# World interaction completion — pre-M4 program **Status:** Slice 1 user-accepted 2026-07-23. Slice 2 Use-hand selection, zero-useability, carried direct-use, AutoWear correction, and the requested shared item-cooldown follow-up passed their connected gates. Slice 3's first connected gate exposed an incorrect IdentifyResponse flag table, a non-retail shared-panel mount, and its missing inscription transaction. All three are corrected and live-confirmed. The creature-page follow-up now has retail's ordered stat rows and animated private preview. The follow-up item report and authored 310 x 400 layout correction are implemented; their connected visual gate is pending. **Milestone:** M4 prerequisite/preamble. **Architecture:** retained gameplay UI over shared selection, object, and interaction state. `GameWindow` remains a composition/callback shell. ## Outcome Close the remaining retail interaction surfaces before the larger M4 quest/emote/character-creation bodies begin: 1. Favorite spell bars expose their DAT-authored overflow arrows and scroll through every server-persisted favorite. 2. The status bar's hand and magnifying-glass controls invoke the same Use and Assess commands as their keyboard paths. 3. Assessing a creature, player, NPC, or object opens its retail information in the independent movable/resizable retail floaty examination window. 4. World picking can resolve visible equipped children, such as a character's wielded weapon, while selection markers remain anchored to the picked child. 5. Vendor use opens the authored vendor surface, publishes its inventory, and supports retail selection/browsing. 6. Vendor buy/sell transactions, quantities, pending-state ownership, and authoritative inventory reconciliation complete the loop. The program reuses the existing retained-window host, `SelectionState`, `ClientObjectTable`, interaction transaction owner, and server-authoritative inventory updates. It does not create parallel panel positions, item tables, selection state, or optimistic inventory outcomes. ## Ordered slices | Slice | Deliverable | Principal owner | |---|---|---| | 1 | Spell-bar overflow arrows | `SpellcastingUiController` + generic retained scrollbar/list | | 2 | Status Use/Assess commands | focused status-bar controller binding to the existing action router | | 3 | Assessment information panel | retained controller in an independent floaty examination window | | 4 | Equipped-child world picking | pure world-query/picking policy plus presentation anchor | | 5 | Vendor browse lifecycle | vendor session/controller plus authored panel | | 6 | Vendor transactions | server-authoritative buy/sell command and reconciliation owner | Each slice begins with named-retail research, produces pseudocode and conformance tests, updates the divergence register if required, and lands as a separate bisectable commit. A visual gate follows each UI-bearing slice. Slice 1 now imports both 23-pixel arrow buttons, their rollover/pressed media, and HideDisabled from the real combat fixture; places their artwork by the authored leading/trailing positions; shares the list's single pixel-scroll model; advances one 32-pixel cell per press; and exposes only actively selected spells. Passive object/endowment refresh preserves a manual offset. The mixed 610/800-pixel DAT anchor chain is solved to a fixed 747-pixel combat root, producing exactly 18 visible 32-pixel favorite cells (nine numbered plus nine unnumbered) without consuming the overflow. The initial 57-test/full-suite gate passed; the corrective 104-test focus set, 3,472 App tests / 3 skips, Release solution build, and 7,845 complete-solution tests / 5 skips pass. The corrected 18-cell bar and both overflow directions passed the connected gate. Slice 2 now drives the toolbar hand from canonical `SelectionState` changes and live `ClientObjectTable` updates. The pure Core predicate ports `gmToolbarUI::HandleSelectionChanged`: combat-use items, armor/clothing/jewelry, and `ItemUses` values without `USEABLE_NO` remain active. Clicking still enters the existing `ItemInteractionController` command path, so weapons use the server-confirmed AutoWield transaction and targeted tools enter the existing use-on-target cursor. Empty selection and explicitly unusable spell components are ghosted and ignore clicks. Empty-selection ghosting is the user's explicit choice over retail's generic `TARGET_MODE_USE` entry and is registered as AP-122. The 14-test Core interaction focus, 49-test toolbar focus, Release solution build, 3,474 App tests / 3 skips, and 7,848 complete-solution tests / 5 skips pass. The connected selection-state gate passed. Its follow-up exposed three deeper shared-policy/delivery defects, now corrected from named retail plus the matching binary. `ItemUses::IsUseable @ 0x004FCCC0` tests only `USEABLE_NO`; reset/absent value zero is usable. `ItemHolder::UseObject @ 0x00588A80` sends a carried direct-use item immediately; acdream had incorrectly routed that packet through a world approach lookup, where Blackmoor's Favor has no spatial entity and was silently cancelled. Owned Use now bypasses that query while world-object Use retains the registered AD-27 approach seam. AutoWear applies `CPlayerSystem::AutoWearIsLegal @ 0x0055EF40` through the same double-click and toolbar-hand path, resolves the overlapping worn object