V5+ games write room names to the upper window (status line) via
select_window(1). Since select_window was a no-op, status line text
leaked into the main output buffer, causing ">Outside" on prompt lines.
Track the active window and only buffer writes to window 0 (lower).
- op_save_v5: generates Quetzal save, stores 1 on success / 0 on failure
- op_restore_v5: loads Quetzal save, stores 2 ("restored") via store byte
- _try_restore: detect V5+ in-game saves (0xBE 0x00 before PC) and process
the store byte with result 2, matching the V3 branch-on-restore pattern
- _find_story() now compares path.stem.lower() so "lostpig" matches "LostPig.z8"
- Server no longer writes its own prompt in IF mode (game handles prompting)
- Suppress phantom game output on restore (saved PC past sread causes garbage)
- Route .z5/.z8 files to EmbeddedIFSession now that V5+ is supported
Implements the opcode set needed for Lost Pig (V8):
- op_aread: V5+ input with text at byte 2, char count at byte 1,
stores terminating character
- op_save_undo/op_restore_undo: stub returning -1/0 (undo not yet
available, game continues without it)
- op_log_shift/op_art_shift: logical and arithmetic bit shifts
- op_scan_table: table search with configurable entry size and
word/byte comparison
- op_tokenize: re-tokenize text buffer against dictionary
- op_copy_table: memory copy/zero with forward/backward support
- op_set_font: returns 1 for font 1, 0 for others
- op_print_unicode/op_check_unicode: basic Unicode output support
Lost Pig now runs to completion: 101K steps, 61 unique opcodes,
3 full input cycles with room descriptions rendering correctly.
New opcodes: op_call_vn, op_call_vn2, op_call_vs2, op_catch,
op_check_arg_count. All call variants delegate to existing _call().
ZRoutine now tracks arg_count for check_arg_count.
Fixed zopdecoder double-byte operand parsing for call_vs2/call_vn2:
the old code called _parse_operands_byte() twice, but this method
reads both type byte AND operands together. The second call would
read operand data as a type byte. Refactored into _read_type_byte()
+ _parse_operand_list() so both type bytes are read before any
operand data.
Also fixed the double-byte detection: was checking opcode[0:7] (7
bits = 0x7A for call_vn2) instead of opcode_num (5 bits = 0x1A).
The check never matched, so double-byte opcodes were always
mis-parsed.
The old code inserted the object into the new parent first, then
tried to remove it from the old parent. This corrupted the sibling
chain because the object's sibling pointer was already modified.
The Z-spec says "if O already has a parent, it is first removed."
Now delegates to remove_object() before inserting.
The 0xBE prefix byte triggers extended opcode parsing. Reads the
opcode number from the next byte, then parses operand types using
the same format as VAR opcodes. Required for all V5+ games.
V8 uses the same format as V5 (object model, opcodes, stack) with
two differences: packed address scaling (×8 instead of ×4) and max
file size (512KB instead of 256KB).
zmemory: add V8 size validation and packed_address case
zobjectparser: accept version 8 alongside 4-5 in all checks
zstackmanager: allow V8 stack initialization
V6-7 remain unsupported (different packed address format with offsets).
quetzal.py: typo in _parse_umem() — cmem.dynamic_start should be
cmem._dynamic_start. Would crash on uncompressed memory restore.
zcpu.py: op_erase_window had if/if/else instead of if/elif/else.
When window_number was -1, the second if fell through to else,
calling erase_window(-1) after the correct reset path.
The Quetzal spec stores the PC pointing at the save instruction's
branch data. On restore, this branch must be processed as "save
succeeded" to advance the PC to the actual next instruction. Without
this, the branch bytes were decoded as an opcode, corrupting execution.
Detect the save opcode (0xB5) immediately before the restored PC to
distinguish in-game saves from out-of-band saves (which don't need
branch processing). Also improve error diagnostics: pop_stack now
raises ZStackPopError with frame context, and the instruction trace
dumps on all exceptions.
return_pc for each frame belongs on the caller's program_counter (the
resume address when this routine exits). varnum is the store variable
that receives the return value, stored as return_addr on the frame
itself. Also handle flags bit 4 (discard result → return_addr=None).
