multiplayer

[winged badger]

it has to have a name stable for networking, given that its not loaded from the level, or spawned from replication

and you also need to override IsNameStableForNetworking

[jolly siren]

Why do I need to override it?

And shouldn't I expect that log spam not to stop if it wasn't working?

[winged badger]

whats your use case here?

and from that code snippet, there is no way unreal can map server actor instances to client ones

FActorSpawnParameters SpawnInfo;
SpawnInfo.Name = SomeDeterministicNameThatsSameOnAllMachines;
SpawnInfo.bDeferrConstruction = true;

then you don't have to use that awkward function to spawn it either

[jolly siren]

Players can build custom maps. Some of the actors they build with don't need replication. For example, a static wall. These actors are all spawned at the beginning of the game. In order for the client/server to stay in sync I'm using net addressable for the ones spawned independently.

[winged badger]

you have more landmines in that field

what happens to an Actor with dynamic NetGUID if it becomes non relevant?

clients destroy it

and there is no code to spawn it back in when it enters relevancy range again, unless its replicated

so you also need to lie to the engine by setting bNetStartupActor and bNetLoadOnClient true

[jolly siren]

We aren't specifically dealing with maps large enough for that, but yeah that's a good point

[winged badger]

that will make the engine treat it as if it was loaded from the level

but in essence - if you want to network actors spawned separately on client and server (replicated or not) that has to be deterministic

names have to match, exactly

[jolly siren]

okay right, yeah that makes sense. I was thinking that was probably the issue

[winged badger]

i am not 100% sure that overriding IsNameStableForNetworking to return true is required

as i never tried not doing it

[jolly siren]

okay yeah it already looks at bForceNetAddressable

[winged badger]

our actors we spawn like that come from Prefabs, and they are all Tagged, so the function returns true if the tag is there

[jolly siren]

bool AActor::IsNameStableForNetworking() const
{
    return IsNetStartupActor() || HasAnyFlags( RF_ClassDefaultObject | RF_ArchetypeObject ) || bForceNetAddressable != 0;
}

In what cases should I not care if the server/client know the actor is the same?

I'm not sure if everything needs to be netaddressable

[winged badger]

when doing the setting bNetStartupActor and bNetLoadOnClient true additional problem is

server won't start replicating anything to a client before it loads the level

but it also assumes that client can Ack all static net Actor NetGUIDs when it does

so replicating Actors spawned like that before Client spawned them as well results in a NetAddressable Actor whose ActorChannel doesn't work for replication

[winged badger]

if you need to send a pointer to it over the network, or replicate it you care

otherwise you don't

[jolly siren]

Right, makes sense. I think a lot of these I don't even need to be net addressable, so I should filter those out for sure

[vagrant grail]

Anyone for my problem please ?

Does anyone know why my timer isn't perfect please ? Like My Starting_Night_Timer variable is at 23 (11PM) and Starting_Day_Timer is at 6 (6AM) and Day Duration variable is at 10 (IRL minutes) and Night Duration variable is at 5 (IRL min) and I used a timer in Windows to check if my timer is correct and it's not, like 10 min on the windows timer my game timer reached 22:50 (10:50 PM) instead of 23:00 (11 PM) so it's off by 10 sec in game somehow.

maybe the tickrate of the timer event isn't perfect ?
Any way to solve this and make it perfectly be at 11PM after 10 min ?

Here's my Code : https://blueprintue.com/blueprint/s0-g7rd9/

[winged badger]

your timer runs on server alone

and clients can't send a server RPC through GameState, as they don't own it

if they could, every client would call BeginPlay, then tell server to start the timer (again)

and server would start one on its own

this is not a network problem in any way, all code that actually executes here runs on single machine

to fix your network part, gate StartTimer behind HasAuthority switch

and remove runs on server from all your events

it won't fix your timer, but it will stop the output logs throwing warnings about no owning connection

[vagrant grail]

My timer is on Server but the variables Hours, Minutes and Days are replicated and I can get them in my HUD by getting the game state and getting those variables. And I'm doing this in the game state first using the begin play just for testing purpose, later I will call my StartTimer in my Game Mode as the server is the one handling the game rules.

And where do you see wrnings ?
So now I just need to understand why my timer is 10 sec off, what could be the reason ?

[winged badger]

i would never use timers for this to start with

i'd store the StartingGameTime on BeginPlay by getting GetWorldSeconds

then just have Tick check how much time has elapsed

and calculate day/night/current time from there

[vagrant grail]

Nope I can't do that, because the timer don't start when the world is spawned and the timer could be paused that's why I didn't choose to go that way

[winged badger]

you can track elapsed time also

Tick - if (NotPaused) { ElapsedTime += DeltaTime; }

also far more simple then multiple timers

and simpler equals more robust

[vagrant grail]

I need to convert that to a Day:Hours:Minutes format anyway

but I already spent 3 weeks on this timer, I need to move forward by just fixing the 10 sec missing

[winged badger]

i think the ElapsedTime approach implemented from scratch is faster then debugging this

also, a bit of a weird scaling there, 10 minutes for 17 hours of daylight and 5 minutes for 7 hours of night?

[vagrant grail]

but using a Tick() isn't as performant as using a Timer By Event I think

[winged badger]

what do you think engine does on Tick?

[vagrant grail]

That's on purpose

[winged badger]

checks timers if they should fire

😄

among other things

[vagrant grail]

Well I'm gonna stick for the timer I made for now but next time I need a Timer I will try that way, but I still need to understand why it's 10 sec off

[jolly siren]

It looks like CMC relies pretty heavily on the client/server understanding that level actors are the same between the two. Getting a lot of floating bots when I don't set them as net addressable. So I'm going to have to figure out how do do this properly. The main issue is that when a player first joins before it spawns these actors I get that actorchannel spam.

[winged badger]

well, if your Base component is not net addressable

you're fucked

(what CMC calls the component you're standing on)

[jolly siren]

right, yeah that's what I thought

So I was right when I thought these all needed to be net addressable

So now to figure out how to fix the issue on first join before the actors have spawned clientside

[winged badger]

PostSeamlessTravel doesn't start players whose PCs haven't reported they are done spawning stuff

but player reporting its done with the spawning makes a check if it should start them up

alternatively you can override PlayerReadyToStart

[jolly siren]

You mean HasClientLoadedCurrentWorld?

[winged badger]

no

that happens immediately

client loads world

[jolly siren]

I don't see a PlayerReadyToStart

[winged badger]

calls NotifyWorldLoaded then ServerNotifyWorldLoaded

at this point, if the server already loaded, it will go start the player up

if not, AGameMode::PostSeamlessTravel will start all controllers waiting

    UFUNCTION(BlueprintNativeEvent)
    bool PlayerCanStart(APlayerController* NewPlayer);```

that one

[jolly siren]

ah okay

hm I don't have that either

[winged badger]

but if the clients need to spawn a floor they are about to start on

you can't use normal engine startup

you have to delay it until the floor has been spawned

[jolly siren]

Is that function custom to your engine?

[winged badger]

and if each machine spawns their own, then clients need to RPC the fact they are done doing that to server

before server can safely start them

maybe

https://gyazo.com/755a756e607528eb403779ccd09ffd2d

but basically, thats the use for it

[jolly siren]

ah okay, thank you

What does your PlayerCanStart body look like?

Just curious

[winged badger]

you can't spawn pawns until PlayerStarts and what they are standing on are already there

[jolly siren]

This just handles spawning the pawn though, I think I would still get these uchannel replication errors without a pawn being involved though

I can test to make sure, but pretty sure this is specific to the net addressable actors

And nothing to do with the pawn

[twin juniper]

Dear community, so we are quite new to UE and were wondering what the "scope" of the game instance is

so if you think of the server handling round time in a multiplayer game

where would that be handled / stored on the server?

e.g server updates all clients round time / end of round / restart round etc

a point in the right direction would be enough 🙂

[fathom aspen]

Game State is the answer. It exists on both server and clients and it's the class for these kind of behaviors(its name speaks for itself)

[twin juniper]

I assume "Game State" is right

lol yea

@fathom aspen would you have a moment to chat about that ? 5 minutes should be enough

[fathom aspen]

Yes feel free man

I got a question to the community though: If I have a pointer variable to a repliacted property, do I need to make it replicated too in order to preserve repliaction of it too?

[peak sentinel]

Object needs to be replicated

[fathom aspen]

Actors?

[peak sentinel]

Actors are objects, so yeah

[frail crystal]

Do i have to do networking on every blueprint system for multiplayer?

[fathom aspen]

Oh yeah, I keep forgetting this, thanks!

[peak sentinel]

@fathom aspen Check this out: https://cedric-neukirchen.net/Downloads/Compendium/UE4_Network_Compendium_by_Cedric_eXi_Neukirchen.pdf

[fathom aspen]

Yeah I've read it all, just wanted to make sure if pointers to replicated properties need to be replicated too

[twin juniper]

thanks @peak sentinel !

[tranquil eagle]

I've read that for some games, RTS for example, it is important that floating point operations are calculated deterministically. But why without additional effort, they are not the same for different hardware architectures or for different compilers? Is it because some compiler might produce slightly different assembly and different hardware might have registers of different sizes? If that is the case, why this problem is not also with operations on integers?

Or, for example, if I made a custom MyFloat class, that behaves just as normal floats with mantissa and exponent kept as raw bits in some raw uint8[] buffer, and implemented some maths functions with it, like arithmetic, trigonometric functions, etc., would that be deterministic?

[vagrant grail]

Intax do you have an answer for my question above please ?

[twin juniper]

Is it possible to take the Shooter game demo's lobby menu and transfer it over to mine? Since I really don't feel like designing a lobby menu from scratch.

I just can't seem to find the widgets for the menu

[lean surge]

Hi everyone here, so I have a variable that is set to replicate using a function. Code below:

UPROPERTY(ReplicatedUsing = OnRep_CurrentHealth)
float CurrentHealth;```

UFUNCTION()
void OnRep_CurrentHealth();```
Inside my GetLifetimeReplicatedProps():

DOREPLIFETIME(ATestEnemy, CurrentHealth);

So a function which is being called on a listen server that decreases theCurrentHealth variable, yet the OnRep_CurrentHealth function is not being called on any device, not clients nor the server itself. Actor is replicated.

Anyone know what I'm doing wrong?

[tawny nova]

are you certain that all of that code is getting run?

[lean surge]

I am certain, I have used logs to check and the function that decreases current health is being called

[tawny nova]

are you one episode further into the tutorial i'm doing? 😄

[fathom aspen]

Are you calling OnRep function on server manually?

The server has to call the OnRep function manually because it does not participate in replication updates since it is the one sending them.

