverC_GossipInfo.GetActiveQuests/GetAvailableQuests); a quest greeting does the same by index. Quests whose objective is collecting items are handed in: those items are the objective, not a cost. Accept summon (#3):CONFIRM_SUMMONout of combat confirms at once. Leatrix waits ten seconds so the player can decline; that wait would need a timer calling the game, which this addon does not do, so the description says it accepts at once. Accept resurrection (#4):RESURRECT_REQUESTaccepts unless the caster is in combat, and hides all three resurrect popups Classic Era defines (GameDialogDefs.lua:1036,1072,1108). Release in PvP (#5):PLAYER_DEADin apvpinstance callsRepopMeunlessHasSoulstoneorC_DeathInfo.GetSelfResurrectOptionsoffers a self-res. Sell junk (#6):MERCHANT_SHOWsells every quality-0 slot withouthasNoValue(C_Container where present, the old globals otherwise) and says the total fromGetItemInfo's sell price; a vendor can refuse part of a burst, so eachBAG_UPDATE_DELAYEDwhile the window stays open sweeps again, up to six sweeps, stopping at the first sweep that finds nothing or atMERCHANT_CLOSED. Repair (#7):CanMerchantRepairand a cost above 0, guild funds first (RepairAllItems(true)) when the new "Use guild funds first" tick is on (default on) andCanGuildBankRepairsays yes, thenRepairAllItems()for anything left. The game calls the file reads were added to.luacheckrcand.luarc.json. New specTests/qol_automation_spec.lua. Not seen in game. Fixed during the build (self-audit 7589454a): a quest greeting window took the first available quest even when grey, unlike the gossip path; it now skips anyGetAvailableQuestInfo(i)flags as trivial (TBC onwards,Classic/QuestFrame.lua:377-379; Era has no flag). And with Block shared quests on, Automate quests leaves a non-friend's shared quest alone, since both listen toQUEST_DETAILand accepting first would win. Fixed after Peer Review (thread 7589454a, findings A and B): the quest page accepted grey quests that reach it without a list (from an item, or shared by a group member);QoL.DetailQuestIsTrivialnow checksC_QuestLog.IsQuestTrivial(GetQuestID())where the client has it (only the Retail and Forever trees do), and the description now says that on Classic a grey quest from an item or a group member is still accepted. And the sell-junk total counted every slot it tried, so a refused sale was counted even if the window closed before it sold;SellJunkPassnow returns the slots it tried, eachBAG_UPDATE_DELAYEDcounts only those no longer holding that item, and the total is said once when the sweeping ends (a sweep finds nothing, the six sweeps run out, orMERCHANT_CLOSED). The QoL specs now put back every WoW global they stub once their file has run (file-scopeteardowninqol_automation_spec.luaandqol_social_spec.lua;qol_frames_spec.luaalready restored per test andqol_system_spec.luaper file), so a later spec's feature detection cannot find a leftover stub. The Quality of Life tab can now be resized (operator, 2026-11-13: "the window resizer is missing on the new QoL tab"):GUI/QoLTab.luadeclaredWINDOW_SIZE = { ..., locked = true }, soMainWindow:ApplyTabSizeturned the grip off; with nine sub-tabs that was too small. It now declaresminWidth = 560, minHeight = 420, and its content already fills the window (a FillTabGroupover full-sizeScrollFrames with relative widths). - Social (#8-#16),
Modules/QoL/Social.luaandGUI/QoLSocialExtras.lua. One rule decides who is a friend,QoL.IsFriendly(name, guid): the friends list and Battle.net friends' WoW characters always count; guild members only with Guild members count as friends (#15) on, character-community members only with Community members count as friends (#16) on (C_Club); both GUIDs known means the GUIDs decide, so a same-named player elsewhere is not a friend. A secret name or GUID skips the event, blocking and accepting nothing. Block duels (#8,DUEL_REQUESTED->CancelDuel), Block party invites (#9,PARTY_INVITE_REQUEST->DeclineGroup), Block requested invites (#10,GROUP_INVITE_CONFIRMATION->RespondToInviteConfirmation(..., false)for join requests and suggestions; queue warnings left alone), Block friend requests (#11,BN_FRIEND_INVITE_ADDED;C_BattleNet.GetFriendInviteInfoon Retail/Forever,BNGetFriendInviteInfoon Classic, and any waiting when switched on), Block shared quests (#12,QoL.ShouldDeclineSharedQuestonQUEST_DETAILfrom a grouped player), Party from friends (#13,AcceptGroupunless queued for a battleground (GetBattlefieldStatus"confirm"/"queued") or listed in the group finder), Invite from whispers (#14, keyword default "inv" and an optional friends-only tick, invites withC_PartyInfo.InviteUnitor a Battle.net friend's game account, only when ungrouped or leading/assisting and not queued). Not built: Leatrix's clicks on Retail's party-sync dialog, and a friends-list refresh before each check (it is a server request). New specTests/qol_social_spec.lua. Not seen in game. - Chat (#17-#29),
Modules/QoL/Chat.luaandGUI/QoLChatExtras.lua. Each per-window option also reaches whisper windows opened later through onehooksecurefunc("FCF_OpenTemporaryWindow"). Use easy resizing (#17: with the window locked, drag a tab to size it from the top, saved withFCF_SavePositionAndDimensions), Hide the combat log tab (#18,ChatFrame2Tabscaled to 0.01 and mouse off while docked, re-checked onUPDATE_CHAT_WINDOWS), Hide chat buttons (#19) and Hide social button (#20,QuickJoinToastButton/FriendsMicroButton), both reparented into a hidden frame and put back on switch-off, Unclamp chat frame (#21, clamp insets 0, kept throughFloatingChatFrame_UpdateBackgroundAnchors), Move editbox to top (#22, anchors saved and restored), More chat font sizes (#23, sizes 22-40 added to the tab menu throughMenu.ModifyMenu("MENU_FCF_TAB")), Disable sticky chat (#24,ChatTypeInfosticky cleared for whispers and channels; switching it off asks for a/reload, because writing a value back taints), Use arrow keys in chat (#25,SetAltArrowKeyMode(false)), Disable chat fade (#26,SetFading(false)), Universal group color (#27, party colours copied to raid and instance chat withChangeChatColor; the old colours kept indb.global.qol.groupColourPrior), Recent chat window (#28, Ctrl-click a tab for its last 200 lines in a copyable AceGUI window; secret lines left out and counted), Restore chat messages (#29, the newest 100 lines per window saved onPLAYER_LOGOUTtodb.global.qol.chatRestore, restored for the same character within 15 seconds, secret lines never stored). New specTests/qol_chat_spec.lua. Not seen in game; Wrath and Cata have no source tree here, so their chat frames were not checked. - Frames & Text (#30-#34, #40-#42, #44, #45, #51-#58),
Modules/QoL/Frames.luaandGUI/QoLFramesExtras.lua. Each option is offered only where its frame exists and puts back exactly what it changed. Hide error messages (the same switch as the game's/uierrorsoff; inventory full, quest log full and a few others still show), Hide portrait numbers, Hide zone text, Hide keybind text, Hide macro text, Show volume slider (under the minimap), Show durability status (a character-window button with per-slot durability, and a chat line on death or resurrection), Show raid button, Show borders (screen-edge bars, each side 0-300), Show player chain (elite/rare dragon on the player frame), Manage widget top (move and scale the top-centre widget), Class colored frames, Hide alerts, Hide clean-up buttons (Retail/Forever), Hide boss banner (Retail/Forever), Hide event toasts (Retail/Forever), Hide stance bar (out of combat only), Hide rested sleep animation. New specTests/qol_frames_spec.lua. Not seen in game; the placement of the slider, durability button, raid button and chain are worked out from the XML, not looked at. - System (#60, #61, #63, #64, #66, #68-#72, #74, #77),
Modules/QoL/System.luaandGUI/QoLSystemExtras.lua. CVar options keep the player's prior value and put it back (QoL.SetOwnedCVar/QoL.RestoreCVar), each CVar used only if the client has it. Disable screen effects (ffxDeath,ffxNether), Set weather density (slider 0-3), Silence rested emotes (Sound_EnableEmoteSoundsoff while resting), Keep audio synced (restarts the sound system when the output device changes, not during a cinematic), Mute mount sounds (airships, horsesteps, mechsteps, mechstriders, zeppelins: five newmount = truegroups intools/build-mute-sounds.py, rebuilt intoMuteSoundData.luawith 37/25/15/9/22 files, kept out of Mute game sounds inQoL.WantedSoundsand its tick list), Media player (a minimal one: the player lists sound file IDs and plays them in order, the next starting onSOUNDKIT_FINISHED; Leatrix's zone-music browser needs a shipped table of thousands of tracks and a timer, so it was not built), Remove raid restrictions (SetAllowLowLevelRaid), Disable loot warnings (bind-on-pickup and roll confirmations only; refund-window warnings are left because they give up a refund), Combat plates (enemy nameplates on in combat, off out of it), Easy item destroy (types the delete word into the box; the player still presses Delete), Remove transforms (cancels the ticked cosmetic transforms, waiting for combat to end), Set edit mode layout (where the client has Edit Mode). New specTests/qol_system_spec.lua. Not seen in game.
- Automation (#1, #3-#7),
Bug Fixes
- Show flight times never saved a flight (Discord support thread, 2026-11-16: the flight time "always" shows as unknown on routes already flown).
