multiplayer

[sinful tree]

The instructions in the video I've sent you are specifically for Steam. The Steam Session Ticket is unique to Steam's system. If the function provided in the video returns "0" that means the computer the player is on is not logged into Steam.

[vital heron]

Quick Question: This is mainly curiosity but computers that host your game's server, is it all just one big server that is cut up into 12 player lobbies or are they are all individual servers.

This is a very stupid question I know but curiosity is a powerful thing 😅

[hollow eagle]

A single server process is usually just one lobby, doing more than that makes things way more complicated.
For the actual hardware, most servers these days are virtualized - one physical host will be running multiple virtual machines (or sometimes containers, though I'm unsure how common that specific bit of tech is for games). Whether a single VM runs multiple server processes is highly dependent on the game.

[vital heron]

Oh so a more complicated game will usually host less server processes on a single VM?

[hollow eagle]

no

It's a cost thing - how much can you divide up a single VM before impacting performance

[vital heron]

Oh

[hollow eagle]

it's not as simple as just adjusting the number of processes running either

because you have multiple VM sizes

I'd say it's probably more common to run a single server per VM as it's easier to guarantee resources that way

but it's complicated

[vital heron]

Why do they use VMs though? If it is all run on a single hardware device why not just run multiple services on the PC itself.

[hollow eagle]

because no one wants to build out a datacenter (or buy out sections of one/colocate) these days because that's expensive and not scaleable

why buy the physical hardware when you can rent VMs on a cloud

or if you do happen to own the datacenter, why build out hardware for one specific game when you could support any number of your games (because you're big enough to have multiple if you have a datacenter)

and scale them automatically

[vital heron]

Oh I see

That does make way more sense

[hollow eagle]

If you're an "independent" AAA you probably don't have the capital to build out the infrastructure worldwide so cloud is basically the only real option that will scale no matter how many (or few) players you have. If you're a large AAA like EA then you probably use a hybrid approach but most stuff is probably still virtualized. If you're microsoft... you own the cloud, why not use it.
And if you're an indie you're definitely not building out your own infrastructure.

[vital heron]

Yea true

This clears up a lot of confusion thank you!

[sinful tree]

Usually when you're dealing with hosted VMs there are really beefy physical servers behind them, so beefy that it's probably already running instances of other VMs on it - and what happens is you just get to use a slice of the computer power available on whatever physical server your VM runs on. If you went with AWS for example, you pick between different sizes of instances, which have different CPU & RAM availability, and whenever you activate it, it then starts running on whatever hardware is available at the time for your instance size to use.

Even if you request a VM host somewhere for a really powerful VM, with tons of RAM, there's still a good likelyhood that other processes will also be running on the same physical server. It's only if you get dedicated hosting (they have server hardware you're effectively renting + it's dedicated for your use only), or co-hosting (you own the physical server, they're just a place for your server to be plugged in with redundant power, proper cooling & extremely high-speed internet access) that you're guaranteed you're the only thing running on the server hardware.

[hollow eagle]

Yep, this is also important to know about VMs

[vital heron]

First of all, wow. The fact you typed all this out is insane. Second of all, thank you this really explains server hosting and VMs really well!

This will be really helpful for looking for cloud server hosting!

[vital heron]

Not gonna lie, they should pin this for more people to see.

[rain quest]

Im new to multiplayer, why is all the clients swapping weapons fine and seen by everyone but the host only swaps their weapons on their own screen but sees everyone else's fine.

[dark edge]

@rain quest can't really read your screenshot my phone, but for swapping weapons, the input should just trigger a run-on server event that swaps the weapons around. The current weapon should be replicated and in the rep notify, that is where you would play your animations etc.

You shouldn't even need a switch has authority, the flow will be the same whether you are hosting or not.

[soft shell]

QQ; let says I have a replicated variable.. can I set that on the server to give it an init value (and never touch it again) but then let the client work out it without worrying about the server touching it again?

[bitter oriole]

Why does it need to be replicated if the server won't change it ?

[soft shell]

Because only the server has the initial value it should be set to

My concern with using a multicast is that this is set during the spawning of the actor; and as far as I know events can be problematic / missed if issued on an actor that has just spawned (correct me if I'm wrong)

[winged badger]

never use multicasts for stateful changes

only for fire and forget stuff

your variable should be replicated with InitialOnly condition

[soft shell]

Ah! thanks

but just for the record, if the server never touches it again the client would be free to use?