[twin juniper]

OnRep are not called on server in c++ 😄

[lean surge]

ah thanks, but if I understand correctly (I couldn't find a definitive answer just by searching the exact term quickly) are OnReps called when the values are changed on the server side (no communication, I mean just literally changing the value) or is there somehing else to it?

[twin juniper]

OnRep are called "On Replicated" so when the variable replicated on the client

so yeah basically when the server changes it

to another value.

[lean surge]

ok yeah, thought so thanks but it isn't being called on any of the clients on change, so am I missing something?

[twin juniper]

show code, how do you set your variable, are you sure that you set it on the server etc

[lean surge]

so I have a weapon script that sends a server RPC as such:

  Server_ApplyDamage(OtherActor, AttackDamageArray[CurrentAttackStage - 1]);

The server then receives this, and uses the built in damage system in unreal to call the ApplyDamage function on the other actor:

void USwordComponent::Server_ApplyDamage_Implementation(AActor* ActorToDamage, const float& Damage)
{
    UE_LOG(LogTemp, Warning, TEXT("Damage applied")); // Being called
    
    UGameplayStatics::ApplyDamage(ActorToDamage, Damage, Cast<APlayerController>(GetOwner()), GetOwner(),                  UDamageType::StaticClass());
}

THEN the other actor receives this (in this case the enemy) and does this function:

float ATestEnemy::TakeDamage(float DamageAmount, FDamageEvent const& DamageEvent, AController* EventInstigator,
AActor* DamageCauser)
{
    UE_LOG(LogTemp, Warning, TEXT("Damage received")); // Being called
    CurrentHealth -= DamageAmount;
    
    return DamageAmount;
}

oh and the server RPC is declared as shown:

UFUNCTION(Server, Reliable, WithValidation)
void Server_ApplyDamage(AActor* ActorToDamage, const float& Damage);
void Server_ApplyDamage_Implementation(AActor* ActorToDamage, const float& Damage);
bool Server_ApplyDamage_Validate(AActor* ActorToDamage, const float& Damage);

And basically the OnRep isn't being called

anyone know a solution?

definetely certain

[lost inlet]

how certain? show ATestEnemy constructor

though if it derives from APawn or ACharacter, it'll be replicated

lol I always forget about HR

[lean surge]

ATestEnemy::ATestEnemy()
{
     // Set this character to call Tick() every frame.  You can turn this off to improve performance if         you don't need it.
    PrimaryActorTick.bCanEverTick = true;

    SetReplicates(true);
}```

Constructor here

oh right I also actually checked in editor though

let me try again rq

[lost inlet]

and the hot reload thing?

[lean surge]

using live coding

[lost inlet]

well you can't make ctor changes with live coding

so I would suggest closing down the editor, compiling, and trying again

[lean surge]

ok thanks everyone, absolutely NO IDEA what I was thinking when I assumed that it automatically compiles .h files in live coding lmao, got it working

Just one more question, anyone know if you have to explicitly call OnRep functions on listen servers or are they called automatically (since it's a weird blend between client and server)?

[lost inlet]

yes because it's a server

[lean surge]

ok thanks!

[twin juniper]

In a NetMulticast RPC should I check for HasAuthority() I'm aware that it can only be replicated from the server and if called on the client it will only perform on itself.

[meager spade]

unless BP

[twin juniper]

But if I don't check HasAuthority does that mean its getting called on the client twice?

[meager spade]

BP is dumb and calls the RepNotify by default (only in BP)

why would it get called on the client twice?

only server should call that function

if a client is calling a multicast function, you have a design flaw

[twin juniper]

Well the way I'm doing it is, client presses key to shoot, server receives it, server performs spawning of projectile but when it comes to animation and sound, once I'm in the function which spawns the projectile,

That's where I'm unsure if HasAuthority needs to be called

For the multicast

[meager spade]

you just need to call the multicast on the server to play the animation

if you want to predict it, you need to do some checks

[twin juniper]

Oh my mind went black for a moment, okay it makes sense now. I'm doing it correctly.

[verbal tendon]

@twin juniper if the projectile isn't really going to be travelling through the world you dont need to spawn it

You can just do a trace and do the visuals on the client

[copper portal]

Hello! Im having a little trouble with some weapons.

I have a BP spawn a couple weapons at beginplay, and then I have a control panel that players can access to use those weapons. But I'm having trouble figuring out where I'm replicating things incorrectly?

The BP comes with a camera to change the players perspective to the weapons, and then I was using the camera's rotation to determine the weapons rotation.

Two things happen. 1: If I replicate the weapons, it spawns two of them in the same spot and the firing shoots from each gun's barrel. It also gives me an error that set actor rotation isn't finding an actor (a gun) to rotate when I call it with a SRV->Multicast RPC. 2: If I don't replicate the weapons, things seem to rotate and shoot fine from a client's perspective, but there's no sound playing and I can tell things are only happening on the server.

What am I replicating incorrectly??

[leaden scaffold]

you replicating rotation on a multicast RPC on tick??

[copper portal]

yes

[leaden scaffold]

It seems like you have a good grasp of the fundamentals since you put this together, but I would approach it quite differently. I never replicated movement values through RPCs since UE can automatically replicate movement and physics values of actors.

Depending on the actor type you'll want to review the Replication section in class defaults. Ensure you have ReplicateMovement checked. Depending on the owner of the actor you'll want to review the Net Use Owner Relevancy setting

Multicast RPC on tick will only lead to world wide armageddon

[copper portal]

It used to be on rotation tick, but I changed it when that wasnt working.

but both the control panel class and the Weapons master class are set to replicate and to replicate movement

[spark stag]

how can i tell when a newly connected client has finished syncing state from the server? i want to delay the connected players pawn from spawning until that happens, and put up a loading screen

[dark edge]

Or just do what I do, and have a projectile system that covers all the bases. And my system, the only difference between a really slow rocket and an instant laser is velocity

[karmic jacinth]

If I fire an RPC from the client to update a replaced variable, will it try to override the value on the clientside? I'm wondering if I should be using two different variables for that otherwise (a separate client variable and replicated variable).

[pulsar river]

is it possible to spawn an actor on the server from the client side

[leaden scaffold]

if you perform an RPC from the client to server

[pulsar river]

okay sorry to be more specific...
ActorA get spawned on the server and binds keypress i to everyone to spawn ActorB
however; when I try to press "i" it would only spawn on the client side. and the server doesn't ( server's not reading the input i on the client )
how should i stucture it so that pressing "i" on the client spawns it for the server/otherclients (other client part is easy just getting it read on the server is the hard part)
Can't RPC because I don't own ActorA

[hollow eagle]

then RPC through something you do own

[verbal tendon]

then you need an RPC on an actor that you own

and it tells the server to spawn the actorB 🙂

[pulsar river]

ah... so there really is no way besides having to RPC through another actor...

okay gotcha thankyou frds

[sinful tree]

If you RPC to the server and change a replicated variable, yes, that new value will override the client's version of the variable.

[sinful tree]

Depends on what you mean by "syncing state" as that's a bit too ambiguous. Syncing the player state? Syncing your own external data on the player from your databases? Syncing the entire procedurally generated world you've spawned?

[spark stag]

yeah, fair 😉 we have a lot of world state that, after the world has been running for 30-45 minutes spawning a bunch of stuff, takes 10-20 seconds to finish syncing after the player connects. i don't really understand how to detect when the client has "caught up" -- right now i can do it by looking at stat game and waiting for time spent on replication to dwindle

the visible effect is that a lot of player related stuff (the pawn itself, the hotbar) take a while to receive their state

i'm not even sure how to detect that this player-related stuff has received state from the server -- for example, how do i tell the difference between when the hotbar is uninitialized or just empty?

[twin juniper]

Can each level have its own level state? Or do I have to rely on a single game state to manage all levels?

[shy aspen]

yes, you set the game state to use for each level within the game mode

each level can have its own game mode and the game mode defines which pawn, controller, player state, game state, etc. to use

[twin juniper]

Oh thank you,

Where does the game mode define which player state it uses?

[shy aspen]

in the class defaults for the game mode

[twin juniper]

Awesome!

Thanks

[shy aspen]

np 🙂 just don't forget to let your level know which game mode it needs to use in the World Settings

[twin juniper]

Yah I realised you can set that in the Gamemode Override

Thank you so much

[pastel heath]

I have a particle I am trying to play on clients that is an idle particle. I've set it up as a particle component on the actor. What is the best way to have this particle play on clients since it really isn't "activated?"

[chrome zealot]

I'll ask it here

So a question about plyer tags

Say I have 2 teams each with the same character. What would be the best way to notify one is an ene my and one is a teammate or yourself

Basically like should I give the characer an enemy and a player tag or

[pastel heath]

You might consider using the built in Team functionality if this is AI or will ever be AI - like bots. I haven't tried extending the team idea to a PvP game, but you could probably do it. https://denisrizov.com/2017/05/08/ai-perception-in-unreal-engine-4-how-to-setup/

[chrome zealot]

For sur

I'm just tryingn to figure out the basis of basic attacks and moving into range to attack

Someone gave the idea of player tags and that madr sense so I was gonna go in that route then realized it could be an issue in the future, also I'm using my current player as a test dummy

[shy aspen]

does anyone know if there's a setting that controls how long an actor continues to replicate after it's no longer relevant?

Or how to go about forcing replication to stop? When the NetUpdateFrequency is lowered, the "cleanup" of a replicated actor is also delayed by quite a bit.. I've been looking at DataChannel, DataReplication, NetDriver, etc. to no avail.. just curious if someone could point me in the right direction

[hollow portal]

ok so im trying to do a line trace in multiplayer
but for some reason i cannot get player 2's camera pitch but i can get the yaw
i can get player 1's camera pitch and yaw just fine

[leaden scaffold]

You can replicate it from the player controller

I believe it gets replicated by default if you have the pawn inherit the player controller's rotation

but likely player 1 is the server in this instance

so looks fine there, but bad data for remote client

Can also set bUserClientSideCameraUpdates in the camera manager

https://docs.unrealengine.com/4.27/en-US/API/Runtime/Engine/GameFramework/APlayerController/ServerUpdateCamera/

[shy aspen]

you can use Get Base Aim Rotation - its replicated already

[hollow portal]

i managed to fix the camera issue
but i just tested a packaged game and ive got a new issue
sprinting feels fine on the host but really weird on the client

[shy aspen]

you need to set the sprint speed on the calling client as well

[eternal canyon]

Input doesn’t run on the server

And walk speed needs to be set on the server

[shy aspen]

Unless the player is the host - so I guess we need more info about how you plan on playing your game 😊

[vague fractal]

From all what i've read now it's not possible to make solid sprint system in Blueprints only as you need to use the client-side prediction(Hope it was that xD) where you only have access to in C++

[spark owl]

I'm having a problem with begin overlap on a box collision in an actor. When an object is moving at high speed towards this collision overlap and through how the gameplay works for my game, the ball gets redirected into another direction, but sometimes it will trigger the overlap event even though it really didn't go into the area where it should overlap

it's a football/soccer game, the issue is that on penalty kicks sometimes the gk will save the ball before it triggers the goal overlap event even tho it doesn't go in the net

this is a clip of it happening

so basically I need a super precise way to detect if a ball crosses the line without errors like this in multiplayer. does anyone have a recommendation for this?

