Loot & Baloot

v3.2.9 (2026-05-15)

Full Changelog Previous Releases

• docs(CHANGELOG): add v3.2.9 release notes
  v3.2.9 entry covers:
  • Host-only stale visual state during 4-human bidding/contract
    flow. Bid-phase analog of v3.2.8's play-phase fix; same root
    cause (SendAddonMessage no-loopback + dispatcher Refresh
    bypassed for host-direct mutation paths).
  • Three sites fixed in N._HostStepBid:
    • action == "next" — mid-bid turn advance.
    • action == "contract" — contract finalization.
    • action == "round2" — round-2 bid kickoff.
  • No turn logic, networking, timers, scoring, bot decisions,
    protocol, saved-variable, .toc, .pkgmeta, .github, or
    packaging changes.
    Deferred: audit findings #4/#5 (N.LocalPlay / N.LocalBid
    non-host optimistic refresh). Pending playtest of whether the
    4th-play-of-trick 2.2s lag is user-visible.
    Verification: harness 1301/0 (+3 BP source-pins from v3.2.8
    baseline of 1298/0), H1=11/0, H7=9/0. Pre-fix BP.1-3 all FAIL
    at 1298/3 (wire-proves all three Net.lua hunks are the cause).
    Committed locally only — no push, no tag, no release until
    Codex reviews.
    Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com
• fix(Net.lua): host bid/contract/round-2 local-refresh (bid-phase v3.2.8 analog)
  Companion to v3.2.8's _HostStepPlay host-turn visual refresh fix.
  Same root cause, bid phase instead of play phase: in 4-human
  games the host advances S.s.turn via S.ApplyTurn + broadcasts
  MSG_TURN via N.SendTurn, but C_ChatInfo.SendAddonMessage doesn't
  loopback to the sender, so the host's own MSG_TURN handler never
  fires and HandleMessage's dispatcher Refresh at Net.lua:1280 is
  bypassed for host-direct mutation paths.
  Three sites in N._HostStepBid all had the same shape and all
  shipped the bug:
  action == "next" (mid-bid turn advance):
  S.ApplyTurn + N.SendTurn + N.MaybeRunBot + return.
  Host UI's bid-glow stales on prior bidder when next bidder
  is host itself.
  action == "contract" (contract finalization):
  Various branches end with N.MaybeRunBot (post-bid Bel
  dispatch). Host UI's post-contract affordances
  (Bel/Triple/Four buttons, bid-card highlight clear,
  phase banner) don't redraw until next message arrives.
  Preempt and overcall sub-branches already returned above
  with their own Refresh or phase-driven UI flow; unaffected.
  action == "round2" (round-2 bid kickoff after all-pass redeal):
  S.ApplyTurn + N.SendTurn + N.MaybeRunBot + branch end.
  Same shape as action == "next".
  Fix: add if B.UI and B.UI.Refresh then B.UI.Refresh() end at
  end of each elseif branch (3 insertions), mirroring v3.2.8's
  defensive pattern. Plus a docblock at the top of _HostStepBid
  explaining the root cause and pointing back to v3.2.8 for
  context.
  Single-site root cause: _HostStepBid is called from _OnBid
  (wire path, dispatcher-covered), LocalBid (host UI click,
  uncovered), MaybeRunBot bot-bid (direct call, uncovered), and
  _HostTurnTimeout AFK (fallback Refresh at Net.lua:5489
  already covers it). Fixing inside _HostStepBid covers all
  direct callers.
  Tests: new BP section in tests/test_state_bot.lua source-pins
  the three v3.2.9 refresh markers. Pre-fix BP.1/BP.2/BP.3 all
  FAIL (verified via stash push: 1298 passed, 3 failed); post-fix
  1301/0. H1/H7 smokes clean (11/0 + 9/0).
  Findings #4 (LocalPlay non-host) and #5 (LocalBid non-host)
  from the multi-human audit are MED severity and deferred to a
  separate slice pending playtest verification of whether the
  WoW UI animator masks the 2.2s 4th-play visual lag in #4.
  Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com