Loot & Baloot

v3.2.8 (2026-05-14)

Full Changelog Previous Releases

• docs(CHANGELOG): add v3.2.8 release notes
  v3.2.8 entry covers:
  • Host-only visual turn stall fix in 4-human games: when the
    player before the host played, the host UI could keep the
    turn glow on that previous seat. Root cause: host's
    authoritative N._HostStepPlay <4 plays branch advanced
    state via S.ApplyTurn + broadcast MSG_TURN to other clients
    but never called B.UI.Refresh() on the host's local UI.
    Fix: add a single defensive UI refresh call after the
    authoritative mid-trick turn advance, mirroring the
    existing pattern in the 4-play resolution branch.
  • Bundled internal cleanup: stale comment line refs
    (Bot.lua:4842+/4886+/2322+/~2474) replaced with named
    subsystem anchors in Bot.lua and Bot/PlayPrimitives.lua.
    Comment-only change, no behavior/test changes.
  • No protocol, saved-variable, scoring, .toc, .pkgmeta,
    .github, or packaging changes.
    Verification: harness 1298/0 (+1 BO.1 source-pin from v3.2.7
    baseline of 1297/0), H1=11/0, H7=9/0. Pre-fix BO.1 FAILS at
    1297/1 (wire-proves the Net.lua change is 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
• cleanup(Bot.lua): replace stale extraction line refs with named anchors
  Four runtime comment sites referenced absolute Bot.lua line
  numbers that became stale during the v3.2.0 cleanup batches
  (especially Batch 5C which extracted helpers to
  Bot/PlayPrimitives.lua, shrinking Bot.lua from ~8,428 to ~6,200
  lines and shifting line numbers by ~1,150). The v3.2.0 audit
  at .swarm_findings/v3_2_0_botlua_comment_audit.md §1.5 flagged
  these as the highest-value lowest-risk cleanup target, then
  sat untouched for ~8 tag cycles.
  Fixed sites (all comment-only, no behavior change):
  • Bot.lua:417 Bot.lua:~2474 in baitedSuit reader
    → the baitedSuit reader in pickLead's fzlokyAvoidSuit consumer
  • Bot.lua:819 bottom-up "want" sender (Bot.lua:4842+)
    → bottom-up "want" sender (Tahreeb "want" arm in pickFollow, Sun-only)
  • Bot.lua:823 T-4 dump-larger sender (Bot.lua:4886+)
    → T-4 dump-larger sender (Tahreeb T-4 arm in pickFollow)
  • Bot.lua:3674 Receiver decoder at Bot.lua:2322+ (tahreebClassify)
    → Receiver decoder (tahreebClassify in Bot/PlayPrimitives.lua) — the function was
    extracted to PlayPrimitives in Batch 5C and
    no longer lives in Bot.lua at all.
  • Bot/PlayPrimitives.lua:291-294 (mirrors Bot.lua:819/823 fix)
    Design rationale: named subsystem anchors over numeric line
    refs prevent re-staleness from future extractions/shifts. Cross-
    file refs are now explicit so a reader can grep for the function
    and land in the right file immediately.
    Verified zero test source-pins on any of the stale substrings
    (grep "4842|4886|2322|2474" tests/ → no matches). No behavior
    change. No CHANGELOG entry (per the design recommendation:
    land untagged unless bundled later with a real runtime release).
    Harness: 1298 / 0 (unchanged from baseline). H1/H7 smokes
    clean (11/0 + 9/0). Diff: +25 / -19 across two files (slight
    wording rebalancing across line wrapping).
    Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com
• fix(Net.lua): host-turn visual refresh after authoritative turn advance
  User-reported bug: in a 4-human game, when the seat immediately
  before the host plays a card and the next turn should be the
  host's, the host's local seat-glow remained stuck on the prior
  player. Other clients showed the correct turn advance.
  Root cause: N._HostStepPlay's <4 plays branch at Net.lua:
  2691-2698 advanced state via S.ApplyTurn and broadcast MSG_TURN
  via N.SendTurn, then called N.MaybeRunBot and returned without
  ever invoking B.UI.Refresh() on the host's local UI. Non-host
  clients redrew on the MSG_TURN wire callback, but the host's
  authoritative loop skipped the redraw step. The 4-play
  resolution branch at Net.lua:2720 already had the symmetric
  B.UI.Refresh() call after _HostStepAfterTrick(); this mid-trick
  <4 plays branch was missing it.
  Fix: add a single defensive if B.UI and B.UI.Refresh then B.UI.Refresh() end call after N.MaybeRunBot() in the <4 plays
  branch. The defensive nil-check mirrors the existing pattern at
  Net.lua:2720 so test stubs without B.UI continue to work
  unchanged.
  Scope check: turn logic (S.ApplyTurn / N.SendTurn / MaybeRunBot)
  is untouched; networking is untouched (MSG_TURN broadcast
  unchanged); timers (the 2.2s C_Timer for trick resolution) are
  untouched; scoring is untouched; bot decisions are untouched.
  The only change is a single redraw trigger on the host loop.
  Tests: BO.1 in tests/test_state_bot.lua source-pins the v3.2.8
  marker comment in Net.lua. Pre-fix harness is 1297/1 (BO.1
  FAILS); post-fix is 1298/0. No behavioural test added because
  the harness doesn't currently provide a B.UI mock; the
  defensive nil-check guard means the new line is a no-op in the
  test harness, so source-pin is the simplest regression guard.
  H1/H7 smokes clean (11/0 + 9/0).
  Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com
• docs: commit v3.2.0 Bot.lua comment-audit design doc
  This is a v3.2.0-era audit / inventory pass on Bot.lua's inline
  comments (651 lines, written 2026-05-12 when main was at
  674de22 and v3.1.14 was the latest tag). It has been preserved
  untracked across several v3.2.x releases via an inherited
  "leave alone" rule; the rule has no documented rationale and the
  doc itself matches the shape of the other committed v3.2.x design
  docs.
  Snapshot caveats:
  • Stats in §0 are v3.2.0-era and are not refreshed here.
  • §2 source-pin table is a historical snapshot; newer BH-BN pins
    have since increased the live count.
  • §3 line numbers may have shifted after later v3.2.x changes.
    Still useful:
  • Appendix A contains a reusable reproducer script for future
    comment-audit passes.
  • The stale extraction-breadcrumb finding remains a valid future
    cleanup candidate.
    Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com