[winged badger]

note that you need to set it on server before it replicates, like immediately after spawning an Actor

or clients would have the wrong value

[soft shell]

Rite, so anything initialOnly needs to be set immediately 👍

[winged badger]

it is bad practice for clients to operate on replicated variables

so best have 2 of them, 1 replicated, one not, but set from OnRep_FirstVariable

[soft shell]

Thanks for the clarification, I'll do that

[atomic acorn]

Does anyone know if it’s possible to replicate a render target, or the data in it and rewrite to a render target?

Or would I be better off using scene capture?

[chrome bay]

You can read the data back and replicate it, but it's extremely expensive

[atomic acorn]

I thought that might be the case

[meager spade]

best bet is to let client also generate it

[chrome bay]

Yeah definitely

[atomic acorn]

this is what i'm working with essentially

ah okay, thank you

[meager spade]

what is that?

[atomic acorn]

it's just a whiteboard kind of set up

using UV's

I'm wondering if I send where to write to everytime someone writes might be the way to go

[chrome bay]

Sending deltas is probably a better way indeed

[atomic acorn]

cheers for the help guys! I will try that instead

[chrome bay]

For players that join late you'll still need to send the initial state ofc

[atomic acorn]

so send the whole rendertarget data on joining basically?

I'll start with the replication first but I'll try get that working also

[glossy veldt]

Hello, i have the following problem: When creating a widget from my Character BP (After being possessed by a player controller) the widget gets added to every clients viewport and not just the one that owns the character

[whole urchin]

how do i possess a vehicle in multiplayer ?

[dark edge]

@whole urchin call possess on it.

[whole urchin]

are you serious ?

i did that, it doesn't work, hence my question here

[chrome bay]

Controller->Possess(TheVehiclePawn);

Assuming you are using the engines' vehicles, and not your own approach

You only call Possess Server-Side, btw

[whole urchin]

i did that, it doesnt work

[chrome bay]

Probably something project specific then

[whole urchin]

in blueprint please

[chrome bay]

Same deal

[dark edge]

Is the vehicle a pawn?

[whole urchin]

yes

[dark edge]

Are you calling possess from the server or client side?

[whole urchin]

server side

[dark edge]

Screenshot your blueprint so we can take a look at it

[whole urchin]

does it make a difference if i do it in controller blueprint or in some other BP ?

[dark edge]

Your options are the controller or your currently possessed Pawn pretty much

[chrome bay]

The only thing that matters is whether you are calling it on the Server

If you are, and it doesn't work, then it's going to be something project specific

[whole urchin]

this is controller BP

[chrome bay]

Looks fine

[whole urchin]

when i play on server, i possess

when i play on clients, it doesn't possess

[chrome bay]

Check the values of the variables

[whole urchin]

i will

[chrome bay]

Vehicle being replicated is a strange choice for example, usually you'd pass the vehicle pointer with the Server RPC

Since presumably the Client is the one who decides what they want to possess

[whole urchin]

it returns nothing, why

[chrome bay]

you mean Vehicle is null?

[whole urchin]

it was a foolish attempt, ignore it please

[whole urchin]

yes

[chrome bay]

There ya go then 😄

[whole urchin]

but why

[chrome bay]

Well obviously you haven't set it

[whole urchin]

i have set it

[chrome bay]

If it's null, you haven't set it (on the Server)

Again though, why the Vehicle variable is replicated and not being passed via the RPC I don't understand. The setup is odd.

[whole urchin]

i did it at BeginPlay

ignore the variable replication, it was a stupid test

[dark edge]

@whole urchin show where you are setting that vehicle reference

[whole urchin]

so you're telling me to get the reference for the vehicle in the server RPC ?

[dark edge]

Does the spaceship exist on the map already?

[whole urchin]

ya

[dark edge]

Try get all actors of class and print the length of that array. Should be one right?

I've never seen that node get actor of class

[whole urchin]

i got the vehicle reference in server RPC, and it possessed

succeess

but i can't move it

any suggestions ?

[chrome bay]

By default vehicles have no network functionality, it's worth pointing that out

I think the engines' Wheeled Vehicle Movement has some very basic code, but otherwise nothing else does (e.g, floating pawns etc.)

[whole urchin]

but isn't FloatingMovementComponent (or any move component ) already replicated ?

