YippYapp Helper - Changelog

v3.4.1 - The Keys tab steps aside (2026-09-21)

Removed

• The Keys tab on the Group Finder window. Reported on CurseForge: it collided with other addons that put their own tabs there. It was only ever a shortcut to the Mythic+ page, which is still one click away from the minimap button or /yh.

Fixed

• Swapping a piece could leave the old piece's upgrade track on the slot. A Champion trinket replaced by a Myth one kept reading Champion on the upgrade vendor panel and the Gear Upgrades page until something else cleared the cache. Cached track reads are now tied to the exact item they were read off, so a swap is always read fresh.

v3.4.0 - Pins to the quest givers, and what the weeklies pay (2026-09-13)

Added

• A pin button on the weeklies that lead somewhere. Asked for on CurseForge: something to click that puts an arrow on the screen, the way Azeroth Pilot Reloaded does. It does not draw one -- the game already has an arrow, a minimap marker and a map waypoint, and all three follow whatever is super tracked -- so a click hands the game a target and gets out of the way. Click it again to let go.

  It opens the world map too, on the map the target is actually on. A waypoint you cannot see is a direction without a distance -- the arrow says which way, the map says how far and what is in between, and "where do I pick this up" is asking the second one. Not in combat: the pin is still placed, the map stays shut.

  The button sits at the right-hand end of the row and is only there on rows that lead somewhere. A crest cap is not a place you walk to, so it has no button rather than a button that does nothing.

• Five quest givers ship with the addon, so the pins are there on a fresh install rather than after a week of play. Lady Liadrin, Halduron Brightwing, Vereesa Windrunner, Archmage Aethas Sunreaver and Warleader Abdumati, from Wowhead's datamined g_mapperData -- which carries its own uiMapId, so the map is not being inferred from a zone name. Each NPC was reached through a quest id the addon already ships rather than by searching a name, and the Silvermoon map id was confirmed a second time against a live client. Anything the client learns overrides them.

• The addon learns where each quest giver stands. The four Silvermoon rows carry a family of quest ids -- sixteen for Lady Liadrin, eight for Halduron -- one of which is this week's, and nothing in the client says which. Rather than answer that question, the addon stops asking it: you have to be standing next to the NPC to accept or hand in a quest, so the player's own position at that moment is the giver's, and it is recorded against the ROW rather than the quest. Take any one of Lady Liadrin's sixteen and her row knows where she stands from then on -- every week after, and on every character on the account, because an NPC is in the same place for all of them.

  Nothing is written down in the addon and nothing is read off a website. Only while a quest window is open, too: QUEST_ACCEPTED also fires for quests the game hands you for walking into a zone, and those would record wherever you happened to be standing as a quest giver's spot -- a confident pin in the wrong place, which is worse than no pin.

  Three ways in, so the page is useful before you have done the week's chores rather than after. Best observation wins, and none of them overwrites a better one:

  • you say where it is -- /yh where set liadrinweekly on the NPC, or /yh where set liadrinweekly 47.9 51.6 from anywhere in the zone, which takes a published coordinate and reads the map id off the client. Percentages and fractions are both accepted and the result is echoed back, because every database writes one and the game wants the other;
  • you accept or hand in one of the row's quests, as above;
  • the world map already knew where the offer was, taken for free on walking into the zone. A coordinate outside 0..1 is dropped rather than converted -- map positions come as fractions and as percentages, and guessing which is how a pin ends up wrong half the time.

  /yh where forget throws the lot away if a record is ever wrong; they rebuild themselves the next time each quest is picked up.

• A weekly that leads somewhere but has no pin yet says so on hover. An empty right-hand edge is otherwise indistinguishable from a crest cap, which is not a place and never will be.

  The question behind the request was never "which quest is it", it was where do I pick this up, so a row points at whichever of these it can:

  • the quest, when it is in your log -- super tracked by id, so the arrow follows the objective rather than the giver;
  • the NPC who hands it over, when it is not. Lady Liadrin stands in the same spot whichever of her sixteen she is offering this week, which is exactly why the giver is the answer that works;
  • the client's own quest-offer pin -- the blue exclamation mark -- on a row whose single outstanding quest can be identified.

  A family of quest ids is never guessed at. Halduron offers one of eight dungeons a week and nothing in the client says which, so the offer-pin route refuses a row it cannot narrow to one rather than picking whichever sorted first. The giver route is what carries those rows, and it carries them without needing to know the answer.

• Weekly rows say what they pay, with the game's own icons. The other half of the same request. Hovering a row that names a quest now lists its rewards under the objective: the currency first, because on a weekly the currency is usually the whole point, then items in the game's own rarity colours, then the pick-one list under a heading of its own -- a weekly paying a thousand reputation with a faction you choose is five lines that each mean "or", and an unmarked list of five reads as a quest handing over all five.

  Each line carries the reward's icon, its name in the game's rarity colour, and the count in a right-hand column of its own rather than glued to the end of the name -- the count is the part you compare between two weeklies, and ninety crests against one spark should line up. Large numbers go through the client's own separator, so a thousand reputation reads as 1,000.

  Read off the client rather than written down here, so a quest retuned this patch reads correctly the day it is retuned, and a reward that scales reads as what this character would actually be paid.

  And before you accept anything, which is when the question gets asked. Most rows resolve no quest at all until it is in your log, so the rewards used to appear only once it was too late to be deciding. A row offering one of sixteen now simply says what it pays -- "Spark of Tides" -- because that is a fact about the row whichever variant comes up, and the row's own description has said so all along.

  The moment a quest IS in your log the client's answer takes over, with its real icons and amounts.

  Archmage Aethas Sunreaver's row shows two lines, because his set pays two different things: a Cache of Amani or Quel'Thalas Treasures for the dungeon, delve and Timewalking weeklies, and Conquest with Honor for the battleground and arena ones.

• The world boss now sits with the other chores, above the Vaults of Atal'Utek weekly, so the weeklies you pick up from a quest giver -- and their map pins -- form one block instead of being split in two.

Changed

• The weekly row's own click is unchanged -- it still ticks the rows that are ticked by hand, and nothing else. Pointing at a quest lives on its own button because it applies to a much smaller set of rows than ticking does, and putting both on one click made most of the page answer a click by doing nothing.
• The pin lights up while the game is pointing there, so the page agrees with the arrow out in the world.

Fixed

• Veteran Mistcrests could read from a dead currency row. Every crest tier exists twice in the client under the same name, and only one row is live. The addon picked whichever row held MORE -- a rule from the week it had been pointed at the empty block -- but a dead row is not always empty. One Priest held 80 on dead 3438 and 50 on live 3443; Blizzard showed 50, the addon showed 80, and the affordability and upgrade suggestions were computed from the 80.

  The resolver now takes the row the character sheet's Currency tab lists, which is the game's own answer, and otherwise the verified order. A balance is never treated as evidence. What the listing found is remembered, so collapsing a header in the Currency tab -- which hides its rows from the list -- cannot flip the addon back. All five tiers go through the same path; Veteran is where it showed.

  The unlisted twin is still reached when the game lists it instead of the usual row, which is what a patch swapping the live block would look like. It is never reached any other way: with nothing listed the answer is the verified id, even if that id reads nothing. A live row showing zero is visibly wrong; a twin showing 80 is not.

• Reward items read "Item" on the first hover. A quest's reward list and an item's name are separate loads, so the tooltip could know which item a weekly pays before knowing what it is called -- and nothing redrew it when the name arrived. The name is now filled from the item cache where it is already known, asked for where it is not, shown as "Loading..." meanwhile, and the open tooltip redraws itself the moment the name lands.

• A pin that opened the wrong zone. There was a third route that super-tracked the client's own quest-offer pin for a weekly you had not accepted. The only map it can name for an un-accepted quest is the QUEST's -- where the objectives are, not where the giver stands -- so the one row it reliably fired on was the one it sent to the wrong place, on a line reading "Pick up Vereesa's weekly in Silvermoon".

  It has been removed rather than patched. What it was for is exactly what the giver does, with a position somebody has actually checked, and it was the only route never seen working in game.

  A row's pin and its label now always agree: while the quest is in your log the row reads "Complete Midnight: Prey" and the pin follows the objective; before that it reads "Pick up Lady Liadrin's weekly" and the pin is Lady Liadrin. A row that says "pick up" and points at a delve is lying, and that cannot happen now.

Removed

• A dead "Map waypoints" block in Core/Data.lua -- its body went in "Sweep the code nothing calls" and left behind a header comment and two unused constants. The map-pin art it named is now used by the thing that actually drops a pin.

v3.3.2 - Durability off the channel everyone is already on (2026-09-10)

Fixed

• The minimap button could take itself out on load. LibDBIcon and LibDataBroker are optional dependencies -- another addon may supply them, or nothing may -- but they were being asked for in the form that throws when they are missing, once for LibStub itself being absent and once for LibStub not having the library. Both are now silent lookups, and with them gone the addon simply has no minimap button; /yh, the key binding and every page are unaffected.

• The Dur column filled for nobody. Durability was shared over this addon's own YYHDUR prefix, so a raider's percentage only ever appeared if they also ran YippYapp -- and the column looked like it worked, because your own row always did. It also never filled at all in a dungeon-finder or LFR group: those need INSTANCE_CHAT, and the broadcast only ever went to RAID or PARTY.

  Moved to LibDurability, the shared library on the LibDRBLT prefix that BigWigs and Method Raid Tools both embed. Its own frame answers a ready check from file load, without the host addon having to do anything, so a raid full of BigWigs users answers -- which is every raid. This is the same fix the Guild keystone tab got in v3.2.0, for the same reason.

• A broken weapon reads green no longer. One dead item among seventeen healthy pieces still totals in the nineties, so the number alone said everything was fine. Broken items are counted separately now and turn the cell red whatever the total says, with the count in the tooltip.

Changed

• The durability cell has a tooltip. It used to fall through to the buff-column wording and hover as "out of range or not applicable" -- off a cell showing 84%. A blank cell now says the honest thing, which is that the player is running no addon that shares durability, rather than pointing at a range problem that is not there.
• Opening the window by hand now asks the group for durability instead of only announcing your own. It used to sit full of dashes until somebody started a ready check.

v3.3.1 - The ring that paid for the other one (2026-09-02)

Fixed

• A second copy of a ring is not a second ring. A ring or trinket slot remembers a line it takes two pieces to build, and they have to be two DIFFERENT pieces. Away from the upgrade vendor the client answers nothing, so the addon falls back to what it can see -- worn gear plus anything bound in the bags -- and it counted a spare copy of the ring already worn as the second half of its own pair. One Hero ring at 311 with its twin in the bags, and the Champion ring on the other finger was told "Free upgrade! All remaining ranks are free". The vendor charges for those ranks.

  Counted once per item now. The client itself was right all along: with a 311 ring and a 295 ring worn, it reports 295 for the finger line -- the good ring alone buys the other finger nothing.

• "1 free via Champion" was never free. The grey note beside a piece sitting in its track's overlap band claimed the first ranks cost nothing. Those ranks are free only as a PROMOTION -- take a Champion piece to 6/6 and it arrives at Hero 2/6 without a Hero crest. A piece that dropped at Hero 1/6 has no promotion left to collect and pays Hero crests for the next rank like any other; the vendor prices by the item's own track, not by the item level the rank lands on.

  It now reads "(1 rank Champion also reaches)", which is the true part and the part the row's own "Wasteful crest spend" status is built on.

• An enchant no longer makes a piece "crafted". Crafted was detected by looking for the profession quality atlas anywhere in the tooltip text, and an enchant carries one -- "Enchanted: Enchant Helm - Empowered Rune of Avoidance" brings the enchant's own quality rank with it, so an ordinary helm read as crafted. On the character this was found on, ten of sixteen slots came back crafted; two of them were, and the other eight were the enchanted ones.

  It cost more than a label: a crafted piece counts as the craft already done, so every slot with a good enchant quietly dropped off the list of things to make. Crafted is now read off the item itself, through C_TradeSkillUI.GetItemCraftedQualityByItemInfo, and the tooltip scan is only a fallback for a client that cannot answer.

• /yh marks prints every bucket the client keeps. It looped over the sixteen this addon was written against, so a bucket the client gained since could not appear on the one screen that exists to show what is actually there.

v3.3.0 - Trinket tier lists, the upgrade-mark fixes, and the Battle Res Timer out (2026-09-02)

Removed

• The Battle Res Timer is gone. EllesmereUI draws one, so does BigWigs, and all three read the same pool the game itself publishes -- so for most raiders this was a second copy of a number already on screen. It was also on screen during a pull, which is where there is least room for one.

  Out with it: Features/Raid/BattleResTimer.lua, its two options (Battle Res Timer, and Show between pulls), the "Battle Res Timer" settings section, and /yh brez. The roster's "who can combat res" column is a different thing and stays -- that lives in Features/Raid/SpecMeta.lua and feeds the raid page.

  Its saved settings and its Edit Mode position are cleared on next login, so nothing is left behind in the SavedVariables file.

  Edit Mode now has nothing to place. The timer was its last frame: every other window in the addon moves itself, dragged where you want it while it is open, and only the timer -- up mid-pull, when dragging is the last thing you want to be doing -- needed a mode of its own. The "Appearance and position" section and its Open Edit Mode button have gone with it, and /yh edit says so instead of opening an empty manager. The integration itself is left in place: it is machinery rather than a feature, and one table entry brings it back.

Added

• Best in Slot now survives the Catalyst. Feeding a piece to the Catalyst destroys it, so a row you had done exactly what the page told you to do with flipped from collected back to missing at the moment it was most finished -- the item ID it was tracking no longer existed.

  What survives the conversion is the secondary stats: class set armour inherits them from whatever went in. So a set piece worn in one of the five class-set slots, whose secondaries match the armour that slot's row names, is now read as that row, converted. It counts toward "collected", carries the Catalyst's purple and a cat tag on the doll, and the slot summary gains a "2 converted" segment on characters who have any.

  It is shown as an inference rather than as a tick, because that is what it is: a set piece that dropped directly rolls its own secondaries and can land on the same pair by chance. The tooltip says what was actually seen -- "you are wearing class set armour at 311 with this item's secondaries" -- rather than claiming to know which piece was fed in. Anything the client can state outright wins: wearing the row's own item reads as equipped, a copy in the bags reads as bagged, and a row already tagged +cat is never inferred at all, since there the item ID answers on its own.

• Trinkets now carry the guide's verdict, not just the sim's. Hovering a trinket has always shown where bloodmallet ranks it. It now also shows the letter grade your spec's Wowhead guide gives it, the content it drops from, and the author's own note on it -- "syncs perfectly with DRW as San'layn", "not great, honestly, careful while using it" -- the kind of thing a throughput number cannot say.

  The two are shown side by side rather than blended. A percentage and a letter are different claims: the sim ranks by damage against one dummy, and the guide ranks knowing which on-use wants pairing with which cooldown, which drop is Vault-only, and which pair nobody would wear together. Averaging them would produce a number neither source stands behind.

  Augmentation Evoker has trinket advice for the first time. bloodmallet does not sim it and QE Live covers only healers, so the spec had nothing at all. Its guide grades trinkets like every other, and that is what the new tab and the tooltip read.

  A Guide tab on the Trinkets page, beside My Spec and Loot Council. It draws the ladder in bands rather than as a numbered list, because within a tier the guide states no order and numbering three S-tier trinkets 1, 2, 3 would invent a ranking the author declined to give. Every band is drawn, never a top-N: the lower bands are where the "do not bother" information lives, which is most of what a tier list is for. The pairing advice sits above it, and each author note under the trinket it is about. No bars, no percentages, no item level stepper -- none of those are what a letter grade is made of -- and the fight style buttons are hidden there, since the guide grades a spec once rather than once per target count.