from the retail-ordered equipment projection, and emits the exact system line `You must remove your to wear that`. Research: [`../research/2026-07-23-retail-item-use-and-autowear-pseudocode.md`](../research/2026-07-23-retail-item-use-and-autowear-pseudocode.md). The focused 25-test Core and 85-test App sets pass, as do the warning-free Release solution build, 3,476 App tests / 3 skips, and 7,857 complete-solution tests / 5 skips. The carried-use delivery correction adds an end-to-end App pin from Favor activation through wire dispatch and authoritative UseDone busy release. The Release build retains the 17 tracked test-project warnings; 3,477 App tests / 3 skips and 7,858 complete-solution tests / 5 skips pass. The user confirmed Blackmoor's Favor now activates correctly. Before Assess, the user requested the adjacent retail item-cooldown presentation. `PublicWeenieDesc` now preserves the optional shared cooldown id and duration, assessed property updates reach the same object fields, and `CEnchantmentRegistry::OnCooldown @ 0x005943C0` plus `UIElement_UIItem::UpdateCooldownDisplay @ 0x004E1E20` are ported through the canonical Core registry and one pure display projection. Every retained item list shares one controller and the exact ten DAT-authored radial sprites `0x060067CF..0x060067D8`; items with the same group display the same server-authored cooldown. Research: [`../research/2026-07-23-retail-item-cooldown-pseudocode.md`](../research/2026-07-23-retail-item-cooldown-pseudocode.md). The focused parser/Core/UI/production-DAT tests, warning-free Release solution build, 3,482 App tests / 3 skips, and 7,875 complete-solution tests / 5 skips pass. AP-123 records only the retained toolkit's procedural-child adaptation; the visible assets, ordering, timing, and shared-group behavior are exact. The user accepted the live cooldown presentation 2026-07-23. Slice 3 preserves the existing toolbar/keyboard Assess command and ports the retail response owner around it. One shared UI-busy reference covers the latest pending GUID, stale replies are rejected, the accepted reply becomes the current appraisal, and combat-time creature/player refreshes occur every 0.75 seconds only while the examination window is visible. Complete `IdentifyObjectResponse` parsing now uses ACE/retail's exact flag values and positional order, includes the three-word HookProfile, and rejects truncated gated payloads rather than continuing from a corrupt cursor. The first live monster gate found that `0x0100` had been mislabeled WeaponProfile; the parser dropped every creature response before response acceptance, leaving the one busy reference held. A literal ACE `0x0100` creature fixture now protects that packet-to-controller path. LayoutDesc `0x2100006B`, root `0x100005F2`, supplies the complete 310 x 400 floaty chrome, title, item/creature/character subviews, close control, and authored scrollbars. `gmFloatyExaminationUI` is an independent top-level window, so Inventory/Skills/Spellbook no longer replace it or inherit its geometry. The imported multiline inscription field now ports `ItemExamineUI`'s public/hook inscribability, inventory ownership and same-scribe permission rules, placeholder/signature presentation, exact failure notices, focus commit behavior, and CP-1252 `SetInscription (0x00BF)` transaction. Basic item and player reports are live. The creature page now resolves its type through retail EnumMapper `0x2200000E`, preserves the authored Character/Level header, creates the exact nine stat rows from template `0x10000166`, and renders a fixed-heading animated clone through a private viewport using retail's bounding-box camera and distant light. The visual-gate correction also ports the separate `0x10000335` damage/critical/resistance rating list, places authored row chrome behind the animated preview and text in front, adds the balanced row inset, and follows the current selection automatically while the examination window is visible. Item-object preview, specialized item/character detail regions, and exact creature appraisal font-state selection remain the narrowed AP-110 residual. The item report now retains `PublicWeenieDesc` hook identity, applies appraisal-only Value/Burden unknowns, suppresses mounted-hook sentinel capacities, preserves retail line/paragraph boundaries, and selects the authored white/green/red item font-color entries. Research: [`../research/2026-07-23-retail-appraisal-ui-pseudocode.md`](../research/2026-07-23-retail-appraisal-ui-pseudocode.md). The focused parser/router/request/controller/fixture tests pass, as do the Release solution build with 17 pre-existing tracked test warnings, 3,496 App tests / 3 skips, and 7,913 complete-solution tests / 5 skips. The creature presentation follow-up adds real-DAT row-template and EnumMapper fixtures plus pure conformance coverage for row order/formatting, enchantment semantics, failed assessments, stable clone identity, live animated mesh updates, hydrated mesh bounds, and retail camera fitting. The warning-free App code, Release solution build, 3,506 App tests / 3 skips, and 7,923 complete-solution tests / 5 skips pass. The connected visual result remains the closeout gate. The rating/layering/selection-follow correction adds 18 focused green tests, passes the Release solution build with the 17 warnings already tracked by #228, 