The Quetzal save format only stores the number of local vars the
routine declared, but the runtime indexes local_vars[0..14] for any
variable access. Restored routines had short lists, causing IndexError
on the first command after restore.
QuetzalParser._parse_stks() creates a new ZStackManager and sets it on
zmachine._stackmanager, but ZCpu._stackmanager and ZOpDecoder._stack
still pointed to the old empty stack. After restore, all stack ops
(local var reads, routine returns, stack pops) used the wrong stack,
causing the interpreter to crash on the first command.
Replaces the async _do_restore() (called after thread launch) with a
synchronous _try_restore() called before the thread starts. This
eliminates the race condition where restore mutates z-machine state
while the interpreter thread is running.
The restore prefix message is now part of start()'s return value
instead of being sent separately in play.py.
EmbeddedIFSession runs the hybrid interpreter in a daemon thread,
bridged to the async MUD loop via threading.Event synchronization.
.z3 files use the embedded path; other formats fall back to dfrotz.
- MUD ZUI components: MudScreen (buffered output), MudInputStream
(thread-safe input), MudFilesystem (quetzal saves), NullAudio
- save/restore via QuetzalWriter/QuetzalParser and :: escape commands
- state inspection: get_location_name(), get_room_objects()
- error reporting for interpreter crashes
- fix quetzal parser bit slice bug: _parse_stks used [0:3] (3 bits,
max 7 locals) instead of [0:4] (4 bits, max 15) — Zork uses 15
Implement V3 restore opcode:
- Add QuetzalParser.load_from_bytes() to parse save data from memory
- Wire op_restore to call filesystem.restore_game() and parse result
- Validate IFhd matches current story (release/serial/checksum)
- Restore dynamic memory, call stack, and program counter
- Branch true on success, false on failure/cancellation
Fix IFF chunk padding bug:
- Add padding byte to odd-length chunks in QuetzalWriter
- Ensures proper chunk alignment for parser compatibility
Add comprehensive tests:
- Branch false when filesystem returns None
- Branch false without zmachine reference
- Branch true on successful restore
- Verify memory state matches saved values
- Handle malformed save data gracefully
Implement full save functionality for V3 z-machine:
- Fixed QuetzalWriter._generate_anno_chunk() to return bytes
- Added QuetzalWriter.generate_save_data() to produce IFF container
- Updated QuetzalWriter.write() to use new method and binary mode
- Added zmachine reference to ZCpu for QuetzalWriter access
- Added _program_counter property to ZCpu for Quetzal access
- Implemented op_save to call QuetzalWriter and filesystem
- Updated tests for op_save (success, failure, IFF validation)
- Added filesystem mock to MockUI for testing
- Added _call_stack to MockStackManager for QuetzalWriter
All tests pass. Save now generates valid IFF/FORM/IFZS data with
IFhd, CMem, Stks, and ANNO chunks.
The IFhd chunk contains 13 bytes of metadata identifying the story
and current execution state:
- Release number (2 bytes) from header
- Serial number (6 bytes) from header
- Checksum (2 bytes) from header
- Program counter (3 bytes) from CPU state
This allows save files to be validated against the story file.
The input detection handled both chr(8) and chr(127) but the echo
logic only checked chr(8). Most modern terminals send chr(127) for
backspace, so the cursor never moved back visually.
Property parsing had 4 classes of bug, all producing wrong addresses
for object property data:
1. Off-by-one in shortname skip: addr += 2*len missed the length byte
itself, causing property table scan to start 1 byte too early.
Affected get_prop_addr_len and get_next_property.
2. V3 property number slices bf[4:0] extracted 4 bits not 5.
Property numbers 16-31 were read as 0-15.
3. V3 property size slices bf[7:5] extracted 2 bits not 3.
Properties larger than 4 bytes got wrong sizes.