QoL.OnTakeTaxiNodeputQoL.FlightTickon the flight bar'sOnUpdate, and the bar is hidden until takeoff. The client runsOnUpdateonly for a shown frame, so the tick that notices the takeoff never ran: the bar never appeared andsaveFlightTimewas never reached. Every route stayed "not flown yet". The tick now runs on its own frame (flightDriverinModules/QoL/QoL.lua), which is always shown and has no size, and the bar only displays. The spec had calledFlightTickdirectly, which skips that client rule. Itstickhelper inTests/qol_spec.luanow firesOnUpdateonly on shown frames, as the client does, and the "stopped" checks assert that no frame is ticking. Not seen in game. - Flight times allow for WoW Forever's Frequent Flier talent (Discord support threads, 2026-11-16, with a screenshot of the in-game tooltip: Legacy Tree > Adventure > "Frequent Flier", "your flight path mount flies 20% faster"). With one account-wide time per route, characters with and without the talent overwrote each other's times. How much the talent shortens a flight is not known: 20% more speed gives 1/1.2 of the time, a 20% cut gives 0.8, and the poster and an earlier draft of this change disagreed. So nothing is scaled on that guess. The account keeps one measured time per route for each speed (
flightTimeswithout the talent, the newflightTimesFastwith it), and each landing replaces that speed's time. A character sees its own speed's time (QoL.EstimatedFlightTime, the map tooltip and the countdown bar). Only a route flown at the other speed alone is scaled, byQoL.FlightSpeedRatio: the average without/with ratio over routes measured both ways, orQoL.FREQUENT_FLIER_SPEED(1.2) until there is one.QoL.HasFrequentFlierreads the Legacy Talents tree (QoL.LEGACY_TREE_ID1118,Blizzard_GenericTraitUtil.lua:105-106in the forever tree) throughC_Traits(GetConfigIDByTreeID,GetTreeNodes,GetNodeInfo,GetEntryInfo,GetDefinitionInfo,SharedTraitsDocumentation.lua), true when an active node's talent is named "Frequent Flier" (itsoverrideName, else its spell's name). Any other client, a missing tree, or a read that raises counts as without, so Classic Era is unchanged and existing saved times stay as the without-talent times. Known limit: the talent is matched by its English name, so on another client language every flight counts as without it. Whether the character has it is read once at payment and kept for that flight. Two new specs inTests/qol_spec.lua. Not seen in game.
Improvements
- The minimap button bag's order list shows addon names, not their internal keys (operator, 2026-11-16, with a screenshot: "it should be Ultimate Tools not TOGTools anymore"). The list under Quality of Life > Minimap button bag printed each button's LibDBIcon registration key, which is often the addon's folder name.
QoL.BagButtonLabelinModules/QoL/QoL.luanow gives the button's own data-objectlabel(ours isaddon.DisplayName, "Ultimate Tools"), else theTitlefrom the TOC of an addon with that name (C_AddOns.GetAddOnMetadata,AddOnsDocumentation.lua:153in the classic_era tree, or the bare global; in apcall, since the key need not be an addon name), with colour codes and icons stripped, else the key.GUI/QoLTab.luashows that; the key is still what the saved order uses, so existing orders are kept. The label comes first because our own TOC Title carries a grey flavour suffix ("Ultimate Tools Vanilla"). One new spec inTests/qol_spec.lua. Not seen in game. - Minimap button bag is UT's own, with MinimapButtonButton loaded too (operator, 2026-11-15: "all it's doing is opening the minimapbutton button window, this needs to be a stand alone function in UT"). MinimapButtonButton pins each button it collects by putting do-nothing
ClearAllPoints,SetPoint,SetParentandSetScaleon the button itself (MinimapButtonButton/Logic/Main.lua:105-108).QoL.UnpinButtonclears those four fields (rawgetfirst, so a frame's real methods, which live in its metatable, are never touched), andtakeButtonreparents the button into the bag;layoutBagre-takes every button on each open, because MinimapButtonButton re-collects on a new icon and on its own button's click. MinimapButtonButton is therefore no longer aQoL.ButtonCollector; only Leatrix Plus's bag is, andQoL.OpenBagFromEdgeno longer calls MinimapButtonButton'sOnMouseDown, which also removes Peer Review's N1 on thread 8a0afde6 (an Alt-click there started a drag of MinimapButtonButton's button that never ended). The free-standing bag button now shows whatever holds the buttons and, like the edge button, opens Leatrix's bag while that is on (Peer Review F1). Known cost: MinimapButtonButton's own bag is left empty, and its layout can re-anchor a button inside an open UT bag if a button shows or hides while open (not seen in game). The spec's fake LibDBIcon button now keeps its methods in a metatable, as a real frame does; the MinimapButtonButton spec now asserts the take-back. - Peer Review of self-audit 784df1d3 (thread 1dd85edd). N1:
ProfCapture.ReadCostsread cost slots 1 toGetMerchantItemCostInfo's count, but Blizzard's merchant frame reads every slot up toMAX_ITEM_COST(3) when the count is above 0 and skips empty ones (classic_eraVanilla/MerchantFrame.lua:3, :386-389), so a cost in a later slot after an empty one would have been lost; it now reads the same slots. Sparse slots are not seen in game. One new spec, and the spec's cost stand-in answers nothing for an empty slot, as the client does. N3: switching Square minimap off put the previousGetMinimapShapeback even when another addon had replaced ours since; it now restores only while ours (a namedsquareShape) is still installed. One new spec. Finding 2:QoL.ROUND_MASKwasTextures\MinimapMask, which no Blizzard tree names; it is nowInterface\CharacterFrame\TempPortraitAlphaMask, the mask Leatrix Plus sets for its round minimap on Classic Era (Leatrix_Plus.lua:10487). Not seen in game. N2:tools/build-rangecheck.pynow records that the patch leaves the library'sisMidnightunused, kept so the copy differs from upstream by one line. - Minimap button bag opens from a button on the minimap's edge, not a right-click (operator, 2026-11-15, with a screenshot: "right click already does something on the minimap ... we need a button that stays on the minimap border, that can't get picked up by things like minimapbutton button that we can click on to open up the minimap buttons"). The bag no longer wraps the minimap's
OnMouseUp(bagClick,bagOriginalClick,bagHookedandQoL.BagClickremoved), so every minimap click is the game's again.ensureEdgeinModules/QoL/QoL.luamakes a 31-pixel button parented toMinimapwith LibDBIcon's Classic sizes and textures (tracking border, minimap background, the backpack icon), placed byQoL.PlaceBagEdgeatdb.global.qol.bagEdgeAngledegrees (defaultQoL.BAG_EDGE_ANGLE, 200) on the circle 5 pixels out, or on the box whenGetMinimapShape()is "SQUARE" (QoL.BagEdgeOffset); dragging it moves it round the edge (QoL.BagEdgeAngleTofrom the cursor), and Square minimap re-places it on switching. It is deliberately unnamed and not a LibDBIcon button: MinimapButtonButton collects LibDBIcon and LibMapButton buttons, whitelisted global names and NAMEDMinimapchildren matching its patterns (MinimapButtonButton/Logic/Main.lua:126-341, names fromGetName()inCore/Utils.lua:46-48), and Leatrix Plus and SexyMap only touch LibDBIcon buttons.QoL.UpdateBagEdgeshows it whenever the bag option is on, andQoL.OpenBagFromEdgeopens whichever bag holds the buttons: with MinimapButtonButton loaded it runs that addon's own button's leftOnMouseDown(MinimapButtonButtonButton,Logic/Main.lua:411, :420-426), with Leatrix Plus's bag on it runs the right-click Leatrix installed as the minimap'sOnMouseUp(Leatrix_Plus.lua:10248), otherwise it toggles this bag. The first build hid the edge button along with the bag whenever a collector was loaded, and the operator, who has MinimapButtonButton and Leatrix Plus enabled on every Forever character, saw no button at all (2026-11-15: "i don't see a button");QoL.BagTickkeeps a minimap-opened bag open while the mouse is over the edge button too.QoL.BagParts()returns the bag's frames for the spec, whose bag tests now open the bag from that button and assert the minimap's click handler is never replaced; two new specs, and the spec's fake frame gainsSetFrameLevel,CreateTextureand a size. Not seen in game. - One secret-value check,
addon.IsSecret(self-audit 784df1d3, sideways finding). The check was written six ways:Compat.luaisSecret,QoL.IsSecretand a secondisSecretinModules/QoL/QoL.lua,secretinModules/QuestDB/QuestDB.luaandModules/Range/Range.lua,isSecretinModules/ProfCapture/ProfCapture.lua, plus an inline test inModules/Diagnostics/Diagnostics.luaplain. Four of them passednilstraight toissecretvalue, whose argument the Forever docs markNilable = false(FrameScriptDocumentation.lua:295); whether anilactually throws there is not verified. The new dependency-freeSecret.luaholds the one check (feature-detected at call time,nilnever passed on), loads second in all seven TOCs, afterVersion.lua, and every one of those sites now calls it;QoL.IsSecretstays as an alias for the QoL feature files' 22 calls. Left alone on purpose: Gratz, Smack, GatheringLog, WhisperLog, NamePrefix and IdHitThat captureissecretvalueinto a local at load time, and their specs set the global before loading them. The 15 spec loaders that load an affected file now loadSecret.luafirst; new specTests/secret_spec.lua. - Both sub-tab branches of the Quality of Life tab now run offline (Peer Review thread 8a0afde6, F5: neither branch of
Tab:DrawinGUI/QoLTab.luahad executed outside the game). New specTests/qoltab_spec.lualoads the tab against a recording stub AceGUI, handed to the file through its environment'sLibStubso the real registry is untouched, once withLAGW-TabStripregistered and once without. It pins that the strip branch draws the starting sub-tab itself (the strip'sSelectTabis silent,LibAceGUIWidgets-1.0.lua:7489-7491), that the TabGroup branch draws it once throughOnGroupSelected(which itsSelectTabfires,AceGUIContainer-TabGroup.lua:341-342), that the remembered sub-tab reopens and an unknown one falls back to the first, that picking a sub-tab remembers it and redraws only that group, and the no-options line on both branches. Code unchanged; layout is not asserted.
