# CharacterMovementComponent: the "received this frame" test in ServerMoveHandleClientError compares a float with a double, so later moves of a frame skip the position check

**URL:** <https://forums.unrealengine.com/t/charactermovementcomponent-the-received-this-frame-test-in-servermovehandleclienterror-compares-a-float-with-a-double-so-later-moves-of-a-frame-skip-the-position-check/2843083>\
**Category:** Issues and Bug Reporting (Unreal Engine)\
**Tags:** unreal-engine\
**Created:** [October 20, 2026, 10:01pm UTC](https://forums.unrealengine.com/t/charactermovementcomponent-the-received-this-frame-test-in-servermovehandleclienterror-compares-a-float-with-a-double-so-later-moves-of-a-frame-skip-the-position-check/2843083 "2026-10-20T22:01:12Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Scandalis](https://d3kjluh73b9h9o.cloudfront.net/original/4X/2/8/5/28502927f1d97ba1051b3b47ccfe17233bf41ef0.png) [@Scandalis](https://forums.unrealengine.com/u/Scandalis)\
**Post date:** [October 20, 2026, 10:01pm UTC](https://forums.unrealengine.com/t/charactermovementcomponent-the-received-this-frame-test-in-servermovehandleclienterror-compares-a-float-with-a-double-so-later-moves-of-a-frame-skip-the-position-check/2843083/1 "2026-10-20T22:01:12Z")

</div>

### Summary

In UCharacterMovementComponent::ServerMoveHandleClientError, the test meant to let more recent moves received in the same server frame still be checked compares FNetworkPredictionData\_Server\_Character::LastUpdateTime (a float) with UWorld::TimeSeconds (a double). The float is almost never equal to the double, even in the same frame. So after a correction is decided in a server frame, every later move the server processes in that frame returns early through WithinUpdateDelayBounds and is never position-checked.

### What Type of Bug are you experiencing?

Networking

### Steps to Reproduce

1. Open Engine/Source/Runtime/Engine/Private/Components/CharacterMovementComponent.cpp at UCharacterMovementComponent::ServerMoveHandleClientError and find the “received this frame” test (line 10204 in 5.8.3): if( (ServerData-\>LastUpdateTime != GetWorld()-\>TimeSeconds))
2. Note the types: FNetworkPredictionData\_Server\_Character::LastUpdateTime is a float (CharacterMovementComponent.h:3295), and UWorld::TimeSeconds is a double (World.h:2010). LastUpdateTime is set from GetWorld()-\>TimeSeconds when a correction is decided (about line 10460).
3. Run a dedicated server with a networked Character, and have a client whose moves disagree with the server while several of its moves are processed in the same server frame (for example, a client sending moves faster than the server ticks).
4. Add a log or breakpoint in ServerExceedsAllowablePositionError, and log the time since LastUpdateTime for each move that reaches WithinUpdateDelayBounds.

### Expected Result

As the comment above the test says, more recent moves received in the same frame as a correction decision are still checked: the test sees the same frame, skips the early return, and ServerExceedsAllowablePositionError runs for those moves.

### Observed Result

The float is almost never equal to the double, so the same-frame branch effectively never applies. Every later move processed in that frame goes into AGameNetworkManager::WithinUpdateDelayBounds, where the time since the update is about 0 s, under CLIENTADJUSTUPDATECOST / CurrentNetSpeed, so the function returns before the position check. Those moves are never position-checked, and the pending adjustment keeps naming the frame’s first disagreeing move.

How we saw it: a server-side safeguard of ours runs inside ServerExceedsAllowablePositionError. It fired only when the move it needed was the first in its server frame: in our logs, once in 514 correction holds on one run, never in 676 holds on later runs. Logging the time since the engine’s last adjustment decision showed about 0 ms for the skipped moves.

### Affects Versions

5.8

### Platform(s)

Linux  
Windows

### Additional Notes

Engine: UE 5.8.3 (source build); the same code appears in earlier 5.x versions. We saw it with a Linux dedicated server and Windows clients; the code itself is platform-independent.

The three lines that show the comparison:

```cpp
float LastUpdateTime; // CharacterMovementComponent.h:3295 (FNetworkPredictionData_Server_Character)
double TimeSeconds; // World.h:2010 (UWorld)
if( (ServerData->LastUpdateTime != GetWorld()->TimeSeconds)) // CharacterMovementComponent.cpp:10204

```

A fix: compare like with like, e.g. `if (ServerData->LastUpdateTime != static_cast<float>(GetWorld()->TimeSeconds))`. Or store `LastUpdateTime` as a double (it is serialized and used in `WithinUpdateDelayBounds`, which also takes a float). We work around it in our game’s override of `ServerMoveHandleClientError` by tracking the frame of the last decision ourselves.
