Detail pages carry a breadcrumb bar built from a per-tab trail in
sessionStorage: following a link from a page on the trail extends it,
going back to one cuts it there, and opening a page directly shows its
natural parents. The property 3D view's back button goes to the previous
crumb, and the property page renames its crumb once its name loads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nothing in the server writes or reads a packet by hand any more, so the
helpers for doing so go:
- CBITSTREAM, CMSGHEADER, CINSTREAM, CINSTREAM_SKIP_HEADER, SEND_PACKET,
SEND_PACKET_BROADCAST and HEADER_SIZE leave dCommonVars.h;
- the free BitStreamUtils::WriteHeader (LUBitStream::WriteHeader writes the
same bytes) and the unused PacketUtils::SavePacket are deleted.
The last raw reads are replaced: WorldServer builds its input stream
directly, the master packet logs read the header with
LUBitStream::ReadHeader instead of peeking at packet->data[1] and [3], and
MessageInspector reads a sent game message's header with the new
NetGameMsg::ReadPacketHeader (the counterpart of WritePacket) instead of
memcmp/memcpy. packet->data[0] is still compared with RakNet's own
connection IDs.
The frozen oracles keep using the macros verbatim through the test-only
tests/dGameTests/LegacyPacketMacros.h; the HeaderSkip tests, which only
tested CINSTREAM_SKIP_HEADER, are removed.
docs/PacketArchitecture.md: "where we are" now describes the final state
and what still touches raw bytes (RakNet IDs, replica headers, behavior bit
streams), and a new section collects the known wire discrepancies found
during the conversion, with client addresses.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every game message still written or read by hand is now a NetGameMsg with
Serialize and Deserialize, in per-domain files:
- MovementMessages: teleport, platforms (resync and its request), orient to
angle, node rotation lock, gravity scale, jetpack mode, control scheme,
respawn checkpoint, rails (set/start, ready, cancel, arrived), mount
inventory ID, dismount complete, possession ack, ghost reference override
and position, camera cycling (eCameraTargetCyclingMode moves here).
- ZoneMessages: player loaded (the old PLAYER_LOADED case), ready for
updates, player ready, restore to post load stats, server done loading,
invalid zone transfer list, zone summary display/dismissed, level
processing complete, object world state, and the localized announcement
WorldMigration wrote by hand.
- PlayerMessages: chat mode, GM level, LEGO score, currency, reputation,
GM invis, pickup currency, zone and player statistics, chat commands, bug
reports, verify ack.
- ObjectMessages: fire event client/server side, notify client
(zone) object, notify object, script network vars, failed preconditions,
terminate interaction, set name, request use, request server object info.
- QuickBuildMessages: notify state, enable, cancel.
- ActivityMessages gains match response/update/request, leaderboard request
and data, shooting gallery score/rotation/fire, activity state change.
- MissionMessages gains MissionDialogueCancelled (a no-op, as before).
The wire structs left in GameMessages.h move to their domains (tooltip and
emote to Effects, loot and item use to Inventory, model build to Building,
behavior sound to Property, skill sets to Skill) and gain the missing
direction. The dismount logic moves to PossessorComponent::OnDismountComplete.
Every inbound message is registered in the GameMessageHandler map; the
switch is gone, and GameMessages.cpp only holds the GameMsg/NetGameMsg base
code. Call sites build the structs (NotifyClientObject, TerminateInteraction,
Teleport, PlatformResync, FireEventClientSide, NotifyObject and
NotifyClientZoneObject get convenience constructors like PlayFXEffect).
Dead senders are dropped: SendSetShootingGalleryParams (no callers, field
order was a guess), SendTeamPickupItem (the struct already existed),
SendRequestActivitySummaryLeaderboardData (the struct covers it).
Verified with RemainingMessagesTests: every old Send* function is frozen
verbatim in Legacy/RemainingMessagesLegacy.h and compared byte for byte
(same bits, destination and broadcast flag) over grids of inputs; every old
Handle* read sequence is frozen as a Read* oracle and compared with the
struct's Deserialize; round trips, truncation and a golden packet.
PlayerLoaded (0x00dc36f0), SetGMLevel (0x00dd6230), MissionDialogueCancelled
(0x00d9cc10) and LocalizedAnnouncementServerToSingleClient (0x00f23c50)
were checked against the client. Behaviour notes: an inbound message that
fails to deserialize is dropped, so ParseChatMessage over MAX_MESSAGE_LENGTH
is dropped instead of truncated, and PLAYER_LOADED / READY_FOR_UPDATES /
MISSION_DIALOGUE_CANCELLED now read their (unused) client fields.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
eReplicaComponentType had the destroyable component's registry type (7) named
BUFF, the real buff component (98) as BUFF_REAL, and a made-up DESTROYABLE =
1000 that DestroyableComponent was stored under. Now DESTROYABLE = 7 and
BUFF = 98, as in ComponentsRegistry and the client.
Undone with it:
- Destructible stats came from whichever of the "buff" (7), quick build and
collectible registry ids was set, so the few objects without a type 7 entry
read a DestructibleComponent row with an unrelated id (the NJ dragon relics
16482-16485 via their collectible id, 125 quick build LOTs when placed with
is_smashable). The type 7 entry is used now; objects without one keep the
defaults (is_smashable objects: 1 health, smashable, factions -1 and 6;
collectibles: an empty destroyable). The client does the same
(LWODestroyableComponent::AllocateComponents / DoObjectComponentLoad).
- DestroyableComponent::Reinitialize, an unused copy of that pick order.
- WriteComponents' destroyableSerialized flags: the components are written from
a list in the client's order, and where the destroyable goes (its own place
after the buff, right before a quick build that has no registry entry for it,
or after the render component) is one function.
- The dashboard's registry 7 -> DESTROYABLE mapping; the destroyable type also
has a name there now (1000 was outside magic_enum's range).