[v0.14.0] (2026-11-13) - Quality of Life tab
New Features
A new Quality of Life tab, built from the Leatrix Plus wish list players voted on. The operator, 2026-11-12, after posting the numbered Leatrix Plus list to Discord: "lets post the list of leatrix plus in discord, and ask the users which ones they want, and we can start with that"; Vishiswaz voted for 19 (tally in
docs/LEATRIX_PARITY.md), and also asked for one tab that grows over time. New engineModules/QoL/QoL.lua(addon.modules["qol"]) with tabGUI/QoLTab.lua, listed in all seven TOCs after Retrack. Every feature is a registry entry{ key, label, desc, available(), apply(on) }.available()feature-detects what the client has, so a flavour sees only the options it can run;apply(false)runs only when the player switches a feature off, never at login, so an untouched feature writes nothing. Every feature starts off (db.global.qol.on). Game settings (CVars) change out of combat only: a write in combat waits forPLAYER_REGEN_ENABLED, and only the latest value is sent. The value a CVar held before the first change is kept indb.global.qol.priorand put back on switch-off, so a value the player had tuned survives; with no kept value the client's default is used (C_CVar.GetCVarDefault/GetCVarDefault), then a fixed fallback. Leatrix Plus was read for which calls and settings each feature needs; nothing was copied. The first features:Automate gossip (wish list #2): with Alt held, a gossip window with exactly one option and no quests picks that option (
C_GossipInfo.SelectOption, present in every flavour's tree). Leatrix also auto-picks a hand-kept list of Forever NPCs without Alt; that list was not taken.Resize quest text (#35): one slider, 10-30 (default 12), sets
QuestTitleFont(+3),QuestFont(+1) andQuestFontNormalSmall, the three fonts Leatrix sets in both its Era and Forever builds. They are shared font objects, so other windows using them change too; the tab says so. The sizes before the first change are captured and put back on switch-off.Disable drunken haze (#59): CVar
ffxGlowto 0, the switch Leatrix uses on Era and Forever. It is the full-screen glow, so that goes too; the description says so.Max camera zoom (#62):
cameraDistanceMaxZoomFactorto 4.0 on Classic Era and TBC and 2.6 elsewhere. Leatrix sets 4.0 in its Era and Anniversary builds and 2.6 in its Retail and Forever builds; Wrath, Cata and Mists Classic are not verified and get 2.6.Show train all button (#43): a "Train All" button on Blizzard's trainer window (
Blizzard_TrainerUI, load-on-demand, so the button is built on itsADDON_LOADEDor at once if it is already loaded) that callsBuyTrainerServicefor every available service; greyed out with nothing to learn, hidden on a training-points trainer, and its tooltip names the count and cost.GetTrainerServiceInfoputs the service type third on the Classic trees and second on the Mainline tree (Retail, Forever), read per flavour fromBlizzard_TrainerUI.luain each tree;QoL.ServiceTypebranches ongv.isClassic. On Classic the money display sits left of Train (Blizzard_TrainerUI.xml), so the button goes right of Exit there and left of Train on Retail and Forever.Show ready timer (#47): a countdown bar under each ready popup the client has. The battleground popup (
PVPReadyDialog, hooked throughPVPReadyDialog_Display; Classic Era has it) counts down fromGetBattlefieldPortExpiration(id)plus one second, as Leatrix does (its reason is not verified), and shows nothing when the game reports no time left. The dungeon-finder popup (LFGDungeonReadyPopup, not in Classic Era) has no expiry call, so it counts down from 40 seconds, Leatrix's figure, not verified against the game;LFG_PROPOSAL_SUCCEEDED/_FAILEDhide it. The bar redraws once a second, never below zero.Simple addon listing (#75): every flavour's
AddonList_Update(Blizzard_AddOnList, loaded at startup, the same code in all five trees) files addons under## Categoryheadings. With this on, ahooksecurefunconAddonList_Updatereplaces the scroll box's tree with one alphabetical list (QoL.SimpleAddonOrder): titles compared without colour codes, each addon keeping itsGroup(children sorted under their parent; a child whose parent is filtered out or missing listed on its own, as the game does), and the search box still filtering by title or group. The node data is the game's own{ addonIndex = i }, so the game's row code draws it unchanged. Switching it off callsAddonList_Updateto redraw the game's own list.Minimap size and Hide minimap addon buttons (#36, Enhance minimap, part one). Size:
MinimapCluster:SetScale, 50-200% in 5% steps (default 100%,db.global.qol.minimapScale), with the cluster's scale before the first change put back on switch-off. Frame changes go through the newQoL.OutOfCombat(key, fn), which runs at once out of combat and otherwise once atPLAYER_REGEN_ENABLED, latest call per key. Hide buttons: LibDBIcon-1.0's ownShowOnEnter(name, true), which fades a button in only while the mouse is over the minimap (libs/LibDBIcon-1.0.lua:431), for every button inGetButtonList()and, through theLibDBIcon_IconCreatedcallback, for buttons created later. A button an addon had already set to show on hover is left alone and is not switched back on afterwards. Only LibDBIcon buttons are affected; an addon that draws its own minimap button is not. Square minimap, on the Classic clients: their round frame art is the oneMinimapBordertexture (Blizzard_Minimap/Classic/Minimap.xml:216in the classic_era, classic_anniversary and classic trees), so the option hides it, setsMinimap:SetMaskTexture("Interface\\ChatFrame\\ChatFrameBackground"), adds a 2px blackBackdropTemplateborder, and replaces theGetMinimapShapeglobal with one answering"SQUARE"(the previous function is kept and put back), all out of combat. The game's own round mask path is in none of the trees, so switching off restores the art, border and shape function and asks for a/reloadrather than guessing the mask. LibDBIcon buttons are not moved at once when the shape changes; not verified when they next reposition. Square minimap on Retail and Forever: their round frame isMinimapCompassTextureplusMinimapCompassTextureUnderlay(Mainline/Minimap.xml:111-119;MinimapBorderis in the Classic trees only, so the art is chosen by which exists, and on ClassicMinimapCompassTextureis the separate rotation ring and is left alone). Each piece's shown state is kept when it is first hidden and put back on switch-off, so Forever's underlay, hidden unless the map rotates, is not shown by mistake. Forever'sMinimapCluster:UpdateMinimapConfig(Camelot/Minimap.lua:31, run on everyrotateMinimapchange) sets the round mask and art again, so ahooksecurefuncon it squares the map again while the option is on; Retail's is empty (Mainline/Minimap.lua:365). Leatrix's Forever build also re-anchors the map and moves the mail indicator into the corner; neither is done here, not verified how the indicators sit in game. Minimap header on hover, every client: the header pieces are found by parent key onMinimapClusterin every tree (BorderTop,ZoneTextButton, and the Classic-onlyToggleButton,Blizzard_Minimap/Classic/Minimap.xml:28-63; on Retail and ForeverBorderTopis a child ofZoneTextButton,Mainline/MinimapZoneTextButton.xml:22), set to alpha 0, and brought back to 1 fromOnEnterhooks onMinimapand each piece;OnLeavefades them only when the mouse is over none of them, so moving from the map onto the zone name keeps it shown. Alpha, notHide, so the zone button keeps its tooltip and click. The hooks are set once and do nothing while the option is off; switching off sets alpha 1. Leatrix only hides the zone bar outright. Minimap button bag: every LibDBIcon button (GetButtonList, and later ones through the sharedLibDBIcon_IconCreatedcallback, now onewatchNewIconsfor both button options) is reparented into oneBackdropTemplatepanel, hidden until a right-click on the minimap. The minimap's ownOnMouseUp(a ping on any button:Minimap_OnClickon Classic,MinimapMixin:OnClickon Retail and Forever) is wrapped, not replaced: only a right-click opens the bag, other clicks go to the game's handler, and switching off puts the game's handler back unless another addon replaced ours since. Opening lays the buttons out sorted by name, one column per four buttons up to ten (QoL.BagColumns, Leatrix's row widths), 30px cells, beside the minimap on the side nearer the screen centre; buttons an addon hid (db.hide) are not shown. It closes two seconds after the mouse leaves both the bag and the minimap (QoL.BagTick, anOnUpdatethat runs only while open). Switching off parents each button back toMinimapand calls LibDBIcon'sRefresh(name), which re-places it and keeps the addon's own hidden state. It and Hide minimap addon buttons each switch the other off, since a bag of buttons faded to nothing would show nothing. Leatrix's excluded-buttons list is not built.Tooltip size, Hide tooltip health bar and Hide unit tooltips in combat (#37, Enhance tooltip, part one). Size:
SetScaleonGameTooltip,ItemRefTooltip,ShoppingTooltip1/2andItemRefShoppingTooltip1/2where the client has them (Leatrix's list also covers Retail-only pet battle and garrison tooltips, left out), 50-200% in 5% steps (db.global.qol.tooltipScale), each tooltip's own scale put back on switch-off. Health bar:GameTooltipStatusBar:SetAlpha(0), so the game showing the bar again for the next unit changes nothing and switching off isSetAlpha(1); Leatrix blanks the texture instead. Combat: oneTooltipDataProcessor.AddTooltipPostCall(Enum.TooltipDataType.Unit, ...)(present in every flavour'sTooltipDataHandler.lua, Classic Era included) hidesGameTooltipwhileUnitAffectingCombat("player")and the mouse is over the world (WorldFrame:IsMouseMotionFocus(), the check both Leatrix builds use, withGetMouseFocus() == WorldFrameas the fallback); Shift shows it. Unit-frame tooltips are untouched. Show unit's target in tooltip shares that one post-call (registered once,hookUnitTooltip):QoL.TargetLinereads the unit fromGameTooltip:GetUnit()and adds a wrappedTarget: nameline fromUnitName(unit .. "target"), skipping a unit token or name thatissecretvalue(feature-detected; Classic has none) reports secret, an empty name or no target; a tooltip hidden for combat gets no line. Show guild rank in tooltip and Class-coloured names in tooltip share the same post-call:QoL.RankLineaddsGuild rank: <rank>fromGetGuildInfo(unit)(in every tree) for a player unit (UnitIsPlayer), since the game's tooltip already names the guild (Leatrix rebuilds the whole guild line instead);QoL.ClassColourcoloursGameTooltipTextLeft1fromRAID_CLASS_COLORSbyUnitClass's class file. Each skips a unit token, player flag, rank or class file thatissecretvaluereports secret. The post-call now reads the unit once and callsShowonly when it added a line. Tooltip follows the mouse: ahooksecurefunconGameTooltip_SetDefaultAnchor(SharedTooltipTemplates.luain every tree) re-owns the tooltip withANCHOR_CURSORwhen the mouse is over the 3D world, Leatrix's "cursor" anchor; tooltips over windows and unit frames keep the game's place. Hooked once, inert while off. Two sliders under it move the tooltip from the mouse, -100 to 100 each (db.global.qol.tipCursorX/tipCursorY,QoL.ClampTipOffset):ANCHOR_CURSORtakes no offsets, so any non-zero offset usesANCHOR_CURSOR_RIGHTwith them, the anchor the game's own code passes offsets with (Blizzard_SharedMapDataProviders/QuestDataProvider.lua:233in the classic_era tree); both at 0 keepsANCHOR_CURSOR. Which way a positive Y moves the tooltip is not verified in game. Reaction-coloured tooltip border shares the unit post-call:QoL.ReactionColourreadsFACTION_BAR_COLORS[UnitReaction(unit, "player")](the table the game's ownGameTooltip_UnitColoruses,Blizzard_GameTooltip/Classic/GameTooltip.lua:70) and sets it onGameTooltip.NineSlice:SetBorderColor(the tooltip's border,TooltipBackdropTemplateMixininSharedTooltipTemplates.lua); a reactionissecretvaluereports secret is skipped. AHookScript("OnTooltipCleared")(fired byGameTooltip,Classic/GameTooltipTemplate.xml:147) putsTOOLTIP_DEFAULT_COLORback the next time the tooltip is cleared, and only when this tinted it, so item and spell tooltips keep the usual border; switching off does the same at once. Leatrix's corner offsets are not built.UnitIsPlayer,RAID_CLASS_COLORSandGameTooltip_SetDefaultAnchorwere added to.luacheckrcand.luarc.json.Mute custom sounds (#67): a multi-line box (