4. V5 property number/size slices bf[5:0] extracted 5 bits not 6.
Also fixed get_property_length reading size from the wrong byte in
V5 two-byte headers (was reading first byte which has property
number, instead of second byte which has size).
Root cause for all slice bugs: BitField uses Python [start:stop)
semantics but code was written as [high:low] inclusive notation.
Same class of bug as the write_global fix in commit e5329d6.
Adds comprehensive test coverage for three newly implemented Z-machine opcodes:
- op_test: Tests bitmap flag checking with all flags set, some missing, zero
flags, and identical values
- op_verify: Tests checksum verification with matching and mismatched checksums
- op_get_child: Tests getting first child of object with and without children
Also extends MockMemory with generate_checksum() and MockObjectParser with
get_child() to support the new tests.
Wrap opcode execution with try-except to catch all ZCpuError subclasses except
ZCpuQuit and ZCpuRestart, which are normal control flow. This ensures we get
trace dumps for divide by zero, not implemented, and other errors.
Spec fixes: implement op_test (bitwise AND branch), add missing branch
to op_get_child, handle call-to-address-0 as no-op in op_call/_call.
I/O fixes: correct keyboard API (keyboard_input.read_line), non-TTY
fallbacks in trivialzui, stdout flush for immediate output. Graceful
handling of unmapped ZSCII characters. Add instruction trace buffer
for debugging.
The V3 A2 table needs newline (ZSCII 13) at code 7, matching viola's
implementation. The previous fix only removed the erroneous '<' but
didn't add the newline, leaving the table at 24 chars instead of 25.
DEFAULT_A2 had 26 chars instead of 25 due to an erroneous '<'. This
shifted all symbol mappings, causing periods to render as exclamation
marks and other garbled punctuation in Zork 1 output.
Python 2 `/` did integer division, Python 3 returns float. Changed to
`//` in zobjectparser (attribute byte offset) and zmemory (address
bounds checks). Zork 1 hit this on the first test_attr opcode.
ZStackBottom lacked attributes that normal stack operations need when
returning from a subroutine to the top level. Zork 1 hit this on the
first subroutine return during startup.
write_word() was routing all header writes through game_set_header()
which enforced overly strict per-byte permissions. Zork 1 legitimately
writes to header address 2 on startup. Now uses the same dynamic/static
boundary check that viola does, matching the reference implementation.
op_verify now performs actual checksum validation against the header
instead of raising NotImplemented. ZLexer is injected into ZCpu and
sread tokenizes input into the parse buffer per the V3 spec.
Implements op_print_addr, op_print_num, op_ret, and op_show_status following
TDD approach with tests first. Each opcode now properly decodes/prints text,
handles signed numbers, returns from routines, or acts as a no-op as appropriate.
Implemented 10 opcodes for object tree manipulation: get_sibling,
test_attr, set_attr, clear_attr, jin, remove_obj, print_obj,
get_prop_addr, get_next_prop, and get_prop_len.
Added 6 supporting methods to ZObjectParser: set_attribute,
clear_attribute, remove_object, get_property_data_address,
get_next_property, and get_property_length.
Fixed bug in insert_object where sibling chain walk never advanced
the prev pointer, potentially causing infinite loops.
Added 16 unit tests with MockObjectParser to verify CPU opcode
behavior. All tests passing.
Copies the cleaned-up zvm source (ruff-compliant, ty-clean) back into
the zmachine module. Adds __init__.py with proper exports and updates
.gitignore for debug.log/disasm.log.
When starting an IF game, check for existing save file and restore
if present. Shows 'restoring saved game...' message and broadcasts
restored game state to spectators.
Also cleaned up redundant tests that didn't properly mock the
auto-save functionality now present in ::quit and stop().
- server.py: broadcast IF output to spectators after each input (skip :: escape commands)
- server.py: broadcast leave message when player exits IF mode
- play.py: broadcast game intro text when player starts a game
Spectators at the same x,y coordinates now see formatted output with
[PlayerName's terminal] header and game text.