3,510 App tests / 3 skips, and 7,927 complete-solution tests / 5 skips. Its corrected connected visual result remains the closeout gate. The item-report/layout correction restores the authored 310 x 400 examination size for the connected profile, leaves retail's ordinary resize range intact, keeps row chrome at its LayoutDesc origin while insetting only foreground creature text, and starts generated item prose at the top of its authored scroll surface. `ItemAppraisalTextFormatter` now owns the decomp-ordered item projection outside `GameWindow` and outside the examination controller. It ports common weapon damage ranges/speed/range/ammunition, armor protection bands, defense/caster modifiers, workmanship, ratings, wield/use/activation requirements, item XP/capacity/lock/mana/uses/creator/rare data, cooldown and imbued special properties, and both short spell lists and full DAT spell descriptions. The conformance follow-up replaces public-value fallbacks with retail's appraisal presence semantics, restores hook/capacity/lock behavior, and carries each `AddItemInfo` fragment's paragraph and font-color index into the retained text shaper. AP-110 now names only the remaining specialized, player-dependent, DAT-display-name, creature-font-state, and object-preview gaps. Focused conformance fixtures cover the geometry layering plus melee, launcher, armor, spell, and special-property reports. The Release solution build passes with the 17 warnings already tracked by #228, 3,514 App tests / 3 skips, and 7,931 complete-solution tests / 5 skips. The connected visual result remains the closeout gate. The item-format conformance correction adds literal hook-tail cursor fixtures, wire/session/object-table propagation tests, the exact Black Phyntos Hive report, capacity/page/lock presence cases, structured paragraph checks, and LayoutDesc color-palette/enchantment-style coverage. The isolated Release solution build passes with the same 17 tracked warnings; 3,522 App tests / 3 skips and 7,942 complete-solution tests / 5 skips pass while the previously launched client retains the normal Release output lock. The corrected connected item visual remains the closeout gate. The exhaustive item-report correction replaces the remaining generic/numeric approximations with the complete named-retail dispatch. Equipment sets use the literal EoR table; ratings, tinkering/salvage averages, coverage, failed weapon unknowns, level restrictions, all three item-XP curves, activation heritage, healer/ordinary boost behavior, rare timers, and magic `~ Name: Description` rows now preserve retail wording and ordering. Description construction ports lifespan prose, workmanship/material/gem decoration and portal/PK restriction bits. `RetailAppraisalNameResolver` follows retail's master `EnumIDMap -> sub-enum 1 DualEnumIDMap` material chain and shares the installed creature mapper for slayers and wield requirements; a production-DAT test pins Ruby and Ghost. AP-110 is narrowed to item preview and the projections that need live player/localization state (effective shield, cooldown remaining, and augmentation-cost `StringInfo`), plus character and creature-font residuals. The focused item formatter suite passes 21/21, App Release passes 3,531 tests / 3 skips, and the complete Release solution passes 7,952 tests / 5 skips. The connected item-report comparison is the remaining Slice 3 gate. The material-title and section-boundary correction carries `PublicWeenieDesc.MaterialType` from CreateObject through the canonical object table and resolves `ACCWeenieObject::GetObjectName(NAME_APPROPRIATE)` through the installed DAT material map. Examination titles now produce names such as `Reed Shark Hide Steel Toed Boots` without duplicating a material already present in the authored base name. Empty `AddItemInfo` calls are retained as real report fragments, restoring retail's intentional blank rows after workmanship, before armor level, around rating/special-property blocks, and at the later use/item-level boundaries. Focused wire, projection, object-table, title, and boots-layout fixtures protect the full path. The Release solution build and 3,533 App tests / 3 skips plus 7,956 complete-solution tests / 5 skips pass; the connected boots comparison is the remaining visual gate. ## Slice 1 — spell-bar overflow arrows ### Retail oracle The authored Magic combat layout contains, inside every favorite-tab group: - horizontal scrollbar `0x100000B5`, 685×36 at the 800-pixel design width; - decrement button `0x10000071`, 23×36; - increment button `0x10000072`, 23×36; - item list `0x100000B6`, inset by 23 pixels on both sides and 32 pixels high. The list references the scrollbar through `UIElement_Scrollable` property `0x71`. The scrollbar references increment/decrement buttons through properties `0x77`/`0x78`, and property `0x79` enables HideDisabled. There is no track or thumb in this specific control: the two authored arrows are the whole visible scrollbar. Named retail references and executable pseudocode are recorded in [`../research/2026-07-23-retail-spellbar-overflow-pseudocode.md`](../research/2026-07-23-retail-spellbar-overflow-pseudocode.md). ### Implementation plan 1. Generalize `DatWidgetFactory.BuildScrollbar` so horizontal and vertical scrollbars both import the referenced decrement/increment children, including their authored positions, dimensions, Normal/rollover/pressed media, and HideDisabled property. 