db.global.qol.muteSounds) for sound file IDs with any notes beside them;QoL.ParseSoundIDstakes every run of digits as an ID, once each (Leatrix strips every non-digit instead, and caps IDs below 20,000,000 for a reason not verified, so no cap here).MuteSoundFile/UnmuteSoundFile(used by Leatrix's Era and Forever builds). A mute lasts only for the session, so the list is muted again at every login; an edit while on unmutes removed IDs and mutes new ones, leaving kept IDs alone. Switching off unmutes every ID on the list, even one another addon had also muted; whether the game counts mutes per addon is not verified. Leatrix's sound finder is not built.Mute game sounds (#65): Leatrix's nine groups (chimes, fizzle, interface, login, ready, trains, vaults, piccolo, yawns), one tick each under the switch (
db.global.qol.muteGroups), muted through the samemuteSetas the custom list so the two never unmute each other's IDs. The IDs are inModules/QoL/MuteSoundData.lua, generated bytools/build-mute-sounds.py(desk tooltogtools-mute-sounds) from the wowdev community listfile with path patterns of our own; Leatrix's tables were read only to learn which sound each group means, and its file IDs were looked up to find the paths. Trains covers every playable race's/trainsound in the listfile (51 files: the Classic races'choochoofiles and the later races'trainfiles); Login covers the login screen music and ambience of every expansion (52 files), where Leatrix's Forever build mutes the one Retail title track. Login is a logout group: muted onPLAYER_LOGOUTand unmuted when the addon enables after the next character is picked (IDs kept indb.global.qol.logoutMuted). No addon code runs on the character select screen, so a logout mute also mutes its music; the Login description tells players so (Peer Review 76f0f99d T3). Whether a mute set atPLAYER_LOGOUTholds on the login screen is not verified. The script gained alookupmode (the path behind a file ID) and now ignores a blank argument, which the desk tool passes whenargsis empty. Not seen in game.Enhance dressing room (#38), every client: eight
UIPanelButtonTemplatebuttons down the right ofDressUpFrame.DressUpModel(aDressUpModelin all three Classic trees,Blizzard_CharacterFrame/Classic/DressUpFrames.xml:148;Blizzard_CharacterFrameisLoadFirst, so it exists before any addon). Undress callsUndress(); Head, Shoulder, Back, Chest, Shirt, Tabard and Weapons callUndressSlotfor slots 1, 3, 15, 5, 4, 19 and 16-18, labelled with the game'sHEADSLOT-style strings where it has them. It is one way: the game's Reset callsDress()(DressUpFrames.xml:143), which dresses the model in your own gear. Retail and Forever dress a model scene instead (DressUpFrame.ModelScene,Mainline/DressUpFrames.luain the live and forever trees), so there the buttons sit on the scene and act onModelScene:GetPlayerActor(), read at each click as the game's own code does; the actor hasUndressandUndressSlot(FrameAPIModelSceneFrameActorDocumentationin the live tree). The animation buttons are Classic only, since the scene's actor has noHasAnimation. Below them,<and>step the model through its animations (QoL.StepDressAnimation, wrapping at both ends) with a label naming the animation's number. Leatrix steps a hand-picked list of IDs; this asks the model instead, callingHasAnimationfor every ID from 0 toQoL.DRESS_ANIM_MAX(1000) once each time the dressing room opens (anOnShowhook onDressUpFrameresets it), thenSetAnimation(both onCharacterModelBasein the classic_era API docs). The client gives no animation names, so the label shows the number; whether 1000 reaches a model's highest ID is not verified. Leatrix's item icons and per-slot toggle back are not built.Show druid power bar (#46), Vanilla and TBC clients only: a
StatusBaronPlayerFrame(BOTTOMLEFT98, 20, where Leatrix puts it) showing mana (UnitPower/UnitPowerMaxof power type 0) with acur / maxlabel, coloured fromPowerBarColor.MANA, shown while a druid's current power type is not mana and they have mana.PLAYER_ENTERING_WORLD,UNIT_DISPLAYPOWERandUNIT_MAXPOWER(player only) decide whether it shows; while shown anOnUpdatere-reads the value every 0.2 s so regeneration moves it. The Wrath, Cata and Mists clients load Blizzard's own bar (Blizzard_UnitFrame_Classic.toc:Wrath/AlternatePowerBar.xmlfor wrath and cata,Mists/for mists), so they are not offered it; Forever and Retail are not covered. Switching off hides it and unregisters the events. Leatrix's status-text option and player-chain border colours are not built.Show range colouring (#48): an action button's icon turns red (1, 0.3, 0.1, Leatrix's colour) while its target is out of range. The game's
ActionBarActionButtonMixin:UpdateUsablealready tints an icon blue for not enough mana and grey for unusable, andActionButton_UpdateRangeIndicatoronly reddens the hotkey (Blizzard_ActionBar/Shared/ActionButton.lua:679-692and1273, the same Shared file in every tree). So ahooksecurefunconActionButton_UpdateRangeIndicatorreddens the icon, ahooksecurefuncon each button'sUpdateUsable(every button fromActionBarButtonEventsFrame:ForEachFrame, and later ones through itsRegisterFrame) keeps it red when the game recolours it, and coming back into range, or switching off, gives the icon the colour the game'sUpdateUsablewould (C_ActionBar.IsUsableAction, elseIsUsableAction). A range answerissecretvaluereports secret is skipped. Leatrix instead switches off the game's button update loop and redraws the attack flash itself; that is not done here, and the pet bar is not covered.Show info cards (#49): a new
/ut card(and/togt card) command, which a player can type or put in a macro bound to any key, reads whatGameTooltipis showing and opens a small AceGUI window (QoL.OpenCardinGUI/QoLTab.lua) with the name, the ID and a Wowhead link selected for Ctrl+C.QoL.CardInfotriesGetItem(the ID from theitem:link), thenGetSpell, thenGetUnit(the NPC ID from aCreatureorVehicleGUID); players get no card, and any link, spell ID, unit token or GUID thatissecretvaluereports secret is skipped. Wowhead paths (QoL.WowheadBase) are the ones Leatrix Plus's own builds use, read from the installed Era, Anniversary, Forever and Retail copies:/classic/,/tbc/,/forever/and none. Wrath, Cata and Mists Classic are not offered it because their paths were not checked. Leatrix's own key setting, its quest, achievement and other card types, and its card layout are not built. The command was used rather than a key binding of its own so no key-capture settings were needed; the card now also has a key binding (#76, below).Show Wowhead links (#50): an
InputBoxTemplateedit box at the top ofQuestLogFramewith the Wowhead link for the selected quest (QoL.QuestURL:GetQuestLogTitle's eighth return, the quest ID, underQoL.WowheadBase), set from ahooksecurefunconSelectQuestLogEntry, which the quest log'sQuestLog_SetSelectioncalls (Vanilla/QuestLogFrame.lua:314). A header or an entry with no quest ID hides the box. Focus selects the text for Ctrl+C, typing puts the link back, Escape lets go. This is what Leatrix's Classic Era build does; its Forever build puts links on the world map and its "go to comments" setting are not built. Offered only where the client has the quest log calls and a known Wowhead path, so Classic Era and Anniversary.Enhance flight map (#39):
TaxiFrame:SetScale100-300% (default 150%,db.global.qol.taxiScale) and theTaxiButtonNflight point icons 16-32 (default 20,taxiIcon), applied from aTaxiFrameOnShowhook becauseTaxiFrame_OnShowcreates the icons, up toNUM_TAXI_BUTTONS(the sameBlizzard_UIPanels_GameTaxiFrame.luain every tree). The sliders apply at once while the map is open. The scale and each icon's size before the first change are put back on switch-off. Leatrix's Alt-drag to move the map re-anchors a panel the game lays out itself, and is not built.Show flight times (#73): Leatrix ships a hand-collected table of flight times; this learns them from the player's own flights instead, so nothing is copied and no table goes stale. A
hooksecurefunconTakeTaxiNodenotes the route (QoL.FlightRouteKey: theTaxiNodePositionof theCURRENTnode and of each hop's destination fromGetNumRoutes/TaxiGetNodeSlot, two decimals, joined start first, so a flight through a different stop is a different route) and the continent (QoL.FlightContinent:C_Map.GetBestMapForUnitwalked upparentMapIDto themapType2 map, because taxi positions are per continent map). AStatusBarnear the top of the screen then waits up to 10 s forUnitOnTaxi("player"), counts the flight down when the route has a saved time (bar filled by the share flown) or up with "(first flight)" when not, and on landing saves the whole seconds indb.global.qol.flightTimes[continent][route]; a flight under 5 s is not saved. AhooksecurefunconTaxiNodeOnButtonEnter(the Classic and SharedTaxiFrame.luain the classic_era tree) adds "Flight time: m:ss" or "not flown yet" to a reachable node's tooltip. Offered where the client has all the flight map calls, which includes Retail and Forever. Those clients usually open the newerFlightMapFrameinstead (Blizzard_Game/Mainline/EventImplementation.lua:735in the forever tree), whose pins build their tooltip inFlightMap_FlightPointPinMixin:OnMouseEnterand never callTaxiNodeOnButtonEnter(Peer Review, thread 76f0f99d). So the mixin'sOnMouseEnteris also hooked (QoL.OnFlightPinEnter, slot frompin.taxiNodeData.slotIndex), at once ifBlizzard_FlightMapis loaded or on itsADDON_LOADEDotherwise. The map copies each pin's method into itsOnEnterscript when the pin is first built (Blizzard_MapCanvas.lua:312), so pins built before the hook (the map opened before the option was switched on) get the line only after a/reload. Not seen in game on either client. The in-flight bar does not depend on the map, since the pins also callTakeTaxiNode. The time saved is the latest flight's, not an average; a/reloadmid-flight loses that flight. Switching off stops a flight in progress without saving.Key bindings (#76, Set keybindings): a new
Bindings.xmlputs three actions in the game's own Key Bindings menu under AddOns > Ultimate Tools, all unbound by default: open or close the window, the info card, and Diagnostics. Each<Binding>calls the new globalTOGTools_RunBinding(cmd)(SlashCommands.lua), which runsAce:SlashHandlerwith the matching slash command ("",card,diag), so a key and its command cannot do different things.BINDING_HEADER_TOGTOOLSand theBINDING_NAME_labels are set in the same file. Leatrix binds keys from its own panel withSetOverrideBindingClickon hidden buttons; the game's menu needs no key-capture settings of our own.<Binding>in an addon'sBindings.xmlis still read on Retail: Blizzard's ownBlizzard_PingUI/Bindings.xmlin the live tree uses it. New specTests/bindings_spec.luareadsBindings.xmland checks every binding has a label and runs a command the slash handler carries out. Leatrix's announce-rare and recent-chat actions are not built.