[pallid mesa]

no, a shape of that will come soon with Dave Ratti's net prediction plugin

[pearl pollen]

Hi everyone! Quick question. I want to add a component to my character that is only relevant on the server and I'd like for it to not exist on the client. HasAuthority somehow wrecks everything because the non-existant component's BeginPlay gets called anyway. Is this approach wrong all together or am I doing something wrong?

[pallid mesa]

I am uncapable of understanding your question

where are you adding this component?

[chrome bay]

Relevancy is done at the actor level, components can't have different relevancy from their owning actor

[pallid mesa]

yeah you should use owner's relevancy

[chrome bay]

I mean you could add a (non-replicated) component at runtime on the Authority only, but that's probably not the most resilient design

[pearl pollen]

Okay so if I want the component on the actor, I add it anyway and just block it from being used on the client, is that right

?

[chrome bay]

If you want it as a default component then yeah, you don't have a choice really

[pearl pollen]

Yeah I want it in there by default

Okay thanks. I'll figure out a way to do that.

[pallid mesa]

maybe you can dance a bit using with_editor pragmas xD

[pearl pollen]

I wouldn't want to mess with that 😛 Too many unexpected crashes

[gritty pelican]

hello, how i can serialize float?

[glossy veldt]

When adding my Widget through the Character it gets added to every players viewport - when there are 2 players there are also 2 widgets on top of each other
Any help?

[empty axle]

@glossy veldt check for IsLocallyControlled before adding the widget to the screen

[glossy veldt]

@thomson so before add to viewport?

[empty axle]

After BeginPlay before GameplayPCRef

[glossy veldt]

Alright will try

now the client only has his hud but the listen server has no Hud

When i do it like this

@empty axle Any idea why?

[glossy veldt]

Nevermind - fixed it by adding a small delay after begin play because for some reason the listen server takes a bit of time to register that he is controlling a pawn

[marble gazelle]

never use delays for these cases!!!!

[silent valley]

Can I add a replicated Actor to a Level?
How does the ownership/replication work?

actually I already KNOW that I can 🙂
I guess what I'm wondering is, if an Actor is loaded as part of the world on a Client, is it possible to know when it has been initially replicated?

[marble gazelle]

AActor::OnActorChannelOpen on client and OnSerializeNewActor on server

[silent valley]

Thank you Sir!

[winged badger]

the Pre/PostNetReceive seems a less intrusive way to go about it

[lament sinew]

@chrome bay in your blog post (https://jambax.co.uk/replicating-uobjects/) you mentioned that subobjects are always created client-side using the Actor as the Outer.
So replication that way is client -> server?
I mean the client would create the object just fine but it would get replaced by the server object.
The one the client allocated would eventually be garbage collected right?

[thin stratus]

Replication always goes from Server to Client, despite ServerRPCs.

[winged badger]

@lament sinew when client instantiates a subobject from replication (AActor::OnSubobjectCreatedFromReplication() callback)

those subobjects always have that Actor as the Outer

no matter what the Outer actually is on server

[covert anchor]

So I've been following tutorials on building a multiplayer server, I wanted to dive deeper into how things work does anyone have any good learning resources?

[steel fox]

I'm making a fps game and the whole "weapon fire" process runs on server, how can I update players hud ammo in clip information? Should I get my weapon ammo amount at each draw?

[rich ridge]

@steel fox is your gameplay Logic coupled with GAS?

[lament sinew]

@winged badger yeah it should be constructed with the GUID from the replicated data, but i thought i needed to create it again on client-side lol

[winged badger]

the point was

[rich ridge]

if no then you need to have one variable like UPROPERTY(ReplicatedUsing = onRep_AmmoCountCallback) AmmoCount;

[winged badger]

if your say, ActorComponent on server instantiated a replicated UObject with itself as Outer

[steel fox]

GAS? Never heard of it, I don't think

[winged badger]

when that UObject replicated, its Outer would be the Actor that replicated if over its channel

[steel fox]

Ok thank you

[winged badger]

not the ActorComponent

[rich ridge]

and whenever AmmoCount is updated by server, client will get callback in that function

GAS -> GameplayAbilitySystem #gameplay-ability-system @steel fox

[lucid vault]

I have an ADS camera attached to my gun. I use set view target with blend when pressing / releasing the aim key.