• The Guide tab leads for specs whose sims are behind. bloodmallet re-sims a season a few specs at a time, and Balance Druid, Feral, Guardian, Devastation and Retribution are still on last tier's numbers -- they have no opinion at all on anything that has dropped since, and the page opened on them anyway. For those specs the Guide tab is now first, so the page lands on advice that is about this season. It is a reorder, not a redirect: My Spec is one click away, still says what it is, and a remembered choice still beats both. Hiding the sim outright would be a bigger claim than "this one is out of date", which is all that is known. The stale-data warning on My Spec now points at the tab that is current instead of only saying the numbers are old.

  909 graded trinkets and 301 author notes across all 40 specs, from Tools/scrape_trinket_tiers.py. The tier list lives on the same /bis-gear page the class-guide scrape already fetches, so keeping it current costs no extra requests.

• A trinket the guide does not list now says so. Previously a hover that found nothing looked exactly like a hover on an item the addon had never heard of. It now says "not on the Fire Mage trinket list" -- but only for a trinket some other spec ranks or grades, which is what proves it is a current trinket worth having an opinion about. An item nobody ranks still says nothing, because the guide's silence there is about its own scope rather than about the item.

• The trinket tooltip answers your question first, and only yours when you are alone. It used to open with five other specs' rankings on every hover. On a character whose own sim is a tier behind, that was the entire tooltip: eight lines about specs you are not playing, a "+29 more specs", and nothing whatsoever about you.

  Your spec now comes first -- the guide's grade, the author's note, and a single Sim #7 line for where the sim puts it -- and the other specs are shown only when you are in a group, cut from five to three. "Who else wants this" is a real question, but it is the loot council's question and it needs a council; the page's Loot Council tab has always held the full list. The duplicate source attribution in the header is gone too: the group headings underneath already said it.

• The consumables popout meets the Auction House. The small popout exists for one place -- standing at an auctioneer with fifteen names to shift-click -- and it did not know when you were there, so every visit started by opening the app, finding the Consumables page and pressing Popout. Now it opens itself when the Auction House does, pinned against the top right of it.

  The top right corner rather than any other: it hangs off the side so nothing is covered, and anchoring the top keeps the window still when its height changes -- Gems is a third the length of Consumables, and paging the arrows from a bottom anchor would have it jump every time.

  Pinned is the default, and a Pin to Auction House box on the window itself is the unlock. Untick it and the window stays exactly where it is -- unlocked in place, not thrown back to wherever it last sat loose -- and drags anywhere you like from there. The box only shows while the Auction House is open, because that is the only time it means anything. Dragging a pinned window is refused outright rather than allowed and snapped back, which looks like the window fighting the cursor.

  Only a window the Auction House opened is closed again when you leave. One you popped out by hand before walking over is left alone.

  Off in Options - AddOns - YippYapp Helper - Consumables at the Auction House if you would rather open it yourself.

  /yh ah walks that whole path and says what it found at each step -- whether the watcher is armed, whether the page is built, whether the setting is on, whether the Auction House frame has ever loaded -- and then runs it. Every one of those failing looks identical from the chair, which is nothing happening.

Developer

• The load harness counts globals. Every global the addon creates is snapshotted around the load pass and diffed. A Lua assignment that forgets local lands in _G, Blizzard's own code reads globals, and reading one an addon wrote carries that addon's taint into the secure path -- the "interface action failed" class, which then gets blamed on whichever addon the player looks at first. Nine exist today and all nine are legitimate: named frames, the keybinding strings, the saved variables table and the bundled libraries' own globals. The tenth will fail the run.

• /yh marks — the whole high-water state, read rather than inferred. Every one of the seventeen redundancy buckets the client keeps, named from the client's own enum, character and account side by side; then which bucket each worn piece answers to, with crafted and off-stat pieces flagged.

  Everything else in the addon reaches these marks one item at a time, which is the right shape for pricing a rank and the wrong shape for answering "why did this not go free" -- it can only ever show one bucket and never the sixteen it sits among.

  It exists because the rules here are not documented anywhere and the player reports contradict each other. Run it, change one piece, run it again: whatever moved is the rule. Two runs settle questions no amount of reading settles.

Changed

• The Auction House list has room, and answers a click instead of the cursor. The window beside the auctioneer was 250 wide with 22px rows, and the names in it are the longest in the addon -- a weapon enchant wrapped onto a second line inside a one-line row, so it was drawn through the name under it. It is wider now, a line taller, and a name too long for the row is cut short rather than wrapped.

  No more tooltip on hover. This window sits on top of the Auction House and the cursor crosses it on the way to something else; a tooltip that opened itself on the way past covered the auction list underneath, repeatedly, for nobody. Now:

  • Click an item and its tooltip opens beside the window -- on whichever side has the room -- and stays there until you click it again, so it can be read without holding the mouse still.
  • Shift-click and the Auction House searches for it: the name goes into Blizzard's own search bar and the search runs. Away from an auctioneer, shift-click still links to chat as it always did.

  A line at the bottom of the window says which is which, and says "search" or "link" depending on where you are standing.

• Steps get a number, not a question mark. A row with no spell behind it used to draw INV_Misc_QuestionMark, on the reasoning that a missing icon and a missing spell id look the same to a reader. That reasoning was wrong about half the rows: "The split", "The quadrants", "The egg fall", "Platform eggs" are not abilities at all. They are steps, done in order, and there was never going to be a spell id for them -- so the question mark was reporting a lookup failure that had not happened.

  Those rows now show a numbered box: the step's place in its phase, after the difficulty filter, so the number matches what is on screen. It reads as sequence, which is what a step is. An ability we genuinely have no id for gets one too -- "here is where this comes" is true, where "something went wrong" was not.

• "Before you pull" answers per difficulty. One flag covered both, so a boss whose Heroic notes existed and whose Mythic ones did not printed "Not recorded" over the top of the Heroic notes it had -- a claim of ignorance the same card disproves two lines further down. Each difficulty answers for itself now: what is written, or an admission for that one alone. Silence still means "nothing changes", which is a real answer and a different one.

• Ula'tek's guide, rewritten against a real source. He was written from two video transcripts because nothing covered him in text. mythictrap has published him now, which adds the half a transcript cannot: real spell IDs for Mother's Wrath, Unchecked Rage, Necrotic Vapors, Anguished Cry and Vicious Echoes, so those five draw the client's own icon and tooltip instead of being matched by name.

  Two phase-one abilities nobody had captioned are in: Unchecked Rage is the reason both tanks hold their targets — let either drift out of melee and the raid takes it — and Necrotic Vapors is the rot running underneath the phase. Both interrupts are named now, one per later phase.

  Heroic changes go on the two mechanics they change. Mythic stays unknown, because mythictrap publishes a Heroic block for this boss and no Mythic one.

  The transcripts held up everywhere else, including the two things they were most likely to have invented — eggs breaking weak or hatching strong, and phase three's marked players being soaked rather than run from.

• Class guide data refreshed (Wowhead, 2026-09-02). 29 of 40 specs moved since the last release, over two passes. Stat priorities changed for Unholy Death Knight, Subtlety Rogue, Enhancement Shaman and Preservation Evoker -- Enhancement's two hero builds no longer agree, so the page now shows Stormbringer and Totemic separately instead of one merged list. Best-in-slot moved in most of the rest, most heavily Restoration Shaman, Enhancement Shaman, Subtlety Rogue, Preservation Evoker and Restoration Druid.

• Consumables data refreshed (Wowhead, 2026-09-01). 16 of 40 specs moved since the last release. Most of it is one story: a new temporary weapon enchant, Rite of the Hash'ey, which 14 specs now recommend over the old oils and stat enchants. Havoc Demon Hunter also dropped from five gems to two and changed its diamond, and Assassination's flask and combat potion became three-way ties rather than single picks.

Fixed

• The Auction House list read item:273072 until you closed it. Walk up to an auctioneer for the first time in a session and the popout came up as a column of red question marks and raw item IDs. Closing it and opening it again fixed it, which is the tell: the client had not sent those items yet, and the names landed a moment later with nobody listening.

  The redraw that runs when an item's data arrives was wired to the Consumables page, and beside the Auction House the page is shut -- so it checked, found the page hidden, and did nothing. The popout now counts as being on screen in its own right, and gets its own redraw when the page is not up to pass one down.

• A free rank that was not free. An off hand at 305 wore a cyan "Free upgrade!" and read "free to 308", and the vendor charged crests for the rank.

  The client answers the high-water query twice over: how high THIS character has been in the slot, and how high the account has. The addon took the larger of the two and called every rank under it free. Only the character's own figure carries a free rank; the account figure is how far the rest of the warband has got, which is a different statement about a different character. That is now the only number a free rank may be quoted from, with the account one kept as a fallback for the client not answering in that shape at all.

  Marks written under the old rule are on disk and nothing about a stored number says which half it came from, so they are cleared once at login and re-read within a few frames.

  Wrong in the safe direction from here on, which is the rule the whole free-rank half of the addon is built on: a mark that is too low hands out too few free ranks, and a row that holds back a rank that turned out to be free costs a walk to the vendor. Too high spends crests that do not come back.

• The raid scan kept everybody you ever grouped with. It is written from INSPECT_READY, which fires for a group member whoever asked for the inspect -- this addon or any other -- it was cleared only by starting another scan, and it was saved to disk on every logout. Measured on a real account: 673 players, 152KB, three times the size of everything else in the saved variables file put together, and nearly all of it strangers from pugs weeks earlier.

  It prunes to the current group on every roster change now, and saves nothing at all when you log out ungrouped. An empty roster reads as "ask again later" rather than "you are alone", so a loading screen cannot blank the page for a raid that is still there.

• The durability cache is bounded. Same shape, smaller: fed by addon messages from other players, one entry per distinct sender name, and nothing ever removed one. Entries past their ten-minute TTL are already ignored on read, so they are now swept once the table grows past a group's worth. It matters mainly because the feed is remote -- a table that only ever grows, filled by input from outside, is worth bounding whether or not anyone is being clever with it.

• The crest discount is re-read the moment it is earned. It is an achievement, and an account-wide one: cross an item level in every slot on any character and every character pays ten crests a rank instead of twenty. The addon checked once at login and cached the answer, so a discount earned with the panel open left every price doubled until a reload -- at exactly the moment the numbers matter most, with somebody deciding what to spend next having just crossed a threshold. ACHIEVEMENT_EARNED clears it now.

• Off-stat gear sets no line. The high-water mark follows gear that is class-appropriate, not everything a character can physically equip -- and the addon was counting anything bound with a high enough item level. A shaman reported spending sixty Hero crests taking an Agility staff from 311 to 321 with the watermark never moving; a shaman can hold a staff, and an Agility staff is nobody's shaman weapon.

  That is the piece a player is most likely to be sitting on, too: the off-stat one kept because the item level is high. Every place the addon works a mark out for itself now checks -- what is worn, what is bound in the bags, a spare worth finishing, a tradeable piece worth keeping.

  Excluded on positive evidence only: this spec's primary stat read off the guide, the item's read off the item, and the two different. Anything unknown still counts, so the rule cannot quietly swallow the pieces that have no primary stat at all -- most necks, rings and trinkets -- or every spec whose guide the addon has not got.

• Your alt was reading someone else's marks. The saved variables file is account-wide -- one table, every character on the account writing into it -- and the high-water mark cache was keyed by slot alone. A mark is a fact about ONE character's slot, so every character after the first read whatever the last one left behind.

  And it read them in exactly the situation the cache exists for: away from the upgrade vendor, where the live query answers nothing and the read falls through to disk. A druid logging in after a demon hunter inherited the demon hunter's weapon line and was told ranks were free that its own vendor charges for.

  Marks and their item links are now filed under the character's GUID -- stable across a rename or a transfer, and unique where two characters on different realms share a name. What was already on disk cannot be attributed to anyone, so it is cleared once at login and re-read.

• Two one-handers share a line, and the addon now knows it. This is the rest of the free-rank bug above, and it takes two one-handers to see: Enum.ItemRedundancySlot keeps the one-hand marks as a ranked PAIR -- OnehandWeapon (14) and OnehandWeaponSecond (15) -- the same shape rings and trinkets have, and there for the same reason. One good one-hander must not hand the other hand free ranks.

  GetHighWatermarkForItem answers about the item's own bucket, which is the higher of the two. Handed to the second hand it is a promise about a weapon the player is not upgrading: a 331 in the main hand answered 331 for an off hand sitting at 305, and every rank to 308 was offered for nothing.

  So the two weapon slots are now a pair -- but only while both hands hold a one-hander. A two-hander, a shield or a holdable sits in a bucket of its own and pairs with nothing, which is why this never showed up on anything but dual-wield. When they do pair, the free line comes from the second bucket, asked for by name, and the bags follow the same second-highest rule rings and trinkets have always used.

• /yh debug can be pointed at a slot -- /yh debug 17 for the off hand, Feet as before with no argument. It asks the high-water query about that slot and prints both figures the client hands back beside what the addon concluded from them, so a row promising a rank the vendor charges for can be taken apart from the chair. Equipped crafted pieces are marked [crafted] in the dump.

v3.2.0 - Crest advice, shared keystones, and a lighter addon (2026-08-24)

Removed

• The Delves tab is gone. A whole page to restate a companion's friendship rank, a key count and a tier ladder -- none of which the game hides, and two of which this addon already says elsewhere. It came out entirely: Features/Delves/, its shell page, its entry in /yh shell <page>, and the client stubs in the harness that nothing else used (C_DelvesUI, the C_GossipInfo friendship pair, C_PartyInfo.IsDelveInProgress, C_Reputation).

  Delves themselves are not gone from the addon, only their tab. The weekly checklist still lists delve chores, the Dashboard still advises "run a tier N delve" when that is the cheapest upgrade going, the Progression page keeps its delve tier section, and the vault's World row is still filled by them.

• The interrupt tracker is gone. It had not worked for the whole of 12.1, and it could not be repaired in place. The tracker never saw which spell a party member cast -- the client stopped handing that over once party spell IDs became secret-tainted -- so the party half of it was inference: note that a teammate cast something, and when an enemy is interrupted within 150ms, assume it was that teammate's class default kick. Every layer of the file was scar tissue from working around that (pcall on every lookup, UnitIsUnit instead of comparing GUIDs because enemy nameplate GUIDs came back secret too), and the guess at the bottom was the part that could not be fixed.

  Making it genuinely work means a different data source -- the combat log's SPELL_INTERRUPT names the kicker and the spell outright -- which is a rewrite rather than a repair, and it would add the one event class this addon has deliberately stayed clear of. OmniCD already does that job well. Removed rather than left on screen lying.

  Its saved settings and Edit Mode position are cleared on next login, so nothing is left behind in the SavedVariables file.

Performance

• The watermark sweep still froze the client, and the test said it didn't. ns:RefreshWatermarks was fixed once already -- it asks the client about one slot per frame instead of sixteen at once -- and it went on producing multi-second frames anyway, most recently 5.3 seconds on a spec change. Only half of it had been spread. To decide which slots were worth queueing it called ns:GetSlotInfo on all sixteen up front, synchronously, and that is the expensive half: GetSlotInfo ends in a real SetInventoryItem tooltip build, and the UNIT_INVENTORY_CHANGED handler driving it wipes the scan cache immediately beforehand -- so every one of the sixteen is guaranteed cold, on item data the client may still be loading. The decision moved into the per-frame runner; the caller now does no slot reads at all.
• ...and the check that was supposed to catch it stubbed the slow half. The harness counted C_ItemUpgrade calls and asserted none landed in the calling frame, while replacing GetSlotInfo with a trivial stub -- so it proved the cheap half was spread and never measured the sixteen synchronous reads at all. It counts slot reads now too, and the assertion is controlled: reinstate the old queue-builder and it reports "read 16 slots in the frame that called it".

A sweep for work the addon does during ordinary play. None of these were features; all of them were the addon spending frames on questions nobody had asked.