Tests/guildkey_spec.luanow also drives the guild filter dropdown on AceGUI's realDropdownfrom the installed Ace3: it installs the harness's widget layer (env.frames) and reloads AceGUI after it, because the real widget reads its template's named child frames, which the plain frame model does not create (WoWAPITesting's answer on inbox thread 62210fcd). The stand-in block that swappedAceGUI.Createfor a fake dropdown is gone: the real-widget block checks everything it did (operator approved the removal, test change 35f196cef027effc).The tab is split into sub-tabs (operator, 2026-11-13: "there is too much stuff. we should break it into a few tabs"): General, Minimap, Tooltips, Flights, and Sound & Camera, a nested AceGUI
TabGroupas the Logs tab uses.QoL.GROUPSorders them andQoL.GROUP_OFplaces each feature;QoL.FeatureGroups()returns only the sub-tabs with a feature this client can run, so a client missing every minimap option shows no Minimap sub-tab. The last sub-tab viewed is kept indb.global.qol.subTab. A spec asserts every registered feature (QoL.RegisteredKeys()) names a real sub-tab, so a new feature cannot fall onto General unnoticed.Fixed before release: every click on the tab jumped back to the top (operator, in game, 2026-11-13). Ticking a feature with its own settings rebuilt the whole tab to enable or grey them, and the new
ScrollFramestarted at offset 0. Each sub-tab'sScrollFramenow keeps its offset in a status table that survives the rebuild (SetStatusTable; AceGUI'sFixScrollre-appliesstatus.offsetafter layout), and a tick rebuilds only its own sub-tab. The fix is not seen in game yet.New spec
Tests/qol_spec.lua.C_GossipInfo,GetCVarDefault,C_CurrencyInfo,GetCoinTextureStringand the trainer calls and frames were added to.luacheckrcand.luarc.json. Not yet seen in game on any client. Locations:Modules/QoL/QoL.lua,GUI/QoLTab.lua,TOGTools.lua(defaults), the seven TOCs,Tests/qol_spec.lua.
Improvements
CHANGELOG.mdarchived under the release-body ceiling. At release prep it measured 122,335 characters, over the 120,000 working ceiling (GitHub refuses a release body over 125,000). The oldest section, v0.9.0, moved whole to the top ofCHANGELOG_ARCHIVE.md, ahead of v0.8.9; the trailing pointer now says "v0.1.0 to v0.9.0".- Gratz recovers achievements whose chat line was hidden in combat (Forever and Retail). The operator, 2026-11-12: "why can't we store it in memory until we're able to send it out?" The hidden line itself cannot be kept and sent: in chat messaging lockdown
CHAT_MSG_ACHIEVEMENT/CHAT_MSG_GUILD_ACHIEVEMENTarrive as secret values, which tainted code may only store or pass, andC_ChatInfo.SendChatMessagetakes secrets only from untainted code (SecretArguments = "AllowedWhenUntainted",ChatInfoDocumentation.luain the forever tree). So a skipped secret line now sets_newsMissed, and once combat ends (PLAYER_REGEN_ENABLED), a restriction lifts (ADDON_RESTRICTION_STATE_CHANGEDstate 0) or the player zones (PLAYER_ENTERING_WORLD), Gratz callsQueryGuildNews(); onGUILD_NEWS_UPDATEit readsGetNumGuildNews/C_GuildInfo.GetGuildNewsInfo(GuildInfoDocumentation.lua:60, no secret flag on the returns) and gratzes eachNEWS_PLAYER_ACHIEVEMENT(1) orNEWS_GUILD_ACHIEVEMENT(0) entry it had not listed before, through the samefireAchievement/fireGuildEntityAchievementas the chat path. Entry fields are read as Blizzard's own news UI reads them (Blizzard_FrameXML/GuildUtil.lua:56-108):newsDataIDis the achievement ID,whatTextthe plain name,whoTextthe member. News carries a date and no time, so the session's first read (requested 3 s after login) is a baseline that fires nothing, and an achievement already gratzed from a readable chat line (_achHandled, keyedname|idorguild|id) is never repeated. While still in combat or lockdown the read leaves everything unread for later, and a news field that is itself secret stops the read and keeps the miss. Only entries the player's own guild news filters let through are listed, so a player who filtered achievements out gets no recovery. Every news call is feature-detected, so Classic clients are unchanged. Forever hides most of its achievement UI (Blizzard_AchievementUI/Camelot), so whether its guild news lists member achievements at all is decided by the server; not verified in game. The party level-up retry also runs on a restriction lift and onPLAYER_ENTERING_WORLDnow, not only at combat end, since names inside an instance may stay secret after combat (not verified). Location:Modules/Gratz/Gratz.lua; specs inTests/gratz_spec.lua("achievement recovery from guild news").
[v0.13.0] (2026-11-12) - Diagnostics shows the full error detail BugSack shows
New Features
Diagnostics keeps the call stack of a blocked / forbidden action and shows it in Explain. A player on WoW Forever got
[ADDON_ACTION_BLOCKED] AddOn 'Ultimate Guild Roster' tried to call the protected function 'UNKNOWN()'.and asked "Why doesn't Ultimate Tools show the full text?" (Discord, via Ultimate Guild Roster, inbox threadb01704dd). Without the stack, UGR could not tell which of its code paths made the call._recordtakes a newstackargument, kept from the first sighting the waymsgis and capped at 3000 characters (MAX_STACK). A repeat fills it in only when the first sighting had none, which happens with an older !!TOGT._ingestRawpassesrec.stackthrough from !!TOGT v0.3.0 (worked on as v0.2.5, which was never released). The fallback capture takes it itself through a newactionStack():debugstack(3)in apcall, nothing stored on a throw, a non-string or a Retail secret value (issecretvaluefeature-detected). That is BugGrabber's fallback level for these events (!BugGrabber/BugGrabber.lua:288), anddebugstackexists in the Forever tree (Blizzard_ScriptErrors.lua:56). The tab'srowRawTextadds the stack under the rebuilt one-line message, so clicking the row shows it in the Explain window's multi-line box, where it can be selected and copied.Interpretstill takes the culprit from "AddOn '...'" first, so the stack does not change who Explain blames. Lua errors still keep only the one-line summary, and nothing keeps locals. The help text and the module header now say that. The row-menu Copy is still the one-line row. Seen in game on WoW Forever (operator, 2026-11-04: "confirmed it's all there"): a forbiddenInvite()showed the stack down to the calling addon's own code,fastguildinvite/FGI_APICompat.lua:45inside itsGuildInviteat line 27, under the companion's handler frame. That needed !!TOGT's_forever_replication fix first: the first test showed only the one-liner because Forever still had the old!TOGT.lua. Not yet seen: anUNKNOWN()block. Four new specs.debugstackwas added to.luacheckrcand.luarc.json. Locations:Modules/Diagnostics/Diagnostics.lua,GUI/DiagnosticsTab.lua,Tests/diagnostics_spec.lua,.luacheckrc,.luarc.json.Diagnostics keeps the full detail BugSack shows, for every error: the stack AND the locals, Lua errors included. The operator, 2026-11-04, after the Discord bot told the player that BugSack "records a little extra detail that Ultimate Tools won't": "why don't we have the full detail?". The entry above covered only blocked actions and only their stack. Now
_recordalso keepslocals, under the same first-sighting and fill-on-repeat rule. Stack and locals are each capped at 4000 characters (MAX_DETAIL, wasMAX_STACKat 3000)._ingestRawpassesrec.localsthrough. The fallback capture's newerrorDetail()takes both from Lua errors as well as from blocked actions. The level is Blizzard's ownGetErrorDataarithmetic (Blizzard_ScriptErrors.lua:48-64, the same in the Classic Era and Forever trees):GetCallstackHeight() - (GetErrorCallstackHeight() - 1)inside an error handler, and 3 outside one.debuglocals(level, true)skips functions and userdata as Blizzard's call does, and|k/|Kbecome plain letters (Blizzard'sRemoveKStrings). A defect in the first cut:actionStackwrappeddebugstackinpcall, which is a frame of its own, so every stack started one level too high, on the capture handler itself. That is the!TOGT.lua:155line at the top of the operator's in-game stack.errorDetailcallsdebugstack/debuglocalsdirectly. Explain now shows the message, then the stack, thenLocals:and the locals, in BugSack's layout (rowRawText, split from the newrowMessage). The interpreter already stops reading callers atLocals:. The tab help no longer says Lua errors keep only the summary. Specs: locals kept and capped, locals on a Lua error, fill on repeat, the fallback handler's level arithmetic and|Kstripping, and missing, empty or secret answers; the global swaps are put back after each spec.debuglocals,GetCallstackHeightandGetErrorCallstackHeightwere added to.luacheckrcand.luarc.json. A repeat by another call path is kept too. The operator compared one forbiddenInvite()row with BugSack in game on Forever (2026-11-04). Ultimate Tools held the first path, the addon's own Scan-tab button. BugSack (5x) held the latest, Blizzard's invite popupOnAccept. Both were real, and keeping only the first sighting hid the second._recordnow keeps every further distinct call path ine.more, each with its own locals and a count, told apart by stack only, since locals change on every call. For the session nothing is capped (operator, 2026-11-04: "we can leave all that data in the live version, it's just whats saved and lives through a /reload that changes"). Only the saved copy is trimmed, byTrimForSaveonPLAYER_LOGOUT. The first two paths keep their locals (FULL_PATHS), every later path keeps only its stack and count, and later stacks share a 16,000-character budget (SAVED_MORE_STACKS). Paths past that budget are counted indroppedPaths, and Explain says how many were not kept rather than hiding them. The newest path leads, as in BugSack (operator, 2026-11-04: "when the stack changes, we keep the 'newest' one like bugsack does ... so when we fix errors, the updated info is what shows"). Every path, the entry's own included, now carries a count and a last-seen time (stackCount/stackLast, andcount/laston each ofmore). The newDiagnostics.Paths(e)returns them newest first. Explain shows the newest path first ("Last seen from (Nx, last time):"), so its caller attribution follows the latest occurrence, and then the older ones ("Earlier, from ...").TrimForSavereorders the entry newest first before trimming, so the newest two paths keep their locals across a reload. An open Explain window now follows its row. The operator, in game on Forever: "i had to close/open the window to get the udpates". The table did redraw on every capture throughonUpdate, but Explain was filled once, from a snapshot of the row's text.openExplainnow takes the entry and keeps it as_explain.entry._recordcalls the newDiagnostics.onEntryChanged(e)on every occurrence, and when the window follows that entry it redraws from the entry as it now is. Redraws are coalesced to one per 0.25 s (C_Timer.After, internal: it repaints only our own window), so a burst of one error redraws once. Typing or pasting into the box drops the link, so the player's own text is never overwritten. One new spec pins the hook; the window itself needs the client. The table's own redraw (onUpdate) is coalesced the same way now, so an error storm with the tab open rebuilds the table once per 0.25 s rather than once per error. Self-audit 050586d2: Explain no longer reads callers from older paths (ErrorInterpreterstackAddonsstops at "Earlier, from"; new spec), and!!TOGT'searlydata_specputs back the eleven globals it stubs once its specs have run. This also covers fastguildinvite's report (inbox 124c584c): a row with a current count and Last seen showed only a stack from code since rewritten, and BugSack showed the path that had actually fired. An interim cut capped paths at three and dropped the rest silently. That cap was never measured; the operator asked why every path would not be listed, and the cap was replaced. Seen in game in that same comparison: locals on a forbidden action (the invited player's name, where BugSack's locals are BugGrabber's own handler variables). That row's stack still began on the old!TOGT.lua:155capture line, because it was first recorded before the fixed !!TOGT reached Forever, and the first sighting is kept. Not yet seen in game: the stack of a Lua error, the stack starting at the caller, and the extra paths. Locations:Modules/Diagnostics/Diagnostics.lua,GUI/DiagnosticsTab.lua,Tests/diagnostics_spec.lua,.luacheckrc,.luarc.json.