[hoary spear]

Check its location

[spark owl]

I tried that before had the same effect

didn't solve the problem

[hoary spear]

Then you did it wrong :p

[spark owl]

ok so how would you do it then?

[hoary spear]

Checking if the vector is within a bounding box , being the goal, with a given tolerance

[spark owl]

what exactly do you mean by bounding box?

[hoary spear]

Cant recall the actual rules of it, its a goal if half the ball is inside ?

[spark owl]

well the technical rules for football are the whole ball has to cross the line

I had it setup like this

[hoary spear]

Right so

[spark owl]

but then as a scuffed fix for this issue I moved the goal hitboxes back some

but this issue still happens

[hoary spear]

ball radius is the tolerance, and the location to check is the balls center vector

Not sure about the exact setup for multiplayer, but what i'd do is on overlap start tracking the ball, and if its vector is within the defined bounds of the goal, then it is a goal

Otherwise its not a goal

[spark owl]

ok so I did this before, but I simply had it check the balls location, see if it's greater than the goal line on X axis and also within the height and width of the goal

[hoary spear]

Cheesing it by reducing the boundingbox by the ball diameter probably works aswell

[spark owl]

is a bounding box different from this?

[hoary spear]

In this context, nah that sounds fine

[spark owl]

but that's what I did before, and the issue still persisted

[hoary spear]

It must be greater than goal line + radius ,

[spark owl]

yeah

but look at the clip

[hoary spear]

Or goal line must be moved the equivalent of the radius

[spark owl]

it didn't even get the the line

I only happens on really high velocity impulses

[hoary spear]

Clip is very far away, and with no collision volume showing..

Not sure what you expect me to be able to see there :-/

I see the ball, the goalkeeper, and the ball being blocked

[spark owl]

I'm wondering if maybe this is an issue with the balls velocity tho

[hoary spear]

You should do some debug setup with a closeup from the side etc

To get a closer look

Showing collision volume, ball vector

It does look like the keeper is inside but hard to tell

[spark owl]

he's on the line but the ball is still not even on the line

[hoary spear]

Try showing it from a different angle

[spark owl]

i don't really have any other clips of it

it happens really rarely

penalties don't happen super often

and even less often the keeper actually saves it

so it's hard to get lots of clips of this

[hoary spear]

Isnt this a project of yours?

Im confused 😅

[spark owl]

yes it is, i'm saying it's not easy to get lots of clips of this happening

since it's a multiplayer game

[hoary spear]

Like, surely you can force the scenario?

In a dummy match vs dummy ai?

Trigger penalty manually?

[spark owl]

I suppose yeah, i'm not sure what info it would give me tho

I would just see that it's not crossing the line but still triggering a goal

[hoary spear]

Youd check your codr

Code

The vector comparator

Their values

Showing the collision volume etc

[spark owl]

well I had done it the way you described before, by checking location, but it had the same issue still

[hoary spear]

So the vectors are wrong

Or the math is wrong

Method should be fine

[spark owl]

what do you mean the vectors? the balls location?

[hoary spear]

Yeah,

Ball location vs goal vector check

[spark owl]

I mean yeah I tried this before but it didn't change anything, I actually was running this on tick

idk if that's a good idea or not, but yeah had same effect as overlap

[hoary spear]

Youd need to check it each frame ,

Or somehow with substeps if this is physics based

Not sure how that'd be done

[spark owl]

it is physics based, it almost seems like to me the overlap is getting triggered 1 frame ahead of when it should

[hoary spear]

Overlap is in wrong tick group?

[spark owl]

like it calculates the ball will be overlapping next frame and triggers the overlap, but then also the gk perries it away s

[hoary spear]

Pre physics vs post ?

[spark owl]

I haven't touched either of those

[hoary spear]

Perhaps its worth trying

Or not idk

On overlap -> enable tick -> check balls location ->

The tick can only capture what is right then and there during that tick, and not what happened during substep afaik

[spark owl]

hmm ok

i'll give that a try

how does post physics work?

[hoary spear]

Not really sure tbh

Im not that into #legacy-physics

Id suggest asking there about the issue if this doesnt help

Sounds like a physics issue anyways, very little to do with multiplayer as for your issue

[spark owl]

well idk it seems like it could be network related

but i guess networking physics

so yeah

thanks for the info

[rocky topaz]

is the player camera manager replicated or not?

correct = it's not?

I just checked it seems like it isn't but idk 🤔

[vivid seal]

So I’ve implemented server-side rewinding of hitboxes to do lag compensation, and it works pretty well, except that the client has to explicitly send their aim vector (camera location and forward vector) and gun socket location (third person shooter so I aim from camera but check line of sight from the gun). If I just use the server’s version of the camera/gun socket, I miss pretty consistently. I THINK this might be because CMC updates rotation and position after my shoot RPC has already processed, but I’m not sure. Anyone know if it’s possible to avoid having to send this?

[pastel heath]

I have a particle I am trying to play on clients that is an idle particle. I've set it up as a particle component on the actor that should just play when the actor is in the world. What is the best way to have this particle play on clients since it really isn't "activated?"

[grand kestrel]

The aim vector should already be sent, see ACharacter::GetBaseAimRotation()

I process firing inputs after CMC PerformMovement (IIRC) so that it completes any strafing frame before firing otherwise it feels a frame behind, wonder if that would solve your problem too

[vivid seal]

How do you delegate the firing after perform movement? Are you sending firing info as part of saved moves?

[toxic flame]

hey all, is there a reason some people use add each player to a gamemode, then send events to all the connected clients there vs. using multicast?

[vague fractal]

I guess cuz of the reason on how multicast's work. As in you may not receive it due to being out of range or similar

[toxic flame]

i see, I had a feeling something like that may be the case

might have to rework a few things to change up the multicasts

[vague fractal]

I think the deal with multicasts was anyways something like "Don't use it when you can" and "Use it only for cosmetic things and not for gameplay related things"

[toxic flame]

right

i'll keep that in mind i guess

[dark edge]

@spark owl back up the trigger. Make it so any overlap or hit = goal

[spark owl]

lol bruh you clearly didn't read everything I wrote

[plucky prawn]

if i have a replicated variable and i want to set it from the client and replicate it to the server, i need a replicated function to do this right? afaik a client changing a variable will not replicate to the server

[hollow eagle]

correct

you need a server rpc

[plucky prawn]

is there a better way of doing this? i need the client to set the value immediately because its used for UI but the server will "override" it anyway. seems dumb but i dont believe theres another way?

void ASBPlayerController::ServerSetReadyStatus_Implementation(EPlayerReadyStatus readyStatus)
{
    ReadyStatus = readyStatus;
}

void ASBPlayerController::ClientSetReadyStatus_Implementation(EPlayerReadyStatus readyStatus)
{
    ReadyStatus = readyStatus;
    ServerSetReadyStatus(readyStatus);
}```

[dark edge]

Just use the replicated value. Who cares if you gotta wait ping milliseconds to see the checkbox. Or use skip owner on the variable.

[plucky prawn]

I don't think SkipOwner is appropriate here since only the server and the player it belongs to know about this variable. No other players need to know. Since this is used for ui though I need it to be instant so I think my solution is.. Adequate

[dark edge]

In some games you can see the ready status of other players, depends on what you want to do though.

[cerulean juniper]

Hey everyone! Quick question.
Why do I need to replicate a "Character Movement > Add Force" But not a "Add Movement Input" Function?

[peak sentinel]

You dont have to replicate Add Force either, but to answer this question properly you should understand how movement is being replicate and how CMC works with that system.

• Actors has a FRepMovement struct

• If bReplicateMovement enabled, with that struct, engine will send location, rotation and velocity across the network

• When server updates those values, clients fetch them via OnRep and set transform of the actor. This way, movement is being 'replicated'.

• When you apply input client moves first then CMC sends the velocity and some other infos to server, then server applies this info and clients (simulated proxies) fetch that via OnRep as explained above. But CMC has prediction out of the box which allows controlled client to move first then send its position to server.
  TL;DR: Client acts before server to move so calculations and replication events happen inside the CMC, Add Movement Input just singals to movement component to make this happen.

Force and impulse also calculated on server but if you are running it on owning client you'll have troubles

[twin juniper]

Does anyone know what an average replay file size would be? Using demonetdriver

And the built-in replay system

I ask because I know that you could theoretically store them online, so using aws I was trying to do some napkin calculations as to how much storage and data transfer it would require

My multiplayer game has 5 players per match and a couple actors in the match that you would have to record in the replay system

[chrome zealot]

question, can i test multiplayer without setting up servers?

[leaden scaffold]

yes

[chrome zealot]

okx.x

oh wait

[leaden scaffold]

https://docs.unrealengine.com/4.27/en-US/InteractiveExperiences/Networking/TestMultiplayer/

[chrome zealot]

i feel like its working but i cant see them?

yeah

it is

so simple but thank you sam :)

[chrome zealot]

just mad at myself for not just googling it x.x

[leaden scaffold]

knowing that to google is half the battle

[chrome zealot]

half of it

[hollow portal]

ok so im having this issue where the host has the correct gamemode but all the other players have no gamemode

[hollow eagle]

that's normal

gamemodes exist only on the server

[hollow portal]

how would i fix this ?

[hollow eagle]

...don't try to access the gamemode from the client?

You haven't mentioned anything that needs fixing.

If there's data that belongs to the gamemode that needs to be replicated to all clients, put it in the gamestate instead.

And maybe read through the network compendium (second pin down in this channel). It explains all of this.

There's a few charts on what types of actors are accessible from clients/servers.

[pallid canyon]

100 times this @hollow portal, getting that understanding is pretty vital

[hollow portal]

i see thanks ill take a look into it

[leaden scaffold]

Compedium is bible

[hollow portal]

result of my 10/10 multiplayer bp coding skills

[leaden scaffold]

nailed it

[chrome zealot]

looks like it works

[floral bison]

in a survival game situation when saving a players data and location is it better to have a single savegame with all the players or a savegame for each player? assuming your server would have 100-200 players

[cosmic plank]

If I have an override Server RPC how can i call Super::MyFunctionOnServer() ? Seems to loop endlessly when I try to call just that.

