multiplayer

[copper raft]

inside my car blueprint i use only run on server event, inside my player character blueprint i use that repnotify isinvehicle

[nocturne quail]

the point is you need to disable collisions/physics before you attach to vehicle

[copper raft]

i already have these nodes

before attaching

i tried with Set Collision Enabled node but nothing changed

[nocturne quail]

these are on server

[copper raft]

yes

inside repnotify too

[nocturne quail]

first set the bool to true so it will call the onrep before you attach to disable collisions/physics , and than attach

[copper raft]

changed it, nothing fixed🥲

[nocturne quail]

remove these from server call

move them to onrep

onrep will called on server/clients both

[copper raft]

removed them, still nothing

[nocturne quail]

change attach rules, target, world, relative

[copper raft]

weld simulated bodies enabled?

[nocturne quail]

you have physics disabled on character, this should not matter

[copper raft]

still nothing fixed, i can't understand why

actually when rotating wheels left or right in movement, the car is still laggy for driver too

do i need to use attach actor to actor maybe? im using to component right now

no, still nothing with actor to actor too

if i run the game with a server client too, on first client and on server all is ok, but on second client have this problem, idk why🤷

on client 1 screen, the character is moving his location and rotation when the car is moving and turning around

[normal fulcrum]

when do you attach? is it multicast event?

[copper raft]

no, i don't use multicast, i have only run on server event inside car blueprint, and a repnotify inside player blueprint

this is what remained inside run on server event, disabling collision and movement i switched full insie onrep

this is inside player blueprint

[normal fulcrum]

Answering your question "how to debug", I would let there be only Attach node on server, and see if it works (without other Posses, replicate, notify etc)
try to attach only and see if it works

[copper raft]

with possess node is working too, the possess node i use only for driver, not for passengers, but when i try to drive my car, my passengers are changing his location and rotation depend of speed of the car

and this problem i can see only from passengers screen, on driver screen i see all passengers are in their right position

[nocturne quail]

disable camera rotation for the character which is rotating

[copper raft]

is rotating the player location and rotation, not camera

when the car is moving

[nocturne quail]

car is moving and the character camera is overriding the vehicle rotation for the character

in result your character don't follow the vehicle rotation

[copper raft]

but why this is happeing only on passenger screen?, on driver screen i see all passengers don't have these problems

[nocturne quail]

because passenger is not posessing vehicle, it is using its own movement stuff

[copper raft]

but i need to rotate the camera around when im a passenger too🤷 😅

how can i look around when im passenger if i will disable camera rotation?

is that correct if when i enter the car, my capsule collision is remaining in that location when i entered the car?

[nocturne quail]

in this case disable usepawncontroll rotation or something like that which effect character

[copper raft]

i don't know if this is true what i made but i think i fixed my problem with set component tick enabled node

[copper raft]

but this created another problems, when trying to enter exit, sometimes not working🥲

[nocturne quail]

just disable these, when you enter vehicle and enable when you exit vehicle

    CharacterMovement->bOrientRotationToMovement = false;
    CharacterMovement->bUseControllerDesiredRotation = false;
    Target->bUseControllerRotationYaw = false;
    Target->bUseControllerRotationPitch = false;

[copper raft]

i tried this, for now i don't see any problems anymore, but i don't know in future😅

[nocturne quail]

if it works it works , good luck

[copper raft]

thanks

@nocturne quail what about suspension is not working smooth when vehicle is moving, do you know how to fix this?

[nocturne quail]

wheels suspension?

[copper raft]

this problem is only when running on Play as Client, when As Listen Server all is ok

[copper raft]

yes

[nocturne quail]

I never encounter this kind of issue

you can record a small clip to show the issue

[copper raft]

https://imgur.com/a/mzrPdiF

[rancid marten]

anyone here know steam sessions? (I have my game hosting and joining on Lan. And I believe I have it hosting proplerly but connecting on steam is not working.) Love to bend your ear or show you a video>

[nocturne quail]

I think this is because of the vehicle physics body or bad wheel configurations

[copper raft]

but why this is happening only on play as client? when play as server or standalone this is not happening

[nocturne quail]

hmm, no idea than

[copper raft]

🥲

[normal fulcrum]

There are some people knowing steam sessions.
You connect from same pc yo the same pc via steam?

[rancid marten]

no two seperate pcs

[verbal ice]

Question for Iris users, can you make a custom UDataStream class?

My usecase would be sending scripts from the server to the client that it needs to execute before doing anything else

My full plan is to stop replication of anything to players until they receive and execute those scripts, which are required for gameplay purposes

This would happen after they connect, and I figured that data stream class sounds useful for this

[grizzled garnet]

I have a level with an actor which loads a sublevel. Currently, clients join the level and the sublevel gets loaded as expected, however upon making the level visible it looks like based on the logs, trying to make the sublevel visible is causing the client to disconnect under MissingLevelPackage?

LogPlayerController: Warning: ServerUpdateLevelVisibility() ignored non-existant package. PackageName='/Temp/Game/maps/gameMaps/baseSubmaps/LaserMaze/LaserMaze_HalfGeometry_LevelInstance_4', FileName='/Game/maps/gameMaps/baseSubmaps/LaserMaze/UEDPIE_0_LaserMaze_HalfGeometry'
LogNet: UNetConnection::Close: [UNetConnection] RemoteAddr: 127.0.0.1:52037, Name: IpConnection_3, Driver: Name:GameNetDriver Def:GameNetDriver IpNetDriver_2, IsServer: YES, PC: UserPlayerController_1, Owner: UserPlayerController_1, UniqueId: NULL:AndyLaptopV4-6EC795694A30C4DFDA34FA8ED45D974B, Channels: 171, Time: 2026.01.20-00.40.07
LogNet: UNetConnection::SendCloseReason:
LogNet:  - Result=MissingLevelPackage, ErrorContext="/Temp/Game/maps/gameMaps/baseSubmaps/LaserMaze/LaserMaze_HalfGeometry_LevelInstance_4,/Game/maps/gameMaps/baseSubmaps/LaserMaze/UEDPIE_0_LaserMaze_HalfGeometry"```

Server side the sublevel is already visible, and client side I can see the sublevel loading in. The client appears to get disconnected after the sublevel on their end has finished loading

Update, I think my issue might be I'm calling the level load/visible bp node on client by accident actually it has to be replicated???

[grizzled garnet]

the issue does seem to be with makign the level visible on clients, or perhaps the level on the server isn't being replicated to clients

I can set to LevelStreaming.DefaultAllowClientUseMakingVisibleTransactionRequests to true which I think would fix the issue but I think a proper solution would be to have the server correctly replicate the streamed level to clients rather than them loading the level manually

actually setting this cvar causes the sublevel to not become visible

[lament flax]

To make level instance working in MP i had to work with unique name overrides

More specifically spawning at runtime new level instances

https://notes.hzfishy.fr/Unreal-Engine/Networking/Networked-Streamed-levels

[grizzled garnet]

What do you mean by this? Set a unique name in OnConstruction so that the streaming level instance is different than on the server? (btw the actor loading the sublevel is in bp rn but shouldn't be difficult to move to cpp)

[lament flax]

I cant check code right now but im talking about the struct/class param when you spawn a dynamic level actor

[grizzled garnet]

Isn't expected functionality of the player state supposed to be that it's restored if a player disconnects and reconnects? In my testing that is not the behavior, as the replaced built in score value at least got reset to zero Actually I think I have code which does this, as the game supports switching teams whihc should reset hte player score to zero

[fiery wadi]

Anyone know why 3 characters would spawn in the level but only 1 controller is spawned please? I have no idea where to even start debugging this as I thought the player controllers were automatically created upon creation of the characters ?? Oddly enough each character can be controlled by the 1 controller (IE... They each jump individually when i press the jump key), Any help is appreciated, Thanks!

As a further addon info here is the print string which pops up from the Top Down Controller. As you can see it doesnt print out Client 1 : and Client 2 it still shows "Server"

[lament flax]

found back my code

void AERGenerationManager::OnReplicatedLevelParamsReplicated()
{
    FERRequestLoadAsyncLevelParams Params = FERRequestLoadAsyncLevelParams();
    Params.Id = ReplicatedLevelParams.Id;
    Params.LevelLocation = ReplicatedLevelParams.LevelLocation;
    Params.LevelRotation = ReplicatedLevelParams.LevelRotation;
    Params.LevelScale = ReplicatedLevelParams.LevelScale;
    LoadAsyncLevel(ReplicatedLevelParams.LoadedLevel, Params);
}

void AERGenerationManager::LoadAsyncLevel(TSoftObjectPtr<UWorld> LoadedLevel, FERRequestLoadAsyncLevelParams& Params)
{
    //if (!HasAuthority()) { return; }
    
    FTransform Transform = FTransform::Identity;
    Transform.SetLocation(Params.LevelLocation);
    Transform.SetRotation(Params.LevelRotation.Quaternion());
    Transform.SetScale3D(Params.LevelScale);
    
    ULevelStreamingDynamic::FLoadLevelInstanceParams LoadLevelInstanceParams(GetWorld(), LoadedLevel.GetLongPackageName(), Transform);
    LoadLevelInstanceParams.bInitiallyVisible = true;

    const FString UniqueName = LoadedLevel.GetAssetName() + "_" + FString::FromInt(Params.Id);
    LoadLevelInstanceParams.OptionalLevelNameOverride = &UniqueName;

    bool bSucess = false;
    ULevelStreamingDynamic* StreamedLevel = ULevelStreamingDynamic::LoadLevelInstance(LoadLevelInstanceParams, bSucess);
    StreamedLevel->SetShouldBeVisible(true);
    
    ER_GENDVLOG("Requested level: §% spawn in progress (Success: §%)", *LoadedLevel.GetAssetName(), bSucess);
}

[meager spade]

is there a way to boost the net tick for start of game

its taking a while to send all replicated actor cause its kinda throttling, but at start of the game, we dont need it to throttle

[nocturne quail]

an external actor exists in world plays Particle if it has no authority and set a timer to deactivate it after 1.5sec delay, if a client leaves before it gets Deactivated, will it crash?

[fiery wadi]

Just replying to my own message as I found out that this option exists so you can see whats on the server world in the world outliner you can switch what world is being displayed in it by choose the "Choose World" option and then seeing whats active in each world. I am still having the issue with the clients not displaying any print strings.

[nocturne quail]

the timer will run for the client that just leaves, and will reach to deactivate the particle for the client that don't exists any more

FTimerHandle DeactivateTimer;
GetWorld()->GetTimerManager().SetTimer(DeactivateTimer, [this]()
    {
        Particle->Deactivate();
    }, 1.5f, false);

[nocturne quail]

if you spawn 3 charactes, only those will spawn controller if they are actually possessed

in your case you are the only one which is possessed

and if you start the game with 3 clients, this will spawn 3 controllers casue each of them is possessed

[fiery wadi]

When i view the Server world in the outliner i see all 3 cointrollers

[nocturne quail]

controllers are only available for local client and server

you can't see other people controller

[fiery wadi]

Wouldnt the server be able to see all of them ?

[nocturne quail]

yes

server has all of them

[fiery wadi]

Yup which i can see in the pic.

Im not understanding why this is printing out on the clients though

When i press jump on each client they jump but the message prints which it should ONLY print if the Input System is not valid i would guess.

[nocturne quail]

put a delay test

its possible the system is not yet init fully when you try to access it

beginplay only calls once when game starts

[fiery wadi]

You can see it here what i mean where should i add a delay before the connection?

[nocturne quail]

your jump key calls at current time where your inputs are valid

[fiery wadi]

ahh ok.