• The login loot prewarm no longer caps the client at 30fps. The Encounter Journal sweep gives the frame back once it has used its budget, and the budget was one whole frame at 60fps -- correct for the scan somebody is watching a spinner for, and exactly backwards for the one that runs unasked a few seconds after a loading screen. The prewarm now takes 4ms rather than 16ms. It finishes later; nothing is waiting on it.
• ...and it no longer runs while you are inside a dungeon. There was already a guard for this, and it had never once executed: it lived inside the closure the load gate only calls under /yh loadtest loot, which is the by-hand path where the guard is least wanted. On the normal path -- the one every session takes -- the gate says "go ahead" and the closure is never called, so thousands of journal calls ran on the way into keys exactly as before. The guard is now where it reads like it was: ahead of the work, on the path that does the work.
• The battle-res timer stopped listening to your own cooldowns. SPELL_UPDATE_CHARGES fires for every charge-based spell the character owns, so a Monk rolling around a world quest woke the handler several times a second -- and the handler re-validates the entire saved settings table and calls GetInstanceInfo before concluding "not in a raid, draw nothing". The event is now registered only while the icon is actually on screen; the events that can put it there were already covered.
• The home page stopped rebuilding for other addons' quests. QUEST_DATA_LOAD_RESULT is global -- it reports every quest anything in the UI asks for, and quest addons ask in the hundreds. Each one with a resolvable title triggered a full home-page relayout, and because a page stays mounted after the window closes, it did so on a window nobody was looking at. Now only the quests this addon asked about count, and only while the window is open.
• Crest IDs are resolved once a second at most, not once per coin. CURRENCY_DISPLAY_UPDATE is a gameplay event -- every turn-in, every boss, every mob that drops anything -- and each one ran ten pcalled currency lookups and handed ten result tables to the collector, to re-answer a question whose answer changes about once a season. Anything that draws a crest still resolves on the spot, so nothing shows a stale row.
• /yh trace costs nothing while it is off. Its frame watchdog is a parentless frame, which the client shows by default, so its OnUpdate ran every frame of every session to read one boolean and return. Trace could never have caught this: it does not wrap its own watchdog. The frame is hidden now until tracing is switched on.
• UNIT_INVENTORY_CHANGED is filtered to the player. The handler already discarded everyone else's -- but in Lua, after being woken for each of them. The client does that filtering for free.

Mythic+

• One redraw per burst of keystone replies, not one per reply. The page redrew on every incoming key, which was survivable while the only repliers were other YippYapp users -- in practice none. On the shared channel a single request into a large guild returns hundreds of replies inside a second, and this asked for hundreds of full page renders to draw a list that changed once. The library throttles what we send; nothing throttles what arrives.
• Keystones now arrive on the channel everyone is already on. Keys were shared over YippYapp's own addon prefix, so the Guild tab only ever listed people who also ran YippYapp -- in practice nobody, and its empty state told the player to go and evangelise an addon to fix it. The addon now embeds LibKeystone, the library BigWigs, DBM and EllesmereUI all carry. Its frame answers a request the moment the file loads, whether or not the host addon displays anything, so a guild or party full of BigWigs users answers. Party keys populate from the same callback. The old prefix still works, so nothing is lost for anyone who does run YippYapp.
  • Needs a full client restart, not a /reload -- it is a .toc change, and the Lua guards on the library being absent so a reload degrades to the old behaviour rather than erroring.
  • M+ rating rides along for free on this channel and is now recorded, though nothing displays it yet.

Mythic+ page

• A full group's keystones no longer run off the bottom of the page. The card grows to the floor and stops; the rows inside it were laid out at one fixed height however many there were, so a fourth key was drawn through the page and into the tab strip underneath. Two row heights now, the way the This Week card beside it already picks between a plain row and a detailed one -- roomy while it fits, tighter when it does not, and a "+N more" line as a backstop rather than a list that silently stops. Never seen before because this list only ever showed other YippYapp users; moving keystones onto LibKeystone filled it up and the layout met a full group for the first time.
• The Guild tab's Refresh button is gone; opening the tab asks. The button was parented to the whole window rather than to the tab's content and nothing hid it again, so once the Guild tab had been opened it stayed on screen over the Home tab's dungeon tiles for the rest of the session. It was also a button for something the addon can do on its own: asking is one throttled addon message, the replies arrive over the next few seconds, and the panel already redraws when they land — so the only thing the click ever added was the player having to know to click it. Throttled to one ask per ten seconds so flipping between tabs does not queue a request per switch.
• The vault strip above the page is gone. The dashboard already draws the Great Vault, in more detail and where a player looks for it, so this was the same three numbers a second time -- charging a whole row of height for them on the page with the worst vertical pressure in the addon. That row goes back to the cards, which is what was overflowing.
• Rating goals you have already passed are hidden. A full bar reading "2000 Done" is a row spent on something you cannot act on, in the card whose whole job is what to do next, and it pushed the focus list under it further down. The last milestone stays when every one is passed, so the card is never empty.

Loot browser

• A pass is refused if the journal ever answered for a boss we did not ask about. Every loot row names the encounter it belongs to, so a row naming a different one is proof the journal's single global selection moved between our EJ_SelectEncounter and our read — which is the actual mechanism behind the identical-loot bug. Checked per row, and tolerant of clients that do not supply the field: absent means "cannot verify", not "wrong".
  • This catches the case the shape check cannot: a selection that moves part-way through a sweep corrupts some rows and not others, which looks entirely plausible and would be cached and served as fact.
• A pass where every boss drops the same thing is refused, saved or loaded. That is not loot — it is one encounter's list copied across every row, which happens because the Encounter Journal has a single global selection and anything that re-selects underneath a running sweep makes EJ_GetLootInfoByIndex answer for whichever encounter won. Prevented upstream by never sweeping while the Adventure Guide is open, but prevention is a guess about every way the selection can move and this is a fact about the result. Checked on the way in as well as out, so a bad pass already written to disk is thrown away rather than only being prevented for players who never had one.
• /yh lootreset throws away all saved loot browser data — on disk as well as in memory — and rebuilds on the next open. The ordinary cache clear deliberately keeps the saved copy, which is right for routine invalidation and exactly wrong when the saved copy is the problem.
• The instance list is written down instead of rediscovered every session. Which raids and dungeons exist, and which bosses are in them, does not change until the game patches — yet finding it out cost a walk through every tier of every expansion in the Encounter Journal, once per login, whether or not anybody opened the panel. It is now saved and stamped with the game's interface version plus the addon's own, so a patch or an addon update looks again and nothing else does. A copy that is stale, malformed or empty is rebuilt rather than trusted, and an ordinary cache clear keeps it — only an explicit forget drops it.
• The loot itself is written down too, so a reload does not refetch it. Which items a boss drops is as fixed as which bosses exist, and fetching it is the bulk of the sweep — the instance walk finds a few dozen bosses, then every one of them is asked at every difficulty. Saved under the same stamp as the instance list, so a patch or an addon update invalidates both together and nothing else invalidates either. Stored whole, links included: the compact form would be item ids with the rest rebuilt on load, and that trades disk for a panel that renders blank until the client returns item data.
• The journal link index is written down too. GetJournalItemLink maps an item id to the journal's own link for it, and building that index is a third walk of the Encounter Journal — lazily, the first time the Best in Slot or Trinkets page asks about a row. Which item the journal lists for a boss changes when the game patches and at no other time. Saved under the same stamp as the instance list and the rows, so all three invalidate together and nothing else invalidates any of them. An empty index is still never accepted, saved or built: the journal is not always populated when first asked, and an empty one served as fact leaves every page showing base items with no way back.
• The login prewarm no longer runs while you are inside an instance. It is a nicety — it makes the first panel open feel instant — and it is thousands of Encounter Journal calls fired five seconds after a loading screen, on a client that is still busy with one. Zoning into a dungeon is the version of that where you can least afford it. It now waits until you are somewhere quiet.
• Redrawing the list no longer hides whatever tooltip you were reading. ReleaseAll called a bare GameTooltip:Hide(), and GameTooltip is one frame shared by the entire UI — so every redraw hid a bag item's tooltip, a spell's, another addon's panel. The list redraws on ITEM_DATA_LOAD_RESULT, which fires once per item as data streams in, so after a loading screen it fired in batches for as long as the client took to catch up. It now hides the tooltip only when the owner is inside the loot browser.
• Half the journal sweep never yielded. The instance cache is built inside a coroutine with a 16ms frame budget, and the dungeon walk hands the frame back at each tier and each instance -- but the raid walk did not yield anywhere at all, so every raid of every expansion, and every boss inside each one, ran in a single unbroken stretch. Neither boss loop yielded either. The budget the scanner is built around simply did not apply to half its work, and this runs a few seconds after a loading screen. Measured: the same sweep now spreads over 141 frames instead of 132.
• Bounded the journal walks. Each was a while true whose only exit was the Encounter Journal returning nil for the next index — the documented shape, and also a hang if the journal ever declines to say no, which is a real risk on the API least settled right after a loading screen. The caps sit far above anything the game ships (verified as having no effect on what is scanned); they are the difference between a bug that stops and one that takes the client with it.

Diagnostics

• /yh trace — times the addon's own functions and says, on screen, when one of them eats a frame. Built for a freeze that only happened after a loading screen: a hang throws no error so !BugGrabber has nothing to catch, and SavedVariables only flush on a clean exit, so anything written for the next login is lost when the client has to be killed. Counting as it happens is what is left. The setting survives a /reload, because the window worth watching is the one right after one. /yh trace report for totals — call counts as well as time, since the problem below was never a slow call.
  • It cannot see inside a loop that never returns; nothing in Lua can. That is what the load gate below is for.
  • The loot sweep is now named. The first real stall this caught was 9.6 seconds and reported "not inside anything we wrap" — true, and useless: the sweep runs in a coroutine driven by an OnUpdate script and the cache it builds is a file-local, so neither is reachable from a list of functions hanging off the namespace. Trace:Section times a block that runs to completion (the coroutine driver, where the frames are actually spent) and Trace:Mark names one that yields without timing it, since a parked coroutine's elapsed time is wall clock rather than work. A stall now says loot: sweep slice, loot: instance cache or loot: journal sweep.
  • The raid path is now watched too. Stalls of 3.6s and 6.1s landed on "Party converted to Raid" and an LFR zone-in rather than a plain loading screen, and nothing on that path was instrumented. Added RefreshRaidPage/Overview/Display/Groups, ScanRaid and the tier data helpers, plus interrupts: rebuild and readycheck: roster — both locals, named at the call site, both firing on every roster change.
  • LootBrowser_RefreshDisplay and RefreshWatermarks added to the wrap list — both redraw off item data streaming in, which is exactly what a bag full of new loot produces.
  • The frame watchdog now reports what share of a stalled frame was ours, instead of which function we were inside. The old version was broken by construction: an OnUpdate runs between frames, after every wrapped call has returned and cleared that field, so it printed "not inside anything we wrap" every single time — and a run of real multi-second stalls got read as the addon being innocent. Every wrapped call now adds its elapsed time to a per-frame total that clears between frames, so whatever ran during the frame that stalled is still counted when the next one arrives.
  • It reports at 250ms rather than 50ms. The frames right after a loading screen are legitimately long, so the lower threshold printed a wall of noise on every load and buried the one line worth reading.
• /yh loadtest — holds back the expensive jobs that normally run themselves a few seconds after a loading screen, gives each one a name, and runs them by hand from a client that is already up. Whichever command does not come back is the one, and it is the whole bisect in a single session rather than a restart per guess. Currently names loot (walks every tier of every expansion through the Encounter Journal), keystones (a guild request now returns a reply from every BigWigs user in it) and window (opening on login builds the dashboard and whatever page it lands on).
  • /yh loadtest only <name> is the mode that actually reproduces a loading screen: it lets exactly one job run where it normally runs, at the moment it normally runs, with the rest still held. One reload per job. Running a job by hand cannot reproduce load conditions and should not be read as if it does -- by then the caches it fills are already full, the journal has settled, item data has arrived and nothing else is competing for the frame.
  • By-hand runs now clear the job's own caches first and wait for its asynchronous tail, reporting the worst frame seen while waiting. The first version timed the handoff rather than the work and reported 0ms for every job -- a diagnostic that clears everything is worse than none.
  • Why a flag rather than a post-mortem: SavedVariables are flushed on a clean exit and a killed client never has one, so nothing a frozen session learns about its own death survives it. A flag set before the freeze does survive, because the /reload that triggers it writes the file on the way out.

Performance

• The six-second freeze, found and fixed. ns:RefreshWatermarks was measured taking 5911ms of a 6314ms frame — 94% of it — on a player zoning between a dungeon and the world. Sixteen slots in a single pass, each up to three C_ItemUpgrade calls, and those are not cheap while the item data behind them is still loading, which is exactly the state after a loading screen. It runs from the UNIT_INVENTORY_CHANGED handler, which fires when you zone.
  • It now asks about one slot per frame instead of all sixteen at once, so even a cold pass costs a slot per frame rather than a stall.
  • And it only asks about slots whose item actually changed since the last answer — which on a zone is usually none of them. A slot's mark is the client's answer about the item in it; with the same item equipped, the previous answer stands.
  • Marks only ever go up and are persisted, so arriving a few frames late costs nothing.
  • And a reload now asks nothing at all. The marks always survived a reload; what did not was the record of which item each one was an answer about, so every session re-asked all sixteen from scratch. A reload cannot change what is equipped, so a mark recorded last session against the piece still in the slot is still the client's answer. Only a slot whose item genuinely changed is asked.
  • A mark can also rise from a piece binding in the bags, which does not change the equipped item — that case is covered separately by BoundBagFloor, which reads the bags directly rather than asking the client.
• One draw of the suggestions panel asked the client for a slot's item over a thousand times. ns:GetSlotInfo is the bottom of nearly every gear question the addon asks, and only the tooltip half of it was cached -- the item link, the ItemLocation and two pcall'd C_Item reads ran every single time, measured at 1285 calls for one refresh. It is now cached on exactly the same lifetime as the tooltip scan it wraps, empty slots included, since both go stale together.
• ...and none of it was being cached in the window where it hurt. ScanUpgradeTrack records a result only when the tooltip parses. Right after a loading screen the item data has not arrived, so nothing parses, nothing is cached, and every one of those thousand-odd reads builds a full SetInventoryItem tooltip -- while several panel draws land in that same window as currency, bag and roster events arrive together. That is a stall that happens specifically after a loading screen and nowhere else. A read taken before the client is ready is now held for the current frame only: a burst inside one frame collapses to sixteen reads, and the next frame asks again and gets the real answer as soon as there is one.