[shell sorrel]

I would like a dedicated server to host multiple instances/maps on demand and at the same time. I was trying advanced-session-plugin but figured out an instance of UE has only one session. The other option is to run multiple UE instances.. but then you need an orchestrator ? Any hints/urls for me please ?

[silent valley]

If your player pawns remain in the game when they are disconnected (eg sleeping) then might as well store in one savegame.

Otherwise it can be advantageous to separate them to keep savetimes down.

[rocky night]

HI there i trz to make that only the owner player see a spotlight /flashlight/ and not other players

Is there a workaround?

[gleaming kite]

I could be missing your question but just handle turning off and on the flashlight on the client, not the server. If you really want to use an authoritative model there are replication conditions in the editor you can set for owner only.

[rocky night]

yes i have the thing that i have one fps arms and a tpp body, so the fs on arms has a spotlight which i need to hide from other players

[buoyant wedge]

Hello. Is it possible to limit the advantage the host has in peer to peer games?

[twin juniper]

Is PostLogin ever called for the Listen Server's PlayerController?

Or is it only called when other clients connect?

[sinful tree]

It's called for the host as well.

[twin juniper]

Strange because I"m trying to set actor->pawn location after post login

But it doesn't seem to apply to the host

Incoming->GetPawn()->SetActorLocation(FVector(TileLocation.X,TileLocation.Y,400),false,&result,ETeleportType::TeleportPhysics);

This is after I call Super::PostLogin

Is it possible that the player controller doesn't have an owned pawn when PostLogin is called?

Because it just spawns where ever PlayerStart is on the level map; however, if I remove Player Start then it starts at a random location

And it isn't set in PostLogin

[sinful tree]

Pretty sure their pawn is spawned in HandleStartingNewPlayer which happens after OnPostLogin.

[twin juniper]

Oooh

I need to do the location setting in HandleStartingNewPlayer

[sinful tree]

void AGameMode::HandleStartingNewPlayer_Implementation(APlayerController* NewPlayer)
{
    // If players should start as spectators, leave them in the spectator state
    if (!bStartPlayersAsSpectators && !MustSpectate(NewPlayer))
    {
        // If match is in progress, start the player
        if (IsMatchInProgress() && PlayerCanRestart(NewPlayer))
        {
            RestartPlayer(NewPlayer);
//.....

[twin juniper]

Ahh thank you

[fleet viper]

Im using this OnRep function of my replicated health variable to call my death function. Problem is that none of the death functions will be called because not even the Server has the authority?? And when it doesnt then it should be able to call a server sided function right?

{
    UE_LOG(LogTemp, Warning, TEXT("UpdateHealth has been called"));
    if (Health < 0.f)
    {
        if (HasAuthority() == true)
        {
            Death();     <-- NetMulticast
        }
        else
        {
            UE_LOG(LogTemp, Warning, TEXT("OnRep_Health has no Authority!"));
            DeathServer();   <-- ServerSide
        }
    }
}```

[twin juniper]

HasAuthority will succeed on the server and proceed to calling your Death function which performs a NetMulticast RPC

Where as if it's a client, it will call DeathServer with a Server RPC

The Server RPC should then multicast the death

So your DeathServer function should perform Death();

[fleet viper]

it does

but still none of the two functions are being called

the server side one wont be called because it has no owning connection, the output log says @twin juniper

[rose egret]

how do I simulate real network by network simulation config or by clumsy ?

my problem is the lag is just initial and I receive further packets immediately

therer seems to be no network saturation

[cerulean juniper]

Yeah! I'm facing this problem. So, is it OK to call a RPC to Add Force? (It is being called by a Move Forward/Right Axis Input for a custom movement)

[cerulean juniper]

And Awesome explanation by the way!! Thank you so much

[twilit radish]

You could for example use the editor settings to simulate network issues. If you click on the little arrow next to "play" you can select advanced settings and in there are some settings you can mess around with 🙂

[rose egret]

this is the problem I have, lag is 1 seconds I have initial delay but receive the packets repeatedly without 1 sec delay, the green circles are replicated locations

isn't it supposed to be one circle draw every seconds

[twilit radish]

If you set the network simulation latency to 1 second then it still sends the packets at the same interval but they are received with a delay, it doesn't send 1 packet per second. So no your circles would not be drawn once a second.

[peak sentinel]

(It is being called by a Move Forward/Right Axis Input for a custom movement
You are calling 'add force' on movement inputs?

Don't do that

You need to implement movement logic into CMC with C++: https://youtu.be/RtQRMcupJs0

[hexed thunder]

Anyone have experience with an issue with child components on certain actors not loading in and sometimes even child actors (this doesn't happen every time) seem to be an issue when loading in a map/gamemode that i am waiting for all players to load before allowing gamemode readytostart?

[twin juniper]

Does it matter if whether GameMode calls HandleBeginPlay on the gamestate or not?

Does GameMode call it automatically or do I have to call it?

[latent heart]

I expect it matters quite a lot. Check where it's called in the code.

[vivid seal]

Hey if you didn’t solve this yet, in c++ OnRep functions aren’t called on the server automatically (in blueprints they are). You can manually call them on the server when you set the health variable though.

[fleet viper]

can you give an example?

[dark edge]

@fleet viper death should be a state probably and just use a repnotify on it

Just call the function manually on server when death happens

[royal vault]

how do i check if a actor is network owned or not

[latent heart]

Check it's role?

[royal vault]

so do i check

if(GetLocalRole() == ROLE_SimulatedProxy)

[latent heart]

Something like that. Or check it's remote role, if it doesn't have one, it's a local actor. I forget the specifics, I haven't worked on mp games in like 10 years.

[royal vault]

alr ty

[kindred widget]

What do you mean by network owned? As in spawned on the server and replicated?

[royal vault]

like anything that has a controller

anything that can fire a server rpc

i want to check if something is able to fire a server rpc

[kindred widget]

If it's a pawn you can just call IsLocallyControlled.

If it's something else, you can run the owner chain til you find a controller and check if IsLocalController

[royal vault]

does this return true if I check a pawn controlled by another client?

[latent heart]

Unless it's not owned by a player?

[kindred widget]

It can only return true on a locally owned pawn. It gets the pawn's controller and if it exists, it calls IsLocalController.

If it's something not owned by a player, then it won't have it's owner variable set to a controller. And if it's on a listenserver, there's only one controller that will return true on IsLocalController. And on clients, other controllers don't exist.

[latent heart]

Seems a very roundabout way of checking for a network spawned actor.

[kindred widget]

He's not checking for a networked spawned actor, he wants to know if it can RPC.

If he was just checking if it was networked spawned, you can check if (HasAuthority() && GetNetMode() != NM_Client)

[latent heart]

True enough. I forgot abotu that requirement!

[royal vault]

    bool CanServerRPC = false;
    //ensure becuase this is a component, so if we don't have an owner that would be... what
    if (ensure(GetOwner())) {

        //check if it is a pawn and locally controlled, only then will we server rpc
        APawn* PawnOwner = Cast<APawn>(GetOwner());
        if (PawnOwner && PawnOwner->IsLocallyControlled()) {
            CanServerRPC = true;
        }
        //check if we have a second owner
        else if (GetOwner()->GetOwner()) {
            APawn* PawnOwner2 = Cast<APawn>(GetOwner()->GetOwner());
            if (PawnOwner2 && PawnOwner2->IsLocallyControlled()) {
                CanServerRPC = true;
            }        
        }

    }
    
    if (CanServerRPC) {
        GEngine->AddOnScreenDebugMessage(-1, 3.0f, FColor::Yellow, TEXT("Yes"));
        
    }
    else {
        GEngine->AddOnScreenDebugMessage(-1, 3.0f, FColor::Yellow, TEXT("No"));
    }

would this work?

[dark edge]

Can't you just check if getowner is local player?

[royal vault]

yea but if the component is a child of a weapon which is a child of a player i need to do it twice

[nimble widget]

Trying to find a solution and haven’t been lucky.

Basically I want to rotate my character whenever the camera moves over multiplayer. Can someone show me how to use Set Control Rotation in blueprints?

I’m using a VR character and have been stuck on this for 2 weeks with 0 answers in the forum.

[pallid canyon]

I manually replicate the HMD position on tick and set the other clients rotation and position of mesh accordingly.

Because you're using the camera it will never actually update the ControlRotation (you can think of it like control rotation is bound to a joystick)

[cerulean juniper]

I thought about it, but, as I have never coded in C++ before, I was kinda running of it
lol
I'll watch this video. Thank you for the suggestion!

[pallid canyon]

and if you dont update that and then apply movement input it will ingest movement without any kind of rotation, effectively snapping your mesh back

[nimble widget]

I tried that with server and multicast RPCs. Components replicated too.

The server is locking the rotation when I tried that with relative/world rotation of the HMD/camera/mesh etc

I saw somewhere I need to use Set Control Rotation because I use the movement component with Use Controller Rotation Yaw.

[pallid canyon]

You have to set world rotation of the mesh, I don't use control rotation at all because of the snapping to whatever rotation it has stored independent of mesh rotation

my forward movement is based on the camera forward vector

[nimble widget]

I did. It didn’t work

[pallid canyon]

right, but again, you can't use control rotation if you're setting mesh rotation manually

like at all, need to deselect the option

[nimble widget]

So I should turn off controller yaw?

Can you take a look at this. Here is my real post with screenshots. The mesh interpolates to the HMD rotation. Which I also tried replicating with Set World Transform.

https://forums.unrealengine.com/t/how-do-i-replicate-rotation-in-vr/492290/6

[pallid canyon]

I have controller yaw enabled (though I dont think I use it except for testing) and Use Desired Controler rotation + Orient Rotation to Movement disabled on the CMC

I'm gonna give you my function for mesh setting with the note that the settings you have checked matter, I went through the same problem as you and it was literally two weeks of clicking options on and off, setting controlRotation manually, all that junk. You don't have to RPC anything but the HMD orientation, logic on the BP can take care of everything else.

I update HMD pos ever .1 second

then on tick check if I can move and use Add movement input to move

add input resolves controlRotation, so if your settings are screwed then it will do the same to you

[nimble widget]

Very interesting. Okay I will give this a try with my setup. I keep thinking that maybe something is checked/unchecked that shouldn’t be. Rotation is tricky. Lol

[pallid canyon]

then on tick for non locals, server has one update function, clients have another

the big ones I remember is that yaw is enabled, orient and use desired cont rot is disabled

but again

I do not use a controller period

it's all hmd, no sticks for rotation

[nimble widget]

Yeah that’s what mine is setup to do. I only use stick to move forwards and backwards.