Multiplayer is so hard to understand XD

[nocturne quail]

it has nothing to do with multiplayer, its a basic call cycle in unreal framework

beginplay calls only once at the time when you press play button

[fiery wadi]

I added a delay after Begin Play to give it time but still get same output

[nocturne quail]

set it to 1sec

[fiery wadi]

nope still same output

[nocturne quail]

show your jump key event

I think you are not using eis

[nocturne quail]

hmm, than I have no idea why its not valid

[fiery wadi]

Ok when i press on each of the Server, Client 1 and Client 2 they all report their jumps respectively in the print string

Client 1 : Jump , Client 2 Jump and Server Jump

This has had me stumped for a while now because im not sure why it prints not valid but then the jump action works.

[nocturne quail]

jump is replicated by default

[fiery wadi]

but if the input mapping wasnt valid it wouldnt know the button was pressed woudl it ?

[nocturne quail]

try to override the OnPossess event and print there

[fiery wadi]

this is the onposses.

[nocturne quail]

print input in onpossess

you are just printing character name

[dark parcel]

What object is this?

[fiery wadi]

Ok i tried the input system on possess and i added this as well

[dark parcel]

Thats not correct

On possess only get called on server.

[fiery wadi]

Its the top down controller.

[fiery wadi]

So i should do this in the Game Mode ?

[dark parcel]

Nope, game mode only exist on server.

Your binding must happend on target machine. Which are clients and listen server if you suppott listen server.

I bind my keys on OnAcknoeledgePossession but thats not blueprint exposed.

[dark parcel]

You probably end up with 3 widgets on server with this code.

Server has everyone controller.

If you play the game with 3 players, there will be 3 controller. You end up spawning the widget 3 times on listen server.

[fiery wadi]

If the server has everyones controller why does it not have the input system connected?

[dark parcel]

Input system is local

[fiery wadi]

Ok that makes sense.

[dark parcel]

Well i mean everyone is running their own instance and code.

[fiery wadi]

So that would explain why I get Input Not OK because the input system is local to the local machine and it deals with it.

[dark parcel]

Not sure with how you get input not ok and in what context.

There should be an event you can use to bind keys

[fiery wadi]

I would guess the server is spawning 3 clients (characters) but it cannot access the Input System local to each player as it only has access to it,s own.

Which seems to make sense.

But it has a socket connection to the controller to each pc for communication right?

[dark parcel]

Communication between server client goes through actor chanel.

Which utilize RPC and replicated variable. Thats about it.

[fiery wadi]

Im dumb.... The node says Enhacned LOCAL input system...

I really appreciate the help 🙂

Its bit of a minefield trying to work out multiplayer but i think once you know how it works and how to do things it,ll become a lot eaier.

[dark parcel]

Tldr most of thr time begin play isnt the right place to initialize stuff ib multiplayer.

Mostly to do with race condition.

[fiery wadi]

I did read that too as begin play can fire different times for different players

If thats true or not i dont know

[dark parcel]

Well ofc. They just fire when ever the actor spawn to the world.

And you need to understsnd everyone has their own instance (world)

Each player run their own code.

[fiery wadi]

true and internet connections are not consistent and some low end pcs/connections will take more time to connect to trigger begin play ?

[dark parcel]

When server spawn replicated actor, it will tell clients to make copies of that actor.

When that actor spawn on player 1 world.

Plsyer 1 computer will call begin play.

When that actor spawn on player 2 world. Player 2 computer will call begin play.

When you want to do some action only for the character that you control. Dont forget to use IsLocallyControlled.

[fiery wadi]

So IsLocallyControlled will only perform that action on the local machien ?

[dark parcel]

Like you dont want player 4 actor spawning in your world. Calling begin play and bind the keys.

You only want to bind the key for the character you possesed.

[fiery wadi]

that makes sense,. Thank you

Would you happen to know about ChoosePlayerStart I have the default pawn set up in the game mode and am trying to use this function to determine where each player starts as you can see i remove a start position location from the array but Im finding that sometimes the players are trying to spawn on the same location and i coouldnt see a way to add a tag like "Occupied" to it to prevent a player spawning on it.

[dark parcel]

You can make this much simpler. Just scramble the start position.

I think the node is called shuffle.

Then you just go from start to end.

[fiery wadi]

I have in the Game Mode this to get all the positions (I figured Begin Play in the Game Mode would mean that the server's Begin Play would always trigger before any connected clients as it owns the world essentially) but i could be wrong in that thinking

[dark parcel]

Cache it and shuffle it.

And you dont need to call random (and you still have the issue of picking the same number)

[fiery wadi]

Thats why i thought if the server deleted one available spawn option when its selected by a player then by process of elimination there would be less and less options remaining in the server?

[dark parcel]

How do you want to go about this? Do you want to guarantee that the player dont spawn in the same place?

[fiery wadi]

yes

[dark parcel]

Then i would just randomize the array. Go from start, then increment the index per spawn.

When reach the end, just get all the plsyer start and randomize again.

This way no one will spawn at the same place unless all the spawn location has been used.

[fiery wadi]

Ok so use a local var SpawnIndexCounter and everytime it selects a spawn point then increment this by one until it equal to the length of the player start rght?

I thought removing an option from the array would work in a similar way but maybe it doesnt.

[dark parcel]

Yeah but not local variable. They are function scoped.

Just make one in the game mode.

[dark parcel]

Yes you can do it this way too.

Grab an array -> shuffle -> get index 0, remove.

Repeat.

When length is 0 just repopulate the array.

Actually your idea is better.

[fiery wadi]

so this will return an empty start if the counter is equal to the length of the array else it,ll increment the counter by 1 each time and use that to select an element ie...0,1,2

ok that looks like its working 🙂

not sure why the other way did sometimes fail.

[dark parcel]

@fiery wadi actually your idea is better. No need with index and check. Just shuffle the array then remove as you spawn.

[fiery wadi]

I think my way might have been failing sometimes due to timings and the player was connceted whilst there was still 3 options in the array

thank you so much for the help !

i,ve been stuck on that for about 2 days xD

[copper raft]

heyy guys, how to get vehicle RPM node? my blueprint class is Wheeled Vehicle Pawn, the component is default auto created Vehicle Movement Component, but to get the RPM it asks to change component to ChaosWheeledVehicleMovement, i tried but i cant remove the old component

i just added this chaos movement component but can't remove vehicle movement component

i tried to change blueprint class but can't found this chaos vehicle

the plugin is enabled

[fallen fossil]

that is not multiplayer question, post whole screen in #ue5-general , I guess its inherited from class so it can not be removed

[copper raft]

@fallen fossildo you know how to change?

[copper raft]

i already tried

there is no any chaos wheeled vehicle class

[fallen fossil]

no I don't, unless its inherited u can't remove it. and I never used chaos plugin
maybe ? #chaos-physics

[copper raft]

i fixed just by adding a cast to chaos wheeled vehicle from simple vehicle movement component😅

[fallen fossil]

well, it can fail xD

[nocturne quail]

add these to your vehicle class

for cpp and optionally for blueprints

FORCEINLINE const TObjectPtr<UChaosWheeledVehicleMovementComponent>& GetChaosVehicleMovement() const { return ChaosVehicleMovement; }
UFUNCTION(BlueprintCallable) FORCEINLINE UChaosWheeledVehicleMovementComponent* GetChaosVehicleMovementBlueprints() const {return ChaosVehicleMovement.Get();}

[copper raft]

why?

[copper raft]

ok will try thanks🙂

[nocturne quail]

also should be in your

UCLASS()
class AMyVehicleBase : public AWheeledVehiclePawn

advanced vehicle template also use AWheeledVehiclePawn

its a good start to make vehicle system based on this class

[fallen fossil]

It depends on Class,

Casting only works if Class inherits from Parent class, then it can be casted, otherwise it fails.
There is also concept of casting pointers Parent pointers to child classes.
but if classes do not share stuff like A->B, then it will fail ;d

but again, I never used chaos car, so I can't tell class hierarchy

[copper raft]

i don't know i made a cast and all works fine without any fails

[fallen fossil]

then I guess its parent is that class somewhere ;d

[copper raft]

maybe😅

i don't know what changes epic games made on UE 5.7🤷 😅

[verbal ice]

bump

[silent valley]

I don't know about Iris, but we used a derived AOnlineBeaconClient to send scripts etc when clients joined, outside of the usual actor replication system

[verbal ice]

That would also work but I'd rather use data streams if it's possible

Since AFAIK they run outside of the usual channels and don't affect regular replication bandwidth

(might be wrong)

I also see UChunkedDataStream which would be awesome for sending files

[exotic wasp]

The name of a level must be consistent between client and server. Usually if it's a level baked into your game the name is already unique and there's no issue

If you're dynamically placing levels at runtime then you need to load the level on both client and server and ensure the name is unique and matching

The level an actor is part of is used to describe its "name" more or less

[exotic wasp]

They have the minimalapi specifier

[verbal ice]

Can't derive from them?

[exotic wasp]

I believe so, yeah

[verbal ice]

oof

well that's annoying

[exotic wasp]

Wait sorry, the virtuals seem to be manually marked for export

Some of them anyway

[haughty ingot]

Snipe the vtable, be a man

[harsh lake]

is there a more complete documentation on unreal net driver config ?

i found several config marked properties that still commented with @todo documentation.

[daring narwhal]

hey! Iam trying to solve a puzzle which way is the best to keep save data for dedicated server for loading a players profile. I did a system that check player id and after login to a server it checks it have a profile. The thing is not always showing widget at first join .Which drives me cause it happens with the second login - widget appear and creating a profile happens. So the thing is how i can push that to be always reliable or make a class structure to have that perfect in slot. Tried different things and code is not the problem i think timing is crucial but cant get place that stops me to fix this issue It must be a message from the server to client to check if he have valid profile to his ID.
So i think i should change structure a bit to have things more controllable. Someone maybe have a graph where the logic of identification should be? Now i have gamemode + player controller and rpc that handle that + save object that data is pushed when certain conditions are met.
So the player state and game instance are now i focus but would like to ask how to do that better (dedicated server)

last piece left to me and iam really annoyed why i cant push condition to first join somehow engine wont react. LOL

[silent valley]

We have the main logic in GameMode::Login. Use the PlayerUID from the PlayerController to lookup the relevant savedata binary object (if it exists).

Serialize the savedata into the PC, and allow replication to handle the transfer to the client as necessary.

[daring narwhal]

Thank You very much for that tip iam not good since iam c++ starter but that actually helped me to fix the issue All the best!

[nocturne quail]

it logs TestActor is invalid on clients!
How to deal with such condition without using delay?

bTest = false;
OnRep_Test();
TestActor = nullptr;
    
void ATestClass::OnRep_Test()
{
    FString RoleString = HasAuthority() ? "server" : "client";
    
    if (TestActor )
    {
        UE_LOG(LogTemp, Warning, TEXT("Valid on: %s"), *RoleString);
    }
    else
    {
        UE_LOG(LogTemp, Warning, TEXT("In Valid on: %s"), *RoleString);
    }
}

TestActor is also a replicated variable

[fallen fossil]

Very high risk and Safety convernes , sending executables between clients is uuh :d

[fallen fossil]

what do you mean without delay?
OnRep shoudl already have valid replicated variable 🤔

[nocturne quail]

we are talking about TestActor is null cause its sets to null directly, if I set it to null with a 0.01sec delay its fine

[fallen fossil]

Im confused, is OnRep function assigned to replication ?

Where you put delay? :d
How you replicate?

[dark edge]

What do these scripts do?