Gear

• The advisor knows you can craft the thing. Several specs are told by their own best-in-slot list to MAKE a piece rather than kill something for it, and every rule in the upgrade advisor priced the vendor route only -- so it would happily talk you into spending the exact 80 crests the craft you are saving for needs. A slot the guide says to craft now reads "Craft instead", with what it costs, what it lands at and why that beats anything crests can do to what is in the slot today; the other rows on that tier say where the 80 went rather than quietly reporting less to spend. Gated on a Spark of Tides actually being in your bags -- sparks are the half that does not refill, and advice to hold crests for a craft you cannot start is just advice to stop spending. One craft held for at a time, and the tier it aims at is the best one this season's cap can still reach, not the first one you happen to be able to afford tonight.
  • A bare slot the list says to craft used to say nothing at all: the improvements list drops empty slots, which is exactly where a craft is the whole answer.
• The strip of per-wallet sentences above Improvements is gone. Three paragraphs about crest tiers sat between the player and the list of slots they were about -- on a page read for the next thing to click. Everything they said about a slot is on that slot's row and its hover. What is only true of a whole wallet -- how far the tier gets, how many weeks the shortfall is, what a craft has taken off the top -- moved onto the crest tiles beside the paper doll, where the balance already lives. Hover one.
• A crest tier is now counted over slots, not over the pieces you are wearing on it. Every demand figure in the addon priced the pieces currently on a track, which is the wrong set: a Champion crest is spent on every slot that will ever hold a Champion item, including the ones still wearing Veteran, still wearing nothing, or waiting on a drop. Counting only what was worn made the bill look small, the wallet look scarce, and every rule downstream reach for "hold". ns:GetSeasonDemand prices it forward instead, off the watermark: ranks at or under a slot's mark are free, so the cost of taking a slot to a track's cap is the ranks above its mark -- whatever is in it now, and whether or not anything is. A Myth piece takes a whole slot off the Hero bill, because the mark it sets is past the Hero cap. Ten Myth pieces leave six slots at five ranks each, and 600 Hero crests covers the lot.
• "Enough available to max all 6 slots without worrying about running short." When everything a tier could ever want fits inside what you can still get, the tier is not scarce, and the rules that ration it stand down: the reserve keeps nothing back, the hold-for-a-drop bet stops firing (with crests for every slot, both happen), "Hold crests" becomes "Upgrade later -- this one is covered too", and the overlap warnings -- find a Champion piece to skip the first Hero rank -- go quiet, because that trick saves crests and there is nothing to save. The sentence says which promise it is making: in hand is spend it tonight, available is you will not be short by the time the drops land.
• A slot's bill is priced off the mark it will have, not the one it has. A slot wearing a Champion 1/6 reads as five Hero ranks, but that piece gets taken to the Champion cap first -- the addon says so on its own row -- which marks the slot at 308 and leaves four. Ten Myth, five Hero and one Champion at 1/6 is 500 + 100 Champion + 80, not 600; the addon had been disagreeing with the player's arithmetic about an upgrade it had already recommended.
• A track you own no pieces of can now be advised on. Wearing nothing on a track ended the answer, which silenced exactly the case worth hearing: 600 Hero crests and no Hero piece yet means the crests are already sorted and only the drops are missing.
• The improvements list now says what a whole wallet can do, not only which slot is next. Three lines head it — the highest crest tracks with something left to buy — each answering the question a capped player actually opens the panel with: how much this track wants, what is in the wallet, and whether the season cap still allows the rest. The arithmetic has been in ns:GetTrackPolicyLine since it was written and nothing had ever called it.
• Held crests and earnable crests are stated apart. The budget they add up to was printed as one number described as income -- "120 to finish them all and 240 coming" -- which counted crests already in the bags as though none of them were there. It now reads "you hold 40 and the cap allows 200 more".
• "Need 80 more Champion" now says where the 80 comes from. Every rank in the spend plan carries a flag for whether the season cap reaches it, and nothing read it, so a slot two keys from affordable and one the cap cannot cover this week gave the same line. Three answers now: farmable this week, waiting on a reset, or -- only on a track the content has outgrown, where crests arrive as overflow from capping something higher -- wanting a drop instead.

v3.1.1 - Closing the window in combat (2026-08-19)

Fixed

• The X closes the window mid-fight. Once the Mythic+ or Teleports page has been opened, the window holds secure action tiles and a frame holding a protected frame is protected itself, so the client refuses to hide it and the close had to wait for the fight to end. The X now carries a secure handler, which is allowed to hide a frame it holds a reference to in combat -- the same mechanism that hides an action bar on a state driver.
• A window with no secure page open no longer waits either. Before Mythic+ or Teleports has been mounted nothing about the window is protected and an ordinary hide is legal mid-fight, but the close was deferred anyway, on the assumption rather than the fact.
• /yh, the minimap button and anything else script-driven still close at the end of the fight, because a snippet only runs off a real click and clicking a protected button from code is blocked in turn. The notice says where the X is rather than only promising to close later.
• Escape is unchanged and still does nothing mid-fight. Making it work means an override binding, and those cannot be cleared in combat either, so a window closed during a pull would go on swallowing Escape -- and the game menu with it -- until the fight ended.

v3.1.0 - The raid guide, the boss trainer, and healer trinkets (2026-08-19)

Raid

• A boss guide for the current raid, /yh guide. Every boss rewritten against a real per-boss source rather than paraphrased generalities: the Tidebound Grotto added, Nymrissa made playable, Nek'zali written by someone who actually watched the fight, Vashnik's three phases stated plainly as the guide's own sections, and the Soulcoil Well and Jawae's Echoes named as the fight names them. Laid out by a solver, so a card grows with its text instead of being guessed at a fixed width.
• A boss trainer, /yh train [boss]. Practise the mechanics without the raid: eight fights run end to end, the two-boss encounters get their second boss, the Sentinels' stasis is a solvable puzzle rather than a coin flip, alternating soaks alternate, Frostfire Volley carries one element and clears in the other, and the Coiled Altar's orbs survive the push. Difficulty comes off one dial rather than forty. Nothing fires or spawns before the countdown says GO, ranged hold station instead of running into melee, and every instant mechanic names itself as it happens. Per-frame garbage cut by 44%, measured.

Trinkets

• Healer rankings, from QE Live. bloodmallet sims damage and publishes nothing for the seven healing specs, so those now come from questionablyepic.com instead -- raid and dungeon, ranked at each trinket's own ceiling with the item-level curve that goes with it. Coverage goes from 31 specs to 38 of 40; Augmentation and Brewmaster are in neither source.
• The loot council list splits damage specs from healers. The two sources divide by different things to reach "percent behind your best" -- a share of total DPS against a share of the best trinket's healing -- so one sorted column ranked by which project simmed you rather than by who wants the item, and buried a Holy Priest's fourth best trinket under an Arcane Mage's sixteenth. Each side now has its own subhead saying what its percentages measure, with yours first. The colours follow the same split: the healer bands are the cutoffs that put the same share of rows in each colour as the damage bands do.
• Trinket tooltips split their five lines between the two sources, so a healer hovering a trinket is no longer shown five damage specs, and the attribution names whichever project ranked what you are reading.
• Re-scraped from bloodmallet. 21 specs came back with fresh runs. Assassination gained a single-target list it did not have, and Assassination AoE, both Outlaw lists, both Demonology lists and Protection Warrior's AoE moved onto this season's numbers. Eleven specs still carry last season's ranking for at least one fight style and say so, rather than being dropped.

Delves

• A Delves page. Your companion and where its next upgrade sits, the week's coffer keys, and the tier ladder. Everything is asked of the client at runtime -- no companion, faction, curio or season ids are written down, because a stale one does not error, it reads as zero and turns the page into a confident lie.

Elsewhere

• The windows that open on their own now share a home. Ready check, dungeon utility notes and the after-key summary can be dragged wherever you like whenever they are open, and each remembers its own place.
• A key binding to toggle the window, under Blizzard's own bindings.
• Stat priority is answered in one place. The Best in Slot page can afford to show two hero builds disagreeing; the character column has four rows and has to pick the build you are actually specced into. Both read the same parse of the same guide.
• Mythic+ utility notes per dungeon and class, generated rather than hand-kept. The mechanic-to-utility mapping is derived from Mythic Plus Utility's curated tags (by Nikyou); the prose is ours.

v3.0.6 - Post-tuning re-sim, and trinket shortlisting (2026-08-15)

Data

• Trinkets re-scraped after the 14 August trinket tuning. Twenty-four trinkets were retuned and the rankings moved a long way with them: Gebbo's Bottomless Bag (secondary effects -29%) falls from 1st to 6th for Shadow and 2nd to 7th for Fire, Hex Lord's Dooming Idol climbs to 2nd for Fire off its rework, and Vexhul's Everflowing Gland goes 17th to 4th. Gaze of the Alnseer (-20% primary) drops out of the list entirely for some specs. Thirty-four blocks carry today's sim.
• Zul'jin's Guillotine Technique now ranks for tanks, following the hotfix that let them roll Need on it -- top pick for Vengeance and Protection Paladin, second for Blood.
• Class guides re-scraped: a stat priority reorder, corrected boss sources, restructured Preservation Evoker builds and new bonus IDs.
• Consumables re-scraped: gem, food and health potion changes.

Trinkets

• Right-click a trinket to add it to your list. Shortlisted trinkets carry a star in both tabs. It is the same list the Loot Browser's favourites star writes to, so the two pages cannot drift -- and a shortlist rather than a slot assignment, because "I want this" is true of several trinkets at once when two of them are within a percent of each other and only one is going to drop.

Fixed

• Favourites were stored against the wrong specialization. The store read the Loot Browser's own dropdown, so a mark made from anywhere else landed under whichever spec that page happened to be showing, or went nowhere at all if it had never been opened. It now takes the spec explicitly, and pages outside the Loot Browser mark against the spec being played.

v3.0.5 - Season 2 trinket sims for 17 specs (2026-08-13)

Trinkets

• Season 2 rankings, single target and 5-target. bloodmallet has re-simmed 17 specs for the new season so far: all three Death Knight specs, Vengeance, all three Hunters, all three Mages, Protection Paladin, Shadow, Subtlety, Elemental, Enhancement, Affliction and Destruction. The rest are still on Season 1 numbers upstream and say so per spec rather than being dropped.
• Blood, Vengeance, Fire, Protection Paladin and Shadow gained a 5-target list they did not have, so the AoE tab no longer sends them back to single target.
• Scaling bars, and an item level stepper. Each row now carries a bar split at every item level that trinket was simmed at, the way bloodmallet draws it: length is the gain over an empty trinket slot, each segment is what the next item level added, and hovering a segment names the level, the gain and where the trinket drops. A stepper above the list walks the item levels and re-ranks everything at the one picked, so "this beats that at 311, and loses to it at 331" is a question the page can answer instead of only ever showing the best case. It is a stepper rather than a dropdown because the useful move is walking the ladder and watching rows overtake each other -- picking a number from a menu means guessing which number to pick.
• Ranking at a chosen item level only lists trinkets simmed at exactly that level, which is often a short list. Filling it out from each trinket's nearest step would rank a 331 against a 318 and present the item level gap as trinket quality. Item-level detail is kept for the top 15 per spec and fight style, and specs still on Season 1 have none -- that detail was never scraped and only a re-sim brings it back.
• Each trinket is ranked at its own item level. bloodmallet stopped simming everything over one shared range -- a crafted trinket caps lower than a raid drop, and a few items exist at exactly one item level -- so the scraper's old rule of comparing at the highest level every trinket shares matched nothing and returned an empty list for all 40 specs. Each trinket is now taken at its own ceiling, which is what bloodmallet's own chart ranks by, and that item level is printed on each row: two trinkets a hair apart can be a fair fight or a 13-item- level head start, and only the number beside them says which.
• Fixed: the ranking disagreed with its own numbers. Rows could read "#7 -2.7%, #8 -2.1%", ranking a trinket below one it beats. The order came from bloodmallet's ranking, which sorts by each trinket's own top item level, while the percentage was computed at the one item level they all shared -- two different questions. Affected every spec still on Season 1 data; on Balance Druid it had Crucible of Erratic Energies sitting at #22 on a number that belonged at #6.
• Loot Council lists this season's dungeon and raid drops only. It listed everything in the file: last season's trinkets, crafted ones, conquest gear and a world boss drop -- 73 items in single target where 37 of them are things a council actually hands out. Crafted and PvP trinkets sim fine and some of them are strong, but none of them is going to be linked in raid chat with five people typing "need". Everything else still answers a tooltip: hovering one, you already know it exists, and which specs ranked it is still worth saying.
• Trinkets from older content are hidden, with a checkbox to show them. bloodmallet goes on simming last tier's trinkets, and a couple from expansions ago -- Emberwing Feather, Algeth'ar Puzzle Box, even a Legion one -- so being in a current run does not mean an item drops now. What separates them is the upgrade ladder: a trinket from this season is simmed from 292 up to 331, 334 or 344, while anything carried forward sits at the single item level it capped out at. Ten of the 37 single-target entries went that way. "Show older trinkets" brings them back, marked, for anyone still wearing one.
• Trinket rows show the item that actually drops, not its base entry. Rows were built from the bare item ID, which resolves to the base item -- Rare quality, +7 Intellect, "Item Level 28" under a row ranking it at 334. They now use the Encounter Journal's link, the same source the Best in Slot page uses, so a trinket renders Epic with real stats and its own upgrade line. The item level is then rewritten to the level the page ranks it at and the rank named in place, "Upgrade Level: Myth 6/6", rather than tacked on at the bottom. Falls back to the bare item if the journal has not indexed yet -- it builds lazily and can miss the first ask.
• Raid trinkets sim at 344, which is above the top of the Myth track (334), so there is no rank to name and none is invented. Those say "above Myth 6/6" instead, rather than leaving a rewritten item level with nothing to read it against.
• Loot Council rows show the item level each trinket is ranked at -- its own ceiling, the top of the Myth track for wherever it drops, so 321, 334 and 344 all appear. Fixed rather than adjustable: that view answers "who should roll on this", and a ranking that reorders under a council mid-discussion is answering a different question badly. The item level stepper stays on My Spec, where the question is "which of these is better, and from what point".
• Expanding a trinket separates the specs that have been re-simmed from the ones that have not. Ranked specs are listed as before; underneath, every spec still awaiting a re-sim is named as "Not simmed for this season yet". Read as one list, a spec missing from the ranking looks like a spec that passed on the item, when the truth is that nobody has asked it yet. That group is the specs themselves, not the ones that ranked this particular trinket: this season's items were never in last season's lists, so the per-trinket version would come out empty for 28 of the 34 trinkets where the caveat is worth making. Browse order and the "best: #N" summary now come from the best re-simmed spec too, rather than from a placing made against last season's set of trinkets.
• Fixed: stat variants rendered as identical rows. Some trinkets are simmed once per stat -- "Drum of Renewed Bonds [Haste]", "[Crit]", "[Mastery]" -- and share a single item ID, so once the client resolved the real item name three separately-ranked rows became three identical ones. The suffix is kept.
• Fixed: the attribution line was drawn over a leftover icon. Rows come from a pool and the icon was only cleared by the callers that remembered to; once the header and the "Simmed ..." line moved left to align with the icon column, whichever trinket had used that frame last left its icon sitting under the text. Cleared centrally instead.
• The page keeps its proportions at any panel width. The panel is 700 to 960 units wide depending on the screen, and the row scale was clamped so it could never go below native size -- proportional on a wide screen, slightly oversized on a narrow one. It now scales with the panel throughout, so the bar, the name column and the item level occupy the same share of the page on every setup.
• Fixed: a scaling bar grew past the gain it represents. Sim noise puts some item levels slightly below the one before them -- 23 such steps in the current data. That step was drawn as a sliver, correctly, but the drop was then handed back to the next segment, which drew wider than the item level actually gained. Bars are meant to be proportional to one another, so a trinket with a few noisy steps read as stronger than one with the same final gain and none.
• Fixed: Loot Council rows could not be expanded. The overlay that scopes the item tooltip to the icon and the name was swallowing the click meant for the row underneath it. Enabling mouse motion on a frame still makes it a hit-test target, so the click lands there and is dropped unless the frame is explicitly told to pass it on.
• The item tooltip is now raised by the icon and the name only, and the region is measured from the name actually drawn rather than from the column it sits in. It used to hang off the whole row -- which included the bar, so crossing a segment replaced the segment's own tooltip with the item's -- and then off the full name column, which on a wide window is several hundred pixels of blank row.
• The ranked rows are drawn to suit the frame. Ten rows left half a wide window empty, which is the half that would have made the bars readable. Those rows now magnify to the width of the frame, up to 1.75x, so the bars, item icons and the numbers beside them grow together. The caveat, the spec header and the attribution stay at the size the rest of the addon uses -- they are prose, and magnifying them only makes the paragraph at the top shout. Loot Council is a lookup rather than a chart and is left at normal size throughout.
• The magnification comes from the frame and never from the content, so searching, switching spec, stepping an item level or pressing Show all leaves it exactly where it was: a list that resizes while you read it is worse than a small one. Resizing the window is the only thing that moves it, and it never shrinks below normal size.
• Bars follow the panel width rather than sitting at a fixed 130px in a window several times that wide, and are around 30% wider again.
• The item level and "vs best" column is left-aligned, so the item level is in the same place on every row instead of starting wherever the text before it happened to end. Bars stop clear of it rather than running into the number.
• Long trinket names are cut with an ellipsis instead of wrapping into a second line the row is not tall enough to show -- which cut them just as short, but with nothing to say they had been cut.
• Season is tracked per spec and fight style, not per file. A new season's sims arrive a few specs at a time over weeks, so re-scraping mid-transition used to delete every spec that had not been re-simmed yet -- trading a stale ranking for no ranking, which is worse. Those keep their Season 1 list, labelled as such in the spec list, on the loot council rows and in the item tooltip. The page-wide warning now fires only when nothing in the file is current. A spec can be Season 2 on single target and Season 1 on 5-target, and is marked per tab.

v3.0.0 - Mr. Yeeper, Best in Slot, and a season's worth of gear maths (2026-08-12)

Mr. Yeeper

• The Upgrade Summary is now a character. He reads your gear, crests, vault, keystones, raid lockouts and guild activity, works out which single fact is worth saying, and says it -- rudely.
• He is a reasoning layer, not a quote list. A snapshot gathers facts, a set of rules recognises situations and computes, and the personality is applied last to a conclusion that was already reached on evidence.
• Priority ordering decides what he says. Something being actively wasted (idle crests, a capped week) always outranks entertainment, because a joke is worthless while 200 crests rot in a bag.
• Actionable messages carry tips underneath: the "so what do I do" that a remark on its own leaves hanging.
• Season aware. Keystones go live a week after the patch, so during the first week he will not send you at Mythic+ content that does not exist yet, and points at Mythic 0 and Heroic instead.
• Socially aware. Remarks about never grouping with guildmates are gated behind actually having a guild with people in it. There is no version of that joke that lands when the honest answer is "I do not have one", so those lines are never generated rather than softened.
• Jokes are sourced from the community (punsnjokes.com, laffgaff.com, punorbit.com) and keyed to your class, race and level. His reluctance to deliver them is the second layer.
• Lines are picked at random with a short memory, so a strict rotation cannot make the running order predictable.
• He knows where keys stop paying. +10, +11 and +12 all drop 311 and all vault 318, so above +10 a higher key buys rating and nothing else. The game never says this and people grind +12s believing the vault is still climbing. The ceiling is derived from the key table rather than written down, so correcting a row moves the answer instead of leaving a stale number behind it, and he only raises it with players who have actually gone past it -- said to someone sitting at +5 it is not an observation, it is a reason to stop trying.

