Align production environment section with Major Domo's format:
container name, remote log command, co-hosted services, tea PR workflow.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove generated architecture docs, vague data flow sections, and boilerplate.
Keep commands, key patterns, and development notes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The gha cache backend silently fails on Gitea Actions due to Docker
networking issues between the Buildx builder container and the
act_runner cache server. Registry-based caching stores layers on
Docker Hub, which is more reliable for self-hosted runners.
range() objects were used directly in list membership checks instead of
being unpacked with *, causing all pitcher error ratings in range values
(roughly e27+) to silently fail during x-checks. Also fixed two range
boundary mismatches on dice 12 and dice 6.
Closes#12
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When a half-inning ended with a CS (or pickoff), the batter who was at
the plate was incorrectly skipped in the next inning. The side-switch
code unconditionally advanced the batting order by 1 without checking
whether the last play was a plate appearance. Now checks opponent_play.pa
before incrementing, matching the existing non-side-switch logic.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Early returns in log_chaos, log_sac_bunt, and log_stealing left play
locks permanently stuck because the lock was acquired but never released.
The new locked_play async context manager wraps checks_log_interaction()
and guarantees lock release on exception, early return, or normal exit.
Migrated all 18 locking commands in gameplay.py and removed redundant
double-locking in end_game_command.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Prevents PositionNotFoundException from crashing mlb-campaign when a
player is placed at a position they cannot play (e.g. an outfielder
listed at Catcher in the Google Sheet). Adds early validation in
get_lineups_from_sheets and proper error handling at all read_lineup
call sites so the user gets a clear message and the game is cleaned up.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
HOTFIX: Production bot failed to start due to circular import.
Root cause: utilities/dropdown.py importing from command_logic/logic_gameplay.py
while logic_gameplay.py imports from utilities/dropdown.py.
Solution: Created play_lock.py as standalone module containing:
- release_play_lock()
- safe_play_lock()
Both modules now import from play_lock.py instead of each other.
Error message:
ImportError: cannot import name 'release_play_lock' from partially
initialized module 'command_logic.logic_gameplay' (most likely due
to a circular import)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
CRITICAL BUG FIX: Play locks were never released on exceptions, causing
permanent user lockouts. Found 13 stuck plays in production.
Changes:
1. Added lock_play parameter to checks_log_interaction() (default True)
2. Removed unnecessary locks from read-only commands:
- /settings-ingame (game settings, not play state)
- /show-card defense (read-only display)
- /substitute commands (just show UI, lock in callback)
3. Added safe_play_lock() context manager for automatic lock release
4. Added play locking to substitution callbacks:
- SelectBatterSub.callback()
- SelectReliefPitcher.callback()
5. Global error handler now releases stuck locks automatically
Architecture:
- Commands that display UI or read data: No lock
- Commands that modify play state: Lock at last possible moment
- 3-layer defense: manual release, context manager, global handler
Resolves race condition from concurrent play modifications.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
