Cerebro

v0.10.33 (2026-03-21)

Full Changelog

• chore: Bump version to 0.10.33
  Fix RCLC import pipeline (date filter, timestamps, dedup), add dry-run simulation command; hide voided transactions in History tab; targeted sync PUSH to divergent peer only; scope Sync Health display to active online peers
• fix(ui): use consistent name format for self-skip in Addon Compliance
  Use GetFullPlayerName(UnitName("player")) instead of GetPlayerFullName()
  for the myName self-skip guard. This matches the same normalizer used for
  guild roster names (GetFullPlayerName(gName)), preventing self from being
  double-counted due to name format mismatch.
• fix(ui): unify Addon Compliance with GetAddonState, remove dead code
  Replaced the entire manual addonStatus lookup + age + version comparison
  in the Addon Compliance tooltip section with v2:GetAddonState(full) — the
  same function that powers green/yellow/gray name colors in the player list.
  Removed ~30 lines of dead code: v2INACTIVE, maxVer, v2modRef, addonStatus
  variable — all now handled internally by GetAddonState. Self is counted
  unconditionally before the loop; the loop uses full ~= myName guard.
• fix(ui): always count self in Addon Compliance synced count
  Self was being missed when GetPlayerFullName() and GetFullPlayerName(gName)
  produced slightly different name strings, causing full == myName to fail.
  Fixed by counting self unconditionally before the guild roster loop, and
  skipping self inside the loop via full ~= myName guard.
• fix(ui): normalize peerHashes short name to full Name-Realm for GetAddonState
  AceComm passes sender as short name for same-realm peers, so peerHashes
  is keyed by short name (e.g. "Bearlybleeds") while addonStatus uses full
  Name-Realm format ("Bearlybleeds-Stormrage"). GetAddonState was always
  returning "unknown" for same-realm peers, causing count to be 1/1.
  Fix: normalize with GetFullPlayerName() before GetAddonState() in both
  ComputeSyncHealth and ShowSyncHealthTooltip.
• fix(ui): tighten active-peer filter to synced/outdated only
  Previously included "inactive" peers (seen before, now offline) in the
  Sync % and tooltip peer counts. Changed condition from != "unknown" to
  == "synced" or == "outdated" so only players with an actively running
  addon are counted as live peers.
• fix(ui): fix active-peer name key mismatch in sync health display
  peerHashes keys didn't match addonStatus keys, causing all peers to be
  excluded (showing 1/1 instead of 3/3). Replaced addonStatus[name]
  lookup with v2:GetAddonState(name) ~= "unknown" in both ComputeSyncHealth
  and ShowSyncHealthTooltip — same proven logic used by Addon Compliance.
• fix(ui): scope Sync % and tooltip top rows to active peers only
  ComputeSyncHealth and ShowSyncHealthTooltip now filter peerHashes to
  peers seen within INACTIVE_THRESHOLD (90s). Offline/unknown peers no
  longer deflate the Sync % label or the top 3 tooltip rows (Sync Health,
  Roster Hash, Config Hash). Addon Compliance section is unchanged and
  still shows all guild members.
• fix(sync): targeted WHISPER push on raider mismatch instead of GUILD broadcast
  When an officer detects a mismatch from a non-officer (raider/stale alt),
  send a targeted WHISPER to only that peer instead of broadcasting to
  RAID/GUILD. This prevents disrupting already-synced officers who would
  temporarily drop from synced state while re-processing the broadcast.
  • Reuses existing _lastSnapshotTo per-sender cooldown (same as officer path)
  • Removes _lastRaidBroadcastAt scalar (no longer needed for this path)
  • Net result: stale raider convergence is now invisible to synced peers
• fix(ui): hide voided transactions in History tab; silence ghost stub debug spam
  • MainFrame/UISandbox History tabs now skip voided transactions
    (previously showed duplicates after RCLC import void+replace cycle)
  • GetFilteredPlayersForSync: remove per-heartbeat debug log for ghost
    stub skipping (Vrock-Eitrigg spam); expected behavior, not worth logging
• fix(rclc-import): Fix import pipeline and add dry-run simulation
  • Add ParseDateArg (MM-DD-YYYY → YYYY/MM/DD RCLC format)
  • Remove duplicate ParseDateArg at EOF that returned Unix timestamp
  • Fix date scope filter to compare RCLC date strings (not servertime)
  • Add full Unix timestamp to NormalizeEntry from date+time fields
  • Fix AddTransaction to honor overrideTimestamp from options
  • Fix VoidTransactions to convert sinceDate string to Unix for comparison
  • Fix audit log bug (date() called with string instead of number)
  • Add SimulateImport dry-run (read-only, per-player GP report)
  • Register /cb rclc dryrun MM-DD-YYYY command
  • Bump version 0.10.32
• fix(sync): fix persistent roster/selection hash divergence between officers
  Two bugs caused officers to never fully converge after PR37:
  Bug 1 — MergeSelectionData timestamp self-escalation (CommunicationV2.lua):
  When incoming selection was accepted, sf:SetSelection() internally called
  SaveSelection() which overwrote customEP.lastUpdate with time(). This ratcheted
  the local timestamp forward on every successful merge, making the local client
  perpetually "newer" than all peers. All future selection merges from those peers
  were REJECTED (theirs < mine forever), producing the continuous REJECTED spam seen
  in logs. Fix: restore incoming.lastUpdate after SetSelection() clobbers it.
  Bug 2 — RepairDatabase lastUpdate bump on login (Database.lua):
  The coreAdmittedAt backfill (data.coreAdmittedAt = 1) also bumped
  data.lastUpdate = time(). RepairDatabase runs on every login at different
  wall-clock seconds per officer, producing different lastUpdate values for
  identical EP/GP/status records. Since lastUpdate is included in ComputeRosterHash(),
  officers with identical data got different roster hashes → endless FETCH/PUSH loops.
  Fix: remove the lastUpdate bump — this is a schema migration, not a data mutation.
  Every officer independently applies this migration on login; LWW propagation
  is unnecessary and harmful."