Best in Slot (new page)

• Every spec's best-in-slot list, scraped from Wowhead's class guides, laid out as a paper doll: the icons sit where the character panel puts them, so "how far off am I" is one glance rather than reading a list and rebuilding the shape in your head.
• Items resolve through the Encounter Journal, so they render as what they actually are -- Epic quality, real stats, the right item level. A modern item's base entry is Rare at item level 28 with "+5 Intellect", which is a real number answering a question nobody asked.
• Item level and upgrade rank come from the same GEAR_TRACKS the Loot Browser uses, so both pages can never disagree about what a rank means.
• Bags count. A drop you have not equipped yet is still a drop you got. The doll shows four states: missing, in your bags, have it at a lower rank, and have it at the target -- because "have it" and "have it at the rank this list means" are different answers.
• Stat priority per spec, and an honest caveat that this is a starting point rather than a substitute for simming.

Crests

• Fixed: the addon was reading the wrong currency. There are two complete sets of Mistcrest rows in the client and the addon pointed at the dead one, so a character holding 80 Champion Mistcrests read as zero -- and affordability, the waste warning and every upgrade recommendation were computed against an empty wallet. Both rows of a pair share a name and a description, so nothing in the game distinguishes them.
• Crest IDs now re-resolve whenever the wallet changes rather than once at login. Resolving once was worst exactly when it mattered: on a fresh season you have none of either row, so the pick was a coin toss that then never got revisited.
• Warns before a wasteful crest spend, not only after. Hero 5->6 and Myth 1->2 cover the same item level band at the same price, so spending the scarce crest there is dominated by spending the one that goes dead in three weeks.

Teleports

• Reorganised so every teleport has one home: the expansion its dungeon came from. The current season is a view over those same entries, not a copy, so a returning dungeon appears twice and cannot read as unlocked in one place and locked in the other.
• Fixed: returning dungeons read as locked. A dungeon brought back for a new season can be unlocked from either its original expansion or the current season, and only the season's spell was checked -- so a Ruby Life Pools port earned in Dragonflight showed as missing.
• Rolling to a new season is now editing one list of names.

Trinkets

• Top ten by default with a search box and autocomplete, instead of every trinket at once.
• Fixed: clearing the search did not reset the list. The clear button empties the box programmatically, and the handler ignored anything that was not typed -- so the text vanished while the filter stayed, with no way back short of a reload. Clicking a suggestion failed the same way.
• Fixed: the "Search" placeholder drew on top of what you typed.

Consumables

• Guides now scroll instead of running off the bottom of the panel.
• Tooltips follow the cursor rather than pinning to the far edge.

Fixed

• Blizzard settings panel threw on every module row. The module toggles it reads were never implemented, so opening settings errored -- and the two features that honour them were silently always on.
• Views only refreshed on the home page. Spending crests, upgrading an item or equipping something left the Gear Upgrades page unchanged until you navigated away and back. Every state change now reaches whichever view is actually on screen.
• The Great Vault filling was not signalled to anything, so a finished key left the Mythic+ page showing an empty slot.
• Omnium Folio's unlock check tested the first quest in its chain instead of the last, so it reported locked for anyone who had finished it.

Edit Mode

• Every movable frame -- interrupt tracker, battle res timer, ready check overview, Mythic+ summary and utility advisor -- now positions through Blizzard's Edit Mode via LibEditMode, with native selection outlines, "Click to Edit", grid snapping and proper settings dialogs.
• The old per-frame Unlock/Lock buttons are gone. They could not work: the settings window is not movable and covers the screen, so you unlocked a frame and then could not see it.
• A toggle panel anchored to the Edit Mode manager picks which frames are conjured for editing, since showing five at once buries the screen.
• Positions are stored per Edit Mode layout, so a raid layout and a solo layout can place the same frame differently.

Settings

• Rebuilt on Blizzard's Settings API. Real checkboxes, the Defaults button, keyboard navigation and options search -- inherited rather than reimplemented. Saved variables are untouched; every entry is a proxy over the existing getter and setter.
• One page instead of four subcategories. Once appearance and position moved to Edit Mode, what remained was a dozen switches.
• Appearance settings live in Edit Mode, where you can see the change as you make it. Each settings section points at it.

Gear: crest-efficiency warnings

• The addon now warns you before a wasteful crest spend. Every Season 2 track overlaps the one below it by two ranks, so Hero 5/6 -> 6/6 and Myth 1/6 -> 2/6 both land on 321 and both cost 20 crests. The game charges the same either way and never mentions that only one of those two currencies still has a season of use in it.
• Myth crests buy ranks 3-6 and nothing else does. Hero crests go dead the week every Hero-track slot is maxed, around week 3-4. Spending the scarcer crest inside the overlap is strictly dominated -- it buys an item level the cheaper crest was going to give you anyway.
• The same shape one tier down: Hero crests on Hero 1/6 -> 2/6 duplicate Champion 5/6 -> 6/6 at 308. That one only starts mattering once you are farming +10 keys and Champion crests have gone abundant, so it is gated on the profile rather than shown to everyone from day one.
• Slots get a red "Wasteful crest spend" verdict naming the crests at stake, the item level band, and which lower-track piece to put the cheap crests into instead. With nothing left on the lower track to spend on it softens back to the old "use cheaper crests" note -- the advice is still true, but there is no move to make.
• Mr. Yeeper reports it across the whole character, just under idle crests in priority. Several wasteful slots on different tiers are totalled per crest track, and the suggested sinks are grouped by which crest actually pays for them.
• Tracks whose crests are abundant for your profile are no longer nagged about at all. Veteran ranks used to draw the promotion note even though Veteran crests are free and there is nothing to save.

Fixes

• M9 great vault was wrong: 318 corrected to 315 (Hero 4/6). Myth 1/6 does not start until M10.
• That M9 correction only landed in one of the two key tables. The Progression page reads its own copy and went on showing 318, so the page and the upgrade advisor disagreed about the same key level. Both now say 315, with the duplicate pointed at the original as the source of truth.
• Mr. Yeeper claimed the vault was the only way to replace last season's gear. Dungeon and raid drops, delves, world content and crafting all do, and the tips now say so.
• "Biggest single jump" described an item's entire remaining track rather than one upgrade step, overstating it badly.
• The interrupt tracker never re-showed pooled slots after a rebuild, so the first rebuild worked and every one after it produced an empty frame.
• Test mode persisted across reloads, stranding the tracker showing five dummy bars with no switch left to turn it off.
• The tracker drew a bare bordered rectangle when it had nothing to show. It now hides when empty, and no longer draws an outer border at all -- each bar already has one.
• Frames could end up off-screen: the Edit Mode handoff discarded the relative anchor point, so a frame anchoring its TOPLEFT to UIParent's CENTER moved by half a screen.
• The settings panel could not be opened from anywhere -- the category ID was being overwritten with a string, so OpenToCategory silently failed.
• Progression's subtitle leaked through the app frame's own page title.

Out of beta

• The Interrupt Tracker, Mythic+ summary and Utility Advisor are no longer beta-flagged. They have been in use across a full season.

Also

• A jump counter on the character bar, live-updating. It does nothing.
• This "what's new" notice, shown once per version. /yh whatsnew reopens it.

v2.12.0 — Omnium Folio page + PTR shakedown (2026-08-09)

New: Omnium Folio page

• Tracks the five-week Rune chain from your real quest state (C_QuestLog), not a checklist you maintain by hand — so it is correct on every character and cannot drift.
• Shows live objective progress when a step is in your log, and falls back to counting the drops already in your bags when it is not.
• Calls out the unlock chain separately when you have not started it, and points at Magister Umbric in the Lycaneum for the weeklies.
• A Runes tab lists all five rows and their choices, greyed until the step that unlocks them.
• Clickable waypoints. Coordinates are buttons that drop a real Blizzard map pin and super-track it, rather than a /way string you have to copy. Shared helper, so any page can use it.
• Card-based layout with status stripes and pills, five progress pips, and hoverable item chips with real tooltips for the collectibles.
• The "Week N" labels are Blizzard's quest titles from when the chain released one step per reset. Anyone starting now is catching up and can usually take several at once, so the page says so rather than implying you have to wait a week between steps.
• Included because the Runes last the rest of Midnight: an unfinished Folio is permanent lost power going into Season 2.

New icon, and a 49 MB Media folder down to 82 KB

• New mark everywhere: a gold four-point compass rose on a runed purple plate, hand-painted in Blizzard icon style. Minimap button, app header, dashboard and the AddOns list all draw the same art.
• Four points, not eight: below ~24px each point needs roughly 3px of width to survive downsampling, and eight of them around a circle that small collapses into a cog.
• The minimap button now fills its ring. LibDBIcon draws the icon at 18x18 inside a 31x31 button and shows only the middle 90% of the texture, which left the mark floating small inside the tracking border. It is now 20x20 with the full texture visible. Press feedback is kept but inverted — the icon dips inward while held rather than sitting permanently cropped and springing outward on click.
• One full-bleed texture everywhere, with each caller deciding its own spacing by sizing its box. A cut with a built-in margin was tried for the inline icons, but once those boxes grew the padding only made the mark look small — spacing belongs to the layout, not the texture.
• The header icon is 32px, up from 22, and the header itself is 2px shorter with 6px less dead space beneath it. It is centred on the band between the window edge and the content panel rather than on the header alone, which is what made it look like it was riding high.
• The 8% texcoord crops on the app header and dashboard icons were removed; against full-bleed art they sliced the octagon's flat sides.
• /yh icon [size] [x] [y] tunes the minimap icon's size and offset live, since ring fit can only be judged on screen.
• The icon texture was 844x880 (not a power of two) and 2.8 MB; it is now a 128x128 32-bit TGA at 64 KB.
• bmc-logo-yellow.tga was 3200x3200 and 39 MB, displayed at 26x26. Resampled to 64x64 (16 KB).
• Removed icon.tga, minimapbutton.tga and bmc-button.tga — 7.3 MB that nothing in the addon referenced.
• Art/icon-master.png is the source of truth; Tools/make_icon.py re-cuts the whole set from it. Both are excluded from the packaged zip along with the CurseForge art.

Dashboard

• Upgrade Summary now counts slots holding gear with no upgrade track (last season's pieces). They were previously invisible — neither upgradeable nor maxed, just missing from every count.
• Upgrade Summary reports a crest shortfall when you cannot afford everything pending, excluding free watermark upgrades.
• Tonight's Plan surfaces the Bonus Roll threshold. Three filled vault slots unlocks it, counted across all rows combined, so it is easy to be one slot away without any single row hinting at it.

Everything below came out of testing against the live 12.1 PTR. Most of it is corrections to data or assumptions that only broke once real Season 2 content was in front of them.

Item levels corrected against the game

• Raid drops are rank 4/6 of their track, not rank 1. Verified from in-game tooltips (Raid Finder reads "Veteran 4/6 / 289"; Mythic reads "Myth 4/6 / 328"). LFR 289, Normal 302, Heroic 315, Mythic 328.
• Dungeon keystone levels re-verified against the community sheet's Dungeons Drops column. They walk one upgrade rank per step from Champion 1/6 at M0 to Hero 3/6 at +10, anchored on M0 = 292 which an in-game tooltip confirms.
• Where tracks overlap (305 is both Champion 5/6 and Hero 1/6) the higher track wins, matching what the game itself reports.

Loot Browser

• Difficulty selector. Pick the content level you actually run — Normal through M+12 for dungeons, Raid Finder through Mythic for raids — and every drop is priced at it. Previously every dungeon drop showed the Mythic 0 value forever.
• The selected difficulty now drives the item link too, so Blizzard's own tooltip agrees with ours. On keystones, where the journal has no scaled item to link, the tooltip's item level and upgrade-rank lines are rewritten in place and tagged with the difficulty.
• Tooltips show the upgrade rank ("Champion 3/6") alongside the item level, and the Great Vault value for that difficulty.
• Icon borders show item quality, not difficulty. Difficulty colouring was redundant once difficulty became a single global choice, and it read as an item-quality lie — epic raid loot rendered green on Raid Finder. Quality is read from the link, since the base item is Rare and the difficulty versions are promoted to Epic by their bonus IDs.
• Favourites is its own star toggle beside the slot dropdown instead of a row buried inside it, with a lit/dim state and a starred count.
• Removed the "Class:" / "Spec:" / "Slot:" / "Stats:" captions — every control already shows its own value.
• Housing decor is filtered by item class (20) instead of a tooltip string. Blizzard moved decor to its own class in Midnight and changed the tooltip wording, so a decor item was showing up under Sszorak.
• Fixed loot icons that rendered uncropped and could not be hovered: boss portraits share the icon pool and left EnableMouse(false), a wrong texcoord and a wrong size behind on recycled frames.

Gear Upgrades

• Season 1 gear is no longer read as upgradeable. Last season's items have their upgrade track stripped, and the code fell back to guessing the track from item level — so a locked S1 Mythic 272 piece was being offered as "Adventurer 3/6 → 282". The tooltip's track line is now the only accepted source; no line means not upgradeable.
• Added an off-track guard: if an item claims a track but its item level does not match that rank, it is flagged rather than silently trusted. Slots show "no track" or "off-track" instead of a bare row.

Fixes

• MAX_ACCOUNT_MACROS is nil at login — it lives in the load-on-demand Blizzard_MacroUI addon. Comparing against it threw and aborted creation of the /yh macro, which is why the drag-to-actionbar handle had nothing to pick up. Macro creation no longer consults it.
• Season 2 Mythic+ teleports added to the Mythic+ page, which had none. Entries accept multiple spell IDs so a dungeon reissued for a new season works whether you earned the old or new teleport.
• Progression page: removed the Season 1 raid wing tabs (Voidspire / Dreamrift / Quel'Danas) — Season 2 is one raid. Fixed delve crests rendering as "999-0" when no amount was known.
• Season name comes from one constant; two pages still said "Season 1".

