feat(multiplayer): Phase 4 tasks 4.1/4.2 - local prediction history ring

Adds LocalPredictionHistory, a client-owned seq-tagged ring recording
predicted ship state per input sequence, plus wiring in
NetworkedMatch to record predictions on send and compare them against
authoritative snapshots on arrival. Ships stay frozen/interpolated
until 4.3 lands actual correction logic; this round only builds the
comparison machinery and its data.

Includes fixes from two review rounds: resync_required now
self-clears once acknowledgements catch back up (mirrors
InputJitterBuffer's stalled flag), NetBodyState gained a copy()
method to stop diagnostic accessors aliasing ring-owned state, and
corrected comments that had described the local ship as being
force-simulated pre-4.3 when it is still driven by interpolated
transform writes.
This commit is contained in:
Josh Creek
2026-08-20 19:29:13 +01:00
parent cf73074e27
commit 3d3024ae8a
4 changed files with 454 additions and 0 deletions
+173
View File
@@ -0,0 +1,173 @@
class_name LocalPredictionHistory
extends RefCounted
const NetBodyState = preload("res://scripts/net_body_state.gd")
# Client-owned local-ship prediction history (multiplayer-todo.md §4.3).
# This is deliberately independent of NetworkedMatch and the scene tree so
# sequence/ring behaviour can be tested from scripted traces. Each entry is
# tagged with its full sequence number: an old value in a wrapped slot is
# never accepted as a prediction for a newer sequence.
#
# Acknowledge and record are separate producer/consumer clocks. The input
# sender can continue producing while snapshots stop arriving, so record()
# explicitly marks overflow once more than RING_SIZE unacknowledged sequence
# positions exist. It still retains the newest representable window, but
# callers can see that an authoritative resync is required instead of
# mistaking a wrapped overwrite for a valid comparison.
#
# resync_required is a live condition, NOT a latch: compare_authoritative()
# clears it again once acknowledgements have genuinely caught back up (see
# that method). This mirrors input_jitter_buffer.gd's `stalled`, which
# likewise drops back to false the moment a normal tick is consumed again.
# A latched flag would mean one transient ~2s stall anywhere in a match
# permanently pinned every later tick into "needs a hard resync", which is
# exactly the behaviour soft correction exists to avoid — and it would also
# cap overflow_count at 1 forever, since a second episode could never
# observe the flag going false again.
#
# Two things a "matched" result does NOT guarantee, flagged for whoever
# builds task 4.3's actual correction logic on top of this:
#
# 1. A "matched" result can still be reporting stale data. The slot-tag
# equality check in get_prediction() guarantees a match's payload
# genuinely belongs to the queried seq (never wrong-seq data mislabeled
# as right), but nothing in the "matched" status itself says HOW OLD
# that entry is. Under sparse recording (record() is not called with
# strictly consecutive seqs — see the record() comment below), an entry
# from well over RING_SIZE ticks ago can still report "matched" for a
# query landing on its untouched residue. resync_required correctly
# stays true in that case (the span guard below is exact), but the
# comparison payload itself carries no matched_stale/age distinction. A
# caller wanting to reject "matched but ancient" needs to separately
# check newest_recorded_seq - seq itself.
#
# 2. record() can be called twice for the same seq with a DIFFERENT action,
# when input_lead_controller's release path resends a duplicated seq
# (delta == 0) — the later call silently overwrites the ring slot, so
# the stored action becomes whichever of the two calls happened last.
# This matches what the WIRE ends up sending for that seq (the resend
# replaces the redundancy history's front entry — see
# networked_match.gd's _send_local_input), but if the SERVER had
# already consumed the seq from the first packet before the resend
# arrived, the server's applied action and this ring's stored action for
# that same seq can disagree. Narrow (release only fires after 120 ticks
# of sustained surplus depth, when the server is least likely to be
# right on the edge of consuming that exact seq) but real; a future
# replay-based catch-up (task 4.5) built on this history should not
# assume the stored action is provably what the server actually applied.
const RING_SIZE := 128