Component types are not stored or sent as enum numbers anywhere besides the
CDClient's own values, which now match. migrations/cdserver/4 is unrelated (it
restores LOT 12916's registry rows that migration 0 overwrote) and stays.
Verified: dGameTests ReplicaComponentOrderTests serialize players, enemies,
smashables, quick builds and collectibles with and without a registry entry,
NPCs, pets, vehicles, models and an entity with every listed component with
the new code and a frozen copy of the old WriteComponents, and expect the same
bits for construction and serialization (a deliberately wrong destroyable place
fails them). Full ctest passes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The pet name tables (Pet Names page and the review queue) get the pet's
LOT from its owner's save and its CDClient name, shown with the icon in
a new Pet column. Each owner on the page is read once.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The ActivatePhysics trigger command was a TODO. The client activates or
deactivates the object's physics component
(LWOPhysicsSystemComponent::msgActivatePhysics, 1.10.64 0x00ccf970);
the server now does the same to phantom physics: switching it off takes
the volume out of the physics world (dpWorld::DetachEntity, without
deleting it) and makes whatever was inside leave, switching it on adds
it back and whatever is inside enters on the next step. This lets
trigger driven volumes like the monument lasers turn on and off.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Volumes on the server touched far more than in the client:
- An enemy's body in the physics world was a sphere the size of its aggro
radius, so trigger and damage volumes caught enemies from far away
(Cavalry Hill enemies taking damage on spawn). It is now the enemy's
own radius and collision group from its physics component.
- Trigger volumes ignored their collision group and caught everything.
dpEntity now filters with the client's collision filter
(PeCollisionFilter, 1.10.64 0x00fb6940, group table from 0x00fcf9a0):
POI walls ignore enemies, threat clearing walls ignore players, and
so on. The aggro sensor keeps seeing only players.
- Rotated boxes were tested as the axis aligned box around them, which
for a turned wall covers a big square (the AG survival boundary).
Sphere and point tests now use the box's own axes.
Proximity monitors take an optional collision group like live's
SetProximityRadius; the AM shield generators use live's (10 finds
enemies, 1 finds players).
Fixes#1127
Refs #1971
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The client only simulates knockbacks on the character it controls
(LWOControllablePhysComponent::msgKnockback); enemies and NPCs are drawn
where the server puts them, so a knockback on them did nothing and
scripts stunned them instead.
MovementAIComponent::Knockback now flies the object the way the client
flies its character: a vector longer than 5 lifts it 0.5, throws it with
that velocity under WorldConfig gravity (times its gravity scale) until
it lands on the navmesh, no sooner than 250ms later; a shorter one moves
it by the vector. Walls taller than a step stop sideways motion, the
landing is put back onto the navmesh, and pathing and the combat AI wait
until it lands (destinations set mid air are walked to afterwards).
Position and velocity go out in the normal serialization every tick.
KnockbackBehavior builds the vector like the client's Cast (strength
capped at 300, angle as elevation, relative, caster and ignore_self) and
knocks back server moved targets that aren't immune, both when the
server casts and when a client's skill hits. The blocked bit it writes
now also answers for the target's knockback immunity, so a player whose
client hasn't caught up with a Personal Fortress isn't knocked out of it.
The AM shield generators knock enemies back like live instead of
stunning them.
Fixes#257
Refs #185
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bind_ip in sharedconfig.ini picks the local IPv4 address every server's
RakNet sockets listen on (auth, chat, master, world, the dashboard's
connection to master), separate from external_ip, the address players
are sent. Empty or 0.0.0.0 keeps listening on all interfaces; localhost
means 127.0.0.1. Anything that isn't an IPv4 address stops the server
with an error, and each server logs what it bound to.
Fixes#225
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every DataTable saves its order and page length to localStorage, keyed by
dashboard user, page (numeric path segments folded) and table id, and
restores it on load. Stale sorts on missing or unsortable columns are
dropped; storage failures are ignored.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts the combat messages to NetGameMsgs in CombatMessages.{h,cpp}:
Die, Resurrect, SetResurrectRestoreValues, SetPlayerAllowedRespawn
(SendToClient, as before), Knockback, SetStunned, SetStunImmunity,
SetStatusImmunity, AddBuff, RemoveBuff, AddRunSpeedModifier,
RemoveRunSpeedModifier and DeactivateBubbleBuffFromServer, and the
received RequestDie, RequestSmashPlayer, RequestResurrect, Resurrect,
ActivateBubbleBuff and DeactivateBubbleBuff. Smash and UnSmash move
here from GameMessages.h (Smash's ghostCapacity is ghostOpacity, the
client's name) and gain a Deserialize.
The callers (DestroyableComponent, BuffComponent,
ControllablePhysicsComponent, QuickBuildComponent, RacingControl,
RailActivator and the scripts) build the structs. SendResurrect's
respawn timer moves to DestroyableComponent::Resurrect. The received
messages are registered in GameMessageHandler's map and their switch
cases are deleted, along with the unused HandleRequestDie overload and
the Send* functions nothing called (SendSmash, SendUnSmash, the run
speed modifiers, SendActivateBubbleBuffFromServer). Handler logic is
unchanged.
No wire change and no change in recipients. Verified byte for byte
against a frozen verbatim copy of the old functions over an input grid
(every optional field, bool patterns, empty and long strings), received
messages compared with the old handlers' read sequences and truncated
payloads rejected, hand computed golden bytes, round trips, and a
deliberate width mutation made the tests fail.
Known difference from the client, not changed here: DLU writes
SetStatusImmunity's nine flags in its own order; the client reads them
alphabetically (GameMessage::SetStatusImmunity::Serialize @ 00d8f140).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts the skill messages to NetGameMsgs in SkillMessages.{h,cpp}:
AddSkill and RemoveSkill (SendToClient, as before), EchoStartSkill,
EchoSyncSkill and DoClientProjectileImpact, and the received
SelectSkill, StartSkill, SyncSkill and RequestServerProjectileImpact.