The blend works nicely if you let it complete fully. The problem is that, if you stop aiming during the transition to the ADS camera, which sets the view target back to the original camera, it snaps instead of blending

I guess you can't blend to the current view target. Is there any solution to this problem that doesn't involve manually moving my camera for ADS?

bLockOutgoing won't work, because the camera will get offset by player movement

[wise bridge]

Hello everyone, I still have an issue with my aim offset replication, so here's a few questions to help me fix this one time for all :
-What value should I use to replicate the pitch accross the network? Control Rotation?
-What variable should keep this value? Default camera?
-After that, I just use my Default camera rotation for the AimOffset right?

Here's a picture of my BP_character. And last question : to trace a line, I just have to use my default camera location for the starting point, and my default camera getForwardVector multiplied by a value for the end point right?

[steel fox]

Does anyone know how to replicate a variable in animinstance to all players? I know it's a UObject, but I have that integer variable that changes the animation pose and I was trying to set it when I need to instead of getting it at each anim update, using UPROPERTY(Replicated) and setting it on Server or Multicast changes the value only on the server and not on the other clients(on listen server the new pose is visible while on clients it isn't)

[dark edge]

@steel fox why does the state need to be replicated inside the anim instance?

@wise bridge control rotation is owned by the controller.

[steel fox]

I can just get it from the player pawn, but I was curious to know if it's possible to replicate it, I need it to know wich animation pose to play(AnimationBlendspace)

[wise bridge]

So, what should I use ? Default camera for replication?

[meager spade]

@steel fox animinstance is local only. No replication at all. not even RPC's.

[steel fox]

That's because it's a child of UObject right? UObjects really have no way to have replicated things?

[meager spade]

no

its cause its local only

UObjects can replicate just fine

if setup properly

but animinstance should and is local only

[pure frigate]

@twin juniper That's not a bug, that's just latency. If something is spawned on server, it has to replicate back to the client which takes time. Once it replicates, it has already traveled X distance on server. You need to spawn a local projectile on the client that doesn't have damage authority, and does what the server's projectile does.

[unkempt wing]

Anybody know how to make certain components of actors only visible to certain players?

[paper lava]

hey guys, I really need some help.. I've been at this the entire day - no joke..

I can't figure it out. I have a custom replication graph and when it is active, I can't do multicasts anymore.

The problem is also there in ShooterGame.. I created a custom player controller and added this: https://i.imgur.com/iJfCpFJ.png

As soon as that multicast hits, the player controller dies.

This is my output:

LogBlueprintUserMessages: [BP_PlayerController_C_0] Server: MULTI! ---------------------------------------------------------------------
LogBlueprintUserMessages: [BP_PlayerController_C_0] Client 1: MULTI! ---------------------------------------------------------------------
LogBlueprintUserMessages: [BP_PlayerController_C_1] Client 2: MULTI! ---------------------------------------------------------------------
LogNet: Received invalid swap message with child index -1
LogNet: Warning: UNetDriver::ProcessRemoteFunction: No owning connection for actor BP_PlayerController_C_0. Function Servertest will not be processed.
LogBlueprintUserMessages: [BP_PlayerController_C_1] Client 2: PC got destroyed

LogNet: Received invalid swap message with child index -1 this line appears, soon as I do the multicast and later, the PC has no owning connection anymore.

It seems to me for some reason there is a swap in player controllers and that fails because INDEX_NONE is sent. ( https://github.com/EpicGames/UnrealEngine/blob/release/Engine/Source/Runtime/Engine/Private/NetConnection.cpp#L3708 )

I've tried to have PlayerController replicated in every which way I could come up with but it doesn't change anything.

I really need this to continue with my project. Help would be highly appreciated

[lethal oracle]

Seems related: https://answers.unrealengine.com/questions/706243/view.html

[paper lava]

oh man, damn

@lethal oracle thanks so much

it's weird though, I don't have these problems when the replication graph is not active

[lost inlet]

you wouldn't use multicast on a player controller though

[lost inlet]

^

[paper lava]

@lost inlet yea, I got that part now.

I'm relaying over the GameState now

[lost inlet]

that or player state are likely the appropriate places for whatever you're doing. only the owning client knows of their player controller, you don't know about anyone else's

[paper lava]

right, yea, that makes sense.. somewhat

[ivory lintel]

Hello Friends,
I have some question about MP in UE4. Games like Fortnite have dedicated server hosted on the Epic Games. They handle with multiple matches at the same time. When X player check ready to play, the master server group they and setup a new sub server and/or session to this players. Each match have X clients at start playing the same game with the same level with the game mode and game state of the specific match.The questions:

1 - Each match have a dedicated server process (.exe) running to control the game or the same dedicated server process can handle multiples sessions of the same level at different times?
2 - What is the best way to develop this solution to control and handle multiples matches of the same level?
3 - Any tutorial or documentation of how to do this?