Improvements
CHANGELOG.mdarchived under the release-body ceiling. At release prep it measured 122,635 bytes, over the 120,000 working ceiling (GitHub refuses a release body over 125,000). The two oldest sections, v0.8.8 and v0.8.7, moved whole to the top ofCHANGELOG_ARCHIVE.md, ahead of v0.8.6, with the editing tools; the trailing pointer now says "v0.8.8 and earlier".- Every list's numbers now sit under their header. The operator, in game, 2026-11-04: "the error numbers don't line up with the column header like the other columns do".
RowList.TranslateColumnmapped a tab'sjustify = "RIGHT"onto the library'salign, which right-aligns the cells, but LibAceGUIWidgets left-justifies every header label so the sort arrow follows the text (LibAceGUIWidgets-RowList.lua:824-826). Thirteen columns in nine tabs (Diagnostics #, Addon Load Memory and Load Time, Gathering and Guild Bank Qty, Gratz and Smack Count, Smack %, Item DB and Quest DB ID, Quest DB Level and Objectives, Mail Money) had their figures at the far edge of a left-aligned header.justifyis now dropped, so every cell is left-aligned under its header.Tests/rowlist_spec.luapins it. Location:GUI/RowList.lua. wow-version-replication.ps1now replicates into WoW Forever's new_forever_install folder (operator, 2026-11-04: "add the new forever folder to your ps1 replication script, so the updates get to that folder")._forever_has its ownWow.exenext to the older_classic_beta_, which stays in$WowVersionswhile it is installed. The dry run lists_forever_\Interface\AddOns\TOGToolsas a target, and after a restart the watcher copied all seven TOCs there. Dev-only; the script does not ship. Location:wow-version-replication.ps1.
Bug Fixes
addon.AceGUIFrameScriptsleft a script on a pooled widget whenever that event had no script before. It saved each prior script insaved[evt]and restored by iteratingsaved, but a nil prior is no entry in a Lua table, so every event that started empty was never put back and the next owner of the recycled frame inherited our handler, the leak the helper exists to prevent. It reaches every Dropdown and EditBox tooltipGUI/UI.luaattaches. The event names are now kept in their own list and every one is restored. Found by the helper's first spec, written for the Compat coverage gap. Also:addon.GUI.MakeColumnHeaderdeleted, with no caller since the lists moved to LibAceGUIWidgets' RowList;GuildFilterDropdown's entry / order / selection logic split intoaddon.GUI.GuildFilterChoices; Compat.lua is now at 247/247 lines covered (was 212/259). The dropdown's spec swapsAceGUI.Createfor a stand-in, because the harness cannot yet build AceGUI's real Dropdown (it readsUIDropDownMenuTemplate's named child textures); requested from WoWAPITesting (inbox 62210fcd). Locations:Compat.lua,Tests/guildkey_spec.lua.- A Gratz guild ding was blocked on WoW Forever with
ADDON_ACTION_BLOCKED"UNKNOWN()". Seen by the operator 2026-11-12: the stack ran LibGuildRosterOnLevelComm-> GratzonMemberLevelChanged->tryFire->sendToChannels->queuedOrSend->addon.Chat.Send(Compat.lua:421) ->C_ChatInfo.SendChatMessage. On Forever / Retail the client refuses addon chat whileC_ChatInfo.InChatMessagingLockdown()is true ("API security restrictions regarding chat messaging are in effect",ChatInfoDocumentation.lua; the wiki names encounters, keystone runs, PvP matches and dungeon / raid maps), and nothing checked it. Newaddon.Chat.MessagingLocked()(feature-detected, pcall'd).addon.Chat.Sendno longer attempts a send while locked and returns false, so Smack and every other caller stop provoking the block and do not count a fire. Gratz'squeuedOrSendholds the line in the combat queue while locked,flushCombatQueuekeeps holding while still locked (combat ending inside a dungeon), and the newADDON_RESTRICTION_STATE_CHANGEDhandler flushes when its state is 0 (Inactive,RestrictedActionsConstantsDocumentation.lua); the event also fires before a restriction activates, which is why only a lift releases. Which restriction was active when the operator saw it is not verified. Four new specs. Locations:Compat.lua,Modules/Gratz/Gratz.lua,Tests/gratz_spec.lua. - Gratz threw "attempt to index local 'name' (a secret string value)" on WoW Forever. Reported on Discord by a Forever player (inbox ff9f7597, 2026-11-10), stack
Gratz.lua:104inshortName<-isSelf<- theUNIT_LEVELhandler. On Retail-engine clients (Retail, Forever;issecretvalueis in the Forever tree)UnitName/UnitLeveland chat payloads can be secret while execution is tainted in combat, andname:matchon one throws. Gratz had no secret guard at all, unlike Smack, Whisper Log and Gathering Log. NewisSecret(feature-detectedissecretvalue, so Classic is unchanged):shortNamereturns nil andisSelffalse for a secret; theUNIT_LEVELhandler, its deferred re-read andrefreshPartyTrackerskip a secret name or level; the achievement handler returns early on a secretarg1/arg2, before thesenderName == playerGuildcompare andarg1:match. A secret is skipped, never guessed at. A skipped party ding is retried when combat ends (Vishiswaz on Discord, inbox 7cf1a18d: "if it can't read the names, shouldn't it try again later"):PLAYER_REGEN_ENABLEDnow also runsrefreshPartyTracker, and since the tracker still holds the pre-combat level, a higher readable level fires the ding then, once. That relies on the name being readable outside combat, which is assumed from the error text ("while execution tainted") and is not verified. An achievement line cannot be retried: the event's text is the only copy, so a secret one is lost. Five new specs inTests/gratz_spec.lua(a paired readable baseline, secret name, secret level, secret achievement text / sender). Not mutation-tested against the old code; not yet seen in game. Location:Modules/Gratz/Gratz.lua. - The two copies of the stack-and-locals capture are now pinned to one behaviour. Self-audit 050586d2 finding 5.
errorDetailexists twice, in!!TOGT/!TOGT.luaand in TOGTools' fallback capture, because the two addons share no code, and nothing asserted they agree. The newTests/errordetail_parity_spec.lualoads the real!TOGT.luafrom the sibling folder beside the real Diagnostics module and drives both through the same API stubs, by the error handler and by a blocked-action event, in nine cases: Blizzard's level arithmetic inside an error, level 3 outside one or without the height API,|k/|Kstripping, empty, secret and non-string answers, nodebuglocals, nodebugstack. Each answer is checked against the expected value and against the other addon's. Changing TOGTools' level by one turned eight of the nine red, then it was put back. Locations:Tests/errordetail_parity_spec.lua,Modules/Diagnostics/Diagnostics.lua(comment). - Diagnostics could grow memory without end from one error whose stack differs on every call. Self-audit 050586d2 finding 2.
addPathkept every distinct call path for the session, so an error raised from a runtime chunk with a generated name added a new path per occurrence. A new flood guard,MAX_LIVE_PATHS = 100extra paths per entry, drops the least-recently-seen extra path when exceeded (dropStalestPath) and counts it indroppedPaths, the counter the saved trim already uses; the entry's own path and every occurrence's count are kept. The operator's "leave all that data in the live version" still holds for any ordinary error: 100 distinct paths for one message is a flood, not a bug with several callers. Explain's dropped-path line now reads "were seen but not kept", since a drop can now happen without a reload. New spec. Locations:Modules/Diagnostics/Diagnostics.lua,GUI/DiagnosticsTab.lua,Tests/diagnostics_spec.lua. - An out-of-order Guild Bank bucket would never be put back in order on Classic Era. Peer Review F1 on the v0.12.0 self-audit (inbox thread 2067f94e). v0.12.0's
InjectEntrybinary-inserts each row and runs the retention prune only when the bucket's last row has expired, so both assume the bucket is newest-first, whileLogs:PruneEntrieswas deliberately made order-agnostic. On Era nothing ingests, so if a bucket ever got out of order no path there sorted it again: rows would go in at the wrong place and an expired row in the middle would never be pruned. Peer Review traced every released write path and found they all sort, so no stored bucket is known to be out of order; this is a safeguard.injectIndexnow checks the order in the pass it already makes over the bucket and sorts it newest-first when a row is out of place, at no extra per-row cost. New spec: a stored bucket that is out of order and has an expired middle row loses that row and is newest-first after oneInjectEntry. Location:Modules/GuildBankLog/GuildBankLog.lua,Tests/guildbanklog_spec.lua.