The one-off StartSkill, EchoStartSkill, SyncSkill, EchoSyncSkill,
RequestServerProjectileImpact and DoClientProjectileImpact classes are
deleted; SkillComponent, BehaviorContext and InventoryComponent build
the structs, and the dashboard's message decoder reads them.
The received messages are registered in GameMessageHandler's map and
their switch cases are deleted; the handlers call SkillComponent as
before. The echoes still go to every client except the caster, through
the new NetGameMsg::BroadcastExcept. SelectSkill still accepts any
payload, since the old case read nothing. Handler logic is unchanged
(the SyncSkill case's unused hex dump of the payload is dropped).
No wire change. Verified byte for byte against frozen verbatim copies
of the old classes and functions over an input grid (every optional
field set and unset, empty, short and long behavior streams), received
messages compared with the old classes' read sequences and truncated
payloads rejected, the echo's broadcast-except destination compared
with the old send, hand computed golden bytes, round trips, and a
deliberate mutation made the tests fail. Messages that fail to
deserialize are now dropped instead of being handled half read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts the vendor, donation vendor and trade game messages to
NetGameMsgs in VendorMessages.{h,cpp} and TradeMessages.{h,cpp}:
VendorOpenWindow and VendorTransactionResult (SendToClient: the old
functions never broadcast), VendorStatusUpdate, ServerTradeInvite,
ServerTradeInitialReply, ServerTradeFinalReply, ServerTradeAccept,
ServerTradeCancel and ServerTradeUpdate, and the received
RequestVendorStatusUpdate, BuyFromVendor, SellToVendor,
BuybackFromVendor, AddDonationItem, RemoveDonationItem,
ConfirmDonationOnPlayer, CancelDonationOnPlayer, ClientTradeRequest,
ClientTradeCancel, ClientTradeAccept and ClientTradeUpdate. A trade
offer entry (the client's inventory item layout, optional fields and
config block included) is TradeItemEntry, shared by both trade updates.
Selling and buying back move into VendorComponent (SellToVendor,
BuybackFromVendor), adding and confirming donations into
DonationVendorComponent, and VendorComponent builds its own status
update and transaction results. Only the VENDOR component sends its
stock, as before. The received messages are registered in
GameMessageHandler's map, the switch cases and the old functions are
deleted. Handler logic is unchanged.
No wire change and no change in recipients. Verified byte for byte
against a frozen verbatim copy of the old functions over an input grid
(to one client and broadcast), received messages compared with the old
handlers' read sequences (every optional field of a trade entry,
raw and compressed config blocks) and truncated payloads rejected, hand
computed golden bytes, round trips, and a deliberate width mutation made
the tests fail. ConfirmDonationOnPlayer still accepts a message without
its vendor ID, since the old handler read nothing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Property worlds now give their property reputation for the time other people spend on it, which
fills the property lists and the news screen's Today's Top Properties. Visitors are counted per
account; the owner's account, accounts linked to it and staff don't count. A visit earns nothing
for the first property_reputation_min_visit seconds (WorldConfig's propertyReputationDelay), then
each active minute (the visitor moved) earns reputationPerMinute times a multiplier, for a capped
number of minutes per visit. Repeat visitors earn less the more recent days they already gave
reputation, and each visitor and each property have a daily cap. Every parameter is a setting;
what each account gave each property per day is kept in property_reputation_visits. The rules are
pure functions with unit tests.
Fixes#636Fixes#637
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Off unless property_rent_enabled is on. Each property world's rent comes from its PropertyTemplate
row (minimumPrice every rentDuration x durationType; Block Yard is free), unless the new Property
Rent dashboard page sets another price or period (property_rent_manage). Rent is taken from the
owner's coins shortly after their character loads, with a mail receipt; unpaid rent is mailed and,
after property_rent_grace_days, makes the property private until it is paid, like live. The
property management component refuses public or best friends privacy while rent is overdue and a
property world that loads overdue makes itself private. Property game messages are unchanged.
Fixes#943
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Staff list contraband items on a new dashboard page (item search, reason, flag or flag and remove;
contraband_manage to edit, reports_view to see). World servers check every inventory when a
character loads and every item a player receives: each find is an economy flag of the new kind
Contraband, shown with the other flags and in the character's related data. Items marked for
removal are taken away, with a character snapshot kept first so they can be given back, an audit
entry and a mail or chat message telling the player why. Staff are skipped unless
contraband_ignore_staff is off. Worlds reload the list when it changes (RELOAD_CONTRABAND, added
at the end of ePlayerAction).
Fixes#1563
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every character gets a save generation in charxml. A world bumps it when it loads the character for
play, and every save from that world only goes through while the stored generation is still the
one it loaded or last saved (and moves it on). Dashboard edits, restores and maintenance writes bump
it too. A world that lost the character to another world (a disconnect noticed late, a zone
transfer, an instance migration) or to a dashboard edit can no longer overwrite the newer data: the
save is refused, logged and audited as stale_save_refused, the world stops saving that character
and a player still connected to it is disconnected with the save failure reason so they reload.
Fixes#639
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Opening a zone the first time built its terrain, scene objects and manifests
and ran ImageMagick on the web thread, stalling the dashboard for seconds.
- Workers: the shared pool plus Workers::Reply (answer at once when built,
else from a worker via Web::Defer).
- terrain_chunks/terrain_layers/scene/paths/scenery/flairs (world3d, property
and showcase routes) and terrain textures go through it; results are built
once in OnceCaches, the .raw is read once per zone for chunks, layers and
flairs, deflated bodies are cached thread-safely.
- ImageMagick conversions are deduplicated and written under a temporary name.
- Workers don't query the CDClient, read settings or call mongoose: ZoneTable,
render components, flairs, object names, LOT kinds and terrain texture names
are read at startup; client_location is read once; base64 is plain C++.
