fix #416 #415: the retail button state/media machine — roster hover highlight clears; probe wait verbs bind without an artifact dir

#416 (char-select roster highlight never cleared on hover-leave): three
decomp-grounded mechanisms replace the media-keyed _availableStates
approximation.
- UIElement_Button::UpdateState_ @0x00471CF0: the button machine commits
  ONLY states authored on the button's OWN ElementDesc (AccessStateDesc
  gate); unauthored requests no-op, preserving custom semantic states.
- UIElement::SetState @0x00464E70: an unauthored state id is coerced to
  state 0 (the unnamed base state) and committed — ported into
  UiDatElement.TrySetRetailState with the base-descriptor PassToChildren
  cascade arm.
- The SetState media rule @0x004651c0: a committed state replaces the
  playing media ONLY when its media array is non-empty. UiButton now keeps
  per-face-segment media states under that rule (segments model retail's
  PassToChildren children), and LayoutImporter records the raw MediaCount
  including the File=0 draw-nothing images the drawable filter drops —
  the roster bar children's base state is exactly such an image, and it is
  what clears the bar.
The row template truth (probe, installed DAT): the row authors EMPTY
Normal/rollover/Highlight descriptors with PassToChildren; the three bar
children author rollover/Highlight media, NO Normal state, and a File=0
base image. An empty-media Normal_pressed still never blanks a Normal-art
button (the media rule keeps the previous art — the exact behavior the
old gate approximated), and the Appearance spins' property-only Highlight
now genuinely commits: label recolors, arrow art lingers — the retail
split AP-222 approximated with a requested-keyed label hack, now retired.
Live-verified at char select: hover +alex shows the grey bar, moving off
clears it, the selected row keeps its amber bar.

#415 (probe wait world-* verbs dead): the filed snapshot-reset diagnosis
was wrong — the automation bridge simply never bound without
ACDREAM_AUTOMATION_ARTIFACT_DIR. A facts-only
WorldRevealFactsAutomationRuntime now binds whenever the retained UI
exists; checkpoint/screenshot verbs still require the artifact directory
and now report that instead of a generic timeout.

App tests 5568/3 skips, Runtime 1756/0, UI.Abstractions 926/0.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

docs/ISSUES.md

@@ -24,22 +24,57 @@ What does NOT go here:
 - Every session: scan OPEN issues at start; promote/close anything we touched during the session before ending.
 - Promoting to a Phase: mark as `DONE (promoted to Phase X)` + commit SHA where the Phase entry landed.
 
-## #415 — UI-probe `wait world-visible` verb is dead after reveal completion (automation bridge reads the reset snapshot)
+## #416 — Character-select roster hover highlight never clears (sweeping the roster leaves every row highlighted)
 
-**Status:** OPEN (filed 2026-08-17 at the #414 cursor repro). The script
-runner's `wait world-visible` polls
-`WorldLifecycleAutomationController.IsWorldViewportVisible`, which reads
-`_getReveal().WorldViewportObserved` from the LIVE transit snapshot. The
-reveal's `event=complete` retires that generation and the snapshot resets
-`WorldViewportObserved: false` (`RuntimeWorldTransitState.cs:583`), so the
-verb only observes true during the sub-second window between
-`event=world-visible` and `event=complete` — in practice it times out even
-though the world revealed (proved by the #414 repro logs: `wait
-world-visible 60000` timed out with `event=world-visible` present in the
-same run's `[world-reveal]` stream). Fix shape: the automation controller
-should latch world-visible per generation (or the verb should accept
-`IsWorldReady`-style completed state), not read the transient snapshot.
-Test apparatus only — no player impact.
+**Status:** ✅ FIXED 2026-08-17 (same round as #414; fix + tests in the same
+commit as this entry). **Symptom:** hovering a roster row highlights it, but
+moving off leaves the highlight on — sweep all rows and every one stays lit.
+**Root cause chain (decomp-grounded):** the roster row template
+(`0x21000004/0x100003A5`) is a media-less button whose three bar children
+(`0x10000481/82/83`) author `Normal_rollover`/`Highlight` media but NO
+`Normal` state — their BASE state instead authors a **File=0 draw-nothing
+image**. Retail clears the bar through two mechanisms our port approximated
+away: (1) `UIElement_Button::UpdateState_ @0x00471CF0` gates the machine on
+`AccessStateDesc` — any AUTHORED state commits (the row's empty
+`Normal` descriptor included), unauthored requests no-op; (2)
+`UIElement::SetState @0x00464E70`'s tail (`@0x004651c0`) resets the media
+machine ONLY when the committed state's media array is NON-EMPTY — an
+empty-media state keeps the previous media (why an empty `Normal_pressed`
+never blanks a Normal-art button), an unauthored state commits state 0
+whose BASE media applies, and the bar children's File=0 base image is what
+draws-nothing. Our media-keyed `_availableStates` gate refused the row's
+empty `Normal` commit outright, latching the rollover forever. **Fix:**
+`UiButton` now ports the machine gate (authored-on-own-desc) and the
+SetState media rule (per-face-segment media states with the non-empty-array
+reset gate; `LayoutImporter` records raw `MediaCount` including File=0
+entries); `UiDatElement.TrySetRetailState` gained retail's
+unauthored→state-0 arm with the base-descriptor `PassToChildren` cascade.
+Retires the AP-222-era requested-keyed label hack (the spins' property-only
+`Highlight` now genuinely commits — label recolors, arrow art lingers, the
+exact retail split). Live-verified at char select: hover `+alex` → grey bar;
+move off → bar clears; selected row keeps its amber bar. Tests:
+`UiButtonTests` (pressed-state face linger via draw capture, property-only
+Highlight commit, machine no-op preservation), `UiDatElementTests` (state-0
+commit/cascade/#408 guard), live-DAT spin pin updated to the commit truth.
+
+## #415 — UI-probe `wait world-*` verbs are dead without `ACDREAM_AUTOMATION_ARTIFACT_DIR` (unbound deferred automation wrapper)
+
+**Status:** ✅ FIXED 2026-08-17 (same round as #414/#416; fix in the same
+commit as this entry's flip). Filed as "reads the reset snapshot" — that
+diagnosis was WRONG: the completed reveal KEEPS `WorldViewportObserved`
+(only the next `BeginRevealCore` clears it). The actual cause:
+`FrameRootComposition` binds the `WorldLifecycleAutomationController` (and
+therefore the probe runtime bridge) ONLY when
+`ACDREAM_AUTOMATION_ARTIFACT_DIR` is configured; without it the
+`DeferredWorldLifecycleAutomationRuntime` wrapper stays unbound and every
+`wait world-ready/world-visible/materialized` verb silently reads false
+until timeout — even while the `[world-reveal]` stream shows the awaited
+edge (the #414 repro logs proved exactly this). **Fix:** a facts-only
+`WorldRevealFactsAutomationRuntime` now binds whenever the retained UI
+exists and the full controller is not composed — the wait verbs need only
+the reveal snapshot, which every launch has; checkpoint/screenshot verbs
+still require the artifact directory and now say so explicitly instead of
+failing with a generic timeout. Test apparatus only — no player impact.
 
 ## #414 — Mouse cursor disappears at character select after the in-world logoff (teardown fly-mode fallback raw-captures the cursor)