Helper function sends messages to all players at the same x,y coordinates
as the source player, skipping the source player themselves. Used for IF
spectator broadcasting.
TDD implementation of IFSession that manages a dfrotz subprocess.
IFResponse dataclass follows the editor pattern with output/done fields.
IFSession handles spawning dfrotz, routing input, and detecting the prompt.
Escape commands (::quit, ::help) are handled without sending to dfrotz.
Phase 6: spawn command creates mobs at player position from loaded
templates. Server loads mob templates from content/mobs/ at startup,
injects world into combat/commands module, and runs process_mobs()
each game loop tick after process_combat().
Phase 5: process_mobs() runs each tick, handling mob attack and defense
decisions. Mobs pick random attacks from their move list when IDLE,
swap roles if needed, and attempt defense during TELEGRAPH/WINDOW with
a 40% chance of correct counter. 1-second cooldown between actions.
Training dummies with empty moves never fight back.
Phase 4: when combat ends, determine winner/loser. If the loser is a
Mob, despawn it and send a victory message to the winner. If the loser
is a Player fighting a Mob, send a defeat message instead.
Phase 3: look command now collects alive mob positions using the same
wrapping-aware relative position calc as players, and renders them as *
with the same priority as other players (after @ but before effects).
Phase 2: do_attack now searches the mobs registry after players dict
when resolving a target name. Players always take priority over mobs
with the same name. World instance injected into combat/commands module
for wrapping-aware mob proximity checks.
Phase 1 of fightable mobs: MobTemplate dataclass loaded from TOML,
global mobs list, spawn_mob/despawn_mob/get_nearby_mob with
wrapping-aware distance. Mob entity gets moves and next_action_at fields.
The variant handler now supports prefix matching for directional variants.
This allows 'pa hi' to match 'parry high', 'pa lo' to match 'parry low', etc.
Implementation:
- First tries exact match on variant key
- Falls back to prefix matching if no exact match
- Returns unique match if exactly one variant starts with the prefix
- Shows disambiguation message if multiple variants match
- Shows error with valid options if no variants match
Tests cover exact match, prefix match, ambiguous prefix, no match,
single-char prefix, case-insensitivity, and preservation of target args.
Defense moves now asyncio.sleep for timing_window_ms instead of using
a cooldown field. Input queues naturally since the per-player loop is
sequential. Outside combat shows "parry the air!" flavor text.
- Set variant defense registration to mode="*" (both attacks and defenses)
- Strengthen telegraph switch test to verify new move's telegraph text
- Remove unused punch parameter from four idle timeout tests
- Use single time.monotonic() call in attack() method
Encounters track last_action_at (updated on attack and defend). If 30
seconds pass with no actions, combat fizzles out with a message to both
players and combat mode is popped. start_encounter initializes the
timestamp so fresh encounters don't immediately timeout.
Defenses now work outside combat mode with stamina cost, recovery lock
(based on timing_window_ms), and broadcast to nearby players. Lock
prevents spamming defenses — you commit to the move. Stamina deduction
moved from encounter.defend() to do_defend command layer. Defense
commands registered with mode="*" instead of "combat".
Attacker can change their move mid-telegraph or mid-window without
resetting the timer. Old move's stamina is refunded, new move charged.
Defender gets a fresh telegraph on switch. Feedback says "switch to"
instead of "use" when swapping attacks.
resolve() returns ResolveResult dataclass with attacker_msg, defender_msg,
damage, countered, and combat_ended fields. process_combat is now async
and sends messages to both participants on resolve. Counter, hit, and
slam messages give each player their own perspective on what happened.
The DREAMBOOK always described "punch right/left [target]" as one command
with a direction argument, but the implementation had separate TOML files
and multi-word command names that the dispatcher couldn't reach (it only
matches the first word). Aliases like "pr" also couldn't pass targets
because the shared handler tried to re-derive the move from args.