[peak sentinel]

Some tutorials claim doing some basic prediction logic in BP is enough for movement, it's not true. Sadly C++ is unavoidable for proper replicated movement in C++

[chrome zealot]

yall

im trying to test multiplayer and i think the host is losing its player controller or something

anyone know the reason as to why?

[pallid canyon]

There's a network compendium that's in the stickies for this channel that you should read first Owyn

[plucky prawn]

is there a proper way of checking if the player controller is ready on the client? like its replicated and i can send stuff to the server? im guessing i cant do this in the constructor so i should use BeginPlay or something?

[lost inlet]

that would be a good candidate

[winged badger]

the engine sends RPCs to the PCs in GenericPlayerInitialization, ClientSetHUD

depending on the setup, this can be a long time before BeginPlay

that happens directly after PostLogin/HandleSeamlessTravelPlayer

as for sending from the client to the server

during seamless travel, PC sends ServerNotifyLoadedWorld, immediately after loading the world

note this will be the the PC of a class from departing level, server will instantiate a new PC in response if the PC class changes for that level

or you can just override APlayerController::PostNetInit, that one will fire before BeginPlay (if BeginPlay is delayed, it can be a long time between those 2), and after its replicated variables are set, and OnReps fire

i personally prefer everything being ready to go before the worlds start dispatching BeginPlay, as order of those can't be controlled

[past seal]

how would u get the info of other players in the ui, not just owning player?

[kindred widget]

Depends on what you need. If you're tracing, you can use their character. If you need general data about every player, usually you use the PlayerState to access what you need. You can get all player's playerstates on any machine through GetGameState->PlayersArray. If you need something on their pawn, like health, you can do PlayerState->PawnPrivate->CastToPlayerPawnClass

[past seal]

Yeah, i need to get all playerstates connected to the game. Not sure how to get this "PlayersArray" you are talking about.

[kindred widget]

GetGameState->PlayersArray

[past seal]

oh wait, yeah it works, my bad. Thanks alot

[twin juniper]

Do I have to manually call RestartPlayer on each player controller later on in GameMode if I have bStartPlayersAsSpectators = true; in the constructor?

And when do I call StartPlay()? Is that done on its own in GameMode?

[pastel heath]

I have an item I drop in the world or spawn, that item has an FX that plays on it to grab the attention of the user. Like any pickup PFX idle particles. What is the best way to replicate these particles to make sure clients see them? I can set the component to replicate, but it doesn't seem to fix the issue.

[sinful tree]

Is there something specific that you want to happen to the item before the FX starts playing or should the FX start playing as the item is spawned?

[pastel heath]

As soon as it is spawned.

[sinful tree]

Then just attach your particle system (if you don't have it set up on your item actor to begin with) and on beginplay activate it. No replication should really be required.

[pastel heath]

The particle system is currently a component on it with a hard reference. Probably already bad, and I could convert it to a Data Asset for a soft reference.

[pastel heath]

Is it that component that is causing me the issues?

[glad viper]

@neon ether can i pm you to ask a few questions?

[neon ether]

yeah of course

[glad viper]

Cool

[neon ether]

So me and a group of other students are making a game and the game is a pvpve arena shooter. We have a player one and a player two that both are children of the same player character. This player character has a custom component for the inventory. The inventory holds references to mods which the player can fire. Then inventory has reference to the player owning that inventory of course and when the player shoots or picks up a mod that mod's instigator is set to the player holding the inventory thats shooting the mod. This way the mod is able to know what player is owning it and trying to shoot it so that the projectile doesnt overlap with the player shooting it. However for some reason over networking the mods seem to get incorrect references sometimes, meaning, lets say, the mod owned by player 1 will have its owner reference set to player 2 somehow. I've done lots of digging and whatnot but havent been able to figure anything out. Just wondering if anyone has any hunches, if anymore detail is required to help determin the possible issue I would be happy to provide that info

[latent heart]

Does it happen that, for instance, all owners on a client are set to the client's player?

[glossy wasp]

Need some direction on how to debug why server broadcasting hands / HMD transform to clients jitter, but client to server is fine though.

Saw on AnswerHub if you have a component marked to replicate that might be the issue, but none of mine are. On Tick I'm checking local controlled, then setting rep'd transforms and broadcast those transforms skipping owner, then setting hands to new transform

[kindred widget]

This looks less like a flaw in your code. Both are working fine. I'd check some stuff like replication rates. It looks more like client is RPCing to server on tick, which is updating server fast enough, but server may not be replicating back as fast. The replicating back is also fine, I'd even consider limiting your client RPC to like 30FPS. Then in the tick, also interpolate local location to server authoritive value fairly quickly, it'll smooth out the jittery look.

Smoothing only needs done on non locally controlled panws though.

[latent heart]

IIRC there was a way to make something replicate at a higher priority, but I might just be making that up.

[quasi tide]

Net Update Frequency

You also have Net Priority

[toxic flame]

hey all, could someone help briefly clue me in on what Seamless Travel actually is?
I've got it ticked in my GameMode, and I've set a (completely empty) transition map, and attempting to move maps via ServerTravel
but I don't exactly get anything that's happening

I can't find many resources of it online

also I'm not exactly sure on what the PlayerState is, and what information I should store on it - I know there's a CopyProperties function on it, making me think I should kinda use that instead of a GameInstance to carry variables through (i think it's persistent?)

[toxic flame]

and I'm not exactly sure if it's all working as intended, since I am travelling through maps, but I'm just not sure

[glad viper]

Are there minimum requirements for running a server for a multiplayer ue game? Could i host the servers on like aws?

[chrome bay]

Key thing to note is that Seamless travel does not work in Editor @toxic flame

And yes the PlayerState has a chance to copy properties to the new playerstate in the next level before it's destroyed.

[toxic flame]

hmm

what exactly gets carried through with seamless travel

do PlayerControllers get destroyed?

[chrome bay]

All actors get destroyed and replaced with new ones

[toxic flame]

I see

[chrome bay]

Have a look at FSeamlessTravelHandler::Tick()

Actually I'm slightly wrong, some actors are actually kept and have their outers changed

But I think it largely depends on whether the gamemodes are also compatible

[toxic flame]

i'll do some more digging i guess

if I add a Print String to BeginPlay in the TransitionMap, should it print the string?

because it doesn't work currently (when testing in Standalone or Packaged)

[twin juniper]

Is there a reason why clients spawn as spectators in the level?

The listen server/host still spawns as a character

But clients spawn as spectators

Do I have to call RestartPlayer on each PlayerController in PostLogin?

[pearl fog]

Hello, I have a small architecture design, I have in my gamemode a CanDamage function that determines whether a player can damage another one, but we now have the need to clients to be able to determine if they can damage one player, I don't really know where to put that function 🤔

[twin juniper]

Clients shouldn't be able to determine whether they can damage a player

That seems like critical logic

to be handled by the server

Unless you want the server to make hte final decision

And replicate the decision to the clients

I think ideally what you should do, is have the server decide if a client can damage another player, then perforn an RPC or set a replicated variable to let the clients know.

[pearl fog]

it's for an ability allowing you to see which players you can damage

the damage logic is still handled server-side

and it's seems for me too op to constantly checking whether a client can damage a player and send it to him

and as gamemodes candamage function uses data that is available to the clients well I thought I can just move the function into a accessible object like game state, but i'm not sure exactly if that's the appropriate place

[verbal tendon]

To any class that lives on the client that has access to the data that makes sense

So somewhere in your ability logic

That then calls the shared function ( which gamemode and your ability use )

You want to avoid putting a function in a class just because it lives on both client and server. Check where your shared data lives and how it's accessible, decide in how to best expose access to the data / function using the data so it makes sense

[pearl fog]

Yep you're absolute right, that's why I wondered where to put this, I'm going to put that in my PlayerState. It seems the most logical and also practical way of handling that.

[dark edge]

@pearl fog what are your candamage rules? Like 2 teams or more complex?

[pearl fog]

Yeah only 2 teams for now, but I consider having more complex rules one day

[dark edge]

I would formulate candamage as some function of teams. Even as simple as CanDamage = MyTeam != TheirTeam, with team being an enum or int

[pearl fog]

Yep! that's what I'm doing

at first I considered just comparing the team, but yeah if the gamemode wants somehting more complex I prefer it to be encapsulated into a CanDamage func

[dark edge]

Iva had good results putting Team on PlayerState and in a repnotify putting it on the pawn. I do tests vs the pawn.team so it's consistent with ai

[pearl fog]

ooh, okay

I think I will stole your idea, it seems the most elegant :p

[plush otter]

For a dedicated server game, after creating a session, how to start a level map belonging only to that session?

Functions like dungeon instances

OpenLevel puts all players to the new level without joining the session

[pallid canyon]

my gut feeling is that you have different servers and instancing that's handled with logic outside of unreal

at least that's what I remember from when I was considering the same

[plush otter]

So inside unreal you can't do that?

[pallid canyon]

my recollection is that it depends on how you implement handing players off from one server to another. Load balancing is a similar topic

I'm sure you could disconnect from one session on the server and go to another, but the question is how you know which port on which server you're trying to go to or if you go through a server reconnection process again

how you know if a serve is available, things like that

[plush otter]

I'm asking for my project group. Waiting for another member more familiar with networking to continue the discussion.

[pallid canyon]

probably also depends on your oss

the net compendium that's stickied might help too

[chrome zealot]

hey yall

i just got a quick question about

relevancy in multiplayer

i feel like that alone is kinda irrelevant when building a moba... right?

especially since someone can look all across the map, some may be covered by a fog of war but

because in a moba you can look all over the map so

[pallid canyon]

if you had a fog of war would you try and render everything happening underneath that?

[chrome zealot]

im just curious on if its worth it to set that up, like does every bit of optimization in that sense matter as it is a moba

[pallid canyon]

depends on your platform how much optimization matters, going into it assuming every optimization is worth it is worthwhile because eventually you'll find out it still wasn't enough

including network

[twin juniper]

I can't set the initial spawn location of the host controller in GameMode, but I can set the client locations.