2. Generalize `UiScrollbar` to use distinct authored decrement/increment extents for rendering, hit-testing, track geometry, and dragging. Preserve the existing 16-pixel default for layouts whose button children are absent. 3. Reproduce retail disabled presentation: a model without overflow rejects pointer input; when HideDisabled is authored, the scrollbar also draws nothing and does not claim hit tests. 4. In `SpellcastingUiController`, bind each group's scrollbar to its `UiItemList.Scroll`, enable horizontal scrolling, and retain one independent pixel offset per favorite tab. 5. Match `SpellCastSubMenu::SetSelected`: selecting a spell through keyboard, shortcut, or code scrolls that item into view; passive state refresh does not re-expose it. 6. Carry the combat root's mixed-parent raw-edge policies through the complete imported tree and solve for the retail 18-cell favorite viewport; keep that HUD capacity fixed across desktop resizes. 7. Pin the importer, arrow hit extents, 32-pixel step, no-overflow behavior, controller binding, and selection exposure with focused App tests. 8. Run the App Release suite, solution Release build, and complete Release suite. Then update the roadmap/memory and request the connected visual gate: place more favorites than fit, scroll both directions, change tabs, and verify arrows disappear on a non-overflowing tab. ### Invariants - DAT supplies the controls and their artwork; no new spell-bar texture or overlay is invented. - The scrollbar and list share one `UiScrollable`; there is no second offset. - One arrow press moves one 32-pixel favorite cell, matching `UIElement_ListBox::InqScrollDelta`. - Hidden disabled arrows cannot intercept combat-page dragging or clicks. - Existing stack, combat-power, chat, inventory, spellbook, and external container scrollbars retain their current behavior. - No substantial feature body enters `GameWindow`. ## Slice 2 — status Use/Assess commands ### Use-hand implementation 1. `ItemInteractionPolicy.IsToolbarUseEnabled` is the pure named-retail selection predicate. 2. `ItemInteractionController.IsToolbarUseEnabled` adapts the selected live object without triggering a request or consuming the use throttle. 3. `ToolbarController` subscribes to canonical `SelectionState.Changed` and selected-object add/update/remove notices, then sets the imported button's normal or ghosted state through `UiButton.Enabled`. 4. Enabled clicks retain the one existing activation path: equipment enters `AutoWieldController`, ordinary use enters the normal Use request owner, and targeted items enter `UseItemOnTarget`, which already owns the retail target cursor. ### Slice 2 closeout - The shared item-cooldown visual gate passed. - `gmToolbarUI::ListenToElementMessage @ 0x004BEE90` confirms the existing magnifying-glass path was already correct: assess the selected GUID immediately, otherwise enter one-shot Examine target mode. - Slice 3 owns the response lifetime and retained examination presentation; Slice 2 adds no duplicate command path. ## Slice 3 — assessment information panel ### Implemented ownership 1. `AppraiseInfoParser` owns the complete positional response payload, including HookProfile and strict truncation failure. 2. `ItemInteractionController` owns the pending/current GUID pair and the balanced shared busy reference. 3. `AppraisalUiController` owns response acceptance, item/creature/character subview selection, report projection, scroll preservation, first-response visibility, and visible-combat refresh. 4. `RetailUiRuntime` imports and registers retail's `gmFloatyExaminationUI` as its own top-level window. It deliberately does not enter `RetailPanelUiController`; Inventory, Skills, Spellbook, and effects keep their shared main-panel geometry while Examination coexists. 5. Network workers deliver immutable parsed data through the existing session router; retained state changes remain on the update thread. 6. `AppraisalUiController` owns inscription presentation and optimistic field state; `WorldSession` owns the exact `0x00BF` GameAction send. Authoritative inventory/appraisal data remains the source on the next response. ### Connected gate - Select an item and click the magnifying glass (or press Assess): a separate floaty window should show its name, available properties, descriptions, inscription/signature, retail chrome, and working scrollbars. - Assess a monster, NPC, and player: the correct creature/character page should appear and the busy cursor should clear. - On a monster, verify the animated creature is centered behind the exact Strength/Endurance/Coordination/Quickness/Focus/Self/Health/Stamina/Mana order, with its creature type and level in the authored header. Assess differently sized monsters to verify the retail bounding-box fit. - In combat, leave a creature assessment open long enough to observe a health refresh; closing the examination window must stop refreshes. - Keep Examination open while opening/moving Inventory or Skills and confirm both windows coexist with independent rectangles. - Assess an owned inscribable weapon. Edit and clear its inscription by clicking elsewhere, then reassess it. An item authored by another player is read-only and reports the exact retail permission line when clicked.