[v0.12.0] (2026-10-26) - Shows as Ultimate Tools in game; "script ran too long" when TOGBankClassic reports bank movements
New Features
- Shows as Ultimate Tools in game (operator, 2026-10-26: "lets do what we did with TOGPM and FGI to make it show as Ultimate Tools in game. we don't change the file structure or folders, we just change how it looks in game for now"). Same pattern as Ultimate Profession Master v1.7.0 and Ultimate Guild Invite v2.20.0. New
addon.DisplayName = "Ultimate Tools"andaddon.ShortName = "UT"inTOGTools.lua, used by: the## Titleof all seven TOCs; the main window title and its Help / Settings icon tooltips; the ESC > Options > AddOns entry and the options page title (AddToBlizOptions' display argument; the AceConfig app name staysTOGTools); the minimap button's LibDataBroker label and tooltip; the Explain window title, the Diagnostics row menu title and the scriptErrors tooltip; the Diagnostics error toast, Titan/LDB label and tooltip; the Item DB and Quest DB wipe dialogs; the Event Trace "no stream" line; the Smack key-binding warning; every "ESC > Options > AddOns > ..." hint in the Logs, Graphs, Gathering and Whisper help. Chat lines start with an orangeUT:instead ofTOGTools::Ace:Print/PrintfinTOGTools.luareplace AceConsole's, which prefix with the AceAddon name (TOGPM's override, copied). The[TOGTools]tags on the Login Digest and Mailbox login lines are now[UT], and the Debug, Retrack, Name Prefix, Profession Capture, LibAceGUIWidgets-outdated and UI warnings printUT./utis registered beside/togt, which keeps working;/ut helplists every command as/utwith one line saying/togtstill works, and the in-game hints that named a command (/ut mail,/ut glcheck,/ut gbcheck,/ut settings) now say/ut. Deliberately unchanged, because changing them would lose data or settings: the folder and file names,TOGTools_DB, the AceAddon / AceConfig / LibDataBroker / LibDBIcon / VersionCheck names (the saved minimap position hangs off the LibDBIcon name), frame global names, the!!TOGTdependency and the CurseForge project. The minimap button, the AddOns-list icon (new## IconTexturein all seven TOCs), the Diagnostics error toast and its Titan/LDB entry use the operator's newtextures/UT_MMB_Icon.tga(operator, 2026-10-26: "change that to be the image used in game") instead ofToGTools_PH_MMB. The first cut copied TOGPM's square crop, and the operator saw the coin stretched in game ("the icon seems streched?"). The newTests/icon_spec.luareads the TGA and measures the coin: that art had it squeezed to an oval (38 wide by 59 tall), so the operator re-exported it ("the image was bad, i replaced the image"). The oval-fitted crop then squashed the round coin ("the icon is now smooshed"). The final art has a round coin at x 13..50, y 12..50 of 64x64. UPM's and UGI's crop,{ 0.135, 0.825, 0.105, 0.795 }, was tried next at the operator's request ("do the slight adjustment UPM and UGI did"), but it is wider than this coin and left about 3 pixels of the art's white background showing ("it seems like it's smaller?"). So the minimap button'siconCoordsand the toast'sSetTexCoordare a square on the coin's own edge,{ 12.5/64, 51.5/64, 12/64, 51/64 }, then shifted 0.02 up and left, the same nudge UPM got in game (operator: "ours is offset like UPM was before the moved it slightly"), which draws the coin down and right in the slot. The spec pins that crop and the nudge, that the window is square and the coin round, that the coin stays inside the window (the nudge trims at most a pixel and a half) with at most two pixels of background showing, and that the coin is round in the file, since the AddOns-list icon (## IconTexture) cannot be cropped. It prints the measured bounds when they disagree, so a future image change shows the new numbers. The oldToGTools_PH_MMB.tgais no longer referenced by any code. README and the CurseForge page are not changed in this pass. New specTests/brand_spec.lualoads the realTOGTools.luaandSlashCommands.luaover the installed Ace3 and pins the prefix, both slash commands and the help text;Tests/toc_spec.luaandTests/mailbox_spec.luanow expect the new names and that the old title is gone. Locations: every.toc,TOGTools.lua,SlashCommands.lua,GUI/MainWindow.lua,GUI/Settings.lua,GUI/MinimapButton.lua,GUI/DiagnosticsTab.lua,GUI/ItemDBTab.lua,GUI/QuestDBTab.lua,GUI/EventTraceTab.lua,GUI/LogsTab.lua,GUI/LogGraphsSubTab.lua,GUI/ProfCaptureTab.lua,GUI/RowList.lua,GUI/SmackTab.lua,GUI/UI.lua,Modules/Diagnostics/Diagnostics.lua,Modules/LoginDigest/LoginDigest.lua,Modules/Mailbox/Mailbox.lua,Modules/NamePrefix/NamePrefix.lua,Modules/Retrack/Retrack.lua,Modules/GatheringLog/GatheringLog.lua,Modules/WhisperLog/WhisperLog.lua,Modules/GuildLog/GuildLog.lua,Modules/GuildBankLog/GuildBankLog.lua.
Bug Fixes
GuildBankLog.lua:198: script ran too longon Classic Era, raised inside TOGBankClassic'sLog:Deliver. User-reported 2026-10-26.GuildBankLog:InjectEntryrebuilt its dedupe set by hashing every stored row, full-sorted the bucket, ran the retention prune over it and redrew the Guild Bank sub-tab, once per row. TOGBankClassic pushes a whole batch in one timer callback, so a batch of k rows into a bucket of n rows cost k x n key hashes plus k sorts, all in one frame, and a long-lived Era log (where every row arrives this way) tripped the client's watchdog. Line 198 was just where the counter ran out, insidebaseKey. Now: the dedupe set and per-tuple highest numeric ordinal are cached per bucket (injectIndex, weak-keyed, valid only for the sameentriestable at the same length, and dropped explicitly by ingest, the three one-shot repairs, a prune that removed rows, andClearAll); the row goes in at its place in the newest-first order by binary search instead of a re-sort; the prune pass runs only when the bucket's oldest row is past retention (newLogs:RetentionCutoff(), whichPruneEntriesnow uses too, so the two cannot disagree); and the redraw is coalesced to one per batch (C_Timer.After(0), synchronous with noC_Timer). Stored rows, ids and dedupe results are unchanged. 10 new specs, including a 300-row batch into a 3000-row bucket under a fixed VM-instruction budget. Not yet verified in game. Locations:Modules/GuildBankLog/GuildBankLog.lua,Modules/Logs/Logs.lua,Tests/guildbanklog_spec.lua,Tests/logs_spec.lua.
[v0.11.0] (2026-10-16) - Tracking comes back after you die
New Features
Retrack: after you die and come back, your next mouse click turns your last tracking back on. Feature request from Discord (Rakish, 2026-10-11: "After dying and resurrecting, automatically activate the last tracking spell that was active"). Tracking is remembered per character (
db.char.retrack={ spellID, name, owed }) on everyMINIMAP_UPDATE_TRACKING. Whether a death owes a restore is decided atPLAYER_DEAD, because the client's order of "tracking off" against the unit reading dead is not known: an off report while alive only notes when it went off, and if it had been off for 2 seconds or more when the death arrives, you turned it off yourself, so it is forgotten and stays off. OtherwisePLAYER_DEADsetsowed, cleared only when the tracking is seen on again or the restore gives up, and only an owed restore is ever armed.PLAYER_ENTERING_WORLDre-arms an owed restore, so a/reloadbetween coming back to life and the click does not lose it. If the remembered tracking is not in the list yet at login, the restore waits rather than giving up, and the nextMINIMAP_UPDATE_TRACKINGarms it (never betweenPLAYER_DEADand coming back to life, when the unit may not read dead yet). Re-entering the arm while already armed (zoning) keeps the tries already spent. Classic Era reads and restores tracking differently, and the first draft got it wrong. The first in-game test on Era did nothing (operator, 2026-10-14: "had herbing on, died to test it, went to my corpse, got back into the game, and nothing"), and nothing was ever saved. The cause:Blizzard_Minimap_Classic.tocloads onlyMinimapTracking_Simpleon vanilla ([AllowLoadGameType vanilla]; the dropdown is[ExcludeLoadGameType vanilla]). It reads the active tracking withGetTrackingTexture()and names it withGameTooltip:SetTrackingSpell(). I only read theC_Minimaplist, which never reported Find Herbs on Era. I first blamed an "always-on non-spell entry"; that was invented, there is no such tracking on Era, and it is withdrawn. Now, when the list shows nothing active andGetTrackingTexture()returns an icon, the name is read from a hiddenGameTooltipTemplatetooltip (TOGTools_RetrackTip). It is read from the game rather than mapped from the icon, because Track Beasts and Track Humanoids share one. The memory getscast = true. Turning it back on there means casting the spell, which an addon can only do through a secure button. So that path's click-through sheet is aSecureActionButtonTemplatebutton (TOGTools_RetrackCaster,RegisterForClicks("AnyUp"):SecureActionButton_OnClickacts on the release for a secure mouse press). Its restingtypeattribute is nil, so it casts nothing.PreClicksetstype = "spell"/spell = <name>only for a click that passes every gate below, andPostClickclears it. Both skip in combat lockdown, so a sheet that cannot be hidden in combat still casts nothing.PLAYER_REGEN_DISABLEDtakes it down andPLAYER_REGEN_ENABLEDputs it back, or hides one a disarm left up in combat. The give-up message on this path says to cast it from the spellbook, since Era has no tracking menu. The operator re-tested this path in game on Classic Era on 2026-10-16 ("seems to work"). The list path keeps onlytype == "spell"entries as the remembered tracking, the split Blizzard's ownMinimapTracking_Updatemakes (Blizzard_Minimap/Classic/MinimapTracking_Dropdown.lua). Every decision is traced throughaddon:Debug(Settings' debug switch) so the next test shows where a restore stops. A click within 1.5 s of the lastSetTrackingis ignored, so a burst of clicks spends one try on a cast still in flight instead of reaching the give-up message while the tracking is coming on. (Earlier drafts forgot the memory the moment tracking went off while alive, with a 5-second grace afterPLAYER_DEAD; the self-audit and Peer Review caught that a corpse run outlasts the grace, and an off report racing ahead of the death beats it, so either order wiped the memory and nothing was restored.)GetTrackingInforeturns the sameMinimapScriptTrackingInfotable (name,active,spellID) in the Era, Anniversary, Classic, Forever and live API docs. OnPLAYER_UNGHOST/PLAYER_ALIVE(alive, not a ghost), the module arms a full-screen, click-through overlay, which is FastGuildInvite's Wingman pattern (operator, 2026-10-13: "should we use an overlay like we do for FGI, so any mouse click turns them on?"). The overlay'sOnMouseDowncallsC_Minimap.SetTracking(index, true)inside the player's own click, the same call Blizzard's tracking menu makes from a click (Blizzard_Minimap/Classic/MinimapTracking_Dropdown.lua). The click still passes through to whatever was under the cursor. Other Classic tracking addons only manage this "best effort" from an event, so this module never calls it from an event or a timer. The file has no timers at all. Each step is guarded:- The overlay is never shown without verified click pass-through (
SetPropagateMouseClicks, protected in combat, checked withCanPropagateMouseClicks). On a client without pass-through it prints the tracking-menu hint instead. - In combat it waits for
PLAYER_REGEN_ENABLED. - A click within 0.3s of arming is ignored, so the click that accepted a resurrection does not count.
