Broker: MidnightEvents

v0.9.1-beta (2026-05-24)

Full Changelog Previous Releases

• docs: changelog entry for v0.9.1-beta bugfix refresh
• feat(diag): /meprobe POI quirk probe for ad-hoc investigation
  Generic POI-by-name probe. Sweeps continent + MIDNIGHT_ZONE_IDS with
  GetEventsForMap, GetDelvesForMap, GetAreaPOIForMap and reports every
  match for a case-insensitive name fragment, capturing IsAreaPOITimed,
  GetAreaPOISecondsLeft, atlas, tooltipWidgetSet, iconWidgetSet, plus
  the corresponding ns.Events.GetActive() entry after dedup.
  Result saved to Broker_MidnightEventsDB._probe with a chat summary.
  Default fragment 'voidburrow'; retarget via /meprobe <fragment> for
  the next POI anomaly. Marked throwaway in the header comment.
• fix(events): widget-text timer fallback for POIs reporting isTimed without secondsLeft
  Mining Voidburrow (poiIDs 8675, 8526) reports IsAreaPOITimed=true but
  GetAreaPOISecondsLeft returns nil. The map-tooltip countdown is rendered
  from a TextWithState widget in the POI's tooltipWidgetSet as a localized
  'Time Left: X Hr Y Min' string with Blizzard chat-escape codes embedded.
  Adds ParseTimeLeftSeconds + StripEscapes + SecondsFromWidgetSet, wired
  into ResolvePoi and BuildMapEntry as a fallback when isTimed=true and
  the standard seconds-left getter returns nil. enUS locale assumption
  documented; non-English clients degrade silently to current 'active'
  rendering (no regression).
  Escape-stripping handles |c color spans, |r reset, |4SINGULAR:PLURAL;
  plural escapes (critical — |4Hr:Hr; contains a literal '4' digit the
  parser would otherwise lift as the hours value), |n newlines, and
  |H...|h hyperlinks.
• fix(picks): clear cached pick on abandon, not just turn-in
  User flagged two follow-up cases on the lifetime-flag picks fix:
  1. What if the player drops/abandons the weekly? The cached pick
     should clear, but the previous logic only cleared when
     IsQuestFlaggedCompleted(prev) returned false. Given lifetime-
     persistent flags (the Liadrin pool case), that check returns true
     permanently for any prior-week completion of the same pool member,
     so abandoning a re-accepted previously-completed pick would falsely
     preserve the cache.
  2. Bonus Event Weekly uses the same picks mechanism, so it suffers
     from the same potential issue (not yet verified but architecturally
     identical).
     Fix distinguishes "removed via turn-in" from "removed via abandon" at
     the event level rather than relying on the unreliable flag check:
  • QUEST_TURNED_IN handler stamps recentTurnIn[questID] = GetTime()
    * QUEST_REMOVED handler checks for a recent (≤5s) turn-in:
    • paired with QUEST_TURNED_IN → real turn-in, preserve cache
    • bare QUEST_REMOVED → abandon, clear matching picks
      Session-local recentTurnIn map; doesn't need to persist (any /reload
      guarantees there's no in-flight transition to track).
      5s window is generous — QUEST_REMOVED typically fires within a frame
      of QUEST_TURNED_IN. Wider than needed but safe; the worst case is a
      QUEST_REMOVED arriving ≥5s after its QUEST_TURNED_IN, which would
      falsely clear the cache and downgrade the row to no-annotation —
      strictly less wrong than the previous behavior.
      Applies uniformly to all picks pools: Liadrin, Bonus Event Weekly,
      Void Assault zone, and any future picks-flagged rows.
• fix(picks): drop "any pool flagged" fallback; restore cache preservation
  Probe on Artherio (2026-05-24) returned IsQuestFlaggedCompleted = true
  for ALL 9 Liadrin pool member quest IDs simultaneously. Conclusion:
  these completion flags are achievement-style and persist across
  weekly resets — once you've ever picked a pool member, its flag is
  true forever. So "any pool member flagged" is meaningless as a signal
  for "which one did you pick THIS week," and pairs()-iteration order
  arbitrarily picks the wrong one.
  Two-pass detection now reduced to:
  1. Active quest log scan (definitive — if it's in the log, it's
  the pick).
  2. Cached-and-still-flagged preservation (covers post-turn-in,
  where the quest leaves the log but its flag stays true).
  The "any pool flagged" pass is gone.
  Cold-start consequence: a character who completed a Liadrin pick
  before the feature was deployed AND never opens the addon while the
  quest is in their log will see the row marked done but without the
  "(<choice> picked)" annotation. Acceptable trade — IsWeeklySlotDone
  still flags the row done from the slot's own pool-OR semantics, and
  choosing-no-annotation is better than choosing-wrong-annotation.
  Weekly reset hook in Settings.lua now also clears char.picks. Without
  this, the cached pick would carry from week to week since the
  underlying flags don't clear — last week's pick would persist into
  the new cycle until the player accepts a new one (which might never
  happen on alts that skip Liadrin this week).
  For Artherio's current bug state (picks.liadrin = 94457 Battlegrounds,
  should be 93910 Prey): the new logic keeps the wrong value this week
  because all 9 are flagged and the cached one passes the preservation
  check. Manual recovery this week = SV edit OR /reload past next weekly
  reset (which now clears picks automatically with this commit).