I'm using RestartPlayerAtTransform

[winged badger]

there is a ShouldSpawnAtStartSpot function in GameMode

it forces the cached PlayerStart assigned first time the controller spawned

[winged badger]

micro optimizations are generally not worth the additional code complexity

for most games, small bandwidth optimizations are also... questionable, as few games actually have bandwidth problems, the chokepoint tends to be the server's NetBroadcastTick

[barren patrol]

@chrome zealot it's also about anti cheat. if you don't use netrelevancy properly you would broadcast the positions of enemy players. hacked clients could have "wallhacks" through the fog of war basically

[chrome bay]

On the flipside, if you cull and uncull too often, it's performance heavy. Has to be a balance

Same reason most shooters don't use LOS relevancy, it's prone to error and often too latent by nature

Plus the cost of destroying/opening actor channels repeatedly

[short arrow]

LOS relevancy? what is this referring to, I can't find it on google

[empty axle]

I suppose line of sight

[chrome bay]

yeah line of sight

[ionic scaffold]

Hey everyone! I'm looking for the best resource (either paid or unpaid) to learn about multiplayer, using only blueprints. (Don't have a second to spare to learn c++)

[past seal]

hi fellow slackers, i am trying to figure out how to seperate players states for two teams (A and B). I have figured that UI widgets can get all player states from the gamestate. These end up in an array with 8 indexes, however i want it to be an array for each team 0 to 3. So that i can get player 1 on team A or player 1 on team B.

I tried making two custom arrays of Playerstate teams in the gamestate but it doesnt replicate properly

TeamSelectWidget:

[kindred widget]

Personally. I'd just place a value in the playerstate. You can get your teams for UI from the normal PlayersArray with a function.

[past seal]

Yeah, each playerstate has a teamtag variable. I am just confused on how to make a function like this 🤔 It would have to loop through all playerstates then find which is player1 on team A then player2 or something

Because the indexing of the playerstates just goes 0-7 but that array isnt really sorted

oh wait, sorting hmmm

[pallid canyon]

Yea, but I’m jaded when it comes to optimizing because my first platform was the Quest. If on mobile I feel the recommendation is still applicable, given the depends on platform part.

[kindred widget]

You do exactly this. Make a function. Make it have a local array of pointers. Add to this array of playerstate's team == team you're searching for. At loop end, return local array.

[past seal]

Tysm 😄 I will try right away!

[past seal]

This worked really well! However, i would really preferred if the indexing was sorted by who joined the team first. What happens now is that if Client2 joins team A, the UI reads it as player 1 on team A. However if Client1 joins it becomes the new Player1 since it is indexed before Client2 in the "PlayersArray".

[winged badger]

why? your players won't have their monitors side by side or know any different

[past seal]

It matters in what order they are shown here. So the top one should be the first player to join the team, then the second will be the second to join the team

So it would work as intended if client1 joins then client2 joins the team, but not vice versa :/

[verbal tendon]

That's a super simple logic fix, unless I'm missing something, to have two arrays and fill them as the players join the different teams

You just update the array elements as players join/leave the team

[kindred widget]

That is a lot of arrays to keep track of. 😦 Upkeep sucks. I mean if you really care that much. Copy the CreationTime variable on actors and replicate it. Overrride PostActorCreated, and set it on the server only. Then you can check which one is oldest. Much easier than sorting multiple arrays. And scalable for more teams later.

[past seal]

Thanks for the input guys 🙂

[past seal]

Yes this was my original thought, but then where would i store it?

[halcyon totem]

is it possible to run a dedicated server on my pc and have others connect to my PC or are we forced to use amazon web services?

[quasi tide]

You can run a dedicated server on your machine. You just need to set it up.

[halcyon totem]

do you know where the directions are for that most are for amazon web servies

[quasi tide]

This isn't specific to Unreal.

It's just general server stuff.

So just Google how to set your computer up to be a server.

[fathom aspen]

I know this is a 2 year old conversation, but do you think playing with the source code to make ViewTarget support UActorComponents a possible mission? Would it fix the lag?

[sullen burrow]

how can I replicate this properly?

[sinful tree]

Clients only have access to their local controllers. The server has access to all controllers.
Because of this, you can't reliably use GetPlayerController in multiplayer. Instead, use GetController and cast to PlayerController, and do the calculations on the Server.

I'm also not sure you need to do a multicast since the Capsule component is replicated by default, you may just need to set the value only on the server.
RPC > Server calculates rotation > Set Capsule Rotation.

If not, then:
RPC > Server calculates rotation > Multicast (Input: Rotation) > Clients will set the rotation of the capsule component

[shadow aurora]

Wasn't sure if this belonged here or in #legacy-physics so I'll run it by here as well.

I have setup my game to use the player's skeletal mesh/physics body for collision with projectiles. However, when running a dedicated server and connecting clients to it the physics body only seems to exist around the players' feet... Which means people can only shoot the feet of enemies (not great haha). Any idea why this would happen? Really stuck on this.

To clarify, if I set the editor to spin up a dedicated server & connect two editor clients to it, it works fine. So I'm not quite sure what the trouble is here.

[leaden scaffold]

are you replicating animations on dedicated server? Does your physics body have constraints in place appropriately? Have you investigated the physics shapes it's creating at run time? There's a number of settings the character movement component related to dedicated servers you might want to investigate

[sly kernel]

I've heard that programming multiplayer with Blueprints is a lot easier when using Photon vs native UE4's networking paradigm. Is that true ?

[kindred widget]

Probably depends on your intended game. Didn't even know what Photon was til three minutes ago.

[crude quail]

Photon Quantum is supposedly agnostic

I'm not sure how true that is

[sly kernel]

Photon Realtime is for multi-player and each engine had its own integration

[lost inlet]

is the idea just to say something outrageous and see if you get corrected? I'm not sure where that would be coming from

though I'm pretty hesitant to use BP for anything but the most trivial of replication

[pallid canyon]

It's an internet rule, say something incorrect to get an answer rather than asking a question

[lost inlet]

I’m sure Photon has it uses, but if you’re dealing with server authoritative networking then I don’t know why you’d want to rip it out

But then you’re probably dealing with more p2p or more advanced use cases that require special netcode like fighting games

[sage ivy]

I Using EOS to login players and manage a lobby / invite friends, then I have an aws dedicated server using gamelift. I'm still confuse if I need to use cognito and lambda to use gamelift Flexmatch.

Does someone know some documentation / video for using EOS with Gamelift ?

most tutorials I found just use Gamelift + cognito + lambda

[winged badger]

there is an EOS channel 2 down from this one

should probably start by checking pinned messages there

[sage ivy]

I have EOS working its more a question about cognito i guess and if gamelift flexmatch needs cognito to work ?

[thin stratus]

Hm, so usually when you don't want a player to move, you set their MovementMode to None, on both ends.
Now I want to have the Player move towards a platform via RootMotionSource, but still not allow them to move. MovementMode-None breaks RootMotionSource though. Is there another safe way of doing this?
Blocking InputEvents is local, same as IgnoreMoveInput.

[kindred widget]

Vague memory of reading somewhere in the CMC that root motions are allowed without being possessed. Possible you could create a state where camera remains on character, but remove possession. Would disallow any authority on the pawn for sure.

[rocky kestrel]

What is best way to make health bar in multiplayer?

I have working health system but

[kindred widget]

UI is independent of gameplay. If you have a health system, all you need to do is locally create a healthbar for the actor and pass that actor or component with the health stats to the ui. UI can display it.

[rocky kestrel]

I know that but Im using build in Any Damage event

Atm im calling Owning client event and it works but because it (should?) only be local thing how to make it properly?

[woeful ferry]

Correct me if I'm wrong, but didn't Epic deprecate this system for UE5 to replace with GAS?

[kindred widget]

Where did you hear that? O.o

[woeful ferry]

I'm not sure 😄 Just thought I read it somewhere her

here

[kindred widget]

Hmm. I doubt they will. GAS isn't nearly far along enough to be replacing such a key feature. Even if it was, I doubt they would deprecate it. It's not in the way.

It's not marked for deprecation in 4.27.2 at least.

Feels like another "AHUD is deprecated."

[bitter oriole]

At least that particular line is just confusion between HUD and the very closely related Canvas

The latter actually being deprecated as fuck

[kindred widget]

It's still surprising the amount of people that think UMG replaced AHUD. O.o

[bitter oriole]

Well UMG replaced Canvas which is 95% of people did with HUD before

[kindred widget]

Fair enough. That was a little bit before my time. 😄 I don't even know when UMG was integrated.

[dense sundial]

Quick question, if I want to have speed(m/s) as a user controlled input for a character with a character movement component that is network compatible would this require extending something like add movement or creating a new movement mode, or is there an easier existing solution that I don't know about? (Don't want to jump down the rabbit hole if I don't have to 😅)

[kindred widget]

Sounds like something you'd just RPC to server when client changes it and set the server cmc max move speed. Let it replicate back and set it on client's cmc.

If you want specific levels, you could clamp it down to a byte and compress it from 0-255. Mostly prevents cheating if you don't want speeds too high.

[dense sundial]

Interesting, My initial thought was similar to this but had heard that changing cmc max speed was problematic for high ping environments and would lead to rubber banding (I very well may have misunderstood the warning)

[kindred widget]

Rubberbanding is always problematic in high ping environments. 🤷‍♂️

[rough kestrel]

has anybody done multiplayer on quest 2? Need some help

[dense sundial]

Yes, what's up?

[rough kestrel]

How do maintian a session in case the headset falls asleep or if someone presses the oculus button? The apps seem to get paused when in this situation killing the session @dense sundial

[dense sundial]

like with a ping of say 500 or so, you'd experience terrible rubber banding which may otherwise be able to be mitigated (I'll try and find the source of this info lol)

[kindred widget]

Yeah. If you're playing with half a second of lag, that's on you, not the game developer. You cannot make competitive games that cater to people who play on net worse than bad satellite. Your caring high mark should probably be closer to 150-200. Average latency is usually 50-120 in most gaming servers.

[sly parcel]

hi regarding multiplayer, I currently have an asset that generates a maze, I think it is not replicated by default but when I tried to setup 4 players. I think it kinda works. I do not have experience with multiplayer but I started learning some basic stuff.
Is there anything that I missed out especially related to the meshes involved? other stuff I need to replicate? Thank you

[dense sundial]

So you can run events on when the headset is no longer active such as pausing the game properly so that the session remains intact, I don't remember the exact solution and am about to head out, but that might be a good starting spot. When I return later I'll try and find it

[rough kestrel]

basically something that does not kick out the player for removing the headset for a second or when they tried to take a screenshot/recording. DM me when you get back got an angry boss to deal with.

[dense sundial]

I mean you're definitely right. I imagine I won't be able to help myself though lol

[dense sundial]

https://forums.oculusvr.com/t5/Unreal-VR-Development/Fixes-for-common-UE4-Oculus-Platform-issues-before-submitting-to/m-p/732728 read through this. I think that on quest it shouldn't drop the session on pause or really stop by default so just make sure the relevant settings are set so that it doesn't pause and instead handle it yourself like outlined

[thin stratus]

So, with the Server ticking Anims of Remote Clients differently, is there even a sane way of tying AnimNotify based Traces to attacks?