[nocturne quail]

standard onrep terminology

[fallen fossil]

is TestActor object replicated ?

[nocturne quail]

TestActor varialbe is replicated and the class it belongs is also replicated
ATestClass is also a replicated class

TestActor belongs to ATestActor

[dark parcel]

You never set the value of test actor though other than making it null?

[nocturne quail]

it was set and on server it is valid

[dark parcel]

Where?

[nocturne quail]

ATestClass by server

[dark parcel]

We are talking about ATestActor right? I only see it set as null.

[nocturne quail]

yeah setting it to something is not related

[dark parcel]

If server set ATestActor to something.

Aint no way the client on rep gets null.

[dark edge]

Show the actual code setting it.

[dark parcel]

Also setting it to null probably triggered OnRep on client where ATestActor is null.

[nocturne quail]

will post the whole flow from setting it to an object and than to null in a minute

[nocturne quail]

BeginPlay()
TestInteractionComponent->OnInteract.AddDynamic(this, &ATestClass::OnInteract);

void ATestClass::OnInteract(ATestActor* NewTestActor)
{
    if (!IsValid(Character) || !HasAuthority())
        return;
        
    if (!bTest)
    {
        SetbTestTrue(NewTestActor);
    }
    else
    {
        SetbTestFalse();
    }
}

void ATestClass::SetbTestTrue(ATestActor* NewTestActor)
{
    if (!bTest)
    {
        TestActor = NewTestActor;
        bTest = true;
        OnRep_Test();
    }
}

void ATestClass::SetbTestFalse()
{
    if (bTest)
    {
        bTest = false;
        OnRep_Test();
        TestActor = nullptr;
    }
}
    
void ATestClass::OnRep_Test()
{
    FString RoleString = HasAuthority() ? "server" : "client";
    
    if (TestActor )
    {
        UE_LOG(LogTemp, Warning, TEXT("Valid on: %s"), *RoleString);
    }
    else
    {
        UE_LOG(LogTemp, Warning, TEXT("In Valid on: %s"), *RoleString);
    }
}

[fallen fossil]

show declaration of
OnRep_Test and TestActor

OnRep usually was used by rpc system, not manual call like yours, so there is some confussion

[nocturne quail]

everything is fine with declarations

[nocturne quail]

only on listen server

[fallen fossil]

Well, my understanidng is u used OnRep like RPC for server

[dark edge]

you need to manually call it on server in C++

[fallen fossil]

? UPROPERTY(Replicated, ReplicateUsing=OnRepFunc)

i was pretty sure, this is the convention among unrel devs, thats why I just want to peek declaration, not change them

[lost inlet]

what's with OnInteract early returning if it isn't authority? This will only set the value on a client, and a listen server host has authority

[nocturne quail]

interaction only happen on server

[dark edge]

Isn't that just to bail out, only allow interaction on authority?

seems fine

[dark edge]

That will only fire on actual replication on clients. If you want the server to run the same code you need to manually call it. In BP it fires on both

[lost inlet]

ah yeah

misread

and yeah the manual OnRep call is valid for the server

[nocturne quail]

if I do this everything works fine, setting it to null on next frame

void ATestClass::SetbTestFalse()
{
if (bTest)
{
bTest = false;
OnRep_Test();
GetWorld()->GetTimerManager().SetTimerForNextTick(this
{
TestActor = nullptr;
});
}
}

[lost inlet]

well on the server you should be calling the OnRep after setting values

[dark edge]

Are you getting the weird repnotify right away or after the actor has existed for some time?

[nocturne quail]

right away

[dark edge]

You sure you're not just firing the onrep on begin play or anything? Can we see the full cpp?

[nocturne quail]

TestActor is not repnotify, its just normal relicated

[lost inlet]

also if OnRep_Test is for bTest and not TestActor, then you can't really rely on that

[nocturne quail]

bTest is set to false on server on beginplay without calling on rep, since its not needed there

[nocturne quail]

why not? I can just benifits from it its normal

[lost inlet]

because the OnRep function is dealing exclusively with TestActor

you can't guarantee the client will receive both values simultaneously

[fallen fossil]

I know, I felt into this some day

[nocturne quail]

hmm, so I need to move bTest to TestActor and setting it by server under ATestClass ?

[lost inlet]

well it would be a good start to test the OnRep on the actual value it's dependent on, yeah

[dark edge]

That's your problem

race condition, it'll be much worse with latency.

why do you even have a bTest?

what does that boolean represent?

[nocturne quail]

lets say

seat in the plane, and TestActor is a pilot
So I need the pilot ref to be valid if the seat gets empty, so plane actor ATestClass can do stuff to TestActor before its sets null

[dark edge]

what state does that boolean represent?

You probably have redundant state here

I think you have a problem like if you had a bHasGun and AMyGun*, it's redundant

whether or not the pointer is valid IS the state

[nocturne quail]

this represent if the plane seat is occupied..
if false -> eject the pilot on server/cient [onrep]
if true -> attach the pilot to the seat on server/client [onrep]

[dark edge]

TestActor* can also represent that same state

[nocturne quail]

in both cases the pilot should not be null, and it should get nulled only if the onrep calls on all

[dark edge]

if TestActor* is not null, then the seat is occupied

[nocturne quail]

yeah it can, but it will be an extra route which is not what I want, I am already interacting with the plane, Plane already have the pilot

[dark edge]

It's not extra, you have redundant state.

[nocturne quail]

that's true but the issue is when the seat needs to be un occupied

and the plane still need the pilot to do custom stuff with it

[dark edge]

Just have the plane store who's the pilot and who's sitting in the seat

anyway, that's your problem, good luck.

[nocturne quail]

I am doing it exactly the same, in the end the issue is still the same.... I can't set pilot to null if the onrep has not done its job yet

[dark parcel]

setting the value to something else after calling OnRep on server sounds like code smell but maybe it's just me.

[dark edge]

Show current code

entire cpp file

[fallen fossil]

variables replication do not happen instantly, after set, it can take some time (unless rpc used)

[dark parcel]

strange statement

not sure how RPC is related and how it can take no time / less time if RPC is used.

[nocturne quail]

currently its not usable, I am re designing the whole class

[fallen fossil]

but true. Changed values on server are not instantly sent

[dark parcel]

so wdym by unless rpc is used it take less time / no time?

latency will always exist.

[fallen fossil]

its not latency, its that state replication awake parameters

[dark edge]

rpc will usually be faster as it'll fire either immediately or at least that same frame (don't remember), while replication can take a bit. This isn't counting latency.

[dark parcel]

sure, there is net update etc. But it won't be instant.

instruction / data will always need to travel through the net.

[nocturne quail]

instantly or not instantly has nothing to do with the issue, its frame related stuff... when onrep is called right after this the test actor is set to null, this means the for clients the onrep calls on next frame where the testactor was to null on the previous frame

[dark edge]

Your previous problem was a race condition since your onrep was on one variable but you were reading the state of another. Once you get the class re-written we'll see what the next problem is.

[nocturne quail]

the another variable was not yet changed when onrep was called, we can read the state of any variable in the onrep, they are not limited to the variable for repicated with them

[dark edge]

the bool can replicate and fire the repnotify BEFORE the pointer is replicated

race condition

that was your problem

don't do that

[fallen fossil]

is the problem only on server ?
im confused where is it, on clienr or server

[nocturne quail]

you mean it was not set correctly on the first place?

void ATestClass::SetbTestTrue(ATestActor* NewTestActor)
{
    if (!bTest)
    {
        TestActor = NewTestActor;
        bTest = true;
        OnRep_Test();
    }
}

but it logs valid on both

[dark edge]

Which replicated variable triggers OnRep_Test()?

[nocturne quail]

bTest

[dark edge]

that's your problem

[fallen fossil]

lol I Made seriazlizaion for ubojcet cause I wanted to send orders in same way for everythign

[dark edge]

it's racing against TestActor

just delete the bool and use the pointer for the state

[nocturne quail]

if I made the seat dependent on ptr, I will not be able to access it when it want leave the plane

[dark edge]

what?

You're not making any sense

This is to represent who's sitting in the seat, correct?

[nocturne quail]

Yes, that's why I need it to be valid inside onrep, if I use a multicast it works fine but than for late clients it will be trouble

[dark edge]

just haveing the pointer will be enough data

pointer is WHO is sitting in the seat, and also represents if anyone is sitting in the seat

you don't need bSeatIsTaken and APersonSitting*

just APersonSitting* is enough information

[nocturne quail]

than to reinvoke onrep it has to be set null

[dark edge]

set it null when they get up

that's the whole point

a seat only cares about who (if anyone) is sitting in it

[fallen fossil]

OnReps store previous values... just saying

[nocturne quail]

set it null when they get up , onrep is getting them up

[dark edge]

https://forums.unrealengine.com/t/how-to-get-previous-item-from-onrep-currentitem/340649

That way you can detect the valid -> invalid edge and tell the previously valid char to stand up

[nocturne quail]

so the parameter passed as the testactor will still be valid even if the variable sets to null?

[dark edge]

void ATestClass::OnRep_Sitter(ACharacter* PreviousSitter)
{
    if (PreviousSitter)
    {
        PreviousSitter.StandUp();
    }
    if (Sitter)
    {
        Sitter.SitInChair(this);
    }
}

[nocturne quail]

PreviousSitter will only be valid if Sitter == null?

[dark edge]

PreviousSitter is the old value, if you give your Onrep a parameter of the same type as the replicated type, it'll send it the previous value.

so you have the pre and post replication values

[nocturne quail]

hmm interesting, will test it now
Thanks very much

[nocturne quail]

this now seems to working fine, when the APilot EndPlay gets called, PreviousPilot should still be available for one time access?

void ATestActor::OnRep_Pilot(class APilot* PreviousPilot)
{
    // seat entry
    if (Pilot && Pilot != PreviousPilot)
    {
        Pilot->AttachCharacterToSeat(this, EVehicleSeat::Pilot);
    }
    
    // seat exit
    if (!Pilot && PreviousPilot)
    {
        PreviousPilot->DetachCharacterFromSeat(SeatTwoExit->GetComponentLocation(), false);
    }
    
    // (character swap)
    if (Pilot && PreviousPilot && Pilot != PreviousPilot)
    {
        PreviousPilot->DetachCharacterFromSeat(SeatTwoExit->GetComponentLocation(), false);
        Pilot->AttachCharacterToSeat(this, EVehicleSeat::Pilot);
    }
}

[verbal ice]

Did I say executables :p I wrote a Luau plugin, it's perfectly safe. Best they can do is crash the player's game, oh well.

[verbal ice]

They're just mods

Think Garry's Mod

[verbal ice]

I do need them to run before the clients start processing anything though since mods can have custom character roles etc

[slate phoenix]

Is VaRest a good plugin for API usage? I’m currently using it for API calls, but I’m not sure about it.
Do you have any solid plugin recommendations?
Or what would happen if I use Unreal Engine’s built-in HTTP / HTTPS system instead?

When using VaRest, I create objects but I can’t clean them up properly.
Because of this, I’m getting crashes.

[sterile plaza]

I'm overflowing the net bunch size (64kb) on the player controller during initial replication. Network Insights shows a few components, but none of them are very big. ~300-400 bits. Do I need to do something to see everything in that bunch?

[verbal ice]

Check what you're replicating

Do you have any big arrays

[sterile plaza]

that's what I'm trying to use insights for

[verbal ice]

im not sure if it shows bunches over that size

could be wrong though

[sterile plaza]

this is from official docs

https://d1iv7db44yhgxn.cloudfront.net/documentation/images/63e8672a-c007-4b02-9157-1199d69dcfba/insights-panels.png

it's over 64kb, but I'm seeing sizes similar to this

[fallen fossil]

have you considered making some library functions or objects to be replicated as such "Script" ? not particullary script but some transformed version of it into such abstract stracuture?