I'm glad if someone can help me to understand the logic flow and give a tip of how to replicate this
^^

[hollow eagle]

1. A single process usually only handles one game, there are lots of additional complications that likely aren't worth it if you want to handle multiple games.
2. It's completely up to you to develop something like that, or you can use a service like gamelift or playfab.
3. You're on your own for that unless you're using a service as mentioned above.

[ivory lintel]

ty for your answers😆

so...I'm trying to create a solution of master and sub server

when the master server / lobby got X players ready to play, him create another process of dedicated server running the level and travel all players...it's right?

[hollow eagle]

You definitely do not want a "master" server to be a real dedicated server instance. There's no way that can scale up for the number of players.

Any lobby or matchmaking system will involve a number of web services that the game calls out to in order to find out where to connect

this is an area unto itself, no one here is going to be able to tell you how to make all that because it's a massive undertaking

I suggest looking at the services I mentioned first. If you still really want to build something yourself then at least read up on how those services work.

[ivory lintel]

understand

ty any way^^

[quick flint]

Does OnRep call on other players clients

[lost inlet]

I'm not sure what you mean, an OnRep or what?

a Blueprint OnRep will call on the server and each client the actor is relevant to

in C++, you have to manually call an OnRep function after changing a variable on the server (if desired)

[lost inlet]

probably should be using the reply feature more

[quick flint]

Oh sorry I mean a ReplicatedUsing function

[quick flint]

If an OnRep/ReplicatedUsing function is triggered, who is it triggered on? The owning client only?

[lost inlet]

the replication system calls it on every client where the actor exists

ie "is relevant to"

if the actor is dormant, you won't get it until the actor is awake or you force a net update but that's getting more advance than what you're looking at

[lost inlet]

you'd only get it on the owning client if the replication condition was COND_OwnerOnly

[quick flint]

oh

[vivid seal]

how do u guys typically handle race conditions between replicated objects and the pointers to those objects inside the things that own them? for example i have a component with an array of uobjects, it replicates all of those objects. However in the callback from changing that array (adding a new object on the server), the object will sometimes not be replicated when the array says it has a new item, or vice versa, the object might check something and see it is not yet in the component's array of objects.

is there a BeginPlay equivalent in UObject that I could use to just tell the owning component that the object is ready to go when it's actually done replicating

[winged badger]

fastarrays will fire changed when the object is resolved if its null during add

and typically its not a race condition - same array always fails or always succeeds

[vivid seal]

so in postreplicatedadd I can check if the pointer is null, and if so do nothing, then in postreplicatedchange check if its valid and then actually add it to the owner's array if it hasn't been?

er, it'd already be in the array, but i'd know i can use it

i feel like every time i have a multiplayer question in here its just you popping in and telling me how fastarrays are here to solve all the problems lol

[winged badger]

you can also swap the position of Super in ReplicateSubobjects

make the array replicate after the subobjects

[oblique coyote]

Hello I am having an issue in my VR project I'd like to ask for advice, here's my forum post of the issue https://forums.unrealengine.com/development-discussion/blueprint-visual-scripting/1854196-issue-with-possession-and-setting-vr-hand-owner

the gist of it is that I am trying to take ownership of som VR hands but on the client side, I've only found success by running the setup on tick which is bad

I'd love to hear any ideas about what is going on

[silent valley]

I am making a multiplayer test class and I need All my clients to be able to send information back to server. However I dont want to modify the Pawn. Any ideas for this?
I'm currently thinking spawn an Actor for each client, owned by the client that can RPC to the server, but is there any nicer ways?

[winged badger]

if its a test class

can be just an ActorComponent

RPCs work just fine, as long as they are on owned Actor @silent valley

[silent valley]

yes this is the problem 🙂
there is 1 test class owned by server, but how do each of the clients report back to the server?

I have decided to create a client owned Actor for each client, and use that

[winged badger]

replicated ActorComponents on owned Actors can RPC

standalone Actor works too, as long as its spawned and had SetOwner called on it server-side

[minor jackal]

Hi all for multiplayer, in which class i should implement spawn item logic. I have created logic in game mode and used multicast event for spawn. But i can not see items on each client. I believed spawning should be done only by server so i used game mode BP

[wise bridge]

No one? I'm still stuck on this :(

[silent valley]

That's fine - depending how your items are setup, you will also need to spawn the Items on the server. You can do that in GameMode too after spawning the player pawn.

