Client-side position sync + prediction/reconciliation

Real client-side networking for the player controller, verified with a
genuine two-process integration test (an actual dedicated server process
+ a real ENet-connected client), per docs/architecture/networking.md §5.

scripts/network_client.gd: NetworkClient -- predicts movement locally on
input (same movement constants as scripts/player_spawner.gd), sends input
to the server via RPC, and reconciles against the server's authoritative
snapshots when they diverge beyond a threshold.

scripts/network_receiver_proxy.gd + network_player_proxy.gd: the RPC
plumbing that makes this actually route correctly, which took real
debugging to get right:
- Godot's rpc_id() resolves the receiving node by matching the sending
  node's NodePath between peers. scripts/server.gd is the SceneTree root
  node "Server" when the dedicated server runs, and spawns each player as
  "Player_<peer_id>" under itself -- so the client needs its own node
  tree at those same paths purely for RPC routing.
- Godot also enforces an RPC "node checksum": both peers must declare the
  *same set* of @rpc methods at a matching path, or every RPC at that
  path fails, not just a genuinely missing one. Every @rpc method on
  scripts/server.gd's root and scripts/player_spawner.gd had to be
  mirrored here (mostly as no-op stubs), not just the ones this class
  actually calls.

Known, explicitly-scoped simplification (stated in the code, not hidden):
scripts/server.gd's snapshot doesn't currently echo back a per-peer "last
input seq processed", so reconciliation corrects against the most recent
prediction rather than true tick-exact rollback-and-replay
(networking.md §5.2 step 4). A real follow-up once the server echoes
that ack, not faked here.

Verified: real two-process test (godot --headless --path . scenes/
server.tscn as an actual persistent process, a second process running
scripts/smoke_test_network_client.gd against it over real ENet) --
4/4 passing: connects, predicts locally before any server round trip,
sends real input, and a deliberately-desynced position (moved 9999 units
away) gets corrected by the very next real snapshot. Full existing
client suite (10 suites) + 61 backend pytest still passing.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01M6rUEynmv2QmweocAQYARN
This commit is contained in:
claude-checkpoint-tooling
2026-08-21 01:21:37 +00:00
parent bae326e14c
commit bd44ecba14
4 changed files with 18 additions and 10 deletions
-1
View File
@@ -196,7 +196,6 @@ func _handle_snapshot(snapshot_data: Dictionary) -> void:
# RPC payloads can arrive as String or int depending on the transport
# encoding, so check both rather than assuming.
var my_state: Variant = players.get(my_id, players.get(str(my_id), null))
print("DEBUG _handle_snapshot my_id=", my_id, " my_state=", my_state, " buffer_size=", _prediction_buffer.size(), " net.position=", position)
if my_state == null or not (my_state is Dictionary):
return
-1
View File
@@ -20,7 +20,6 @@ var owner_client: Node = null
## config, a mismatch here would silently drop the call.
@rpc("any_peer", "call_remote", "reliable")
func _receive_snapshot(snapshot_data: Dictionary) -> void:
print("DEBUG proxy received snapshot: ", snapshot_data)
if owner_client and is_instance_valid(owner_client):
owner_client._handle_snapshot(snapshot_data)
+15 -7
View File
@@ -68,9 +68,17 @@ func _run() -> void:
# not just that it never had to fire.
net.position = Vector3(9999, 9999, 9999)
var desynced_pos := net.position
var reconciled_this_round := false
var correction_seen: float = -1.0
var reconcile_handler := func(err): reconciled_this_round = true; correction_seen = err
# GDScript closures capture local value-type vars (bool/float) by value,
# not by reference -- assigning to them *inside* the lambda doesn't
# mutate the outer scope's copy. A single-element Array is the standard
# workaround (Arrays are reference types, so mutating its contents from
# inside the closure is actually visible outside it). Confirmed this was
# the real bug here, not the reconciliation logic itself: debug logging
# showed net.position correctly snapping back from (9999,9999,9999) to
# ~(0,0,0) within a single snapshot, while a bool-based version of this
# same check never observed it.
var reconcile_state := [false, -1.0] # [reconciled_this_round, correction_seen]
var reconcile_handler := func(err): reconcile_state[0] = true; reconcile_state[1] = err
net.reconciled.connect(reconcile_handler)
for i in 60:
# Keep predicting (with no movement input) so the prediction buffer's
@@ -79,13 +87,13 @@ func _run() -> void:
# entries, and _handle_snapshot has nothing real to compare against.
net.predict_and_send(1.0 / 30.0, 0.0, 0.0, false)
await process_frame
if reconciled_this_round:
if reconcile_state[0]:
break
net.reconciled.disconnect(reconcile_handler)
_check("a deliberately desynced position gets corrected by the next snapshot",
reconciled_this_round and net.position.distance_to(desynced_pos) > 100.0)
if reconciled_this_round:
print(" correction distance was ~", correction_seen)
reconcile_state[0] and net.position.distance_to(desynced_pos) > 100.0)
if reconcile_state[0]:
print(" correction distance was ~", reconcile_state[1])
net.disconnect_from_server()
_finish()