I don't know your goal, but I could tell you more about this a bit

[verbal ice]

Networking the scripts and all isnt an issue, I can send the data over the network

i just need to prevent networking for anything else until the scripts and map have been sent to the client and initialized

Scripts are just .luau files

Iris can do it, I just need to figure it out

[fallen fossil]

I don't know how joining session works 🤷‍♂️ sorry

[slate phoenix]

play standole and f1 or "
And write stat net u will see

how many have actor

[slate phoenix]

I experienced the same ping issue as well.
I was spawning the blocks as Actors (around 50k blocks) and I was getting about 300 ping!
After switching to HISM, my ping dropped to around 25.
You can check it the same way by typing stat net.

[sterile plaza]

I will have to try that. I ended up logging inside AActor::ReplicateSubobjects. Found one component replicating 44kb of debug data and another replicating 23k.

[little cairn]

anyone know a way to prevent net corrections from character capsule collisions? especially with other characters.

[fossil spoke]

If you are getting excessive corrections, you are usually doing something wrong.

[halcyon ore]

I'm having a 2 brain cell moment.
Where the heck does the engine handle control rotation reaching server?
I would've thought it was the set control rotation?

Actually, NVM.
Even if I knew that, doing it that way would cause issues. 😛

[little cairn]

just walking into another character causes a net correction. this is the out of the box behavior in unreal engine. do you know a solution?

ofc that is without ignoring pawn collision.

[stiff pine]

I don't know exactly the right question to ask to fix my problem. I suspect that once I know the question I'll be at the solution anyways. Only one of these created widgets shows up. I've tried every function in the Player node (Players are a subclass of the Character class) that can return a Player Controller object and none of them change that.

[halcyon ore]

Thats not how widgets work. 😛

[stiff pine]

Yeah that's what I figured the answer was gonna be.

[halcyon ore]

Each 1 client has to spawn it themselves.
Simply passing in others PC's isn't how that works.

[stiff pine]

This is local multiplayer.

One screen.

4 characters.

[halcyon ore]

OH

Um

[halcyon ore]

Then, whats your add node?
There is that other one, for split screen IIRC

[stiff pine]

Leading directly off the create widget node.

The stuff coming from above is printing out a ton of stuff from my failed debugging attempts, ignore that.

[halcyon ore]

Oh, well then IDK. lol
Out of my depth.
IDK how you even test 1 screen multiplayer with UE, i've never dabbled in that.
Sorry 🙁

[stiff pine]

Augh.

[dark parcel]

I'm not following this, make do you actually want to do?
Btw the default character class already have get controller method.

Just be aware though while server have a copy of everyone controller. Client only know their own controller.

widget is local, they are not reachable by anyone else but the local machine.

[stiff pine]

Make a textbox hover above my characters' heads.

[dark parcel]

involving controller but own's local controller not gonna work.

[stiff pine]

That's it! I don't know why it needs controllers at all!

[dark parcel]

are you following a tutorial?

So do you want a name tag or something?

a text that show up for each player's name?

[stiff pine]

It's a popup telling a player how well they did in a rhythm game.

[dark parcel]

so a pop up msg?

[stiff pine]

Like "good," "ok," "great," "perfect," etc.

[dark parcel]

are you making this text component at run time, or each character already has one and you just want to set the text?

[stiff pine]

I have all the logic for scoring and input handling, I just need these freaking components.

[dark parcel]

lets say they hit 5 beats in 1 seconds

[stiff pine]

At runtime, that's the issue.

[dark parcel]

do they see like 5 "good" in sucsession

[dark parcel]

Run time or not shouldn't be an issue.

just different approach.

[stiff pine]

Only one at a time. You only get one try per window, if you're too early or too late your inputs lock until the next window.

[stiff pine]

You'd think so, wouldn't you.

To be clear, that is hostility towards Unreal, not you.

[dark parcel]

imo it's a pretty trivial problem.

so what issue do you have atm, attaching the text component or displaying the msg to players (networking)?

[stiff pine]

I THOUGHT it would be. Because I thought what I'm doing should work!

I still don't know why it doesn't!

This is the whole thing.

That is too low resolution, I will split it up.

[dark parcel]

#multiplayer message

[stiff pine]

There is no networking.

It is local multiplayer.

The problem is that it doesn't show up for anyone except player 1.

[dark parcel]

I don't know the exact problem from static image (refering to the part where you show the widget).

Split screen?

[stiff pine]

I don't either. I cannot find where it's happening. Every test I have run indicates success.

[stiff pine]

Nope. One screen.

All 4 of these characters are accessed in that loop. I checked.

And no, there aren't 4 of the same widget on the one player. I checked that too.

[dark parcel]

need to see where you pull this from

brb breakfast, will comeback to check again.

[stiff pine]

The player's mesh. The players are what's being looped on.

[stiff pine]

Np, take your time.

[dark parcel]

Can you show it in picture

[stiff pine]

Like, the mesh itself?

[dark parcel]

The code

Cant see where its pulled from

[stiff pine]

You can see it in the first of these images.

It's that one launching off the top and going forwards.

[stiff pine]

The whole thing is visible here but I didn't want you to hurt your eyes squinting at this.

[dark parcel]

@stiff pine

no issue here

print string the player array element, make sure it's not the same object.

[stiff pine]

Different.

[atomic violet]

Maybe one of yall can help. When I add a basic static mesh into my level, my num Network Actors increases. Does this mean they are replicating?

[dark parcel]

If you tick off NetLoadOnClient, does it still increase?

[dark parcel]

I am not doing anything special, those 4 nodes does the job. Does the socket actually exist on every mesh?

[atomic violet]

It decreased, but now my client doesnt see anything. Hm.

[stiff pine]

It's the same mesh. And this happens even if I leave the socket as None.

It working properly for you concerns me.

Makes me wonder if there's something insane going on in the MinigameCharacter, which I did not make.

[dark parcel]

sounds like it doesn't find the socket and just attach to world origin

[stiff pine]

I've tried slapping the camera all over the place and I've never been able to spot them.

. . . do widgets backface cull by default?

And why would it work for the one player? Like I said, they are the same mesh.

[dark parcel]

if you are rendering screen space then it should always be pointing to the camera

[stiff pine]

Actually wait these shouldn't be able to backface cull, yeah.

[dark parcel]

it may not work on any of them probably

just stood in world origin is my guess

[stiff pine]

That player with the display can move, and it moves with them.

Also spawn locations are random.

[dark parcel]

#multiplayer message
Maybe make a function just to test, and work outside your array. Just Get All actor of those character and add the widget component

[stiff pine]

I'm not sure I follow. Doesn't Get All return an array anyways?

[dark parcel]

yeah but since I can't debug from here, that's probably the next test I would do.

[stiff pine]

Fair enough. Lemme check.

[stiff pine]

Well, I have new information. Not sure if it's helpful, but it's new. Nothing changes if all I do is switch out the Player reference for a Get All Players node, but I noticed that in your code, you have nothing attached to the target of the Add Widget Component, whereas in mine I'm attaching the player.

Detatching the Player from that makes NONE of them appear.

Nothing. Loop still runs 4 times on 4 different characters but there is no visual representation whatsoever.

[dark parcel]

well I'm attaching it on the last node with Attach Component To Component. Basically attach the widget component to the head socket in the character mesh.

also are you using Screen position? I have mine set on screen instead of world.

[stiff pine]

You mean here?

[dark parcel]

yeah nothing unusual there

post the function you made.

[stiff pine]

No visible changes from previous approach.

If I detach that target from Add Widget Component, the one that does appear for Player 1 disappears too.

[dark parcel]

you are not attaching to any socket

[stiff pine]

I know. I set it to the default to remove another point of uncertainty. It just moves the center of the widget to the center mass of the player. Still moves with them.

Just for the sake of idiot-proofing, is there some why I could have set up the widget itself that would cause this behavior?

I tried it with one of the default Unreal widgets, so I assume not.

[stiff pine]

I found them in the debugger. I cannot figure out where they say where they're rendered though.

I also cannot find them under the players. I can only find them like this if I filter for their class.

More information that might be useful but I'm not sure how to interpret it. That name does not change all 4 times this node executes, and none of them show up in the debugger after it gets out of the loop.

[stiff pine]

This is going to make me lose my everloving mind. I don't understand what the problem could even be! It won't throw errors and everything I can think of to validate ends up valid, save for the fact that they just don't seem to exist!!!

I went and modified the dang code that SPAWNS the players to make adding these components the first thing they do and it still doesn't work.

It's the EXACT same behavior! The first player gets one, the others get nothing, and I can't find ANY of them in the dang scene tree.

[nova wasp]

can you just run a draw debug string for each player as a sanity check? (their locations in the world at the time of this being called)

I suspect the widget is doing something internally different

I want to see what "just moves the center of the widget to the center mass of the player" actually means

how does it know which player?

[stiff pine]

Already have. Multiple times. I've also run sanity checks on every single gosh dang connection in the loop to make sure it's not missing anything. I have checked the NAMES of the players to confirm they're unique. I have checked the names of the WIDGETS to confirm they're unique. The closest thing to something useful I have found is that the Widget Component that gets attached to the player has a weird name ending in -1, and it's the same name every time. I have not managed to turn that information into anything useful yet.

[stiff pine]

Beats the everloving frick out of me, but it's always player 1.

[nova wasp]

show the code

you said "just moves the center of the widget to the center mass of the player"

you made that

[stiff pine]

It's already here, unless you want the endless permuations of print statements I've gone through.

[stiff pine]

One sec.

That's all that happens.

Spawn positions of the players are randomized, but the one with the label is always Player 1.

There, I have it changing at runtime to display the name of the player it's attached to.

The node that creates these gets reached 4 times. The player object that it gets attached to are different all 4 times. No errors are thrown. And yet, this is the result.

I have positioned the camera all around the scene to make sure the others aren't hiding somewhere. I have tried changing individual names manually on the spawned components after the loop finishes to make sure player 1 doesn't have 4 identical widgets attached to him for some reason. I have tried using different widgets, which yields the same behavior of only Player 1 getting one. I am running out of ideas to sanity test.

At this point I'm sure it's going to end up being something elementary that I don't know about but I just cannot find it!

[twin vessel]

Idk if you tried already, but might be worth testing. Did you try to use world space widget component instead of screen space?
I know screen space widgets have issues with local multiplayer, but i do not remember if it was only for the case of multiple viewports

[stiff pine]

It turned out to be something that is almost certainly a 5.6 bug. Creating widgets using something that has a controller as a target results in the widgets belonging to everyone but the first player being collapsed in a manner that cannot be altered or circumvented by any methods I'm familiar with. You have to spawn them targeted on something random that isn't a player, and then bind the component to the player socket.

Apparently the developers of 5.6 never considered that someone might want to make a local multiplayer game that wasn't split-screen.

If that sounds stupid and convoluted, that's because it is. It took me 5 hours and the help of many people to find that.

[stiff pine]

The reason we're pretty sure it's a bug is because 5.7 has a project setting that is genuinely a toggle for the exact behavior I was experiencing.

[fallen fossil]

I added Custom UObject Permission Manager into player state.

Everything is replicated, but my understaning is that It does not trigger
PostNetReceive on PlayerState, cause its using its own channel? 🤔

[chrome bay]

Depends. If it's a replicated property of the player state, that property changing would trigger PostNetReceive. But if it's not, then no - you would need to implement PostNetReceive on the object itself.