Settings

• The settings panel is now also registered under Escape → Options → AddOns. It is one panel re-parented between hosts, not a duplicate, so your place in it follows you. Toggle which one opens by default under Settings → General.

Diagnostics

• /yh crests — which Mistcrest currency ID resolved per track.
• /yh lootdebug — Loot Browser difficulty state and per-boss data.
• /yh ejtest — probe the Encounter Journal's keystone item levels.

v2.11.0 — Loot Browser rework (2026-08-09)

Season 2 dungeons now resolve at runtime

• The dungeon list was a hardcoded table of Season 1 Encounter Journal instance IDs, so the dungeon view would have gone completely blank on the season rollover. It now reads C_ChallengeMode.GetMapTable() (locale-correct and self-updating every season) unioned with a named fallback list, so it survives the next rollover without a code change.
• Row backgrounds fall back to each instance's own Encounter Journal art instead of a hand-maintained texture table, so new dungeons look right with no lookup.

Grouped by boss

• Dungeon loot is now grouped boss-by-boss under a dungeon header, the same treatment raids already had, with boss portraits.
• Bosses with nothing for the current filter are hidden in dungeon view (a single-slot filter otherwise left ~30 empty rows to scroll past); raids still list every boss so the pull order stays readable.
• The old one-row-per-dungeon overview is still there behind the By boss toggle — it is the better shape for scanning many dungeons at once.

Upgrade marking

• Every drop is compared against what you have equipped in that slot and upgrades get a green +13 badge (new for an empty slot). The tooltip spells out the comparison.
• Paired slots (rings, trinkets, weapons) compare against the weaker of the two — the piece you would actually replace.
• Item levels come from the Season 2 tables in Core/Data.lua rather than the journal's item links, so the numbers are right immediately instead of waiting on item data to cache.
• Equipped item levels are cached per slot and dropped on gear change, so a refresh doesn't rescan your gear once per icon.
• Dungeon Mythic entries note that the journal figure is M0 and that keystones drop 295-311 — the journal has no concept of key levels.
• Toggle with the Upgrades chip.

Myth 9 bosses

• The penultimate and final bosses of the current raid are badged Myth 9 with their 344 item level, since they and Very Rare items are the only above-Myth-6/6 sources in the instance.

v2.10.0 — Season 2 data (2026-08-09)

Gearing rebuilt for Curse of Ula'tek

• New gear tracks. Adventurer 266-282, Veteran 279-295, Champion 292-308, Hero 305-321, Myth 318-334. Cross-checked against the in-game Mistcrest currency descriptions and norumu's community sheet, which agree exactly. Season 2 tracks are perfectly regular, so every track now overlaps the next by 2 free ranks (Veteran was 1 in Season 1).
• Above-Myth item levels are modelled separately in ns.ASCENDANT: Ascended Hero 328, Ascended Myth 341, and Myth 9 (344) for Very Rare items and the last two Mythic bosses.
• Great Vault raid rewards. LFR/Normal/Heroic vault slots now jump to rank 1 of the next track (Heroic vault = Myth 1/6, 318) and the Mythic vault always gives Myth 6/6 (334).
• Mistcrests replace Dawncrests. Blizzard ships two currency IDs per crest tier, so the addon resolves the live one at login instead of betting on a hardcoded ID. /yh crests shows what resolved.
• Progression tables rebuilt: M+, Venomous Abyss (8 bosses), Lairs, Delves, Trovehunter's, Prey, world content, crafting, Venomstones and Great Vault notes.
• Season 2 M+ rotation and its eight teleports.

Consumables

• ConsumablesData.lua is now generated, not hand-maintained. Rescrape all 40 specs with python Tools/scrape_consumables.py.
• Each spec records the season its guide was written for. Wowhead has not published Season 2 enchant/gem guides yet, so every spec currently shows a banner saying the advice is written for Season 1.

Trinkets (new page)

• Trinket rankings for 29 specs scraped from bloodmallet.com.
• My Spec tab: your trinkets ranked, with how far each is behind the best pick.
• Loot Council tab: pick a trinket and see every spec that sims it, best rank first — the "only Frost DKs should roll on this" view.
• Tooltip integration: hovering any trinket anywhere (loot window, bags, chat links, Encounter Journal) lists the specs that want it and where it lands for you. Toggle in Settings → General → Trinkets.
• bloodmallet has no data for the six healer specs, nor for Augmentation, Brewmaster, Windwalker or Assassination in this tier.
• Its data is still tagged MID1, i.e. Season 1 sims, and the page and tooltips say so.

Packaging

• Tools/ is excluded from CurseForge builds via .pkgmeta and ignored by git, so the scrapers never ship inside the addon.

Known gaps

• Utility Advisor has no entries for the Season 2 dungeons yet.
• Crafting item levels are inferred and need verifying in-game.

v2.9.0 — Pre-12.1 strip-down (2026-08-09)

Clearing the decks before patch 12.1 (2026-08-12). Everything removed here was either tied to a Midnight Season 1 encounter that stops existing on the 12th, or was guild-raid management rather than the "look it up in-game instead of googling it" job the addon is for.

Removed — dead on 12.1

• L'ura memory-game helper. Hardcoded to encounter ID 3183 (Midnight Falls). Gone along with Lura.lua, LuraUI.lua, the bundled pentagon symbol textures, the /yh lura command, and its Settings → Raid section.
• Nexus-King Salhadaar interrupt marker. Hardcoded to encounter ID 3179. Gone along with /yh nk and its settings checkbox.

Removed — off-mission

• Power Infusion auto-assign. All six Features/PowerInfusion files plus the PI target picker and PI-target column in Raid Tools. The scoring model needed a manual bloodmallet JSON re-paste every tier, which made it the highest-maintenance feature in the addon for the narrowest audience.
• Raid split analysis and swap suggestions. RaidSwapLogic.lua and the Split Balance / Suggested Swaps UI. The Raid Tools page keeps its group roster, composition counts, raid-buff strip and consumable audit.
• Release Spirit blocker. ReleaseBlocker.lua, /yh release, and its settings row.
• Raid Addon Scanner. AddonScanner.lua, the "Scan Addons" button and /yyhscan.
• Welcome / What's New window. Integrations/Welcome.lua and /yh welcome. It was a feature tour that went stale every patch.
• Xal'atath's Toes. The joke feature, its Settings → General "Fun" section, and /yh toes.

Notes

• Gear-track item levels, crest IDs, the dungeon pool, and the raid progression tables are still carrying Midnight Season 1 values. Those get updated in a follow-up pass once 12.1 numbers are confirmed.
• Removed features leave their old YippYappHelperDB keys behind harmlessly; they're simply no longer read.

v2.8.0 — Wipe-aware release blocker, M+ window movers, leak sweep (2026-05-13)

Bug fixes

• MoneyFrame "secret number" taint fix. The Loot Browser's EJ-suppression helper was calling SetScript("OnEvent", …) / Un/RegisterEvent on the EncounterJournal frame to silence its bulk-scan event chatter. Those writes tainted the EJ frame, which then poisoned MoneyFrame_Update on item-sell-price tooltips inside the Encounter Journal. Drop the script/event manipulation entirely; keep the filter save/restore, which is taint-free.

Mythic+

• Battle Res Timer now shows in Mythic+ keystone runs. Added difficulty ID 8 (Mythic Keystone) to the brez-difficulty whitelist and dropped the IsInRaid() gate, so the icon + cooldown swipe light up the moment you engage in a keystone.
• New Mythic+ settings tab. Completion Popup and Utility Advisor toggles moved off the General tab into their own dedicated tab, mirroring the Raid tab's grouping.
• Lock / Unlock for the M+ windows. Both the Completion Popup and the Utility Advisor now stay locked by default and grow a cyan border + "drag to move" overlay when unlocked — same affordance the Interrupt Tracker uses. Lock hides the overlay and dismisses the preview window so it doesn't linger on screen.

Release Spirit Blocker — wipe-only redesign

• Trigger moved from PLAYER_DEAD to PLAYER_REGEN_ENABLED. Dying mid-fight no longer blocks the popup — you're free to release for a corpse run if that's what you want. The blocker engages only once combat has actually ended and the player is still dead in a raid instance.
• Retry chain on popup-find race. If the death StaticPopup hasn't rendered by the time the block tries to attach, retries at +0.25 s and +0.5 s before giving up. Cancels as soon as the player ungohosts.
• Leader broadcast feature removed. No more YYH_RB addon messages between raid leaders' clients — the setting is now strictly personal. Per-user checkbox in Settings → Raid is the only control surface; the raid-tools header button is gone.
• Guild-majority check replaced with raid-instance check. Simpler and works for pug raids too.

Ready Check

• Cramped down sizing. Window scale 0.95 → 0.85, row heights 26 → 22 (full) and 20 → 17 (compact), header height 38 → 30, paddings + status icon + check icon all trimmed proportionally. Roughly 15–20 % shorter vertically on a 30-raider list.
• Preview / Hide buttons in settings. Test the window with a 24-player dummy roster (or hide it again) without needing to type /yyhrc.

L'ura helper

• Cleaner panel. Removed the duplicated inner title, every description/hint subtext block, and the "Solo Test (no raid required)" sub-section. Panel height dropped from 560 px to 360 px.
• /say debug mode gone. The L:SetSayDebug / L:SayTest / L:IsSayDebugOn helpers and CHAT_MSG_SAY listener are removed; /yh lura say/saytest sub-commands no longer exist.

Nexus-King Interrupt

• Stripped to one toggle. Removed the marker-size slider, color picker stubs, and Preview/Clear UI. Marker is now a fixed 36 px orange triangle. Slash command trimmed to /yh nk on|off.

Settings UI

• Wider content area. SettingsContentWidth() derives the panel width from the live AppFrame size (≈ 760 px on a typical screen) instead of the old hardcoded 500. Existing two-column layouts (Interrupt Tracker) and full-width panels both benefit.
• Interrupt Tracker BETA banner removed. Interrupt bars are no longer beta-flagged in the General sub-tab.

Performance — memory-leak sweep

• Interrupt Tracker slot pool. rebuild() now reuses slots from per-mode pools (bar/icon) instead of CreateFrame-ing fresh ones on every GROUP_ROSTER_UPDATE / zone change. Previously each rebuild orphaned up to 5 frames.
• Interrupt Tracker GUID-state pruning. I.state now drops entries for GUIDs no longer in the party on roster updates, so long pug-M+ sessions stop accumulating one entry per unique player ever seen.
• Pooled UI rebuilds. Progression raid cards, M+ Completion Popup rows, Utility Advisor icons, and Addon Scanner rows are now all backed by persistent pools instead of being rebuilt with fresh CreateFrame calls each time.

v2.7.1 — Upgrade advisor: fix "wait for 6/6" on Myth 2/6+ (2026-04-22)

• Myth 2/6+ now recommends upgrade instead of "wait for raid drop." The upgrade advisor was telling players with Myth 2/6 (or any rank below 6/6) to save their crests because "Myth 6/6 drops from Mythic raid." That's wrong on two counts: raid bosses drop a spread of ranks (3/6, 4/6, 5/6, only endbosses guarantee 6/6), and even if they did all drop 6/6, a 2/6 Myth item is valuable enough to upgrade now rather than sit on precious crests waiting for a specific drop.

• Root cause was a fallback in GetFarmableInfo that assumed raid reliably drops #levels (= 6) in the item's track when no M+ source was available. The same bug affected Hero items at Heroic raid tier (Hero 2/6 said "Hero 6/6 drops from Heroic raid, save crests"). Both now flow to the slot-priority upgrade rules (Rule 6-8).

• What Myth items recommend now:

  Rank	Path	Message
  Myth 1/6	Rule 3 (overlap promo)	"Max a Hero piece → free promo to 2/6"
  Myth 2-5/6	Rule 6-8 (slot prio)	"top priority" / "priority #N of X"
  Myth 6/6	Rule 5 (maxed)	"Fully upgraded"

• No Myth-specific carve-out required. M+ end-of-dungeon loot in Midnight S1 caps at ilvl 266 (M+12), which is below Myth 1/6 (ilvl 272), so the same-track check can't enter a "wait for M+ drop" path for Myth items by construction. If a future season adds M+ keys that drop Myth, the same-track rank-comparison branch would start firing again and recommend saving crests — behavior we'd probably want then anyway.

v2.7.0 — Raid split balancer rewrite + PI manual-only (2026-04-22)

Raid split balancer — group compaction rewrite

• Compaction is now raid-wide, not per-split. The old pass balanced each side (A / B) independently, which in odds mode produced the alternating 5,4,5,4,5,4 layout — side A filled to three fives, side B balanced to three fours. Visually it looked broken because the raid window isn't split-aware; it just shows a 1..maxGroup list. New pass computes a single target across all groups.

• Fill greedy, balance the last two groups. Formula: k = min(maxGroup - 2, floor(N / 5)) groups filled to 5 (or k = maxGroup when N = 5 * maxGroup exactly), then the remainder spread across the tail with ceil/floor(remainder / tailCount). Result:

  N	6-group layout
  22	5,5,5,5,1,1
  23	5,5,5,5,2,1
  24	5,5,5,5,2,2
  26	5,5,5,5,3,3
  27	5,5,5,5,4,3
  30	5,5,5,5,5,5

  And at smaller group counts: N=6, g=2 → 3,3; N=11, g=3 → 5,3,3. The 2-tail shape is a consequence of the g-2 clamp, not a hardcode — it generalizes.

• Side-crossing accepted. In odds mode a compaction move from G5 → G4 crosses sides, so split balance can shift by a body when the tail bridges the two halves. The balance panel rebalances on the next pass; the user's explicit ask was visual density.

Raid split balancer — plan stickiness

• Plan cached across refreshes. Single-row swaps used to call SwapRaidSubgroup, wait 0.5s, then re-run Suggest() from scratch. Because the whole sequence was re-ranked against the new roster state, clicking step 1 could quietly rewrite steps 2 / 3 into a different plan. The plan is now cached on ns._raidSwap, keyed by (mode, maxGroup), serialized by player name so it survives raid-index shifts.
• Reconcile on refresh, don't regenerate. Each refresh walks the cached plan and drops entries whose players are gone or whose swap/move is already satisfied (someone got moved manually). If any live entries remain, they're re-rendered and Suggest() is skipped entirely. Regeneration only happens when the cache is empty or the mode/maxGroup changed.
• Apply-All and Refresh flush. Apply-All clears the cache after dispatch (in case any SetRaidSubgroup silently no-oped on a full destination) so the next pass replans any leftover. The Refresh button is now the explicit escape hatch when PI picks or manual moves should produce a brand-new ranking.

Power Infusion — auto-assign dropped, manual only

• Auto button removed. WoW's inspect API is serial — one NotifyInspect at a time, ~0.6-2.3s per reply depending on the timeout path. In a 20-30-man raid the scan genuinely can't complete inside any reasonable UI budget, and the retry pass stacks on top. Auto-assign was printing "inspect data incomplete" more reliably than it was assigning PI. The scan-orchestration code (PI:EnsureScan, PI:IsDataReady, PI:AutoAssign, PI:Recommend, plus their helpers) is gone.
• Manual picker unchanged. Click a priest's row → pick a DPS target. When the tier / iLvl scan (run by ReadyCheck / the loot ranker) has happened, the picker still annotates rows with the bloodmallet ST score and a "(best)" tag; otherwise it falls back to alphabetical. Assignments still drive the swap balancer's "PI pairing" criterion.

Bug fixes

• ReadyCheck "secret value" taint crash. The per-unit aura scan was comparing a.spellId == 1459 (and four other stat IDs) in an if/elseif chain. In 11.x, aura fields from C_UnitAuras.GetAuraDataByIndex are marked "secret values" and a direct == compare while execution is tainted raises attempt to compare local 'sid' (a secret number value, while execution tainted by 'YippYappHelper'). Converted the chain to a STAT_BUFF_IDS[sid] table lookup — indexing with a secret key is safe, same pattern already used for flaskSet / foodSet / bronzeSet. Same class of crash as the L'ura msg:match fix in v2.6.0, just on the number side. Two unused helpers (UnitHasAuraBySpellID, FindAuraBySpellID) that would have tripped the same bug if wired up were removed.

