Skip to content

P_StandardUpdate.md

Latest commit

 

History

History
126 lines (86 loc) · 8.95 KB

P_StandardUpdate.md

File metadata and controls

126 lines (86 loc) · 8.95 KB

P_StandardUpdate

Direction: Both (C→S movement input; S→C per-tick position broadcast) Numeric ID: 14 Server handler: ServerNet.bb:1761 Server broadcast emitter: GameServer.bb:1078-1112 Client handler: ClientNet.bb:1490

Purpose

The per-tick movement broadcast. The highest-network-volume packet in the protocol — every connected player sends one to the server per movement-tick, and the server broadcasts every visible actor's position to every nearby player. The "standard" in the name is "what every actor needs every tick": position, destination, locomotion state.

No sub-codes. A single fixed-shape packet per direction. The conditional fields appear on the S→C side based on actor properties (mount presence, is-you, Environment_Fly).

Field layout

C → S — "my movement input this tick"

The client sends its intended destination + current sampled position. The server validates with ClampWorldCoord and a speed-hack delta clamp.

Offset Width Type Field Notes
1 4 Float DestX# Where the client wants to be heading (movement input). Clamped via ClampWorldCoord#.
5 4 Float DestZ# Same.
9 4 Float NewY# Vertical sample. Clamped; written directly to AI\Y#.
13 4 Float NewX# Current X (the client's sampled position). Clamped; passes the speed-hack check before being committed to AI\X#.
17 4 Float NewZ# Current Z. Same.
21 1 Int IsRunning Locomotion state. Force-cleared when WalkingBackward = True (anti-cheat).
22 1 Int WalkingBackward Locomotion state.

Total packet length: 22 bytes after the sub-code byte (there is no sub-code byte; all 22 bytes are payload).

S → C — "actor X moved to position Y"

The server broadcasts this for each nearby actor on each broadcast tick (per GameServer.bb's UpdateActorInstances per-tick scheduler). Three tail-shape conditionals:

Offset Width Type Field Always present?
1 2 Int A2\RuntimeID The actor being updated. Yes.
3 4 Float A2\X# Authoritative X. Yes.
7 4 Float A2\Z# Authoritative Z. Yes.
11 1 Int A2\IsRunning Locomotion. Yes.
12 1 Int A2\WalkingBackward Locomotion. Yes.
13 4 Float A2\DestX# Server-side destination. Yes.
17 4 Float A2\DestZ# Same. Yes.
21 2 Int A2\Mount\RuntimeID (or 0) Mount RuntimeID, or 0 if no mount. Yes.
23 2 Int A2\Attributes\Value[EnergyStat] Only when A2 = AI (the receiving player is the actor being updated) AND EnergyStat > -1. Self-energy passback.
23 4 Float A2\Y# (or interpolated YPos#) Only when A2\Actor\Environment = Environment_Fly AND A2 <> AI. Flying-actor Y. For AI in patrol mode, server interpolates Y between OldY and waypoint Y.

The "23 or 23" overlap is intentional: the two conditional tails are mutually exclusive (you only get Energy on your own update; you only get Y on a remote flying actor's update). The client at ClientNet.bb:1538 reads offset 23 as a 4B float ONLY in the Environment_Fly branch; the self-Energy passback was commented out in the client (see ClientNet.bb:1545) so currently arrives but is ignored.

Validation requirements

Float clamp on every position field (C→S)

All five floats (DestX, DestZ, NewY, NewX, NewZ) pass through ClampWorldCoord#, which rejects NaN/Inf via the canonical comparison trick (see encoding.md). PR series #237-#239 unified the prior "Y-only filter, X/Z raw" inconsistency.

A single NaN in any of these would poison every receiving client's spatial code (collision, EntityDistance#, LOD culling, AI targeting).