Save yourself trouble, use an actor component instead.

Many of those functions are native to UObject, so all objects can implement them,

[fallen fossil]

yes, its ActorComponent

but I found the isseu

somehow compiler did let me start project without setting LifetimeReplication for my structs

....
Its not replicated, PostNetReceive is never triggered at all :/

UCLASS(BlueprintType, Blueprintable)
class UCommandPermissionManager : public UActorComponent {
    GENERATED_BODY()

public:
    UPROPERTY(BlueprintReadWrite, EditDefaultsOnly, Replicated)
    FCityCommandInfo BuildPerm{true, UCmd_Core::StaticClass(), "Build"};
// ... More commands 

    virtual void GetLifetimeReplicatedProps( TArray<class FLifetimeProperty>& OutLifetimeProps ) const override;

    UFUNCTION(BlueprintCallable)
    void SetPermissionAllowed( const ESharedCommandsUI type, const bool IsAllowed );
    virtual void PostNetReceive() override;
    void NotifyPS() const;

UCLASS(BlueprintType, Blueprintable)
class MINICITYBUILDER_API ACityBuilderPlayerState : public APlayerState {
    GENERATED_BODY()

    UPROPERTY(BlueprintReadWrite, EditDefaultsOnly, Replicated)
    TObjectPtr<class UCommandPermissionManager> CommandManager;

[tardy fossil]

yea not defining the lifetime function wont cause any compiling errors

[fallen fossil]

UCommandPermissionManager::UCommandPermissionManager()
{
    SetIsReplicatedByDefault(true);
}

Found real issue.

I though marking uproperty as replicated is sufficient oh

but its always pointer, so maybe that was problem

[slate phoenix]

nice

[dark edge]

What's the desired result here?

You want a widget to appear at XXXX for each YYYY when ZZZZZ happens.

[stiff pine]

It was specifically that they need to stick to XXXX of each YYYY. And YYYY are players. YYYY being the players ended up being the cause of this insane behavior.

[dark edge]

use the word pawn

just have each pawn have a widget component, what's the problem?

[dark parcel]

Soo, did Autheur suggestion fix it?

[stiff pine]

I'm not sure if that's specifically the issue which is why I'm being vague. It also might be controllers.

[dark edge]

they can just exist in the BP and have their widget swapped out whenever

[stiff pine]

Yes.

[dark parcel]

@dark edge something to do with local multiplayer.

[dark edge]

I don't see any reason why any of this has to rely on controllers at all

[stiff pine]

Yeah. 5.6 does not seem to have properly accounted for the concept of a local multiplayer game that doesn't use split-screen.

[dark parcel]

5.6 they do some optimization.

Well optimization isnt the right word.

[stiff pine]

It doesn't, but the things I was using as a target when creating the Widget Components had controllers, and that appears to have been causing the components to be collapsed.

[dark parcel]

But basically some widget r hidden for split screen. Im guesing when not owned by the player.

Have to tick some bool in the project setting.

Force something widget for split screen.

[stiff pine]

That does NOT exist in 5.6. That's the most condemning thing.

[dark parcel]

So how do you fix it? Upgrade the engine?

[dark edge]

why are you creating them at runtime anyway, just plop one in your pawn bp

then you can dynamically shove widgets in them

[stiff pine]

Well yes, but since I can't do that right now, the workaround is spawn the components targeted on something random, and then attach them to sockets of the player meshes.

[dark edge]

If that's the problem then yeah it's totally different

[dark parcel]

Since you r not doing split screen anyway, i find giving widget controller kinda redundant.

Just shove all widget for player 0

[stiff pine]

We only need them in this level, and this is a minigame collection. If our solution to everything was to attach new components to the player blueprint whenever one minigame needs it, the character would rapidly gain bloat. But yes, I am aware that in most cases this is the correct solution.

[stiff pine]

I wasn't! I was for a while but I stopped after my initial conversation with you when I saw that your functional version didn't do that.

[dark parcel]

So if u dont plug controller on creating widget it work?

[stiff pine]

If you don't plug something that has a controller. The targets on these component creation nodes are the players, which are pawns that have corresponding controllers. I don't know exactly if it's them being pawns, or having controllers, or something else entirely, but what I DO know is that if those connections are replaced with a random actor in the scene that doesn't have anything to do with multiplayer, it works. Right now they're targeted to the dang skybox and it works just fine.

Ignore the fact that this is an abomination. I'm very aware of how terrible this approach is, I was only doing it as a sanity check.

[dark edge]

Looking like the timeline to Primer over here

https://xkcd.com/657/large/

[dark parcel]

My man just use Get instead looping with break.

[stiff pine]

Does that get a direct reference instead of a copy?

[dark edge]

copy of what

[dark parcel]

Its a copy of the address so its the same thing.

[dark edge]

a pointer?

[stiff pine]

The only reason I'm doing this blasphemy is because I couldn't figure out a better way to get a direct reference out of an array.

[dark edge]

if it's pointers it doesnt matter

get by copy of something out of Players (should be pawns) is the ADDRESS of a pawn, not hte pawn itself. It won't make a new pawn

[dark parcel]

You are getting a copy of the memory address, which is just an intger.

The underlying value it points to is the same.

[stiff pine]

Really? The node description specifically says that changes made to it will not propagate back to the original.

[dark edge]

you need to understand the difference between a pointer and an object

[stiff pine]

If it would've worked this whole time, I made this whole disaster for nothing.

[stiff pine]

I understand it in C++.

[dark edge]

welcome to programming

things this color are pointers

[stiff pine]

I wasted so much time.

[dark edge]

BP calls them references but they can be null so pointers are more accurate

[stiff pine]

Welp. Good to know going forwards!

I'm annoyed at that node description now.

[dark edge]

I'm on my 5th rewrite of a system for my game, started 10 years ago. It happens.

[dark parcel]

Tbf the node description isnt wrong.

Just misinterpreted

[stiff pine]

If the dark blue means a pointer, the Get (A Copy) node does the exact same thing in terms of return value as the For Each loop, which does NOT give you the warning.

[real quarry]

Greetings Multiplayer Posse

Question : the OnlineSubsystemEOS plugin has a README with dire warnings that it is an experimental plugin ( though the plugin doesn't have the experimental tag in the editor Plugins view ).

What is the general consensus : fine to use for prototype at least I presume, but is it ok to build on in the long run? What is the 'correct' path for production?

[woeful ferry]

Not done much session related work since 2020, so I'm a bit rusty and can't get it to work.

How I host the game:

    public void HostGame(APlayerController hostingPlayer, IDictionary<string, string> arguments, FPrimaryAssetId map)
    {
        UCommonSessionSubsystem sessionSubsystem = GetGameInstanceSubsystem<UCommonSessionSubsystem>();

        ActiveHostRequest = sessionSubsystem.CreateOnlineHostSessionRequest();
        ActiveHostRequest.MapID = map;
        ActiveHostRequest.ModeNameForAdvertisement = "BazaarMayhem";
        ActiveHostRequest.UseLobbies = true;
        ActiveHostRequest.OnlineMode = ECommonSessionOnlineMode.Online;
        ActiveHostRequest.MaxPlayerCount = 4;
        ActiveHostRequest.ExtraArgs = arguments;
        
        sessionSubsystem.HostSession(hostingPlayer, ActiveHostRequest);
    }

How I look for sessions:

[UClass]
public partial class UEmberSearchSessionRequest : UCommonSession_SearchSessionRequest
{
    public string LobbyId { get; private set; } = null!;

    public void InitializeFrom(string lobbyId)
    {
        APlayerController playerController = GetTypedOuter<APlayerController>();
        
        LobbyId = lobbyId;
        OnlineMode = ECommonSessionOnlineMode.Online;
        UseLobbies = true;
        OnSearchFinished += OnSessionSearchFinished;
        
        UCommonSessionSubsystem sessionSubsystem = GetGameInstanceSubsystem<UCommonSessionSubsystem>();
        sessionSubsystem.FindSessions(playerController, this);
    }

    [UFunction]
    private void OnSessionSearchFinished(bool succeeded, FText errorMessage)
    {
        APlayerController owningPlayerController = GetTypedOuter<APlayerController>();
        UCommonSessionSubsystem commonSessionSubsystem = GetGameInstanceSubsystem<UCommonSessionSubsystem>();
        
        foreach (UCommonSession_SearchResult searchResult in Results)
        {
                    // Always empty
        }
    }
}

Config:

+NetDriverDefinitions=(DefName="GameNetDriver",DriverClassName="OnlineSubsystemSteam.SteamNetDriver",DriverClassNameFallback="OnlineSubsystemUtils.IpNetDriver")

[OnlineSubsystem]
DefaultPlatformService=Steam

[OnlineSubsystemSteam]
bEnabled=true
SteamDevAppId=480

[/Script/OnlineSubsystemSteam.SteamNetDriver]
NetConnectionClassName="OnlineSubsystemSteam.SteamNetConnection"

[/Script/CommonUser.CommonSessionSubsystem]
bUseBeacons=false```

The session sets up right, I can join it with `open 127.0.01` in PIE, but it fails to find sessions when doing standalone. Anyone see anything that I've missed? 😄

[muted mantle]

Hi! I have a WidgetBP that autogenerates Childs and saves the references from it to an Array.
I need to init a change to a specific Child from the Server-Side.
The Server also has an Index that equals the ChildWidget index.

How can I do this? I dont get it to work.

Blueprint Runtime Error: "Accessed None trying to read (real) property UI_Cells in BP_GridManager_C". Node:  Call UI Cell Has Changed Graph:  Grid_WriteCell Function:  Grid Write Cell Blueprint:  BP_GridManager
Blueprint Runtime Error: "Accessed None". Node:  Call UI Cell Has Changed Graph:  Grid_WriteCell Function:  Grid Write Cell Blueprint:  BP_GridManager

I do have the Widget Blueprints added as reference object into my variables. But it says "Unknown" in the Debugger

[soft relic]

Widgets do not exist on servers unless its a playing listen server, in such cases you need to do workarounds such as replicating structures with data per widget instance etc.

[dark edge]

Widgets don't network

don't do networking within a widget

[dark edge]

What are you actually trying to do here, what's the actual gameplay purpose?

[muted mantle]

Well I try to Display a Grid in 2D and want just say something line tetris. But I Need a way to Reflect changes of every cell in my Grid to the ui

I just started with unreal 4 days ago so im still a big noob

(The Server side Logic works) and gets written to a struct Array that reflects each cell

[soft relic]

you can make an array with structures to represent rows/collumns, then have all clients update their widgets based on that replicated variable

Might be complicated if you are just starting out though, replicating widgets is harder than normal in game replication such as actors, components etc.

[muted mantle]

I do have the struct replicated and the Grid Displays. ( I Test as Client in the dropdown)

So I got everything working beside updating the Ui

[soft relic]

are you replicating in the widget? widgets dont replicate, you can use a component that replicates inside the game state maybe

[muted mantle]

I have the bp gridmanager that holds the struct and does a repnotify.
The Widget just has an ref var to the Grid Manager

[soft relic]

Then the error must be somewhere else. Make sure you have a valid reference to the actor and that the grid manager bp also replicates. If its a grid manager with no components at all in it, make sure to also check "Always Relevant"

[muted mantle]

Im back at the PC now and open the Project, then i can give more detail

It was on replicate but not always relevant

I already started reworking it based on this image i found.
So Right now i do in the OnRep function: Get All Actors of Class -> ForEach -> Call Event in BP_PlayerController.