var _ring_seq: PackedInt32Array = PackedInt32Array()
var _ring_entry: Array = []
var _has_recorded := false
var newest_recorded_seq := -1
var last_acknowledged_seq := 0
var overflow_count := 0
var resync_required := false
func _init() -> void:
_ring_seq.resize(RING_SIZE)
_ring_entry.resize(RING_SIZE)
for i in RING_SIZE:
_ring_seq[i] = -1
# Stores a private copy of both action and state. Returns true when this
# record crossed the unacknowledged-capacity boundary; the caller does not
# need that return today, but it makes the eviction event observable rather
# than silent when reconciliation starts applying corrections in Phase 4.3.
func record(seq: int, action: ShipAction, state: NetBodyState) -> bool:
var overflowed_now := false
if not _has_recorded or seq > newest_recorded_seq:
if seq - last_acknowledged_seq > RING_SIZE:
# Only the LEADING edge of an episode counts: resync_required is
# still true for every subsequent tick of the same stall, and
# counting those would report one outage as hundreds. Because
# compare_authoritative() can now clear the flag, a genuinely
# separate later episode does increment this again.
overflowed_now = not resync_required
resync_required = true
if overflowed_now:
overflow_count += 1
newest_recorded_seq = seq
_has_recorded = true
var idx := posmod(seq, RING_SIZE)
_ring_seq[idx] = seq
_ring_entry[idx] = {
"action": action.copy(),
"state": state.copy(),
}
return overflowed_now
# Returns independent copies so diagnostic/reconciliation consumers cannot
# mutate a retained prediction by accident.
func get_prediction(seq: int) -> Dictionary:
var idx := posmod(seq, RING_SIZE)
if _ring_seq[idx] != seq:
return {}
var entry: Dictionary = _ring_entry[idx]
return {
"seq": seq,
"action": (entry["action"] as ShipAction).copy(),
"state": (entry["state"] as NetBodyState).copy(),
}
# Produces comparison data only. Applying a snap, teleport, velocity delta,
# or visual offset belongs to later Phase 4 tasks.
func compare_authoritative(seq: int, authoritative: NetBodyState) -> Dictionary:
if seq > last_acknowledged_seq:
last_acknowledged_seq = seq
var prediction := get_prediction(seq)
if prediction.is_empty():
return {
"status": _missing_status(seq),
"seq": seq,
"authoritative_state": authoritative.copy(),
}
# A successful match is the only evidence that the acknowledgement clock
# has genuinely caught back up, so it is the only thing allowed to clear
# resync_required — a "missing_evicted"/"missing_not_recorded" result
# proves the opposite, and must leave the flag alone.
#
# The extra span check is not redundant. record() is not guaranteed to be
# called with consecutive sequences: input_lead_controller.update() can
# return 0 or up to 1+3, so the client's seq can skip forward, leaving a
# ring slot holding a tag OLDER than newest_recorded_seq - RING_SIZE
# (its residue was simply never rewritten). get_prediction() would still
# report that as "matched", so matching alone does not imply the
# outstanding window is back within capacity. Gate on the exact inverse
# of record()'s own trip inequality instead, which holds regardless of
# how sparsely sequences were recorded.
if newest_recorded_seq - last_acknowledged_seq <= RING_SIZE:
resync_required = false
var predicted_state: NetBodyState = prediction["state"]
var position_error := authoritative.position - predicted_state.position
var rotation_error_radians := predicted_state.rotation.angle_to(authoritative.rotation)
return {
"status": "matched",
"seq": seq,
"action": prediction["action"],
"predicted_state": predicted_state,
"authoritative_state": authoritative.copy(),
"position_error": position_error,
"position_error_magnitude": position_error.length(),
"rotation_error_radians": rotation_error_radians,
"rotation_error_degrees": rad_to_deg(rotation_error_radians),
"linear_velocity_error": authoritative.linear_velocity - predicted_state.linear_velocity,
"angular_velocity_error": authoritative.angular_velocity - predicted_state.angular_velocity,
}
func _missing_status(seq: int) -> String:
if _has_recorded and seq <= newest_recorded_seq - RING_SIZE:
return "missing_evicted"
return "missing_not_recorded"
+24
View File
@@ -21,3 +21,27 @@ var turbo := false
var thrust_z := 0.0 # -1..1; re-quantised to a 3-bit bin on the wire
var stalled := false
var avel_range := 4.0 # NetCodec.SHIP_AVEL_RANGE; set to BALL_AVEL_RANGE for the ball