I can literally see a diashow on the idle anim on slightly higher pings

Freaking animNotifies trigger randomly delayed for the server too

[toxic flame]

hey all, I'm using Seamless Travel and using ServerTravel to move maps (confirmed via logs), but the CopyProperties function on the PlayerState only seems to work on the Host/Server and not Client?

[ember vine]

@thin stratus i ended up ticking the animation manually and having various states, then its abit easier to extrapolate on ping aswell

[rough kestrel]

this applies even if I do not use the Unreal or Oculus networking code right?

[thin stratus]

Hm, I know the change one can make to tick the pose normally, but not sure that fixes anything

[ember vine]

that and a turncap for damaging phase and you can get the tracers synced up very nicely with some extrapolation

theres something sussy about anim notifies on the server

[dense sundial]

😬 I can’t make that guarantee. In fact, I’d wager that if you’re rolling your own networking solution then that might be the cause of your problem as normally a ue generated session should stick around when the menu comes up, or at least that’s my understanding (take with a can of salt lol)

[rough kestrel]

not my own using a photon plugin for ue4

[thin stratus]

Problem is that I won't have the time to implement something of that scope. Haven't really done that yet

I might go with simply tracing in front of the player for now

And changing the mentioned tick pose stuff I guess

[severe nymph]

Need a little help here

I'm moving logic over from a listen server to dedicated server

part of this logic is driven by "instructions" audio cue

bound to "on audio finished"

this doesn't exist for dedicated server like it does for listen

so I'm having issues with converting it in a way that still functions

If I want to start the match after instructions have completed for the clients I guess I can hard code a time out

which is kind of hacky

but I can't rpc back from the clients that "instructions are done" cause then I would get 1 for every client

vs on listen you can authority check the bound event

any ideas are welcome 🙂

[dense sundial]

I gotcha, I don't have xp with that plugin but I'd still make the same wager as before, I'd guess that there is a specific quirk in how photon does things that is causing your issue, but that's just a guess

[dense sundial]

I mean, idk what you're doing or if this idea fits, but can you check with some flag if all players are done? So like you're saying with "instructions are done" each client then indicates they're done by setting a flag in perhaps the game state and when they're all set you can do x, or maybe it's a float that starts at 0 and when the float == Num Players do x

[humble knot]

Hey, I had a question. Since RPC have to be called on an owned actor, does that mean that you cant have any RPC logic on an actor in the world? To be more specific, if you have an Tree in the world that can be chopped down. When the tree hits 0 hp it would spawn a log. The log spawning logic couldnt be inside the tree actor's class, right?

[shadow bane]

Can anyone point me in the right direction of a tutorial or something obvious I'm missing with this? I have a problem where my listen/server character does what it is supposed to do in regards to replicating health and animations, but the client doesn't send the "take damage" notifications to the server and the montage animations don't play on client or listen/server.

[heady python]

if its not sending the event make sure that your branch checks are going through. my guess is that the replicated "is attacking" variable gets set on the client when it should be the server

[shadow bane]

@heady python To test this, I set the false from each branch to the next execution pink and ensured "is attacking" was set to "replicated". The same problem existed. I then tested the "failed" cast to the Attack interface, which it apparently fails on the client side. Now I'm really confused. The mannequin class requires that interface, but it doesn't work in a client/server relationship?

[dark edge]

@shadow bane Are you trying to do prediction or clientside hit detection or just show on clients what is happening on server?

[severe nymph]

@shadow bane hmm

any reason your not use damage events from the server?

if your using trace results or projectiles or even authority overlaps

you really should be feeding the server take damage event

[shadow bane]

@dark edge No on prediction. I think I just followed some tutorials that showed clientside hit detection and trying to make it work for simple multiplayer. Mistakes are clearly made.

[severe nymph]

and using "on point damage or any damage" to feed the client behavior

[shadow bane]

@severe nymph any tuts on that? I've seen so many mixed examples, I can't tell whats what

[dark edge]

Ok so just make sure the data flow from client to server is, or is enough to derive, "I'm doing this move/playing this animation" and "I think I just hit that guy"

[dark edge]

No tuts. Stop following tutorials and get what you have working. Make sure you understand the flow of who's sending what where.

[severe nymph]

in your blueprint from the damage causer actor

switch has authority

[dark edge]

He's doing clientside hit detection

[severe nymph]

but why?

[dark edge]

Idk, but it's valid.

[severe nymph]

movement component prediction is built into the authority damage check

client auth damage is only for that client

which results in visual desync

one client will think they aren't being hit

while the client performing the trace is overriding

just bad to do it this way with ue4 when they have things implimented to give you a very clean expierence

and can "predict" intra and extralerp off the movement component

[dark edge]

Clients always disagree on what happened no matter how you do it if you have any sort of prediction. That's the cost you pay for responsiveness.

[severe nymph]

not true

projectile components, movement components factor in the difference ... its why the server "who knows this" can make the best decision on how to deal with simulated proxies

[dark edge]

Absolutely is. How big the disagreement is can vary but if you're predicting there's divergence. You need to favor someone. Most games favor the aggressor.

[severe nymph]

for p2p client auth you are correct

for listen server and dedicated ... i'm sorry

its just not true

[shadow bane]

Guys just for reference this is the basic mannequin and paragon character class with a collision box on the sword. not fancy at all if that matters

[severe nymph]

there is over 600 lines in the source on these components that explain why it isn't true

its why add actor input to the movement component feels so smooth for clients with 300ms ping

and how the server understands how to deal vector math to the "server copy" via factoring in the client simulated proxy location

for the damage event

[dark edge]

It's a fundamental problem with ping. Every game with prediction lies to somebody. Good engineering can hide the lies but they're always there.

[severe nymph]

its not really a "lie" its the "best assumption" that can be made... the client in this case isn't and hardly ever is

[shadow bane]

@severe nymph @dark edge Any way I can pull you both into a voice chat for like 10 minutes?

[severe nymph]

also good luck with cheat engine

on client auth damage

[dark edge]

I'm on a phone at a restaurant lol.

[severe nymph]

@shadow bane are you dealing this damage from another player?

or a world damage actor

[shadow bane]

no, this is damage from two players to a spawned npc

[severe nymph]

ok

in this instance

the server is going to be handling the npc

[shadow bane]

paragon is player characters and npc is mannequin

[severe nymph]

the simulated proxy of the npc isn't going to even be moved for the client until after the server issues the event

the location of that actor will be true minus the client sim proxy calc

deal point damage is your best option from the trace result

the server will spawn both the projectile or the trace from the "character client" location that it knows will line up with the sim proxy offset

on the clients end

resulting with a dead on result

on the npc your can do your logic off "recieved point damage"

which will again be server side

this logic can exist solely on the server

minus the "fake event" for the client

[shadow bane]

i don't have a trace or projectile, just simple collision on the sword

[severe nymph]

the fake event will play the effects

[shadow bane]

is that my first step?

[severe nymph]

for collision on the sword its instant

you don't even have to worry

on collison -> has authority -> deal point damage at location

in npc event on point damage recieved "do logic"

collisions fire on all similar to a multicast .... you can just trust the server in this case anyway

your swing I'm assuming is driven by an animation?

for the sword?

[shadow bane]

montage 😦

with notifies

[severe nymph]

ok

so your notify is calling the deal damage event

[shadow bane]

notifies basically only turn on and off the collision for the swing

the deal damage is happening in the player character blueprint on collision

[severe nymph]

ok sure .... you using actual hit or overlap?

[shadow bane]

its overlap

[severe nymph]

jump down in a channel with me

I'll do a screenshare

[shadow bane]

sure which one?

[severe nymph]

lets go hangout

[twin juniper]

Would I need to introduce an anticheat if I wanted to prevent users from doing things like manually settings X,Y,Z values or changing speed? Is there some already written algorithms for the server to check valid character movement?

Or does the server already handle checks like this, predicting where the player should be judging on thier velocity and input and simulating it server side and if the client's state doesn't match , then fixes it?

[hollow eagle]

CharacterMovementComponent already does this to some degree

[twin juniper]

Ah okay, neat

[peak sentinel]

For custom movement abilities you might need to implement your own solutions but most of the time you are fine i guess

[twin juniper]

Sweeet I'll trust the engine

[severe nymph]

it does this

[twin juniper]

thank you, thank you

[shadow bane]

Thanks again @severe nymph @dark edge @heady python Appreciate the help. going to apply those learnings now

[sinful tree]

You probably already know this, but also be sure you're not letting clients tell the server what to set their location or speed to either... The client can request that their speed be changed or that they need to be moved, but don't trust that the client will provide a valid input.

[twin juniper]

Absolutely

[light iron]

Clients would never cheat..

[twin juniper]

Of course not 🤥

[lime iris]

Question: does "mulicast (reliable)" trigger 99.9% or 100%?

[lost inlet]

unreliable is effectively 99.9% unless you have really bad packet loss

reliable means what it says since it resends RPCs if they're not ack'd

[lime iris]

Okay, so what reliable does is, it resends the RPC when the first try fails? I am asking this because in my BP, if I don't check "reliable", it can randomly skip (btw, simply incrementing int value, testing in PIE as listen server), but checking "reliable" won't. Why would this be happening (in the same machine)?

Also, what happens if the order is messed up (e.g. lag or latency)?

[lean tide]

Yo got a question, everything is replicated and working in pie but when I package and run on dedicated server all clients connect and no players can see each other or see any actions taken. Both players connect to seemingly individual sessions.

[sinful tree]

My understanding is that if a non-reliable event is used, the engine can decide to discard the event to ensure that other data that needs to be replicated can be squeezed into the package. The idea is that if you're replicating something that is required for gameplay purposes, you probably want to mark it as reliable. If it's not important and your game won't break by it not happening, then you can use unreliable. An example would be replicating visual effects or sound effects - these don't necessarily need to happen in order for the game to function so they can be unreliable. An event indicating to the players that it's game over and show some wrap up UI or what have you should probably be reliable.

[lime iris]

a good rule thumb, thanks!

[sinful tree]

Here's another good example that mixes that logic up: replicating on event tick your player's head position in a VR game.... While it may be somewhat important to know where their head is depending on the game, it's also not so important as you're frequently getting updates anyway (since it's on tick!) so that should be unreliable.

[lost inlet]

you can overflow the reliable buffer so don't forget that either

[lime iris]

okay, how can i measure that? network profiler?

[lost inlet]

the chances of hitting that limit are usually very slim unless you're doing very irresponsible things, but that'll help, yes

[lime iris]

thanks!