[minor jackal]

@silent valley Thank you, I have item actor class with mesh. Then in game mode i am looping for all target points and spawn actor at target location. So why items are not visible on clients?

Just to test i created this in game mode. but i get output "Server: Hello" .

Is this normal?

[silent valley]

possibly in BeginPlay the clients are not yet connected

[minor jackal]

how would i check if clients are connected in game mode?

[silent valley]

oh and also GameMode only exists on Server not clients

so NetMulticast will not send anything to any clients anyway

[minor jackal]

i see, so this means, i cant spawn items to all clients in game mode?

[silent valley]

hmm I think you're missing something fundamental here

[minor jackal]

i am sure, need to learn

[silent valley]

generally, only Server should create Actors, and any replicated actors created on Server will also be created on all clients

so if you use SpawnActor in GameMode (server) to spawn a Replicated item actor, then it should also be created on clients too

[minor jackal]

i see so i need to tick "Replicated" checkbox inside actor itself

[silent valley]

yes, inside your item actor I guess

[minor jackal]

Done, but same result. @silent valley

[silent valley]

is the item definitely visible on server? and it's just not appearing on client? you using Listen mode?

[minor jackal]

yes 100%

[silent valley]

idk, it should be working based on what you've told me 🤷

I'd put some more logging in and try and narrow it down that way

but I've not really used BP

[minor jackal]

I am going to test it on a fresh game, let me see if i can grasp some understanding. I hope you dont mind if i ping you with DM.

[silent valley]

just ask in this channel please

[minor jackal]

ok, thanks

[ember osprey]

Would the GameState be the best place for AIs to generate random patrol points so every AI on clients will sync their patrol or is there a better place for that?

[meager spade]

AI is server only

so why would clients need to know?

[ember osprey]

So doing it in the AIC would work then?

[meager spade]

well AI is server only, so you don't need to replicate anything to do with its path

just generate it on the AI or make an AI director if you need predefined random paths or w/e

[ember osprey]

AIC is server only but not the pawn correct?

rn each AI creates its own random location on each client version.

so putting random locale on the pawn is not working rn.

This is on the pawn:

//We get a random direction, then set a location based on that and finally we clamp it to the navigation zone
FVector AAI_Bot_Pawn::GetRandomLocationInNavigableZone()
{
    return ClampLocationToNavigableZone((UKismetMathLibrary::RandomUnitVector() * MovingRange + GetActorLocation()));
}

[meager spade]

why would the pawn handle that?

[ember osprey]

and each client does its own version of it

[meager spade]

ai controller should handle all movement

[ember osprey]

problem is I cannot get the AIC from the controlled pawn in multiplayer, so how do I tell it that it needs to generate a new patrol point?

[meager spade]

why cant you

[ember osprey]

I think thats the framework.

I tried to do it and it wouldn't let me obtain the AIC from the controlled pawn in multiplayer.

pawn on client and AIC only on server.

the server pawn could get it I believe, but the client pawns could not.

wouldn't*

not would

[meager spade]

ofc

why would the clients need it?

clients should never access a controller

[ember osprey]

my pawn navigates and then collides and then generates random location to go to.

I tried to get the AIC from the AI pawn to tell it to generate a new patrol point, but I cannot get the AIC from the pawn.

[meager spade]

but that is all server only

so why does clients need to access it?

[ember osprey]

I think I may be spawning the AI in the wrong place.

So its not all server.

[meager spade]

AI should be spawned server only

dont forget, pawn overlaps will fire on client and server

[ember osprey]

This would only be the case if I was spawning through game mode correct?

[meager spade]

you just need to gate it with HasAuthority

or some server check.

[ember osprey]

ok

So you saying...

When collision, use hasauthority check to get new control point. right?

in the AIC

I'm spawning AI in a spawn manager that is in the level already and not spawned.

So I think it exists on the client and server when AI is spawned.

[ember osprey]

This will always fail unless I wrap it in a HasAuthority check correct?:

AAI_Bot_Controller* AIBotController = Cast<AAI_Bot_Controller>(GetController());

[meager spade]

yes

your ai should only spawn on server

and be replicated

so your spawn manager needs to make sure its only running spawns on the server

[ember osprey]

so I should put that in a HasAuthority check in the spawn manager then.