I hope that is even the right thing to do

[muted mantle]

OMG I got it working

(atleast a print string from the widget blueprint)

i have no idea if this even makes sense but to get it working after i did the call from my BP_GridManager to the BP_PlayerController i did this

[dark edge]

that should just be a repnotify probably

OnRep_MyCell

[muted mantle]

But where?

[dark edge]

the cell you're in is state, prefer to use repnotify for responses to state

What's the context here, you want the widget to update when?

[muted mantle]

in my BP_Gridmanager i have a random int so i use it as an index to change a bool isBlocked in my "Struct Array" to true.
This struct already is RepNotify.

BP_GridManager is Spawned per player. So i test in Client Mode to get it right

[dark edge]

What's the actual thing you're trying to do here? This is a grid of what, and CurrentCell represents what?

[muted mantle]

It's this ugly grid thats defined in gridmanager and gets build as a widget.
I try to do some chaos tetris, block stuff - i dont know for sure what its gonna be.
I'm just learning tbh

The Struct hast infos like CellType (Block, Bomb, Empty)
And i want to change celltype right now from empty to block. And then i want to change the color of the cell

From my OnRep for the struct i go to the BP_PlayerController you saw what happend there on the other pic.

[soft relic]

You can also make a dispatcher on the OnRep call and have your controllers bind to it

Above implementation is wrong since each PC should call it locally when the rep happens

[muted mantle]

How would I do that?

[soft relic]

I would suggest looking up how dispatchers/delegates work, will be very handy once you understand it

[muted mantle]

Oh would I just bind the Event dispatch from my UI BP Directly in the onrep of my bp Grid Manager

[soft relic]

the on rep would call a dispatcher which your UI would be binded to, yes

[muted mantle]

I think I tried that First but had all the unknown class errors

[dark parcel]

looping over controller in the OnRep make no sense.

server have everyone controller but clients only know their own controller.

[halcyon ore]

If an actor isn't marked as replicated, but is placed in a map by default.
You can still at least replicate its reference around, cuz its a stable/ static name or whatever.
But, you can't do RPC's through it right?

[nocturne quail]

any idea why this crash happens?

[nova wasp]

that's not even close to enough context to have any way of guessing what happened here and I don't even know which verison you are on

we can only see what you show us

[nocturne quail]

on 5.4, I don't have any more context also

its just crashed

[nova wasp]

if it's a client callstack with no dump that's fair I guess but it would help to include that in the first post

we can only see what you show us

[nocturne quail]

seems like onrep is called twice somewhere but I have double checked and its not called twice

like RegisterNetGUID_Server calls twice in the call stack

[nova wasp]

is it this check? Failed to validate ObjectLookup map in UPackageMap. Object '%s' was not in the NetGUIDLookup map with with value '%s

anything related in the logs?

[nocturne quail]

Yeah exactly but how to find which object it is complaining for

will crash it again now, I think in logs there was no usefull information, but I will try t crash it again cause its not crashing all the time that's why I'm very worried

[nova wasp]

if you can crash it on command just attach a debugger (I recently found out about using -basedir to use cooked content from the ide easily, very useful)

no need to guess

[tardy fossil]

could try turning on LogNetPackageMap for more info

[normal fulcrum]

*as well here is -BaseFromWorkingDir which is the same as -basedir, just takes the IDE's Working path *

[nocturne quail]

I think I found that onrep is called twice

but if this is the case, why its not crashing always?

this with a typo fixed

[swift bay]

Hey there, I have a component on my PlayerState that needs to execute some logic in InitializeComponent but only for the copy of the local player. The logic is a form of replication guard so that import stuff triggers only after all the import stuff has replicated.

Is there a way to properly detect that the code is currently executed on the local player's copy of the component, in the case I start playing in PIE with 2 players?

I was trying to go the ROLE_SimulatedProxy / ROLE_AutonmousProxy route but because the component is part of the PS, it is always ROLE_SimulatedProxy. I also cant compare playerstate's by accessing GetWorld()->GetFirstPlayerController()->PlayerState == ComponentOwner as the PCs playerstate might not be replicated yet (cant really wait for OnRep_PlayerState I think?).

[dark parcel]

@swift bay Player state have method to get the controller that created the player state.

to get the player state that is owned by the local player, simply.

[swift bay]

That controller isn't replicated yet unfortunately

[dark parcel]

Controller is the one that makes the player state it should be valid.

However it would be invalid on non local player state because Clients only have a copy of their own controller.

if you are in CPP, there's another method

GetOwningPlayerController

if (GetOwningPlayerController())
{
if (GetOwningPlayerController()->IsLocallyControlled)
{
do stuff
}
}

absolutely no need for LocalOwnerRole check or NetMode. Get rid of those.

[swift bay]

Ye these are currently there to prevent unnecessary break points (and the dedicated server I guess).
I'll check the GetOwningPC, maybe that works. Just stepped through all 4 calls of PS->GetPlayerController() but none returned the PC. Should I use a different method to do my logic than ::InitializeComponent() ?

[dark parcel]

Don't you want the init code to run, regardless if it's listen server or dedicated server?

I would get rid of those NetModde and LocalRole checks.

[swift bay]

Also 4x null with OwningPC

No its just for the client

[dark parcel]

You can filter with HasAuthority

When is the code run? Why InitializeComponent?

Try begin play of the player state for sanity check.

[swift bay]

I'll try BeginPlay then and get back to you

[nocturne quail]

this could be because of this I think

LogAIPerception: Warning: UAIPerceptionComponent::OnRegister: Perception Component is being registered with BP_Bot_C_0, they are designed to work with AAIControllers!
LogScript: Warning: Script Msg: A null object was passed as a world context object to UEngine::GetWorldFromContextObject().
LogScript: Warning: Script Msg called by: DemoMap_C /Game/DemoMap/UEDPIE_0_DemoMap.DemoMap:PersistentLevel.DemoMap_C_0

well I can't reproduce the crash again, if it crash next time will see

[frosty harbor]

I upgraded to 5.6 and im having some issue with crouching. When simulating network latency (100) the issue I have is that if I crouch & uncrouch fast, there's a little correction at the end of the uncrouching which looks very unnatural. If I play with 0 latency it never happens. I assume it's because the server capsule still didn't "update" so locally you get corrected. Any idea how to fix this? I was thinking of uncrouching only started once server gets there (like the net sync node you have in the gameplay abilities) but there's no such node outside of gameplay abilities.

[dark parcel]

Why are you doing it on triggered?

use started

Triggered runs every frame.

Btw there shouldn't be correction with 100 ms or 500 ms.
Crouching is already handled in the cmc.

[lost inlet]

you can use pressed/release triggers in the IA

there can be gotchas with "Started" but that only really matters for things like delays, actuation amounts, and key chords

[frosty harbor]

Yea I'm using Pressed/Released triggers

[frosty harbor]

Then I have no clue why it only happens on latency

[lost inlet]

then if you are, I would only use triggered and use action value to determine

but is probably not the issue, and I can't say I've seen it happen that often with the default CMC

[frosty harbor]

yea the issue is not with the crouching itself

it's with the uncrouching

[lost inlet]

well you probably should only use triggered

[frosty harbor]

I can show a quick snippet of what its doing

[lost inlet]

because you'll get triggered and completed at the same time on release

[nocturne quail]

don't depend on capsule for crouching, this is a bad practice to start with, only do it if your character recieving damage hits using capsule

[lost inlet]

since there is a release trigger

[nocturne quail]

better do it using character body physics whilch will be smooth

[lost inlet]

that's safe to ignore

[nocturne quail]

I have a custom character class which has no capsule at all, instead I am using a sphere root as a root for the class

[frosty harbor]

https://medal.tv/games/screen-capture/clips/lYhnVCnxuaGeV8MvE

mute it since music in background lol

[lost inlet]

yeah did you make the change yet?

[frosty harbor]

I depend on it due to things like not being allowed to uncrouch in certain areas, etc

[lost inlet]

triggered -> branch on action value -> true exec on Crouch, false exec on UnCrouch

[nocturne quail]

don't rely on inbuilt things, they are done very badly

[nocturne quail]

which areas are you talking about?

[frosty harbor]

When I put print in it's only triggering once for both started & completed

[frosty harbor]

If there's a ceiling above the character for example, you can't uncrouch

[nocturne quail]

you can just give a tag to any area and when you crouch, check if you are not in this area that's all

[lost inlet]

your example didn't use "started" though, it uses "triggered"

[nocturne quail]

you can do it with character physics also, no need to depend on capsule

[lost inlet]

this is typical missty anti-advice

plenty of multiplayer games uses the capsule for movement

[nocturne quail]

it's not necessary at all

[frosty harbor]

Sorry I meant to say:

• I changed to Started / Completed and it still happens, when I print it still only shows a single print in the action value (True for when I press and False when I release). If it was getting double triggered it would print mulitple times instead of only once per action no?

I also tested triggered/completed and you were right, it's triggering twice when I release, so i switched to started but still same issue

[dark parcel]

CMC works with the capsule component.

[lost inlet]

I'm not sure why you didn't do it this way if you are using those IA triggers

but sure, it won't really change the outcome

[nocturne quail]

use CCMC 😄

[dark parcel]

what's that?

[frosty harbor]

I can do it that way, it's the exact same as started/completed essentially no? Don't think it changes anything in that regard

But I think the issue will be the same

[nocturne quail]

Custom

[dark parcel]

Why did you mention gameplay ability system? Is the input event not the only place where you crouch and uncrouch?

[lost inlet]

yeah, though using the action value was pretty handy when you simulatenously want to support hold and toggle crouch

[dark parcel]

I have no issue crouching with 50 ms or 500 ms.

no correction.

[frosty harbor]

Yes, I mentioned it because I thought it was server correcting the local player so I thought about using the net sync node that you ahve in GA's to make client prediction wait for server but yea you guys are saying it's not server correction so im not too sure what could be causing it

And yea I don't use it in abilities

[dark parcel]

they are not relevant.

Client movement prediction data is stored in FSavedMove

crouch is already in it by default.

[frosty harbor]

that does make sense

[nocturne quail]

also if you are feeling the animation is not smooth, it can be because you upgraded to another engine version, try reimporting the crouch animation

[lost inlet]

unless you're doing a true first person system, I'm not sure why you'd have crouch transitions driven by animation

[frosty harbor]

I don't think it's animation issue, if I crouch slowly (as in hold it crouch and onyl uncrouch after a little bit) it works flawlessly

it only has issues when uncrouching quickly after crouching

[lost inlet]

and it does look more like the transition than any sort of server correction

[nocturne quail]

put a cool down, you can't uncrouch if courch is not yet finished

[glass rune]

Hi guys!
I'm taking a multiplayer C++ course on Udemy. There is something that drives me crazy, and I don't know whether this is how the industry does it or whether something is off. The tutor is using A LOT of guard initializations (I don't know what to call them) with ternary operators, e.g.:

MyHUD = MyHUD == nullptr ? Cast<AMyHUD>(GetHUD()) : MyHUD;

I understand that this can help when developing multiplayer games, since initialization can vary a lot between server and client, but this syntax seems really misleading because we have no clear information about whether the pointer is going to be initialized. As a result, later, when we check the pointer, if something doesn't get called, we may not even know why. Personally, I'm trying to find the correct places to initialize these pointers, and it seems to be a much better and more “professional (?)” approach.

What is your take on this? Maybe someone from the industry can explain how it looks in practice.

[lost inlet]

did you implement anything specific for the view transition?