- Logger writes one line at a time (mutex; localtime's buffer is shared).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts AddItemToInventoryClientSync, SetInventorySize,
RemoveItemFromInventory, ConsumeClientItem, UseItemResult,
UseItemRequirementsResponse, ResponseMoveItemBetweenInventoryTypes,
NotifyNotEnoughInvSpace, UpdateInventoryUi, MarkInventoryItemAsActive and
MoveInventoryBatch (11 old Send functions) and the received EquipInventory,
UnEquipInventory, RemoveItemFromInventory, MoveItemInInventory,
MoveItemBetweenInventoryTypes, RequestMoveItemBetweenInventoryTypes,
PushEquippedItemsState, PopEquippedItemsState, ClientItemConsumed,
UseNonEquipmentItem, SetConsumableItem, UpdateInventoryGroup and
UpdateInventoryGroupContents (13 handlers) to NetGameMsgs in
InventoryMessages.{h,cpp}. The received messages are registered in
GameMessageHandler's map and handled by InventoryComponent (new On* methods
holding the old handler logic verbatim); the old functions and switch cases
are deleted. AddItemToInventoryClientSync::SetItem fills the fields that
come from the Item.
No wire change and no change in recipients. Kept as DLU has always sent them
and documented: NotifyNotEnoughInvSpace goes out with the message ID of
VehicleNotifyFinishedRace, and RemoveItemFromInventory always sets the flag
of the fields DLU fills. The old SendMoveInventoryBatch was never called
and wrote one flag bit fewer than the client reads (it had no moveSubkey);
the struct follows the client's layout (1.10.64,
LWOInventoryComponent_Common::msgMoveInventoryBatch at 00ce1310) and has no
oracle. UnEquipInventory still ignores the trailing replacementObjectID the
client sends, as before.
Verified byte for byte against a frozen verbatim copy of the old functions
over an input grid (AddItemToInventoryClientSync with real Items, extra info
and bind flags), to one client and broadcast; received messages compared
with the old handlers' read sequences and every truncated payload rejected;
hand computed golden bytes; round trips; a deliberate width mutation made
the tests fail.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts PlayAnimation, PlayNDAudioEmitter, PlayEmbeddedEffectOnAllClientsNearObject,
PlayFXEffect, StopFXEffect, BroadcastTextToChatbox, Play2DAmbientSound,
Stop2DAmbientSound, UIMessageServerToSingleClient, UIMessageServerToAllClients,
StartCelebrationEffect, DisplayMessageBox, DisplayChatBubble, ChangeIdleFlags,
SetNameBillboardState, ShowBillboardInteractIcon, PlayCinematic, EndCinematic,
SlashCommandTextFeedback, PlayEmote and SetEmoteLockState (21 old Send
functions, 23 with overloads) and the received MessageBoxRespond,
ChoiceBoxRespond, CinematicUpdate and PlayEmote to NetGameMsgs in
EffectsMessages.{h,cpp}. The high traffic messages get a constructor for their
required fields. UI messages own their AMF arguments. Every caller is switched,
the received messages are registered in GameMessageHandler's map, and the old
functions and switch cases are deleted. Handlers are copied verbatim.
No wire change and no change in recipients: functions that always broadcast
whatever address they were given are sent with Send(UNASSIGNED_SYSTEM_ADDRESS),
functions that only did SEND_PACKET use SendToClient. Quirks are kept and
documented on the structs (PlayAnimation's UTF-8 sized name, the always written
null terminator in BroadcastTextToChatbox, StartCelebrationEffect always
writing celebrationID, SetNameBillboardState having no payload).
Verified byte for byte against a frozen verbatim copy of the old functions over
an input grid, to one client and broadcast; received messages compared with the
old handlers' read sequences and every truncated payload rejected; hand
computed golden bytes; round trips; a deliberate width mutation made the tests
fail. The byte-equality helper gains a Broadcast mode for functions that
ignored their address.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every MASTER service packet is now an LUBitStream struct (docs/PacketArchitecture.md,
PR 14) and the master server's switch is a dispatch map (PacketDispatcher), as are the
master handlers of the world, chat and dashboard servers.
- dNet/MasterPackets.h: RequestZoneTransfer, RequestZoneTransferResponse, ServerInfo,
RequestSessionKey, SetSessionKey, SessionKeyResponse, NewSessionAlert, PlayerAdded /
PlayerRemoved, CreatePrivateZone, RequestPrivateZone (passwords still cut to 50
characters when read), WorldReady, WorldReadyInfo (WORLD_READY to the dashboard),
PrepZone, Shutdown, ShutdownResponse, WorldShutDown (SHUTDOWN_RESPONSE to the
dashboard), ShutdownUniverse, AffirmTransferRequest/Response, RequestServerList,
ServerListResponse, DashboardShutdown, ConfigReload, InstanceShutdown. The Send*
functions are gone; MasterPackets::SendToMaster(msg) and SendTo(sysAddr, msg) send a
struct.
- The dashboard and instance migration structs (PlayerAction, DataChanged, Dashboard
messages, MessageCapture, InstanceMigration) are LUBitStreams of the MASTER service now
and moved to dNet/master/, included by MasterPackets.h. Their payloads are unchanged;
master forwards them by re-serializing the struct instead of copying raw bytes.
- InstanceManager, ZoneInstanceManager, MigrationCoordinator, dServer (server info, zone
transfer response), auth (SET_SESSION_KEY), the world (session keys, player added and
removed, world ready, shutdown response, affirmations, prep zone, shutdown universe) and
the dashboard (server list, instance shutdown, config reload, announcements, player
actions, message capture) send and read structs.
- The login stamps are a `stamps` field of RequestZoneTransfer and
RequestZoneTransferResponse (read leniently as before: a message without them reads as
empty); master adds its stamps in the REQUEST_ZONE_TRANSFER handler and when it answers,
as it did.