Changes:
- Merge punch_left/right, dodge_left/right, parry_high/low into single
TOML files with [variants] sections
- Add command/variant fields to CombatMove for tracking move families
- load_move() now returns list[CombatMove], expanding variants
- Handlers bound to moves via closures at registration time:
variant handler for base commands (punch → parses direction from args),
direct handler for aliases and simple moves (pr → move already known)
- Core logic in do_attack/do_defend takes a resolved move
- Combat doc rewritten as rst with architecture details
- Simplify mud.tin aliases (pr/pl/etc are built-in MUD commands now)
Integrates the Editor class into the MUD server's shell loop, allowing
players to enter and use the text editor from the game.
Changes:
- Add editor field to Player dataclass
- Modify shell input loop to check player mode and route to editor
- Add edit command to enter editor mode from normal mode
- Use inp (not command.strip()) for editor to preserve indentation
- Show line-numbered prompt in editor mode
- Pop mode and clear editor when done=True
- Add comprehensive integration tests
- Fix test isolation issue in test_movement_updates_position
Parse MTTS from telnetlib3 writer during connection and store capabilities
on Player.caps field. Add convenience property Player.color_depth that
delegates to caps.color_depth for easy access by rendering code.
Changes:
- Add caps field to Player with default 16-color ANSI capabilities
- Parse MTTS in server shell after Player creation using parse_mtts()
- Add Player.color_depth property for quick capability checks
- Add tests verifying Player caps integration and color_depth property
Extended ansi.py with fg_256, bg_256, fg_rgb, and bg_rgb functions
for generating 256-color and truecolor escape sequences. All functions
include value clamping to valid ranges (0-255).
Parses MTTS bitfield values from telnetlib3 ttype3 into a ClientCaps dataclass.
Includes color_depth property that returns the best available color mode
(truecolor, 256, or 16) based on client capabilities.
Player state is now saved when using the quit command or when the connection
is lost unexpectedly. Ensures progress is preserved even without auto-save.
Adds login/registration prompts on connection, database initialization on
startup, and periodic auto-save every 5 minutes in the game loop. Player
state is now tied to authenticated accounts.
Implements account management with password hashing (pbkdf2_hmac with SHA256)
and constant-time comparison. Includes player state serialization for position
and inventory persistence across sessions.
Scan content/commands/ for .toml files at startup and register them
as commands alongside Python-defined ones. Two flavors: handler-based
(points to a Python callable via module:function) and message-based
(auto-generates a handler from inline text). Includes example MOTD
command, type validation, error logging, and full test coverage.
Each cloud in the trail gets a slightly longer TTL than the one
before it (0.15s stagger). The origin cloud dissolves first, then
each subsequent tile follows. Two consecutive flights produce a
trail where the oldest clouds are already gone.
fly with no args toggles flying on/off. Movement commands (fly east,
etc) only work while airborne. "You aren't flying." if you try to
move without toggling on first. Player.flying field tracks the state.
fly <direction> moves the player 5 tiles, ignoring terrain. Leaves
a trail of bright white ~ clouds that fade after 2 seconds. Effects
system supports arbitrary timed visual overlays on the viewport.
TinTin aliases: fn/fs/fe/fw/fne/fnw/fse/fsw.
Tileable Perlin noise: each octave wraps its integer grid coordinates
with modulo at the octave's frequency, so gradients at opposite edges
match and the noise field is continuous across the boundary.
Coarse elevation grid interpolation wraps instead of padding boundary
cells. Rivers can flow across world edges. All coordinate access
(get_tile, is_passable, get_viewport) wraps via modulo. Movement,
spawn search, nearby-player detection, and viewport relative positions
all handle the toroidal topology.
1000x1000 tile world generated deterministically from a seed using
layered Perlin noise. Terrain derived from elevation: mountains,
forests, grasslands, sand, water, with rivers traced downhill from
peaks. ANSI-colored viewport centered on player.
Command system with registry/dispatch, 8-direction movement (n/s/e/w
+ diagonals), look/l, quit/q. Players see arrival/departure messages.
Set connect_maxwait=0.5 on telnetlib3 to avoid the 4s CHARSET
negotiation timeout — MUD clients reject CHARSET immediately via MTTS.