[lean tide]

Dang, think I may have had oculus plugin on during packaging.... hmmm

Didn't mean to cut in, I know you guys were in the middle of it 🙂

Dang, don't think that was it..

[twin juniper]

hello guy

[lean tide]

Hey there!

[twin juniper]

how to make lan connection like if all the players from different devices and if the connect to same wifi

they have to join into one room

how can i do that?

[lean tide]

I'd look up a tutorial on youtube on setting up a listen server if it's just you and your friends.

Wish I could help more, but I have a similar question xD

[twin juniper]

i tried this

in my pc with different exe opened its working when i add this exe to another laptop its not working

[lean tide]

Are your ports open?

Usually the biggest issue people have is not opening port 7777 on their firewall

I wish that was my issue. 😦

[twin juniper]

its is open

[lean tide]

not working as in, it doesn't connect?

or doesn't open?

[twin juniper]

when i typed ip address its not opening the level

[lean tide]

in console?

[twin juniper]

i dont know where its not connecting or opening

[twin juniper]

in windows

[lean tide]

try connecting through console

set up a key in blueprint on your character for it.

[twin juniper]

ok i will try

[lean tide]

I'm wondering if unique ids are required for a dedicated server.

Everything seems to connect fine.

Replicated right

Log says: Notify Accepting Connection accepted from "ip" on the last entry.

But I can't see the other player at all when the second client connects, and reverse too.

Trying "always relevant" next, but that never works lol

[dark edge]

Does it work in PIE?

[lean tide]

Yeah, no sweat

Wondering if I'm packaging wrong.

I'm running source from a batch file and loading the map via instructions from a tutorial.

I'm thinking this is wrong and the dedicated must be built from the editor.

However I am missing build target server option.

Kinda where I'm at through scouring. 🙂

[dark edge]

Test with just launching as a server.

[lean tide]

Like? Standalone?

[dark edge]

Ye

Without building

You can do it with a bat file

[lean tide]

I have done this once before and it worked, it was a while ago. I wonder if it still works. One moment I'll try it.

Is an entry map required for dedicated? Also starting to wonder they're getting stuck in their own entry maps. I set them all as the default map.

Yep still works from a batch file

Is it the way I'm connecting? Through bp?

Im just getting on both clients and activating "open "ip""

@dark edge

Grrr always get stuck at these advanced points, end up reading 1000 forums for routing ip. -.-

[dark edge]

If it works when you launch and not when you build, I'd look at your build setup

[lean tide]

Oh I thought you were asking me to launch it from the build with a target in the batch file.

The server batch targets source and the project

[dark edge]

I mean launch it like
Engine uproject -server etc

[lean tide]

Yeah

This is what the batch does

[dark edge]

So NO cooking or packaging, just fires it up?

[lean tide]

Server batch: "D:\UnrealEngine-release\Engine\Binaries\Win64\UE4Editor.exe" "G:\Unreal\Unreal Projects\NFTPit\NFTPit.uproject" ThirdPersonExampleMap -server -log -nosteam

Client batch: "D:\UnrealEngine-release\Engine\Binaries\Win64\UE4Editor.exe" "G:\Unreal\Unreal Projects\NFTPit\NFTPit.uproject" 192.168.1.2 -game -ResX=800 -ResY=900 -WinX=0 -WinY=20 -log -nosteam

[dark edge]

And does that work?

[lean tide]

Yes two clients side by side

everything works perfect

[lost inlet]

"nftpit" do I even want to ask

[dark edge]

Have you tested on multiple machines, not just with localhost?

[lean tide]

The client there is accessing source so no

[lean tide]

Haha it's actually pretty cool, but it's up to you xD

[dark edge]

What. I mean fire up server on one machine, client on the other. Or are you doing all this testing on one machine

[lean tide]

I mean because the client is accessing the source build, (109) gb I can't move it. It probably is packaging here.

I keep looking for an option for build target>server and I don't see it in the editor.

[dark edge]

Afaik you need to do a dedicated server build from the solution.

[lean tide]

Does that hint you in to any cluelessness I may be experiencing? xD

To be clear you can't do it from the editor now?

[dark edge]

Step one is to make sure the damn thing works over the network before trying to build dedicated. Have a laptop for host or whatever.

I've never seen a dedicated server build from editor but last time I did one was years back.

[lean tide]

Okay, I think you've clued me in here.

Thank you Adriel.

Also one other question.

Are full source builds even required for a dedicated?

[dark edge]

Last I knew yes

[lean tide]

Okay so we build the solution from the source?

[dark edge]

To package a dedicated. Can test all day with launching a dedicated tho.

[lean tide]

ok

Sorry if my questions sound stupid, I crash coursed this after like 4 years.

Today lol

[dark edge]

There's basically 3 levels or ways of firing up your game. PIE, launch, and packaged. I'd test in that order but I do 99% of my stuff with a host on another computer launching the server.

[lean tide]

Pretty sure I'm using a testing method to launch my dedicated. I need to build the solution for purely the server files like you were pointing out.

Basically

Part 1 and 2 of the documentation lol

The tutorials I was doing I think threw me off.

[kindred widget]

You cannot. If you need that kind of control, usually you just use the pushmodel system and don't mark it for replication unless you want it to.

[chrome bay]

You have to mark it replicated, there's no getting around that. It also has to be present in GetLifetimeReplicatedProps

You can disable property replication at the class level, and at runtime you can turn replication of a single property on or off, but that's it

Have a look at AActor::PreReplication

That's the only function where you can turn properties on or off, and it's called everytime the actor is considered for replication.

The actor class does this to turn on/off attachment vs movement replication depending on whether it's attached or not, for example

[thin stratus]

What's the default rate of ServerRPCs, send by a Client? (CMC related)

And where was I able to see/change that

I read something about 10Hz on UDN

That and MoveCombine would lower the calls even more I assume

(can I see that number live somewhere without a netprofiler?)

And why would a highping (no package loss) even affect this?

[chrome bay]

UCharacterMovementComponent::GetClientNetSendDeltaTime I think?

It get's throttled depending on the netspeed

Presumably to avoid making ping/loss even worse by saturating an already strained connection

[thin stratus]

I assume I won't ever get around that "known" change for TickPose

[chrome bay]

IIRC they're only delayed if it's not considered an "important" move

[thin stratus]

Even UDN has no other idea

I don't get why a 200ping would cause the MoveAutonomous stuff to be called less often

Shouldn't it still be the same amount of RPCs reaching the server?

[chrome bay]

I'm fairly sure UCharacterMovementComponent::GetClientNetSendDeltaTime throttles it based on both ping and loss

[thin stratus]

((Player->CurrentNetSpeed > GameNetworkManager->ClientNetSendMoveThrottleAtNetSpeed)

[chrome bay]

Seems to be relying on lots of external factors though

[thin stratus]

So that

[chrome bay]

yeah

[thin stratus]

ClientNetSendMoveThrottleAtNetSpeed = 10000;

What's the unit of that

ms?

[chrome bay]

bytes per second IIRC

[thin stratus]

Ah

[chrome bay]

or bits per sec maybe

AGameNetworkManager seems to update it periodically

[thin stratus]

Hm, can't hit a breakpoint in that function

[chrome bay]

Maybe it's not that. I wonder if high ping causes the delta-time threshold to be exceeded more often, causing less RPC's to be sent

Because the old moves are getting too stale

[thin stratus]

Hm that could be

But, okay, let's say we have 0 ping, shouldn't it still be limited to less than 60Hz?

Cause the Anims on the Server look kinda fine with that

As if they animate per tick

[chrome bay]

Looks like max is 120Hz and min is 5 Hz

const float NetMoveDelta = FMath::Clamp(GetClientNetSendDeltaTime(PC, ClientData, NewMovePtr), 1.f/120.f, 1.f/5.f);

Hard to really know for sure though.. it's so complex

But yeah when you think the animations don't run at all server-side unless you get a packet from the client, if the client is sending stuff at a low rate, it's going to look crap

[thin stratus]

Ah, now I enter it. Had combine off

[chrome bay]

Yeah maybe that's what it is then, it can't ack moves fast enough and so it's just building up a huge buffer of them

[thin stratus]

So it is indeed throtteling

Given the GAmeState is valid and it's less than 10 players

The CurrentNetSpeed is probably shite then

Hm no also not

what..

[chrome bay]

Hard to know really.. been so long since i looked at it 😄

[thin stratus]

Yeah not really asking for that, just confused

[chrome bay]

Used to have a good understanding but it's so complex you just forget stuff

[thin stratus]

The if for this is NetSpeed > 10000, GameState valid and Players <= 10

Which should be all true, yet it went into the else.......

[chrome bay]

You in DebugGame?

[thin stratus]

I know I should. Don't blame me. brb

[chrome bay]

Development will be useless 😄

Will be jumping all over the place

[thin stratus]

Hm, yeah same weird result

Maybe I'm just blind

Seriously

[nimble parcel]

:no_entry_sign: Flops#4525 was banned.

[thin stratus]

Spam about CMC fixes and I would have crowned you

[chrome bay]

Roll on the replacement code from UE5 where we can enjoy fresh new issues instead 😄

[thin stratus]

SURE

[shy hill]

Hey hey,
I have a P2P multiplayer game, the 'servertravel' command seems to be inconsistent. Sometimes 1 or more clients aren't teleported to the new map. Any idea what can cause the issue or how to fix it ?

[thin stratus]

Steam?

SeamlessTravel enabled?

[shy hill]

yes, steam

[thin stratus]

Steam doesn't like non-seamless travel

So make sure you enable that

[shy hill]

Where can i check Seamless travel ?

[thin stratus]

It's a boolean in your GameMode

Be aware that SeamlessTravel will cause the new map to work a tad different. E.g. PostLogin not calling

[shy hill]

yep seamless travel is checked

[thin stratus]

But it's adviced to always perform Seamless ServerTravels if possible

Then it shouldn't DC on travel, hm

[chrome bay]

Not travelling to the same map are you?

That breaks a lot

[shy hill]

so actually the loading screen doesnt go away

it might be that the client travels correctly to the map, but the loading screen stays for some reason

(happens rarely and inconsistently)

[thin stratus]

That can totally be (:

[shy hill]

Is lag / packet loss causing these issues ? I don't even know how to reproduce it. I dowloaded clumsy to try to simulate lag but I can't seem to make it work 🤔

[thin stratus]

        // Lower frequency for standing still and not rotating camera
        if (Acceleration.IsZero() && Velocity.IsZero() && ClientData->LastAckedMove.IsValid() && ClientData->LastAckedMove->IsMatchingStartControlRotation(PC))
        {
            NetMoveDelta = FMath::Max(GameNetworkManager->ClientNetSendMoveDeltaTimeStationary, NetMoveDelta);
        }