Under the hood

• Inspect retry queue in RaidInspect.lua. Units that fail CanInspect on the first pass (out-of-range at the repair vendor, cross-phase, recently joined) are deferred to a retry buffer and re-tried once after the main queue drains, rather than being silently dropped. Helps the tier / iLvl / gear- quality panel populate more completely without needing a manual Rescan.

v2.6.0 — Tonight's Plan + L'ura revived (2026-04-19)

L'ura Memory-Game Helper — back, and this time it survives combat

• Root cause of the v2.0.x failure identified. Midnight's encounter system wraps CHAT_MSG_RAID / CHAT_MSG_RAID_LEADER payloads as "secret string values" — any index op (msg:match, msg:sub, msg:find, msg:lower) trips taint and crashes the handler with "attempt to index local 'msg' (a secret string value tainted by 'YippYappHelper')". The v2.0.x code did msg:match("YYL[1-5]"), which is exactly the forbidden op.
• Protocol rebuilt to NSRT's shape (with our symbols). Macros broadcast a short code (circle / diamond / t / triangle / x) via /raid; the receiver bakes the texture-path prefix into its SetFormattedText("|T<prefix>%s<suffix>|t", msg) format string, so msg flows through without ever being indexed or concatenated. No taint. Uses our bundled symbol TGAs inside the addon folder — no Interface\ICONS\ file-install hack like the NSRT+LuraMemoryFiles combo needs.
• Raid-leader-only listener. Only CHAT_MSG_RAID_LEADER drives the pentagon — regular raid chatter can't trigger a false render.
• Encounter-scoped by default. Listener is armed only during L'ura's memory-game windows (heroic: 10 / 80 / 150 s, mythic: 33 / 95 / 157 s + phase-4 re-arms). Checkbox flips to "always listen" as a fallback for cross-realm / /reload-mid-pull cases.
• Full UI in Raid Tools → L'ura tab. Enable toggle, scope toggle, Preview / Simulate / Clear buttons, Unlock / Lock mover, Create / refresh macros button, reference table (slot number, symbol, name, macro name), and a /say debug mode for solo end-to-end testing without a raid group.
• Macro generator. Creates 5 macros named YY_Lura_1..5 with body /raid <code> and a shared numeric icon FileDataID (not an addon path — those render transparently on macro slots). Deletes stale old-schema macros (YY_1..5, previous YY_Lura_*) before creating new ones so a broken prior save can't land on your bar.
• Chat payload shrunk from the full path Interface\AddOns\YippYappHelper\Media\Textures\symbol_<name> (~60 chars) to just the short code. The prefix lives in the receiver's format-string literal.

Ready Check dismissal + hover-to-pause

• Three-shape dismissal. Combat start (PLAYER_REGEN_DISABLED) snap-closes with no fade — the fight is live and the panel is in the way. Raid pull countdown fires the fade immediately. Normal finish / everyone-answered paths linger 3 s before fading so the final status stays readable.
• Hover-to-pause, both directions. If the cursor is over the window when a fade is about to start, the fade is held. If the user moves the cursor back into an already-fading window, the fade stops and alpha restores to 1. The instant the cursor leaves the window, the fade plays. Combat snap always wins — hover does not pause a combat-start dismiss.
• Ticker-based hover detection. Frame OnLeave fires when the cursor moves onto a child row, even though IsMouseOver() still reports true for the parent — unreliable signal. A 100 ms ticker polls IsMouseOver() and self-cancels once neither a pending fade nor an active fade remains.
• Position persistence fix. OnDragStop was computing the anchor via GetLeft/GetTop - UIParent:GetLeft/GetTop — brittle math that could flip signs or diverge under differing effective scales. Rewritten to use frame:GetPoint() directly and save point / relativePoint / x / y; show path restores with the matching SetPoint. Legacy anchorLeft / anchorTop still read as fallback.
• "X missing" summary removed. The counter conflated offline members with players missing buffs — in a small test group you'd see 1/2 ready · 1 missing where the "missing" was just the offline teammate. Per-row icons already show exactly which buffs / food / flask / durability are missing. Summary now just reads X/Y ready.

Raid Tools UI

• L'ura tab lives next to Overview with its own purple accent.
• Header row shifted so the Overview / Release Block / Scan Addons buttons no longer slice the inner-panel border.

Loot Browser bug fixes

• Stale-texture pool bug fixed. Icon pool entries cleared their itemID / itemLink on release but not the texture itself. When the pool handed a slot back out and the next row didn't explicitly set a texture, the old item's icon stayed visible — rendered with no tooltip and a dark-gray "unknown difficulty" border, which looked like an errant bag-slot icon. ReleaseAll now calls SetTexture(nil) on every pooled icon.
• Filtered Encounter Journal rows with no itemID. The EJ occasionally returns a loot entry that has a name and an icon but no itemID — unlinkable, untooltippable. Those entries now get dropped at scan time instead of rendering as a no-tooltip blank.

New features

• Tonight's Plan — replaces the static Farm Guide on the home dashboard with a live activity planner. Shows vault progress as three dot-strips (M+, Raid, Delves) and up to three prioritized actions with inline progress bars — "1 more M+ run » unlocks vault slot 2", "2 raid bosses » unlocks raid slot 2", "40 Hero crests left this week". Driven by a new Features/Planner/PlannerData.lua module that reads C_WeeklyRewards.GetActivities and the profile's precious-crest list. Auto-refreshes on currency / inventory update bursts.

• Utility Advisor description rewrite — the dense per-dungeon prose paragraph is now bulleted per mechanic, with color-coded response keywords (interrupt = yellow, purge / dispel = cyan, enrage = red, curse = violet, poison = green, bleed = dark red, stun = gold, fear = violet, slow/snare = sky, grip = tan, incap = pink) and bolded ability names (two-word Title Case auto-detected). The "lead" one-liner still sits above the bullets as an intro.

ReadyCheck rewrite (30-man perf pass)

• Persistent rows — removed the Acquire/Release pool churn; row frames are created once and reused in-place. Icons anchored at creation (previously 300 ClearAllPoints+SetPoint calls per render at 30 raiders). Header labels / column dividers anchored once at load instead of every render.
• Per-unit aura cache — ScanUnitAuras results cached per unit token, invalidated by UNIT_AURA(arg1) or GROUP_ROSTER_UPDATE or the start of a new READY_CHECK session. A UNIT_AURA for one unit now triggers a scan for that one unit; the other 29 hit the cache.
• Pulse animation no longer thrashes — _pulsing flag avoids the unconditional Stop() + Play() that restarted the eating animation every render.
• Dismiss window on time — removed the 3s post-finish delay and shortened the fade to ~1.5s, so the window visibly fades away "right as the 30 seconds end" instead of lingering with a tail.
• Pull-timer dismiss — registered START_TIMER; when a player countdown (/cd, /countdown, or BigWigs/DBM option to use the native countdown) starts, the window fades immediately.
• Cross-realm name keys — readyStatus now uses Ambiguate(name, "short") consistently at all read/write sites; previously mixed raw UnitName with short-ambiguated keys, causing the initiator's seed to be missed on cross-realm players.
• OnHide unregister — manual × close now stops UNIT_AURA tracking too; previously only the finish paths did.

Interrupt Tracker

• Bars no longer dim when on cooldown. The draining StatusBar was already the cooldown signal — the extra 0.5/0.85 alpha just made the bar harder to read.
• Input values from SavedVariables are now type-validated and clamped to sane ranges (barWidth 40–600, posY –8000 to 8000, etc.), so a corrupted SV can't produce a 0-width bar or an off-screen frame.

Settings

• Quick Access sidebar tile — the YippYapp macro drag icon is pinned to the bottom of the settings tab column, always visible regardless of which tab is open. "Drag to your bar, or type /yh" hint inline.
• Slash commands reference — new "Slash commands" section at the bottom of the General tab listing /yh, /keys, /yyhrc, /yyhopts, /yyhinterrupts, /yh toes off, /yh toes mover.

Fun

• Toes of the Harbinger — Xal'atath's feet now hover at the top of your screen. Yes, the feet. Just the feet. No, we will not be answering questions. Drag them somewhere, rotate them upside-down, flip them, resize them until they are an ominous 2× or a tasteful 0.5×. Saved across sessions, which is either charming or concerning depending on your relationship with cosmic horror. Turn them off in Settings → General → Fun when your raid group starts asking things, or type /yh toes off and pretend you never saw them. Inspired by Sakreble's Xaltoes addon, which itself is a structured version of a ModelScene one-liner by @Herotherogue on X.

Safety & perf

• ReleaseBlocker's mover overlay re-parented to UIParent (was parented to the DEATH StaticPopup's release button); insecure + mouse-enabled children on a frame whose OnClick runs protected code is a taint path. OnUpdate also throttled to 20Hz so holding CTRL for the override doesn't burn per-frame CPU during combat.
• RaidInspect GUID scan replaced with a GUID→unit map rebuilt on GROUP_ROSTER_UPDATE / PLAYER_ENTERING_WORLD. Fallback refresh on stale lookup. Previously O(n) scan per INSPECT_READY of up to 30 units.
• MythicPlusData guild-keystone reply uses a 15s cooldown (lastGuildKSReplyAt) so a burst of KSQ requests after a Tuesday reset doesn't schedule 100+ independent C_Timer.After reply timers.
• PVETab defers PanelTemplates_DeselectTab via C_Timer.After(0, …) so it never runs inline with Blizzard's click dispatch.
• Core/Core.lua :SendItemToUpgrade has an explicit InCombatLockdown() guard now (the right-click path into it was bypassing the caller's guard).
• Welcome UISpecialFrames insert is idempotent — fast Show/Hide/Show sequences no longer stack duplicate entries.
• Consolidated CONSUMABLE_SPELL_IDS (food / flask / bronze) into Core/Data.lua — RaidUI used to carry stale Dragonflight flask IDs that no longer matched Midnight S1 consumables.
• Dashboard crest-refresh builds a single track → crestData map per tick instead of nested scans (O(n²) → O(n)).
• ScheduleRefresh in the gear panel coalesces rapid clicks via a pending flag.
• Removed duplicate ns.CREST_COST_PER_UPGRADE constant (was dead code; ns.BASE_CREST_COST in Recommend.lua is the sole source).
• CompletionPopup party class lookup uses party-unit tokens instead of bare names — fixes missing class colors for cross-realm / out-of- range cached keystone senders.

v2.5.0 - Midnight-ready (2026-04-15)

New features

• Interrupt Tracker (BETA) — party-wide interrupt cooldown bars.

  • Per-member bars with class-colored fill, class-colored border option, icon, name, and countdown/elapsed timer.
  • Tracks the player's own casts directly; attributes party interrupts via a time-correlation model (ExwindTools-style) that sidesteps Midnight's secret-value wrapping on friendly spellIDs.
  • Fully configurable: bar/icon mode, texture, dimensions, orientation, grow direction, anchor, multi-context visibility filters, backdrop, and drag-to-move with an unlock overlay.
  • Slash commands: /yyhintlog debug log, /yyhintdebug class rotation.

• Mythic+ completion popup (BETA) — small window on CHALLENGE_MODE_COMPLETED showing your new M+ rating, the party's new keystones in Mythic+ page styling, and a button that navigates the app frame to the M+ page. Row click casts the teleport directly via InsecureActionButtonTemplate (BigWigs-style) so it works even when the popup was opened from a tainted click path.

• Utility advisor (BETA) — on entering a Mythic+ (or Mythic) dungeon, shows a small window with per-class recommended utility spells for that dungeon plus a short description. Spell icons hover to full Blizzard tooltips. Gold border marks talent-dependent picks. Show-once guard so it doesn't spam on reloads or re-entries.

• Raid Addon Scanner — button in the Raid Tools header that pings the party/raid via addon-comm and shows a modal listing who has YippYappHelper installed vs not.

Settings overhaul

• Removed the Blizzard AddOn Options integration. All settings now live in the addon's own Settings page (accessible via the gear button, /yyhopts, /yyhsettings, or /yyhinterrupts).
• Top-level tabs (General / Interrupt Tracker) with vertical sub-nav inside Interrupt Tracker for its many sections.
• Feature-row widget with inline "Preview" buttons so you can test what a popup looks like before enabling it in live play.
• "Show in" visibility is now a multi-select (any of: Always / In a group / Dungeons & raids / Mythic+ / Raid / PvP).

Mythic+ page

• Views are now strictly party-scoped (5-man) regardless of whether you're in a raid — no more raid-member keystones polluting the grid.
• Tabs restyled to underline tabs matching the Loot Browser.
• /yyhmplustest slash command injects 5 fake teammates with synthetic keystones and scores so you can preview the full group layout solo.
• Dungeon-tile tooltips now use an O(1) mapID-indexed lookup instead of rescanning the keystone list per hover.

Raid Tools

• L'ura Runes feature removed. Midnight locks down chat text, addon-comm, raid markers, and custom channels during raid encounters — there is no unprivileged signaling primitive left that works cross- realm in a raid instance.
• Ready Check performance: single-pass aura scanner (replaces 8 separate FindAura calls per member — ~8x fewer GetAuraDataByIndex calls at 30-man), UNIT_AURA debounced to 0.5s, READY_CHECK_CONFIRM throttled through QueueRefresh, CHAT_MSG_ADDON durability updates coalesced.
• Durability broadcast staggered 0–1.5s to avoid CHAT_MSG_ADDON flood on ready-check in full raids.
• Durability name-key mismatch fixed (cross-realm raiders no longer show as "not applicable").
• Early-finish detection: window fades as soon as everyone answers, rather than waiting the full 30s timer.
• "Hearty" and "Well Fed" prefix patterns added to food detection.
• ReadyCheck hide delay reduced 8s → 3s after everyone answers.
• Respond() now fires ConfirmReadyCheck before any UI state touches, plus optimistic local status update so the row flips instantly.

UI consistency

• AppFrame uses a darker, flatter background (0.03 on black at 0.92 alpha) with a 1px black border.
• Every page (Gear, Raid, Progression, Loot, Consumables, M+, Teleports, Settings, Home) gets the same inner bordered panel — unified look.
• Settings window lives inside the AppFrame as a page; Home-dashboard and AppFrame header both have a "Settings" button that routes there.
• Loot Browser and Consumables headers in app-mode dropped their card border in favor of a single 1px bottom divider under the filter row.
• All the in-addon widgets (checkbox, slider, dropdown, tab) rebuilt to match the addon's dark palette — no more WoW default styling.

Performance

• Interrupt slot OnUpdate throttled to 20 Hz and caches settings- derived flags on the slot.
• Interrupt events short-circuit when the feature is disabled.
• Ready Check aura scanner batches 8+ independent aura iterations into a single pass per member per render.
• Mythic+ RefreshMythicPlus pre-builds membersByName and keystonesByMapID lookups to avoid inner-loop rescans.
• Loot Browser tooltipHiddenCache is now bounded (2048-entry cap with wipe-on-fill).
• AppFrame uses UISpecialFrames only out of combat to avoid the ADDON_ACTION_BLOCKED taint we hit at combat end.

Bug fixes

• Interrupt bar no longer snaps back when dragging to a new position (atomic write + manual applyPosition instead of listener cascade).
• Interrupt bar no longer jumps out from under the cursor when GROUP_ROSTER_UPDATE fires mid-drag.
• Active interrupt cooldowns now survive a rebuild — reattaching state from ns.Interrupts.state to freshly-created slots so GROUP_ROSTER doesn't visually reset ongoing bars.
• Interrupt attribution flash no longer hides the slot (was passing showWhenDone=false to UIFrameFlash).
• Loot Browser filter row shifted down in app mode to clear the inner panel top edge.
• Consumables header matches Loot Browser styling in app mode.
• Teleports / Progression / M+ page layouts nudged to fit the new inner panel without gaps or overlaps.
• All the tainted-value handlers we had to add while chasing the Midnight secret-value system are removed now that we've landed on ExwindTools' taint-free attribution model.