HUD initialisation is predictable anyway since the client spawns it

frankly, I find that a bit ugly

[frosty harbor]

could be but why is the transition slower with latency than with no latency how is it relevant

[dark parcel]

It really depends on the intention. Not really a yes or no question.

[frosty harbor]

it should take just as long except starts 100 ms later

[frosty harbor]

yea that's one of the solutions I've thought about but feel like it'll feel kinda bad to uncrouch and then it only becomes responsive after "xx-xxx" ms

[glass rune]

Exactly my thoughts. I usually find a way to initialize those pointers in some init functions with NetMode if checks.

[nocturne quail]

correct way is to put an anim notify in the end of the crouch animation to set iscrouch to true, and when you press uncrouch check if iscrouch is actually true
this way you will make sure to correctly uncrouch and will not spam the server also since if the bool fails locally don't request the server to uncrouch

[lost inlet]

HUD's also a bad example, very seldomly would gameplay code need to be aware of it since HUD elements should be event driven anyway

[dark parcel]

MovementData is send all the time.

"spam the server" is just not the right take.

[lost inlet]

well is this relying on something default in the CMC or did you do something to implement the camera height between crouching and standing?

[frosty harbor]

.< I wanted to avoid change from the default crouch / uncrouch as much as possible but yea if I can't fix these issues I guess i'll have to do something like that

[frosty harbor]

It's using the default cmc stuff

[dark parcel]

@frosty harbor Can you describe these "correction"? Are you just having animation glitch and assumed it's correction or the client actually get Corrected?

You know how to debug movement correction?

[frosty harbor]

All I ever did was add the animbp to play the animation when crouching

[dark parcel]

You don't even need the animation to see the crouch in action. Just visualize the capsule component.

@frosty harbor p.NetShowCorrection

[frosty harbor]

https://medal.tv/games/screen-capture/clips/lYhnVCnxuaGeV8MvE

Basically when I crouch -> Uncrouch quickly / instantly, there's a "up" then "down" "correction" at the end. but ONLY when I have latency. If latency is 0 this DOES NOT happen.

If I Crouch -> Wait 1-2 seconds -> Uncrouch no issues arise.

[dark parcel]

if you don't get any red debug sphere then you are not corrected.

[frosty harbor]

And I do not know how to debug but i'll give it a look

[dark parcel]

I don't even think this is network related.

All crouch does is just changing the capsule component height INSTANTLY.

if you want smooth camera transition, you do have to implement that your self.

from what I see, this isn't even network related.

you might want to check what drives your camera.

[frosty harbor]

back in 5.5.4 what I did was add the camera to a springarm and it made the transition smooth but that got messed up when I upgraded to 5.6

I also had tried all those timeline methods for smooth transitions but there were always some issues

[glass rune]

He is also doing it for example in PlayerState in OnRep_Score function but with Character pointer.

void AMyPlayerState::OnRep_Score()
{
    Super::OnRep_Score();

    Character = Character == nullptr ? Cast<AMyCharacter>(GetPawn()) : Character;
    if (Character && Character->Controller)
    {
        Controller = Controller == nullptr ? Cast<AMyPlayerController>(Character->Controller) : Controller;
        if (Controller)
        {
            Controller->SetHUDScore(Score);
        }
    }
}

It seems ugly for me too so my take is to Initialize Character and PlayerController pointers in PlayerState with OnPawnSet delegate once

[nocturne quail]

from the video I feel like this is collisions blocking each other

set the capsule to ignore camera collision

[dark parcel]

Code smell to me.

why is character being initialized every time a score is replicated....

I have nothing againts trenary operator, I used them all the time. But why are controller and character is set when a score is replicated is beyond me.

[frosty harbor]

Yea I have it on ignore

[dark parcel]

where do you use this RelativeLocationZ? Or where is the code that drive the camera?

[glass rune]

I also really like ternary operators, but this guard initializations is too much for me. I wanted to get some opinion

[lost inlet]

yeah I consider this a bit of an anti-pattern. the HUD element displaying the score should just register a delegate for when the score changes and use that to update

[nocturne quail]

in this case I would just a create a new character with only character mesh and try to reproduce the bug

[lost inlet]

it can differ slightly if you want to use MVVM, but this isn't something that would be directly on the player state either

[glass rune]

Ok, that's what I thought. It's just my first touch with multiplayer and I thought that maybe that's how we do it in multi

[dark parcel]

doesn't matter multiplayer or single player.

You want to keep dependency one way or minimum.

[glass rune]

Great. Thanks. So it's not me that is insane in this case 😅

[frosty harbor]

You can ignore the variables, those are just old variables I forgot to delete when I was testing stuff way back then related to camera

[dark parcel]

Crouching alone shouldn't move your camera

all it does is changing the capsule half height

well if the camera is a child of the capsule then it should change in height (maybe) but it would be instant.

[frosty harbor]

The camera is a child of the mesh, which is a child of the capsule

It's bound to the head bone

[dark parcel]

yeah so issue is with animation then.

if your camera is tied to the head bone.

[frosty harbor]

I see

[lost inlet]

damn so the animation part was right

[frosty harbor]

I guess the issue is something along the lines of the player already asking to uncrouch before crouching animation even begins on server

[lost inlet]

generally attaching the camera to 3p animation is how you make people motion sick

[frosty harbor]

im doing true first person with spine bending and stuff so needs ot be attached to head bone

[lost inlet]

still stands though, it's subjective but I've always found true first person games to have a layer of jank to them

[frosty harbor]

hmmm

I guess the only way to do it really is to add a cooldown before the uncrouch can happen

[dark parcel]

you gotta make sure the transition is smooth to what ever method you decide to do.

[frosty harbor]

actually nvm cooldown wont fix it since its latency issue

[dark parcel]

like if it's tied to the head bone, then you want to make sure the animation doesn't go from going down then snap back up when doing the uncrouch animation.

I don't think it's latency issue.

if you play as listen server u have no issue?

[frosty harbor]

It doens't happen at 0 latency at all, only when I have latency so has to be latency issue

[nocturne quail]

try to disable correction at all and test again to find if this is latency issue

[frosty harbor]

I think it's along the lines of player asks to crouch / uncrouch at nearly same time before server even started crouching

hence the latency

[frosty harbor]

I disabled corrections in CMC but nothing changed, unless there's another way to disable corrections for everything

[nocturne quail]

its a bug in your code this is confirmed cause we don't have this issue, I just tested with 5.4 and in 5.6 this works the same way

[frosty harbor]

I don't really touch anything related to crouch other than the input action as I showed above and in the animbp for the animation

[nocturne quail]

by the way why did you updated to 5.6? consider never ever updating unless you need a new cool feature from the new engine in your project

[frosty harbor]

thats all I do for crouching in ANIMBP too

[frosty harbor]

Needed it because of metahuman , they no longer support anything below 5.6 + Fab assets have a lot more compatibility with metahuman than other skeletons

in terms of armors and stuff

[nocturne quail]

hmm legit reason

[nocturne quail]

in your anim bp when you set iscrouching to true/false , your character crouch/uncrouch smoothly ?

[frosty harbor]

yea it does

[dark parcel]

I just noticed my camera smoothly adjust to the character crouch position.

stole the code from lyra Camera Mode.

[nocturne quail]

try with a new character class, attach camera to head head socket and just do crouch/uncrouch

if this is smooth, you have problem in your previos character class

[nocturne quail]

so it is confirmed they have issue with camera adjustments?

[dark parcel]

With SteelHard/ I have no idea.

as for Crouch, all it does it just changing the capsule height.

it doesn't do anything to the camera out of the box.

[nocturne quail]

maybe they are adjusting camera manually, we don't know yet 😄

[nocturne quail]

are you doing camera adjustments when character crouch?

[dark parcel]

well seems like he is using the head bone.

Crouching just changes the capsule height, nothing else.

i can spam it, and it's still fine.

[frosty harbor]

The adjustment is based on the animation itself since its following the head bone

Since this only happens under latency and never when I have 0 latency I think it's just some issue between triggering uncrouching too fast after triggering crouch, locally we're already uncrouched but on server the animation for crouching just began due to latency so it kinda bugs out is my guess

[dark parcel]

if you add smoothing on the crouch / uncrouch transition, do you still get "corrected"?

[frosty harbor]

Dunno, depends on how you do it. I remember people doing these timeline based transitions but those touched the capsule which I wanted to avoid doing.

Yea if I add a delay of 0.5+ it works just fine

before the uncrouching happens

if i do 0.5 it literally does not work

has to be specifically 0.501 minimum lol

it does feel kinda meh not uncrouching right away tho

Could it be because on server the capsule still hasn't updated / crouched yet due to latency & same for crouching? Dunno its just weird

especially now that it literally requires me to wait a very specific 0.5s before uncrouching for it to not happen

[dark parcel]

Client will always be ahead.

It simulate move on it's own machine (Crouching, proning, w/e).
At the same time sending the move data to server.

Server replay the Move Data see if it's valid, if not valid correct the client.

Meanwhile client went ahead with it's move every tick. The only time it get snapped is when the client receive correction.

[frosty harbor]

I'm looking into Lyra way of doing crouching but I can't seem to find which animbp they're setting up the animations in for crouching

[vocal current]

Not fully answering the question because I'm not at a PC to go look, but basically Lyra's animation setup is that there's one main Anim BP, which then dynamically references child Blueprints using the Anim BP interfaces feature.

Those child Blueprints handle all the anims that change when you equip a different weapon..

There's a base class for those child BP's that has some logic in it .. I think it's probably in there, since there are different crouch anims depending on the weapon you're holding

[frosty harbor]

that sounds quite convoluted

found it but yea since camera is not attached to the head bone its a different thing

.<

[hoary spear]

Is it possible for a property to replicate before the owning actor is finished constructing ?

I would think no, without looking deeply into it

[nova wasp]

"finished constructing" as in running OnConstruction?

[hoary spear]

Not sure what i've been drinking but for a moment thought the InputComponent was setup in the native constructor

Doesnt make sense

Having an issue with ASC and a possible race condition

[nova wasp]

subobjects should at least require their parent object (in the REPLICATION sense) to be present beforehand

[hoary spear]

Its called both in SetupPlayerInputComponentand ÒnRep_PlayerState, yet sooometimes it fails still

[nova wasp]

There's no gaurantee a replicated object will be there for either of those imo

properties are completely capable of being skipped in either case in the bunch afaik

[hoary spear]

yeah that's probably whats up

reminds me of the hud issue i had

[nova wasp]

You are generally going to want to account for the fact the playerstate might not be there yet pretty much everywhere

[hoary spear]

ended up having 3 different paths checking that everything else was ready, so any order of replication would work

[nova wasp]

Generally speaking if you can't think of a nicer way the most simple "dumb" fix is to just set a timer for next tick every time it's missing (or a very short delay) to consider the "initialization" steps required to trigger your "start the game" flows

but this is only appropriate really if

• you KNOW the object will show eventually or you know to cancel retrying if you need to give up
• you actaully re-try instead of just delaying once and praying

It's a bit of a headache for obvious reasons so it might be convenient to have 1 place that kind of acts as the arbiter of "init state"
for Lyra's init states for example are just a tagged if statements

that let you define different "steps" of readiness depending on who cares about what

[hoary spear]

hmmmm

i'm not a fan of any timer solutions, but man sometimes the way around them are painful

perhaps i should just join the dark side

Appreciate your help and time , as always, Mega!