- InstanceManager::GetInstanceBySysAddr takes a const address.
Verified: tests/dGameTests/dNetTests/Legacy/MasterPacketsLegacy.h is a verbatim copy of
the old writers and readers; MasterPacketsTests requires identical bytes for a grid of
inputs, checks the old readers read what the structs write, round trips and truncation,
hand written golden packets, that zone transfers without stamps still read, that the dashboard/migration structs write what
"header + Serialize" wrote, and that the dispatcher drops truncated packets. No wire
bytes changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every packet of the chat service is now an LUBitStream struct in dNet/ChatPackets.h
(docs/PacketArchitecture.md, PR 13), and nothing in the chat server or the chat side of
the world server reads or writes a packet by hand any more.
- World <-> chat: LoginSessionNotify, UnexpectedDisconnect, GMLevelUpdate, GMMute,
Announcement (GM announce), CreateTeam, TeamUpdate (TEAM_GET_STATUS to worlds),
AchievementNotify, ShowAllRequest, FindPlayerRequest.
- Client -> world -> chat (friends, ignore list, teams, general and private chat):
GetFriendsList, AddFriendRequest, AddFriendResponse, RemoveFriend, GetIgnoreList,
AddIgnore, RemoveIgnore, GeneralChatMessage, PrivateChatMessage, TeamInvite,
TeamInviteResponse, TeamLeave, TeamKick, TeamSetLeader, TeamSetLoot, TeamGetStatus.
The 77 bytes the handlers skipped are named fields now (the sender block the client
fills in); the 4 unused bytes after the player ID are kept as `unknown`.
- What the client receives. Chat service (ChatPackets::Client): GeneralChatMessage
(replaces SendChatMessage; SendSystemMessage stays as a helper built on it),
PrivateChatMessage. Client service (ClientPackets, as structs go in the file of the
ServiceType in their header): SendCannedText (replaces SendMessageFail), GetFriendsListResponse, AddFriendRequest,
AddFriendResponse, RemoveFriendResponse, UpdateFriendNotify, WhoResponse,
ShowAllResponse, Get/Add/RemoveIgnoreResponse, TeamInvite, TeamInviteInitialResponse,
and the team game messages chat writes (TeamInviteConfirm, TeamGetStatusResponse,
TeamSetLeader, TeamAddPlayer, TeamRemovePlayer, TeamSetOffWorldFlag) as TeamGameMsg
structs, since chat doesn't link dGame's NetGameMsg.
- WORLD_ROUTE_PACKET is ChatPackets::WorldRoutePacket (its own file, dNet/WorldRoutePacket.h,
since it carries packets of other services): the target and the inner packet.
ChatPacketHandler::SendRouted replaces the per-function route headers and
SendRoutedMsg.
Dispatch: dNet/PacketDispatcher.h is a dispatch map from packet ID to (struct, handler
function); packets that fail to Deserialize are logged and dropped. The chat server's
switch and the world's HandlePacketChat switch are now maps. The handlers take the
structs; their logic is unchanged. World code sends to chat with ChatServerLink::Send
(dGame) instead of writing to Game::chatServer by hand. eChatChannel,
eChatMessageResponseCode and eAddIgnoreResponse moved to dCommon/dEnums so the structs
can use them.
Verified: tests/dGameTests/dNetTests/Legacy/ChatPacketsLegacy.h is a verbatim copy of
the old senders and readers; ChatPacketsTests (25 tests) sends a grid of inputs through
both and requires identical bits, bytes and destination, checks the old readers read
what the structs write, round trips every struct, checks truncated packets are refused,
and has hand written golden packets for general chat, canned text, GM mute and a routed
team game message. Breaking one field width (UpdateFriendNotify) and the TeamAddPlayer
zone flag made the tests fail. No wire bytes changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts build mode, arranging, modular build and brick by brick
messages to NetGameMsgs in BuildingMessages.{h,cpp}: StartArrangingWithItem,
FinishArrangingWithItem, ModularBuildEnd and SetBuildModeConfirmed
(sent), and StartBuildingWithItem, DoneArrangingWithItem,
ModularBuildFinish, ModularBuildMoveAndEquip, ModularBuildConvertModel,
SetBuildMode, BuildModeSet, UnUseBBBModel, BBBLoadItemRequest and
BBBSaveRequest (received, registered in GameMessageHandler's map with
their handlers' logic kept). The BLUEPRINT_SAVE_RESPONSE and
BLUEPRINT_LOAD_RESPONSE_ITEMID client packets become the LUBitStream
structs ClientPackets::BlueprintSaveResponse and BlueprintLoadItemResponse,
also used by WorldServer's level load. The never-called
SendBBBSaveResponse is gone.
SetBuildModeConfirmed keeps writing modeValue's and startPos's default
flags as always set, as DLU did. BuildModeSet and UnUseBBBModel now read
their whole client layout (DLU read only the first fields).
No wire change and no change in recipients. Verified byte for byte
against a frozen verbatim copy of the old functions and inline packet
writes over an input grid, received messages compared with the old
handlers' read sequences with every truncated payload rejected, hand
computed golden bytes, round trips and a mutation check. Layouts
confirmed against the 1.10.64 client in Ghidra.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts the property messages to NetGameMsgs in PropertyMessages.{h,cpp}.
Sent: OpenPropertyVendor, OpenPropertyManagement, DownloadPropertyData
(replaces PropertyDataMessage, now with the client's PropertyData field
names), PropertyRentalResponse, PropertyEntranceBegin,
PropertySelectQuery (replaces PropertySelectQueryProperty with the
client's PropertyInfo), GetModelsOnProperty, PlaceModelResponse and
HandleUGCEquipPre/PostDeleteBasedOnEditMode. Received, registered in
GameMessageHandler's map: SetPropertyAccess,
UpdatePropertyOrModelForFilterCheck, QueryPropertyData,
PropertyEditorBegin/End, PropertyContentsFromClient,
ZonePropertyModelEquipped/Rotated, PlacePropertyModel,
UpdateModelFromClient, DeleteModelFromClient, ControlBehaviors,
PropertyEntranceSync, EnterProperty1, UpdatePropertyPerformanceCost,
ReportOffensiveModel/Property and GetHotPropertyData. The news screen's
NewsSendHotPropertiesInfoToClient moves here with the top properties
lookup; the dashboard's player reports now take the decoded text.