---

v2.0.1 - Polish pass (2026-04-14)

L'ura Runes

• Macros renamed from O D G T X to YY_1 - YY_5 with a blank question-mark icon. Raid leader copies the bundled symbol TGAs into Interface\ICONS\, restarts WoW, then manually picks each symbol from the macro icon browser's Items tab.
• Tab fully redesigned into two full-width sections (Settings & Simulation, Raid Leader Macros) with a side-by-side reference table showing each macro number, its symbol icon, and the filename.
• Macro listener now only responds to CHAT_MSG_RAID_LEADER, not CHAT_MSG_RAID - so only raid leaders / assists can drive the pentagon display.
• Removed the per-slot icon picker; bundled symbol TGAs are the only display textures.
• Description now correctly states runes fill the pentagon right-to-left in press order.

Ready Check

• Response tracking rewritten to capture READY_CHECK_CONFIRM arguments directly into an internal status map (MRT-style) instead of polling GetReadyCheckStatus, which could be stale for a tick after the event fired.
• Blizzard's own ready-check popup is now dismissed reliably (multi-tick safety sweep).
• Footer collapses after you respond so the window shrinks up to the bottom of the roster.

UI polish

• Main app window close button restyled to match the Ready Check / Settings header buttons.
• Buy Me a Coffee icon moved to the home dashboard bottom-right with "if you want to support" label.
• Welcome / What's New window: Unicode em-dashes and arrows replaced with ASCII (no more square glyphs on certain fonts), and bullet wrapping estimator now strips color codes so long bullets stop overlapping.
• L'ura tab reference header reworded from "Reference" to "Pick this icon for each macro:" with explicit Items-tab instruction in the guide.

Other

• /keys / /yhkeys slash command opens the Mythic+ page directly.
• Settings panel two-way sync: toggling the in-raid-frame Release Blocker button now updates the settings checkbox immediately.
• Guild keystone cache: 1-week TTL so stale entries from players who no longer have keys age out on their own.
• Loot Browser cache no longer invalidates while the panel is closed - keeps re-opens instant.
• Deleted orphaned GetTrinketMythInfo / GetTrinketMaxIlvl helpers left over after Trinket Rankings was retired.

v2.0.0 - Raid companion rewrite (2026-04-13)

New features

• L'ura Memory-Game Helper - live rune pentagon display, macro generator (O D G T X), per-slot icon picker, bundled rune symbol textures, and built-in simulation for dry-runs. Works in both Heroic and Mythic with press-order rendering.
• Ready Check Window - opens automatically on any ready check. Shows food, flask, vantus rune, Int/AP/Vers/Stam/Haste/Move, and durability % for every raider. Collapse/expand toggle, countdown timer, Ready/Not-Ready buttons, and live refresh as buffs change mid-check.
• Settings Panel - full entry in Blizzard's standard Options UI. Toggle Release Blocker, Ready Check, L'ura Runes, and the minimap button. /yhopts opens it directly.
• Keys tab on the group-finder window - opens the Mythic+ page without closing the current LFG panel.

Major improvements

• Loot Browser: class dropdown (inspect any class), horizontal-scroll icon rows, housing decor + recipe filter, mount border distinction, and login prewarm so re-opens feel instant.
• Mythic+ Guild Keystones: auto-request on login, manual Refresh button that sits cleanly next to the Vault row.
• Raid Split Balance: added melee vs ranged DPS counts for each split.
• Dashboard panels restyled with accent underlines and consistent typography. Upgrade Summary rewritten into readable "Slot - Action (reason)" lines.
• Release Blocker simplified to work in any raid zone with guild-majority detection.
• New YippYappHelper logo used for the minimap button and in-app header.
• Full folder reorganization: Core/, Features/, Integrations/, Media/.

Retired

• Trinket Rankings (too expensive to keep accurate via sim).
• Raid Tools - Tier Tracker tab.
• Raid Tools - Performance tab.

v1.4.0 — Mythic+, Teleports & More (2026-03-30)

Mythic+ Helper (new)

• Dungeon overview with clickable teleport icons showing your best key level and score
• Group ratings grid with color-coded per-dungeon scores for all party members
• Group keystones section with dungeon icons and class colors
• Great Vault tracker with Blizzard-native tooltips — click slots to open vault
• Rating goals (2000/2500/3000) with progress bars and focus dungeons
• Guild tab for viewing guild members' keystones (shared via addon comms)
• Home/Guild tab navigation
• Accessible from dashboard, app nav, and /yh mplus

Dungeon Teleports (new)

• All Hero's Path dungeon teleports from MoP through Midnight
• Organized by expansion with clickable secure buttons
• Greyed-out icons for teleports not yet unlocked
• Accessible from dashboard, app nav

Release Spirit Blocker (new)

• Hides the Release Spirit button in current expansion guild raids
• Hold CTRL for 1 second to override — progress bar shows hold time
• Does not interfere with combat resurrections
• Toggle with /yh release (enabled by default)

Loot Browser

• Instance-aware filtering: detects current dungeon/raid and shows only that instance's loot
• Slot tabs grey out when no items match the instance + stat filter combo
• Auto-enables instance filter when entering a dungeon

Dashboard

• Upgrade summary and farm guide now auto-refresh on gear/crest changes
• Top 3 actionable recommendations sorted by priority

Trinket Rankings

• Fixed Hunter Beast Mastery spec key mismatch in "Your Spec" mode
• Fixed Rogue Assassination mapped to Outlaw rankings
• Added missing AOE trinket rankings for all Demon Hunter specs

Raid Tools

• Fixed swap suggestions reusing the same player as swap partner across multiple triggers

UI

• All frames use dynamic sizing based on screen resolution
• Responsive dungeon icons scale to fit available width
• Font sizes scale proportionally at different UI scales

v1.3.6 — What's New Screen (2026-03-26)

• Updated What's New popup to show v1.3 and v1.3.5 highlights
• Removed features list from welcome screen, now shows only what's new

v1.3.5 — Polish & Fixes (2026-03-26)

Consumables

• Shift-click items to link in chat or search the Auction House

Bug Fixes

• Fixed sub-panels (Gear, Raid Tools, Loot Browser, Consumables) being independently draggable when embedded in the app shell
• Fixed consumables icon pool contamination causing greyed-out icons
• Added nil-safety to consumables object pool functions

v1.3.0 — Consumables Guide, Loot Browser & More (2026-03-26)

Consumables Guide (new)

• Full enchant, gem, and consumable recommendations for every class and spec (all 40 specs)
• Three-tab layout: Enchants, Gems, Consumables — each with item icons, quality-colored links, and tooltips on hover
• Class dropdown and spec buttons to browse any spec's recommendations (defaults to your class/spec)
• Guide text below each tab with detailed reasoning (flask choices, potion tradeoffs, weapon buff notes, food comparisons)
• Alternative items shown inline with "or" label (e.g. flask alternatives)
• Accessible from dashboard, app nav bar, and integrated as full page in the app shell

Loot Browser (new)

• Browse all dungeon, raid, and world boss loot by slot
• EJ-based scanning with spec and secondary stat filtering (Crit/Haste/Mastery/Vers)
• Two-row slot tabs: armor on top, accessories + weapons on bottom
• Difficulty columns (Normal/Heroic/Mythic for dungeons, LFR/N/H/M for raids)
• Item icons with difficulty-colored borders, tooltips on hover, shift-click to link
• Trinket tier badges (S/A/B) cross-referenced from SimC rankings
• Spec dropdown to browse loot for any spec
• Remembers selected slot across open/close (defaults to Head)
• Accessible from dashboard, app nav bar, and /yh loot

Raid Tools

• Group display reordered: top row shows groups 1, 3, 5 and bottom row shows 2, 4, 6 for split visualization
• DK Grip distribution: suggests splitting Death Knights across raid splits for Mass Grip coverage
• Healer stacking fix: only suggests swaps when source split has 3+ healers (5-healer raids handled correctly)

Profile System Simplified

• 6 profiles → 3: Normal (Casual/Delves/M+ up to 8), Heroic (Heroic Raid/M+ 10), Mythic (Mythic Raid/M+ 10+)
• Old profile IDs auto-migrate (heroic_raider → heroic, mplus_high → heroic, etc.)
• Each profile defines farmable track for smarter recommendations

Recommendation Engine Improvements

• SAVE_FOR_DROP: "Hero drops from M+6-10 / Heroic raid — save Champion crests, wait for replacement"
  • Only triggers for low-rank items (1-3/6) with precious crests
  • Shows source hint (M+ key range + raid tier)
• CREST_CAPPED: "Need 20 Champion — capped until next reset"
  • Detects both weekly caps (Hero/Myth) and season cumulative caps
• Myth 2/6+ always recommended for upgrade (top track, nothing replaces it)
• Myth 1/6 still suggests using Hero crests for free promotion via overlap

Trinket Rankings Refreshed

• Fresh SimC data: 60,000 iterations, 30 specs (ST + AOE), latest nightly profiles
• New spec: Rogue Assassination
• Item level shown on all trinket rows (Myth 6/6 = 289)
• Tooltips now show Myth 6/6 stats; shift-click links the Myth 6/6 version
• Removed legacy trinkets not in current loot pool

UI Polish

• Unified underline tab style across all panels (Raid Tools, Consumables, Loot Browser)
• Dashboard app nav buttons in a 3x2 grid instead of single row
• Magisters' Terrace trinket sources corrected

v1.2.0 — Raid Tools, Data Overhaul & Polish (2026-03-25)

Recommendation System Reworked

• Crests can only be spent on their own track — addon no longer suggests "hold crests for higher track gear"
• Items now show explicit upgrade priority: "Upgrade 1st of 5 Champion", "Upgrade 2nd", etc.
• Priority based on slot value (weapons/chest/legs first, neck/wrist last)
• Crest scarcity calculated from total season budget (current + still earnable), not just inventory
• Overlap zone fix: "Use cheaper crests" only shows at rank 1 (not rank 2 where free ranks are already used)
• Profile changes instantly refresh all recommendations on the dashboard

Raid Tools — Performance Tab (new)

• DPS rankings from Blizzard's built-in Damage Meter API (C_DamageMeter)
• Average DPS per player per split (A vs B) with imbalance detection
• Avoidable damage taken leaderboard
• Death count tracker
• Interrupt count tracker
• Uses current combat segment, falls back to overall

Raid Tools — Overview Improvements

• Two-column layout: groups + swaps on left, composition + consumables on right
• Consumable checker: flask, food, augment rune status per player (spell ID based, taint-safe)
• Missing players listed by name next to each consumable type
• Raid buff list: one per line, cleaner layout
• Composition shown in large font with role breakdown
• Swap suggestions now show player names in bordered badges with class-colored borders
• "Swap All" button executes all suggested swaps via SwapRaidSubgroup API (raid leader/assistant only, out of combat)
• Debuff class distribution: suggests spreading Monk (Mystic Touch) and DH (Chaos Brand) across splits
• PI targets need same split (not same group) — corrected from 40yd range
• Player count balance: suggests moving healer if it fixes both size and healer imbalance
• Auto-refreshes on GROUP_ROSTER_UPDATE (instant update when players are moved)
• Group display always reserves 2 rows (no layout shift with fewer than 4 groups)

Season 1 Data Updated

• Gear tracks: Champion 249/252/255, Hero 262/265/268, Myth 275/278/281 (ranks 2-4 corrected)
• Champion free ranks from Veteran: 2 (was 1)
• M+ crests: Heroic=Adventurer, M0=Champion, +2-3=Champion, +4-8=Hero, +9-12=Myth
• Raid: per-boss crest amounts, Void 6 bonus crest (20 base + 10 next tier), per-wing detail view
• Raid progression: clickable wing buttons (All / Voidspire / Dreamrift / Quel'Danas) with boss-by-boss detail
• Delves: Tiers 1-4 give Adventurer crests, Tier 10=Champion, Tier 11=Gilded Stash (10 Hero + 5 Myth)
• Delve cards grouped by crest type: Tier 1-4, 5-6, 7-10, 11
• Trovehunter's Bounty Map data added (tiers 4-11)
• Prey: Normal=Adventurer, Hard=Veteran, Nightmare=Champion
• Crafting: added Veteran tier, Spark of Radiance (no crests) tier
• M+ cards show ilvl ranges with track names (e.g. "259 - 263 Hero")
• M+ grouped as +2-5, +6-8, +9, +10-12

Profile System

• Clickable dropdown on dashboard to switch profiles
• Profile change refreshes dashboard, gear recommendations, and suggestions instantly
• Hero crests now precious for Mythic Raider and M+ High profiles (weekly cap makes them limited)
• "What's New" window shows on version update with profile selector built in

Trinket Panel — ChonkyCharacterSheet Support

• Trinket tab and panel now anchor to the actual visual right edge of the character frame
• Detects ChonkyCharacterSheet and anchors to CharacterFrameBg instead of CharacterFrame
• Panel inherits CharacterFrame scale when Chonky is active
• OnSizeChanged hook re-anchors on dynamic resize
• "All Specs" button collapses sidebar then opens app (no more panel stealing)

Minimap Button — LibDBIcon

• Replaced custom minimap button with LibDBIcon-1.0
• Works correctly with ElvUI square minimap, SexyMap, and other minimap addons
• Draggable, position saved automatically by the library
• Embedded libraries: LibStub, CallbackHandler-1.0, LibDataBroker-1.1, LibDBIcon-1.0

Visual Polish (inspired by Plumber addon)

• DisableSharpening on all frame borders for smooth edges at any UI scale
• ADD blend mode highlights on hover (subtle glow instead of flat color change)
• Text shadows on all headers and titles for depth
• Consistent color hierarchy: primary (0.92), secondary (0.55), tertiary (0.35)
• All frames adapt to screen resolution (dynamic sizing via GetAppFrameSize)
• All popups clamped to screen (trinket detail, PI picker, profile dropdown)
• Frame z-ordering fixed: sub-panels normalized when inside app, no addon interleaving

What's New Window

• Welcome screen now doubles as a "What's New" popup on version updates
• Shows "Welcome to YippYapp Helper" on first install, "What's New" on updates
• Features list + version-specific highlights
• Profile selector built in for immediate setup
• "Got it" saves the version — only shows again on next update
• /yh whatsnew to re-show

Bug Fixes

• Frost DK and Devourer DH spec keys corrected for trinket rankings
• TIER_QUALITY defined before use (trinket detail popup quality borders)
• Swap badge frames and PI picker rows now pooled (memory leak fix)
• Tier tracker rows pooled (memory leak fix)
• Tainted aura name strings handled via spell ID lookup + pcall fallback
• Dashboard.lua removed call to nonexistent RefreshRaidTrinketDisplay
• quickBtnText dead references removed

---

v1.1.0 — Midnight Season 1 Overhaul (2026-03-23)

Unified App Window

• Replaced the old popup dashboard with a single unified app frame
• All features accessible from one window with back-button navigation
• Home page dashboard with character info, crest overview, upgrade summary, farm guide, and navigation buttons

Gear Upgrades (redesigned in-app)

• Equipment panel scaled up for readability
• Info panel widened with 2-column crest display
• Suggestions redesigned as card-style rows
• New recommendation: "Use cheaper crests" for overlap zones

Trinket Rankings (redesigned in-app)

• 3-column layout: S-Tier, A-Tier, B-Tier side by side
• Class/spec browser with ST/AOE toggle
• Fixed persistent "Loading..." issue
• Trinket data updated with latest SimC nightly

Raid Tools (redesigned)

• Overview tab with composition, buffs, splits, PI assignments
• Tier Tracker tab (scan-on-demand)

Progression Dashboard (redesigned)

• Card-based single-page layout
• M+ breakpoints, Raid by difficulty, Delve breakpoints, Prey, Crafting
• Crest Sources side panel

Welcome Screen

• Draggable macro icon for action bar
• Macro auto-created on login

Quality of Life

• Minimap button opens the unified app
• /yh welcome to reset welcome screen