# Self-referential preload, not get_script().new() — this file deliberately
# has no class_name (same cache-timing reason as test_case.gd and other
# path-`extends`d files in this project), and get_script().new() throws
# "Nonexistent function 'new' in base 'GDScript'" from within the script's
# own body in this Godot version.
const _NetBodyState = preload("res://scripts/net_body_state.gd")
# Same contract as ShipAction.copy() (see its own comment): a distinct
# instance with equal fields, for callers that hold onto a state past the
# tick/comparison it was returned in.
func copy() -> RefCounted:
var c := _NetBodyState.new()
c.position = position
c.rotation = rotation
c.linear_velocity = linear_velocity
c.angular_velocity = angular_velocity
c.frozen = frozen
c.turbo = turbo
c.thrust_z = thrust_z
c.stalled = stalled
c.avel_range = avel_range
return c
+96
View File
@@ -8,6 +8,9 @@ extends GameMode
# everything, including its own ship, from the interpolation buffer; there
# is no local prediction yet (that's Phase 4), so every body on the client
# is FREEZE_MODE_KINEMATIC and driven entirely by incoming snapshots.
# Tasks 4.1/4.2 add the seq-tagged recording and comparison plumbing that
# Phase 4 will need (LocalPredictionHistory below), but deliberately stop
# short of unfreezing or locally simulating anything — that is task 4.3.
#
# No HUD/Arena child in networked_match.tscn — both are built in code, once
# the arena is actually known (the server picks one; the client learns it
@@ -29,6 +32,7 @@ const NetBodyState = preload("res://scripts/net_body_state.gd")
const NetInterpolator = preload("res://scripts/net_interpolator.gd")
const InputJitterBuffer = preload("res://scripts/input_jitter_buffer.gd")
const InputLeadController = preload("res://scripts/input_lead_controller.gd")
const LocalPredictionHistory = preload("res://scripts/local_prediction_history.gd")
const HUD_SCENE = preload("res://scenes/HUD.tscn")
# Minimum plausible interpolation delay even on a same-machine/LAN link —
@@ -101,6 +105,20 @@ var _input_seq := 0 # client only
# 3-packet burst loss still recovers every tick's action via a later
# packet's history. Client only.
var _input_history: Array[ShipAction] = []
var _local_prediction_history := LocalPredictionHistory.new() # client only; 128-entry seq-tagged history (§4.3)
# Latest raw result from LocalPredictionHistory.compare_authoritative(). This
# pass records and compares only; Phase 4.3 will consume it to choose and
# apply the actual reconciliation correction.
#
# Read its error fields with the caveat documented on
# _local_ship_prediction_state(): until task 4.3 unfreezes and locally
# simulates the local ship, the "predicted" side of every comparison is an
# interpolated past-snapshot pose, not a forward simulation. The
# position_error / rotation_error_radians / *_velocity_error numbers
# therefore measure interpolation-vs-authoritative drift, and are NOT
# prediction error. Expect them to be small and largely uninformative, and
# do not calibrate any snap/blend threshold against them yet.
var _last_local_prediction_comparison: Dictionary = {}
var _last_received_snapshot_tick := 0 # client only: echoed back as ack_snapshot_tick
var _input_lead_controller := InputLeadController.new() # client only (§3.3)
var _last_known_input_buffer_depth := -1 # client only: -1 = no snapshot with this field yet
@@ -478,6 +496,12 @@ func _send_local_input() -> void:
# one tick of latency recovered).
var delta := _input_lead_controller.update(_last_known_input_buffer_depth)
_input_seq += delta
# Record this tick's (seq, action, local-ship state) triple. action is
# sampled exactly once above; record() makes its own copy for the
# longer-lived prediction history. See _local_ship_prediction_state() for
# what the "state" half does and does not currently mean.
if _my_slot != null and is_instance_valid(_my_slot.ship):
_local_prediction_history.record(_input_seq, action, _local_ship_prediction_state(_my_slot.ship, action))
# Redundancy (§3.1): carry the last MAX_REDUNDANCY ticks' actions,
# newest-first, so a burst of up to (MAX_REDUNDANCY - 1) consecutive
# packet losses still lets the server recover every dropped tick's
@@ -535,6 +559,15 @@ func _on_snapshot_received(decoded: Dictionary) -> void:
# own slot's server-side InputJitterBuffer.depth() at send time, which
# is exactly what the input_lead control loop (§3.3) needs.
_last_known_input_buffer_depth = decoded["input_buffer_depth"]
# Compare the server state for this client's own fixed slot against the
# entry tagged with the exact input sequence the server applied. Do not
# correct the body here yet: this result is intentionally inspection data
# for the later snap/blend pass, and (per
# _local_ship_prediction_state()) is not yet true prediction error.
if _my_slot != null:
var my_index := _slots.find(_my_slot)
if my_index >= 0 and my_index < bodies.size():
_last_local_prediction_comparison = _local_prediction_history.compare_authoritative(decoded["last_input_seq"], bodies[my_index])
_update_tick_bias(server_tick)
for i in _slots.size():
if i < bodies.size():
@@ -550,6 +583,69 @@ func _on_snapshot_received(decoded: Dictionary) -> void:
_ball_interpolator.add_sample(server_tick, ball_state, reset_gen)
# NOT a prediction yet, despite the name — the name is for task 4.3, which
# is what will make it true. Pre-4.3 EVERY ship on the client, including this
# client's own, is freeze = true / FREEZE_MODE_KINEMATIC (see _apply_match_config,
# which sets that uniformly with no exception for _my_slot) and is moved only
# by NetInterpolator transform writes derived from ALREADY-RECEIVED, past
# server snapshots. Nothing locally simulates the local ship, and nothing ever
# writes linear_velocity/angular_velocity onto it.
#
# So what this samples is "wherever the interpolator had smoothed the ship to
# at packet-send time", NOT "where the action sampled this tick will put the
# ship". The consequences for anyone reading the comparison output:
# - linear_velocity/angular_velocity here are NOT zero — a first pass at
# this comment claimed they were, but FREEZE_MODE_KINEMATIC derives a
# body's velocity from its own consecutive transform writes, so these
# fields genuinely reflect the interpolator's implied motion (confirmed
# live: non-zero, direction-correct velocities while driving). What they
# are NOT is the result of locally simulating the sampled action's
# thrust/rotation through the ship's own force formulas.
# - the resulting position_error / rotation_error_radians measure how far
# an interpolated PAST pose (and its implied velocity) sits from the
# later-arriving authoritative pose for that sequence. That is
# interpolation lag, not prediction error, and on a clean link it will
# read small and largely uninformative.
# - do not calibrate a snap-vs-blend threshold, or benchmark "prediction
# quality", against these numbers.
# They only become genuine prediction error once task 4.3's net_ship_predictor.gd
# unfreezes the local ship and steps it forward locally (multiplayer-todo.md
# §4 / §7 tasks 4.3 and 4.5). The recording/matching plumbing is landed first,
# on purpose, so 4.3 has a tested ring to build on.
func _local_ship_prediction_state(ship: Ship, action: ShipAction) -> NetBodyState:
var state := NetBodyState.new()
state.position = ship.global_position
state.rotation = ship.global_transform.basis.get_rotation_quaternion()
state.linear_velocity = ship.linear_velocity
state.angular_velocity = ship.angular_velocity
state.frozen = ship.freeze
state.turbo = action.turbo
state.thrust_z = action.thrust.z
state.avel_range = NetCodec.SHIP_AVEL_RANGE
return state
# Diagnostic accessor. Same caveat as _local_ship_prediction_state(): the
# error fields are interpolation-vs-authoritative drift, not prediction error,
# until task 4.3 lands.
#
# Dictionary.duplicate(true) recurses into Arrays/Dictionaries but copies
# Objects (RefCounted included) BY REFERENCE — an adversarial review caught
# that this returned a dict sharing its "action"/"predicted_state"/
# "authoritative_state" ShipAction/NetBodyState instances with the stored
# comparison, so a caller writing through the "copy" silently rewrote
# history. ShipAction.copy() and NetBodyState.copy() exist precisely so
# callers holding onto one past its own tick copy it (see ship_action.gd's
# own comment) — this accessor has to honor that contract itself, not just
# assume duplicate(true) does.
func get_last_local_prediction_comparison() -> Dictionary:
var result := _last_local_prediction_comparison.duplicate(true)
for key in ["action", "predicted_state", "authoritative_state"]:
if result.has(key):
result[key] = result[key].copy()
return result
# See the class-level comment above _tick_bias_samples for why this exists.
# bias_ms is how much further ahead to_tick(server_time_est) lands than the
# server_tick this snapshot actually carries — mostly the server's own