Handlers keep their logic and hand the work to PropertyManagementComponent,
PropertyVendorComponent, PropertyEntranceComponent and
MultiZoneEntranceComponent as before. Messages whose payload DLU ignored
(PropertyEditorBegin, PropertyContentsFromClient, ZonePropertyModel*)
now read it with the client's layout. The never-called
SendZonePropertyModelEquipped (which wrote no default flags) is gone.
No wire change and no change in recipients. Verified byte for byte
against a frozen verbatim copy of the old functions, PropertyDataMessage
and PropertySelectQueryProperty over an input grid (to one client and
broadcast), received messages compared with the old handlers' read
sequences with every truncated payload rejected, hand computed golden
bytes, round trips and a mutation check. Layouts confirmed against the
1.10.64 client in Ghidra. Known wire bug kept as is: PlaceModelResponse
writes response where the client reads a rotation quaternion
(PlaceModelResponse::Deserialize 0x00dc0170).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Converts the pet taming minigame, naming, command and bouncer messages to
NetGameMsgs in PetMessages.{h,cpp}: NotifyPetTamingMinigame,
NotifyTamingModelLoadedOnServer, NotifyPetTamingPuzzleSelected,
PetTamingTryBuildResult, PetResponse, AddPetToPlayer, RegisterPetID,
RegisterPetDBID, ShowPetActionButton, BouncerActiveStatus, SetPetName,
SetPetNameModerated and PetNameChanged (sent), and PetTamingTryBuild,
NotifyTamingBuildSuccess, RequestSetPetName, StartServerPetMinigameTimer,
ClientExitTamingMinigame, CommandPet and DespawnPet (received, registered
in GameMessageHandler's map; each Handle hands the message to the
player's taming or active PetComponent exactly as the old handler did).
PetComponent, BouncerComponent and the hydrant/catapult scripts build
the structs; the old Send*/Handle* functions and switch cases are gone.
The never-called SendClientExitTamingMinigame is folded into the
ClientExitTamingMinigame struct. MarkInventoryItemAsActive stays with
the inventory messages.
No wire change and no change in recipients. Verified byte for byte
against a frozen verbatim copy of the old functions over an input grid
(to one client and broadcast), received messages compared with the old
handlers' read sequences with every truncated payload rejected, hand
computed golden bytes, round trips, and a mutation check. Layouts
confirmed against the 1.10.64 client's Serialize/Deserialize in Ghidra.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
WorldPackets now holds what a client sends a world server, one struct per
packet with Serialize/Deserialize: Validation, CharacterListRequest,
CharacterCreateRequest, CharacterLoginRequest, GameMessage,
CharacterDeleteRequest, CharacterRenameRequest, LevelLoadComplete,
PositionUpdate, MailPacket, RoutePacket, StringCheck, GeneralChatMessage,
HandleFunness, UIHelpTop5. WorldServer's switch became a dispatch map of
handlers (each overrides Handle in WorldServer.cpp, logic unchanged:
dashboard hooks, LoadPlayer, chat logging, migration checks, message
inspector capture all still run); a packet that does not deserialize is
logged and dropped. UserManager's create/delete/rename take the structs.
The answers are ClientPackets (the CLIENT service): LoadStaticZone,
CharacterListResponse, CharacterCreateResponse, CharacterRenameResponse,
DeleteCharacterResponse, TransferToWorld, ServerStates, CreateCharacter,
ChatModerationString, MakeGMResponse, HTTPMonitorInfoResponse and
DebugOutput, built at the call sites (world server, UserManager, slash
commands, dashboard actions, components, migration). The world -> chat
forward of a routed packet is ChatPackets::RoutedFromClient. All
WorldPackets::Send* functions, HTTPMonitorInfo and the ClientPackets parse
functions are gone. The architecture doc now states the file rule: a
packet lives in the file of the ServiceType in its header.
No wire change. Verified against frozen verbatim copies of the old code
(tests/dGameTests/dNetTests/Legacy/WorldPacketsLegacy.h): every response
is sent through the old function and the struct over grids of inputs
(all enum values, strings of every length class, IDs, >64 moderation
segments, big XML) and must match byte for byte; every request is read
by the old code and the struct and must give the same values; plus
golden bytes, round trips, truncation checks and a field width mutation
that made the tests fail. Only differences: malformed requests are
dropped instead of handled with partial values, and LevelLoadComplete now
reads the zone ID the client sends after it (lu_packets and captures show
the 1.10.64 client always sends it; DLU ignored it).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The login's Stamps now go with auth's REQUEST_ZONE_TRANSFER to master and
come back in REQUEST_ZONE_TRANSFER_RESPONSE, so master stamps the steps it
performs itself, with its own time: got the request
(WORLD_PACKET_RECEIVED, zone), the world is still starting and the request
waits (IM_LOGIN_QUEUED, instance), answered with a world
(WORLD_SESSION_CONFIRM_TO_AUTH, instance). Auth sends what comes back in
the login response.