- The overlay stands down when
MINIMAP_UPDATE_TRACKINGreports the tracking on. - After three clicks that do not take, it prints "pick it from the minimap tracking menu" and stops.
- The tracking is matched by spell id first, then by name, because the list's order changes as spells are learned.
- Feature-detected on
C_Minimap.GetNumTrackingTypes/GetTrackingInfo/SetTracking, so a client without them gets an inert module.
No tab (operator, 2026-10-13: "i don't think this needs it's own tab ... lets just add it to general settings"). There is one toggle, "Turn tracking back on after you die", in Settings > General, on by default ("have it turned on by default"), stored in
db.global.retrack.enabled. Because it has no tab it is NOT inaddon.modules:IsModuleEnabledfails open for a key it does not know.Ace:OnEnablestarts it directly, and turning the toggle off stands down a pending restore.Tests/retrack_spec.luacovers both paths;Modules/Retrack/Retrack.luais at 100% line coverage. Tested in game on Classic Era only; not verified in game: whetherSetTrackingtakes from the overlay click on the list flavours. Locations:Modules/Retrack/Retrack.lua,TOGTools.lua,GUI/Settings.lua, every.toc,.luarc.json,.luacheckrc,Tests/retrack_spec.lua.- The overlay is never shown without verified click pass-through (
[v0.10.2] (2026-10-07) - CurseForge installs every dependency; Profession Capture fixes on Forever
Bug Fixes
- The CurseForge upload declared LibDBIcon-1.0 as EMBEDDED, so the CurseForge app would not install it, and never declared LibAceGUIWidgets or !!TOGT at all. All three are on every TOC's
## Dependenciesline, so a player who got TOGTools through the app without already having them got an addon that does not load. Found by the harness's.pkgmetachecker (Tests/wowapi/pkgmeta.lua, run for the first time on this repo). Root causes: (1) the Forever-only embedded copylibs/LibDBIcon-1.0.luakept LibDBIcon's first line,--@curseforge-project-slug: libdbicon-1-0@; the packager greps every shipped.luafor that marker and it overrides.pkgmeta'srequired-dependenciesentry, so the v0.9.1 switch to a required dependency never took effect on CurseForge. The marker line is deleted; no code changed. (2)required-dependencieslackedlibaceguiwidgets(the slug every other fleet addon uses) andtogt(!!TOGT, project 1533953, slug read off its CurseForge page). Both added. The checker now reports no FAIL; it is declared as the writ suitepkgmeta(lua Tests/wowapi/pkgmeta.lua --slug !!TOGT=togt). Not verified on CurseForge until the next upload. Locations:libs/LibDBIcon-1.0.lua,.pkgmeta. - Profession Capture on WoW Forever filed every profession's recipes under each profession, and never recorded an unlearned recipe. Reported by ProfessionDB (inbox
c934e6aa) from the operator's 2026-10-07 Forever capture: Mining, Fishing, Blacksmithing, First Aid and Cooking each held the same mixed list, alllearned = true. Two causes, both read inwow-ui-source-forever. (1)C_TradeSkillUI.GetFilteredRecipeIDs()answers for every profession the character has; Blizzard's own list keeps a recipe only whereC_TradeSkillUI.IsRecipeInSkillLine(recipeID, professionID)is true (Blizzard_ProfessionsTemplates/Blizzard_Professions.lua:847-849). The read now applies the same test, against the base or the child skill line (newProfCapture.SkillLineTest). The merge that keeps an earlier read's rows applies it too, so the mixed rows already saved are dropped on the next read. (2) Forever's Professions UI forcesSetShowUnlearned(false)after every list build and offers no menu entry for it (Camelot/Blizzard_Professions.lua:16-25), so the window never lists an unlearned recipe, andGetRecipeSourceText(called only for unlearned rows) never ran. The read now also takes LibProfessionDB's recipes for the open profession and reads each one while the window is open. It asks byparentProfessionIDwhen the client sets one, elseprofessionID, because LibProfessionDB files every recipe under the base id and answers nil for any other. Each known record now carrieslibProfessionIDandlibRecipes(the count LibProfessionDB gave, 0 when it found nothing, unset when the library is absent). The filters are left alone, so the player's list does not change. Every unlearned row the read asks about now carriessourceChecked = trueandsourceTextType(the raw type returned), so a missingsourcereads as "the client answered nothing", never as "not asked"; an empty or secret string is not stored as a source. The operator's v0.10.2 Forever run (2026-10-07) read 662 unlearned rows across five professions andGetRecipeSourceTextgave none of them text (ProfessionDB's report). 9 new specs inTests/profcapture_spec.lua. Location:Modules/ProfCapture/ProfCapture.lua(retailKnown,CaptureKnown). - "Walk recipe sources" told the player Forever's gap was covered when it was not. On a walk that found no source it printed that it had "Saved the game's recipe link ... so ProfessionDB can look sources up offline". ProfessionDB cannot: the link is
skillLineAbilityID, and Forever's SkillLineAbility table has no vendor, drop or quest column (ProfessionDB, inboxc934e6aa). The message now says the walk gets no sources there and to open each profession window instead; the button's tooltip and two engine comments say the same. Locations:GUI/ProfCaptureTab.lua,Modules/ProfCapture/ProfCapture.lua.
Process
.pkgmetahas no comments any more (operator's standing rule, 2026-10-03: comments in.pkgmetacan break packaging). Its notes moved toCLAUDE.mdunder ".pkgmeta". The commented-outcurseforge-project-idline went with them; the id (1533830) is recorded there.Test harness pin moved
3d492c7->b20754a, with its adoption steps.Tests/gatheringlog_spec.luano longer installs its ownGetSubZoneText; the harness now does, returningwow.subZone(harness inbox280f62ed).Tests/smack_spec.luano longer stubsSetOverrideBindingClick/ClearOverrideBindings: the bindings and clear count it asserts now come from the harness's recordedwow.overrideBindingsandwow.clearedBindings(harness61c09b0, inbox366486c9).Tests/gratz_spec.lua's GreenWall stand-in check went red on the move, correctly: the harness (4fd0cbf, measured on Classic Era by GuildRoster) now reports an automated CHANNEL send on Classic as BLOCKED (wow.blocked,ADDON_ACTION_BLOCKED) instead of a silent drop, so the example assertswow.blockedand is renamed "...is blocked on Classic". No addon code changed by the move.
[v0.10.1] (2026-10-04) - WoW Forever recognised again
Bug Fixes
- WoW Forever was treated as Classic Era, so opening a profession window threw
ProfCapture.lua:412: attempt to call a nil value(GetTradeSkillLine). Reported by the operator from a Forever client, opening TOGProfessionMaster's Crafting tab. Root cause:addon.DeriveGameVersiononly recognised Forever whenWOW_PROJECT_ID == WOW_PROJECT_MAINLINE, but Forever reports project id 18 (FastGuildInvite observed "Project ID: 18, Build: 16001" in-client on build 1.60.1.70124,FGI_Compatibility.lua:22-38; the forever tree'sProjectConstants.luanames only 1 and 2). So Forever at Interface 16001 was derived as Vanilla,isRetailwas false, and every Forever/Retail branch in the addon was dead there -- ProfCapture took the ClassicGetTradeSkillLinepath, which Forever does not have. Other consumers that were wrong on Forever for the same reason, among them: Quest DB refused to run at all (QuestDB.lua:71gates onisForever, "only runs on the WoW Forever client");ProfCapture:ReadTrainerreadGetTrainerServiceInfoin Classic return order (ProfCapture.lua:182-198); I'd Hit That skipped Forever's ranged-hit modifier (IdHitThat.lua:275); Smack offered its six combat-log triggers, which it hides onisRetail120Plus(Smack.lua:142-147); and/togt vctried VersionCheck instead of saying it is not available. The load-time call now counts project id 18 as a retail-codebase client too. New spec: "is derived from Forever's own project id, 18" (Tests/version_spec.lua). Not verified in game. Location:Version.lua.
Process
CHANGELOG.mdarchived under the release-body ceiling. With the entry above it measured 121,457 characters, over the 120,000 working ceiling (GitHub refuses a release body over 125,000). The oldest section, v0.8.6, moved whole to the top ofCHANGELOG_ARCHIVE.md, ahead of v0.8.5, with the editing tools. The trailing pointer now says "v0.8.6 and earlier".
Older versions (v0.1.0 to v0.10.0) are in CHANGELOG_ARCHIVE.md.
This mod has no additional files