[frosty harbor]

I found that latency heavily affects how long the delay has to be before the uncrouching, how the hell is latency even affecting the animation or whatever is going on this makes no sense

If I use 0 latency, nothing happens, if I use 20 latency I can get away with this only happening if I instnatly release the crouch button, even holding it for 0.1s is enough for this to not happen, but if I use 100 latency I need 0.5s delay for the issue to not happen

For those not aware of what Im talking about, I'm having an issue where after uncrouching the camera moves up then down when I release it too fast, however, it only happens when there's latency and the higher the latency the longer I must hold "crouch" for for this issue to not happen.

A video showcasing the issue:
https://medal.tv/games/screen-capture/clips/lYhnVCnxuaGeV8MvE

[ashen plume]

It's not recommended to use delay nodes with CMC, the input predicting nature doesn't handle delays in its simulation

The crouch state is getting reconciled there

Crouch in bp flips the bwantstocrouch boolean, from bp the result gets handled next frame

Those factors with latency will have weird results since CMC has its own update frequency

[frosty harbor]

it was the only way I found to avoid the snapping at the end of the uncrouching when doing it quickly

this way it doesn't uncrouch right away

but yes this is a terrible solution

It doesn't even work as intended because 0.5s is for 100 latency, anyone above 100 latency will still have this issue

I want to find the root cause so I can try to fix it but I just cant find what could be causing the issue

For context, the video was recorded WITHOUT this delay system by the way, it was just simple crouch / uncrouch

[dark parcel]

@hoary spear is your ASC lives in player state?

When the pawn is possesed, client would have valid controller and player state.

From memory I think I handle input binding and init ability actor info via OnAcknowledgePosession for client and doesnt seems to ever fail.

[frosty harbor]

So I kinda confirmed that the issue is because it's doing the animation locally then does it again on the server

seems like this happens because hte crouch then uncrouching are BOTH happening before the crouch on server even begins

[viscid obsidian]

unreal specifically mentions this issue you can get if you rely on GAS to send your movement abilities and CMC to send moves separately. Does anyone know their recommended fix? is it to send the movement abilities through the CMC pipeline and manually replicate them server side?

[hoary spear]

Yes it does actually !

[dark parcel]

Hmmmm, I called InitAbilityActorInfo on OnRep_PlayerState. What is the race conidition on your part?

[hoary spear]

Well it seems ASC isnt always ready when the playerstate replicates

So my hotfix was a asc onrep function 😅

[dark parcel]

To init ASC, we call InitAbilityActor info right? that's what set the owner and other important bits.

[hoary spear]

Hmmm 🤔

Cant recall calling that one

Just woke up with the kids, so cant walk throigh teh code atm...

Apparently 2 year olds dont like coding, inherently 😅

Ill look into it once i got the chance tho!

[quasi tide]

Yes - that's the function you call

[dark parcel]

they will soon like roblox 🥹 .

ASC->InitAbilityActor is what essentially sets the Owner of the ASC (In your case the player state) and the Avatar Actor (The pawn you are controlling).

[hoary spear]

But... i cant call that if ASC is not replicated yet ?^^'

Nvm, its ptobably locally avaliable , being a subcomp and all

Just me not referencing it correctly at before the BindAScInput gets called

[hoary spear]

In my case it's called during InitPlayer, which in turn is called from OnRep_PlayerState

but this seems insufficient

It is also called during PossessedBy, for the server side 🤔

[dark parcel]

For server it's called in PossesedBy
For client it's called in OnRep_PlayerState.
🤔

what sort of race condition do you run into?

[hoary spear]

Seems the InputComponent isn't ready

this could be a unique in-editor playtest bug tho

almost as if OnRep_Playerstate is called before client is done possessing

[dark parcel]

void AAGPlayerCharacter::OnRep_PlayerState()
{    // ** Called On Clients ** //
    
    Super::OnRep_PlayerState();
    
    // ** Init Ability Actor Info for the client ** //
    InitAbilityActorInfo();

    // Bind gameplay tag event in Interaction component.

    
    if (IsLocallyControlled())
    {
        // Tell server to add startup abilities and abilities from data assets if locally controlled
        Server_AddCharacterStartupAbilities();
        Server_AddDefaultAbilitiesFromDataAssets();

        // Bind gameplay tag event in interaction component. (Client).
        InteractionComponent->BindGameplayTagEvent(GetAbilitySystemComponent());
    }
    //** Add Input mapping context **//
    SetupDefaultInputMappingContext();
}

The idea of calling Init actor would also be that the Avatar Actor is readily available.

Is the OnRep_PlayerState the ones in the character? I don't think I have issue binding input in the OnRep_PlayerState.

how can Input component not be valid 0o?

is the input component the Input subsystem, or something else?

[hoary spear]

Im confused too, but i guess the input component is created on client possession, which apparently can be after playerstate has replicated ?

But that doesnt make all that mich sense to me either

[hoary spear]

This is the fun part about replication order etc i guess 😅

[echo pasture]

so I'm new to GAS, and want to program this ability where if the input is pressed down, we charge a variable, and if the input is released, we uncharge that same variable. basically something where if you hold the key it'll go up but then if you don't hold it'll go back to zero.

does this look like the correct implementation, given that I'm too lazy to code a "wait for input" solution?

It feels like I'm maybe missing a better way to do it but idk

also for my inventory system which is currently client side, do ya'll think it's worth it to overhaul it to be server side before release

for context it is co-op and probably not gonna be huge so I'm not sure cheating will really a crazy concern, but still idk feels better to not use rpcs a lot just to share items between other players

[hoary spear]

If its already working, id consider just postponing it untill it becomes a problem

[dark parcel]

wonder if it's already tested with multiple clients.
If an item is dropped to the floor, like who gets it if multiple clients simultaneously grab the item? It would be like jesus performing the bread and fish miracle.

My inventory comp are server auth, but design is the biggest motivation, not cheating.

[hoary spear]

hehe yeah im not sure how they'd solve it

perhaps they circumvent it completely, idk

[frosty harbor]

If anyone can please have a look at a thread I opened, still havent been able to find a way to fix the issue >.< https://www.reddit.com/r/UnrealEngine5/comments/1ql9oyr/crouching_uncrouching_issue_in_multiplayer_with/

[lament flax]

you cant do much about that, especially in BP only

welcome to networking issues

[bright summit]

There should be a big warning when engine is launched for the first time saying that if you want multiplayer + custom movement you should learn how to do it in CPP 😛

[frosty harbor]

not an issue, I can do it in c++

I just have no clue what the fix would be

[lament flax]

mh, maybe write a "small" discord message explaining your issue so people here doesnt have to scroll on your big reddit post

[frosty harbor]

Well the issue is basically:
Due to latency, when a player crouches / uncrouches really fast, both of them happen on the client before the server even begins the initial crouching which ends up in causing a visual issue where the camera moves up & down after the uncrouching is finished already locally. However, if I have 0 latency I can spam crouch/uncrouch and it never happens.

If I crouch + hold crouch for a little bit then uncrouch it's also fine (because the server crouch already begun as well before the client uncrouch finished).

For a visual representation of the issue you can check this short clip, easier to understand:
https://medal.tv/games/screen-capture/clips/lYhnVCnxuaGeV8MvE

My camera is bound to the player head bone since I'm doing a true FPS with spine bending.

[bitter veldt]

couldn’t you use the waitforinput task that exists in GAS already?

[echo pasture]

Yeah I’m just lazy, it takes time to set that up

[viscid obsidian]

Is anyone familiar with the issue I'm talking about here?

It essentially boils down to the 2 different network pipelines for GAS vs CMC

[nocturne quail]

you have corrections using cmc?

[viscid obsidian]

I plan to, unless you think I should disable them

[nocturne quail]

if you will have corrections it means you have not implemented the movement correctly

[viscid obsidian]

well you can still get corrections even with perfect movement code due to dropped packets or ordering issues no?

[nocturne quail]

and you don't need to disable it

[nocturne quail]

you can't solve dropping packets issue, its not related to your code its a natural issue

but you do some client side predictions to make those corrections smooth

[hoary spear]

Sounds kinda like a delayed server correction?

[viscid obsidian]

right, i get that, but the issue i'm talking about here is that GAS uses a gas rpc to sync client ability activation with server ability actviation, whereas the CMC has its own servermove rpc. if you use them both separately, then you get jitter due to ordering issues like unreal points out, so I'm wondering what the recommend fix is. For example, not using the gas rpc and communicating solely through the severmove rpc.

[nocturne quail]

why you want to use them both on the first place? trying to sit in two boats at the same time?

[viscid obsidian]

why do i want to use both as in why do i want to use both CMC and GAS?

[nocturne quail]

both rpcs, you can do some refactoring to rely only on one rpc

[viscid obsidian]

yeah that's what i was thinking of doing, I was just kinda looking for some confirmation.

[nocturne quail]

yeah you don't need both rpcs to use for same reason, choose one that fits best for you

[viscid obsidian]

okay thank you

[frosty harbor]

Could be but not sure how I could ever fix this

[hoary spear]

Is it just the camera or everything?

video is down..

[frosty harbor]

just the camera, there's no changes visually when I watch from 2nd client

[hoary spear]

Seems i get the same thing with my raw implementation of "sprint"

it's very glitchy with the server corrections

[frosty harbor]

yea I have no clue how to fix something like this ngl. I thought about making some sort of queue where even if the user asks to uncrouch it will only process once server crouching sends a flag

but I'm not sure how to do that since you generally run these things from the owning client since its an input

[hoary spear]

depends if it reallly matters that much

1000ms isnt really realistic support in my eyes

[frosty harbor]

no no

that was an exaggerated test

even with 10ms it happens

the higher the latency the longer i must hold crouch for it to not happen essentially

so less latency = less likely to happen unless i instantly release CTRL

[hoary spear]

Guess you could block release untill server confirms

also depends on the game and the criticality of the implementation ofc

[frosty harbor]

I thought about blocking it yea but im not too sure how to have the server crouch add a flag whenever it begins

since its a locally predicted mechanic handled by the cmc

I tried turning crouching and uncrouching into gameplay abilities and crouching would add a loose gameplay tag ot actor when it had authority (thus server) and uncrouch would only work when actor had that tag

but the issue was that it wasn't working because the tag was only added on server so i guess the client wasn't picking it up

not too sure what the issue was anymore since i tested it a while ago

[hoary spear]

A mockup would just be a bool dictating the server state

and you'd only be allowed to change crouch state if the server state matches your 'predicted' state

[tough nimbus]

I use GAS in my project, but saw Mover and NPP recently and wanted to know if implementing them together is possible? From what i understand so far is that GAS and NPP have different prediction systems and there isnt really a bridge between them. I wonder if for my project this would be a problem since i use GAS mainly for the attribute system anyway and use only a few abilities like equipping items etc. No need for fancy root motion stuff since its a 2d game.

If someone with experience using the plugins could help me it would be nice, its my first multiplayer game and i am still learning and i wanna try the new plugins and see what they are capable of.

[nocturne quail]

just use your custom crouch logic, have an enum which have states, stand, crouch, prone and update this using server and replicate the change using onrep
than inside onrep notify animation class to crouch anim, set capsule height to half and optionally play a sound

on uncouch do reverse of it

you don't need to use built in crouch if its not working correctly

[frosty harbor]

its just a huge pain in the ass having to do a custom crouch because of all the height checks on terrain and stuff since our game will use that kinda stuff

[nocturne quail]

its not that hard, when you want to uncrouch, first do a line trace upward, if hits anything with a blocking tag just return else uncrouch

like you can give a tag to any lhing which can block crouch/uncrouch

[frosty harbor]

So in a way I sort of managed to fix it by doing this

It works although in very high ping there's still a bit of an issue with the crouchign itself as well

but at 100 ping its perfectly fine so thats good enough