Server to server wire change (message IDs unchanged): both messages end
with a Stamps list (u32 16 * count + 4, then the stamps); it is empty
(4 bytes) when a world server asks for a transfer. Touched:
- MasterPackets::SendZoneTransferRequest / SendZoneTransferResponse: take
and write the stamps (default empty)
- MasterServer.cpp REQUEST_ZONE_TRANSFER: reads them, stamps, keeps them
in the PendingInstanceRequest (new stamps member, InstanceManager.h)
- InstanceManager::ReadyInstance / AffirmTransfer: stamp and send them back
- ZoneInstanceManager: HandleRequestZoneTransferResponse reads them; a
RequestZoneTransfer overload takes stamps and a callback that gets them
- AuthPackets LoginRequest::Handle uses that overload
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The login response's stamps were a fixed list: START and CLIENT_OS at the
start, an ERROR or two, and IM_LOGIN_START / WORLD_COMMUNICATION_FINISH
appended to every response whatever happened. Now a Stamps list (new
dNet/Stamps.{h,cpp}: eStamps, Stamp and Stamps with Serialize /
Deserialize in the login response's layout, so server messages can carry
it too) is created when the login request arrives, and each step adds its
own stamp, with the time, where it happens: the request read (value: the
client's OS), the account lookup (found or not), the password check,
failed checks, the closed server and play key checks, database writes,
asking master for a world, master's answer, handing the session key to
master, and sending the player on. The response serializes whatever was
stamped; the auth server logs the list like the client does.
What the client does with stamps (1.10.64): PacketHandler_MSG_CLIENT_
LOGIN_RESPONSE @ 00b32f90 reads them (LoginResponse::ReadStamps @ 005ed3c0,
count = (size - 4) / 16) and only logs each with its name from
StampLookup @ 017e6e88 (the 39 eStamps names) and its time relative to the
first and previous stamp. Documented in Stamps.h.
Behaviour change: on success, master gets the session key just before the
client gets the response instead of just after, so the handoff can be
stamped (and master knows the key before the client can reach a world).
The client-facing layout is unchanged; tests pin the stamp bytes, check
the steps of a failed login in order, and round trip the list.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adds CommonPackets (ServiceType::COMMON): ClientVersionConfirm (the
client's handshake, with the handler that logs it and answers),
ServerVersionConfirm, DisconnectNotify and GeneralNotify (layout from the
1.10.64 client's PacketHandler_MSG_SERVER_GENERAL_NOTIFY; DLU does not send
it yet). AuthPackets::LoginRequest reads the login and keeps the old login
logic in its Handle; ClientPackets::LoginResponse (with the Stamp list)
replaces the hand written response, and AuthPackets::SendLoginResponse
fills it from the settings as before. dServer::Disconnect writes a
DisconnectNotify. AuthServer's if-chain and WorldServer's COMMON case
become CommonPackets::Handle / AuthPackets::Handle, dispatch maps that
read the struct, drop and log it if it does not deserialize, then Handle.
HandleHandshake and SendHandshake are gone.
No wire change. Verified against a frozen verbatim copy of the old
functions (tests/dGameTests/dNetTests/Legacy): the old handshake and login
handlers and the new dispatchers are fed the same packets and must send
identical bytes to the same address (same RNG seed for the session key),
over grids of versions, service types, ports, response codes, error
messages, IPs and stamp lists; plus hand computed golden bytes, round
trips, truncation checks, and a deliberate field width mutation that made
the tests fail. The only behaviour difference: truncated handshake and
login packets are now dropped instead of handled with whatever was read.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dashboard's web server answers one request at a time, so converting a big
.nif (glom files up to tens of MB) held up every other request, flairs included.
- dWeb: Web::Defer hands a request to another thread; the reply is sent from the
web thread on its next poll (DeferredQueue). A client that leaves first cancels
it and the late reply is dropped. The synchronous route API is unchanged.
- Web::Shutdown closes connections while the state their close events touch is
still alive; the destructor no longer runs handlers during static destruction
(stopping the dashboard aborted in ~WSClient).
- WorkerPool: priority lanes, with one thread only for urgent work (flairs,
small models, textures), and limited background work.
- Scenery: mesh and texture routes (and the showcase's) convert on the pool;
thread-safe memory and disk caches, one conversion per model at a time with
waiters sharing it; zones are converted ahead onto the disk cache while viewed.
- Setting scenery_workers (0: half the cores, 2 to 4).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Its key starts with the zone, so the economy compaction read the whole table twice for every
day it merged into a month. With 900k old rows to merge on MariaDB the task went from 107 s to
14 s. The index holds only the day and the key, so the game's upsert that adds to a row's
events never updates it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Leave staff out of the currency and U-score reports by excluding the few staff characters,
instead of joining every ledger row to its character and account.
- Top earners: total per character first and look up names for the top rows only.
- The places list and the activity report read map events for every kind in one query
(GetMapZonesAllKinds) instead of one query per kind.
With about 1M currency rows and 650k map event rows (90 days) on MariaDB: currency 1.8 s to
0.8 s, top earners 2.4 s to 0.75 s, places 2.6 s to 0.5 s, activity 2.6 s to 1.7 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mysqldump was given --host=tcp --port=//host:port for the tcp:// form the servers accept, so
backups failed. Read mysql_host the way the servers connect with it: tcp://host:port (a trailing
/database is dropped), unix:// as --socket and pipe:// as a named pipe.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The picker's EXISTS on ComponentsRegistry (no index) took 2 to 4 s with the bundled SQLite for the
treasure kinds; the same filter as IN takes under 10 ms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Read activity and map_id as signed: a row with map_id -1 made the page fail with a server error
when sorted by map.
- A search counts and pages through the characters it matches (index on character_id) instead
of joining every row of the log to a name; when the matches are a large part of the log the
page still walks the log in order.
- Sorted by character name, only the row ids are sorted and the page's rows are read after.
On a log of about 800k rows this takes searches with few matches from about 1 s to 0.1 s and
name sorts from about 3 s to 0.5 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On a database NexusDashboard (or a server's own changes) already migrated, 29, 30 and 40 logged
an error for every column that was already there, and 40 and 57 added a second index on
columns MariaDB had already indexed for the inline REFERENCES foreign keys of 0_initial.sql.
Add those columns and indexes only when information_schema says they are missing.