Anti-cheat: speed-hack delta clamp on NewX / NewZ

ServerNet.bb:1796-1827. Per-packet position delta is bounded against the actor's Speed attribute scaled by elapsed milliseconds:

MaxUnitsPerMs# = 0.15 * (SpeedAttr# + 0.5)  ; matches GameServer's 1.5 base unit/tick
MaxDelta# = MaxUnitsPerMs# * Float#(ElapsedMs)
If MaxDelta# < 2.0 Then MaxDelta# = 2.0      ; floor for short-tick lag tolerance
If MoveDist# > MaxDelta# Then ...            ; hold at prior position

Without this, the client could teleport arbitrarily within ClampWorldCoord bounds between updates, defeating PvP positioning and movement-based triggers. The 2.0 floor + SpeedAttr + 0.5 scaling covers lag spikes, collision shove-back, and the 3× tolerance the server uses for AI movement.

The first update of a session (or after a >5s lag spike) accepts and re-baselines LastPosUpdateMs, so the clamp never fires on legitimate re-sync.

Anti-cheat: backward-running prevention

ServerNet.bb:1784 — If AI\WalkingBackward = True Then AI\IsRunning = False. Without this, the client could set both flags and the server would let the actor run backwards (animation+speed mismatch — invisible-frame teleport opportunity).

Rider-mount synchronisation

When AI\Mount <> Null, the mount's position is force-set to the rider's position after every update (ServerNet.bb:1832-1842). The mount is also a server-side ActorInstance; without this, the mount drifts from the rider on every tick.

AI\Rider = Null precondition

ServerNet.bb:1767 — only the rider can update; if AI is a mount with a rider, the rider's P_StandardUpdate will move both. Without this gate, a malformed packet could move the mount independently of its rider.

AI\IgnoreUpdate = 0 precondition

ServerNet.bb:1765 — set by P_ChangeArea and P_RepositionActor while the client is mid-warp. Prevents in-flight updates from clobbering the server's reposition while the client is still rendering it.

Server broadcast scheduler

GameServer.bb's UpdateActorInstances walks FirstOnlinePlayer (engine-wide chain, added PR #283) and for each online recipient walks AInstance\FirstInZone (per-area chain) to broadcast every visible actor's update. Three distance bands:

Band Condition Update?
Close ActorDistance# < UpdateDistance Always
Mid UpdateDistance ≤ ActorDistance# ≤ UpdateFarDistance Only if AlsoUpdateMiddleRange = 1 (timing-gated)
Far ActorDistance# > UpdateFarDistance Never

Mid-range updates skip every other tick to halve bandwidth while keeping the remote visible. AlsoUpdateMiddleRange flips at LastCompleteUpdate boundaries.

Historical bugs / PR references

PR Fixed
#237 – #239 Unified ClampWorldCoord across all five position floats (was Y-only previously)
#270 / #272 Underwater-damage check converted to per-Area chain (touches the same per-tick loop)
#277 Null-guard on stale ServerWater handle inside the per-tick loop (was a zero-sentinel phantom-damage bug, not a crash, due to BlitzForge's non-short-circuit And + zero-sentinel semantics)
#282 FindActorInstanceFromRNID(M\FromID) -- which P_StandardUpdate calls on every inbound packet -- went O(N) -> O(1) via the ActorByRNID index
#283 The server-side broadcast loop walks FirstOnlinePlayer instead of Each ActorInstance / If A\RNID > 0
Speed-hack delta clamp (date pre-#237) The MaxUnitsPerMs# calculation and 5s lag-spike re-baseline rule

Related packets

  • P_ChangeArea — sets AI\IgnoreUpdate = 1 while warp is in flight
  • P_RepositionActor — server-initiated reposition; same IgnoreUpdate gate
  • P_StatUpdate — Energy stat sometimes hitches a ride on the S→C tail but P_StatUpdate is the canonical Energy update channel
  • P_NewActor / P_ActorGone — area-entry / area-exit gating that determines whether a remote actor is in the recipient's per-tick broadcast set

See also