40 also gives pet_names.owner_id values written before 23_store_character_id_as_objectid.sql
the persistent bit, as 23 did for every other character id column, so they match characters.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each skip computed an absolute offset as a 32 bit tellg plus a size
product in 32 bit math (colorMapSize * colorMapSize * 4 and friends),
which wraps for large values and can seek backwards. The skips are now
relative to the current position with the size promoted to
std::streamoff first. Offsets are unchanged for every real terrain file,
so this only removes the overflow the TODO asked about.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Two script TODOs with the live scripts shipped in the client to follow:
- AgPropGuard: accepting mission 768 before touching the orb (flag 71
unset) plays the "MissionCam" cinematic, as L_AG_PROP_GUARD.lua does.
Also no longer dereferences a missing Character.
- BaseFootRaceManager: joins the "FootRaceStarter" group on startup like
L_ACT_BASE_FOOT_RACE.lua, since other foot race objects look the
starters up through that group.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New instances already get their address from external_ip through the
master server's dServer, so the TODO asking to read it from config was
done. No behaviour change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Trigger commands with target="zone" gathered no targets, so they never
ran. Forbidden Valley uses it for two ActivateSpawnerNetwork commands
(qb1 on interact, qb2 on activation) in nd_forbidden_valley.lutriggers,
and the winter large property uses it too. The zone is now the zone
control object, the same entity scripts treat as the zone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Speed buffs (BuffParameters "speed", for example the slow from the Great
Katana of Freezing and the Shurikens of Ice) and speed behaviours go
through ControllablePhysicsComponent's speed multiplier, which only
players' clients used. Enemies move with MovementAIComponent, which
ignored it, so slows had no effect on them.
- MovementAIComponent scales its waypoint speed by the entity's speed
multiplier when one is set.
- RemoveSpeedboost set the multiplier of any entity without a
LevelProgressionComponent (every enemy and pet) to 0 once its last
boost ended, which was harmless while nothing read it for them. It now
goes back to the normal base of 500 (multiplier 1).
Needs an in-game check with the freezing katana charge-up against an
enemy.
Fixes#1179
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The live script (L_EXPLODING_ASSET.lua) does everything inside
"if not bIsHit": smash the attacker if they stand within 10 units, cast
the explosion skill, smash itself and progress missions. The server only
guarded the attacker smash, so another hit before the asset despawned
cast the skill and progressed missions again.
The TODO asking to kill everything in a radius is obsolete: live smashes
only the attacker directly, and everyone else in range takes the
explosion skill's damage, which the server already casts. Removed it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The drill only progressed missions 972, 1305 and 1308 for the player who
started it, so another player breaking the drill (the usual way on a
busy Crux Prime) got no credit for the drill objectives. The live
script (L_SKULLKIN_DRILL.lua, notifyDie) credits the killer, then also
the activator when that is someone else, which is what it does now. It
also no longer dereferences a null attacker. Removes the "Missions"
TODO.
May account for part of #1021 (Crux Prime daily smashable objectives not
tallying); needs an in-game check with two players.
Refs #1021
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ported from the live server scripts that ship with the client:
- scripts\ai\LS\ForceFieldEffect.lua (LEGO Club and Starbase 3001 force
fields): plays effect 3671 "cast" on whatever touches the phantom.
- scripts\ai\NP\L_NPC_NP_JETPACK_GUY.lua (jetpack ambient NPCs): plays
the object's "launch" effect when used.
- scripts\02_server\Map\AM\L_SKELETON_SPAWNER_VOLUME.lua (Crux Prime
Skullkin area): with 10 or more players in the volume the Skullkin
engineer, miner and patroller networks maintain 2/6/6 instead of
1/4/4, with a 60 second cooldown between changes and a catch-up check
when it ends. Only players are counted, and the count never goes
below zero.
WBL_Enemy_Grabbler.lua only sets wander variables for the old Lua
movement AI, so it joins the other alpha wander scripts in the excluded
list instead of logging as missing.
Refs #746
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The faction and immunity change, spawning the spider wave, knockbacks
and reading the animation time are all implemented right below their
TODOs, so the comments were stale. No behaviour change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dropship computer's OnDie activated the "next" spawner network
(pipe number + 1) when a player smashed it and only fell back to the
first network otherwise. The computer has no next network, so once a
player smashed it (easiest with area attacks like the Doom Hammer) it
never came back. The live script (L_DROPSHIP_COMPUTER.lua, shipped in
the client's server scripts) always resets its own network and
activates the first one, which is what it does now.
OnUse also dereferenced GetMission(979) without a null check, crashing
the world when a player without that mission used the built computer.
It now matches the live script: only players with the mission active
get the Nexus Talon data card, once. SsModularBuildServer had the same
unchecked GetMission and now checks the mission state.
Fixes#596
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"client_location = ../client " or a value with a trailing space is a
common setup mistake that surfaces much later as an unrelated error (a
missing client folder, a failed database login). Keys and values are now
trimmed of spaces, tabs and line endings when the ini is read, which
also covers the \r of files saved with Windows line endings. Spaces
inside a value are kept, and lines with an empty key are ignored.
Verified with a new unit test that loads an ini with padded keys and
values.
Refs #1113
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Loading a character whose xml has a malformed flag threw out of
std::stoul, and a character whose xml failed to parse (or has no obj
tag) dereferenced null on load and again on every save.
- Flags are parsed with TryParse and a malformed one is logged and
skipped instead of aborting the load.
- The quick parse checks for the obj and items tags before using them.
- Saving refuses to run when the document has no obj tag, logging that
the character was not saved, instead of crashing and taking every
other player's unsaved progress with it.
- The remaining obj-child lookups go through a helper that returns null
when the root is missing.
Component LoadFromXml/UpdateXml still assume a valid document; refusing
to load a player with invalid xml belongs in the world server's load
path, which is being converted separately.
Refs #1332
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