[meager spade]

yes, and overlaps on pawn run on both client and server, so those will need to be gated

[ember osprey]

the pawns will replicate automatically?

[meager spade]

if they are ACharacter based

yes

or APawn

and your movement comp is replicated

[ember osprey]

ok, thanks I'll get on it. I appreciate it. BTW, I'm not using the movement comp Im using the Network Prediction plugin.

for movement.

[rich ridge]

does this work?

[polar wing]

Woah, just started upgrading my project to 4.26. They really rewrote the CMC's RPC system! Not sure if this was mentioned here already? Probably 🤣 .
"// Use newer RPCs and RPC parameter serialization that allow variable length data without changing engine APIs."
To see the changes, you can run the UCharacterMovementComponent from 4.25 against the 4.26 variant in a diff checker. Same thing goes for the accompanying stuff in ACharacter.
I had to rewrite some parts of the engine API (but managed to get it done in child classes) to make up for the severely-limited compressed flags before, due to a large amount of custom move data. Looks like I'm going to have to do another big refactor to use the new API while I wait for Network Prediction to mature. Just wondering if that's really the best use of my time right now, without knowing when Network Prediction will hit production-ready (I'm still playing around with it, but not in my production project).

[chrome bay]

They have already started on a Net-Prediction driven character movement setup IIRC

There's early parts of that in the Master branch already

[wise bridge]

Hello everyone, I still have an issue with my aim offset replication, so here's a few questions to help me fix this one time for all :
-What value should I use to replicate the pitch accross the network? Control Rotation?
-What variable should keep this value? Default camera?
-After that, I just use my Default camera rotation for the AimOffset right?
Here's a picture of my BP_character. And last question : to trace a line, I just have to use my default camera location for the starting point, and my default camera getForwardVector multiplied by a value for the end point right?

[polar wing]

That's exciting!

[ember osprey]

yes it does. wait....which part?

[rich ridge]

with GAS? as per my understanding it isn't compatible with GAS at the moment. @ember osprey

[ember osprey]

I'm using gas and Network prediction plugin for movement.

[rich ridge]

but the RootMotion tasks are coupled with MovementComponent?

so how does it work?

[meager spade]

@polar wing don't expect network prediction to be stable for at least a year

[chrome bay]

I don't think they've checked anything into that plugin for ~3 months now

Probs busy with UE5

[rich ridge]

https://github.com/EpicGames/UnrealEngine/blob/master/Engine/Plugins/Runtime/GameplayAbilities/Source/GameplayAbilities/Private/Abilities/Tasks/AbilityTask_ApplyRootMotionJumpForce.cpp

It uses MovementCompoent

[ember osprey]

I have to port all my AI navigation code over to the Network Prediction code.

I only had my player movement setup through NPP.

[ember osprey]

wdym?

[chrome bay]

Shouldn't have to, the point of network prediction is to write the simulation code only

And not have to think about the networking side (as much), let alone AI side

[ember osprey]

I used GAS to set movement parameters on my player's movement that is setup using NPP.

[polar wing]

Gotcha! I will do my little rewrite to the CMC, then.

[rich ridge]

is this for me?

[chrome bay]

Nah, 8ball

[ember osprey]

the same way you would with CMC but with NPP

[ember osprey]

???

[chrome bay]

I'd be careful writing any code based on the NPP right now, it's going to change a lot still

[ember osprey]

oh IK

lol

[rich ridge]

this i can do. But the AbilityTasks are already tightly coupled with MovementComponent not NPP

[meager spade]

Dave said the core is pretty much set though

[ember osprey]

I tried to update to 426 and its different.

i'm using 4254 rn.

[meager spade]

in 4.26

[ember osprey]

lol

too late

[rich ridge]

does this mean the abstract APIs are frozen for NPP?

[ember osprey]

I would have to refactor everything to work with 426 NPP

[meager spade]

i assume that is what he means

[chrome bay]

Not going to bet the stability of my project on that though 😄

Not until they go full official anyway

[meager spade]

though, virtual FVector ComputeVelocity(float DeltaSeconds, const FVector& InitialVelocity, const TNetSimInput<CharacterMotionStateTypes>& Input, const TNetSimOutput<CharacterMotionStateTypes>& Output, const FVector& LocalSpaceMovementInput) const;

specifically const TNetSimInput<CharacterMotionStateTypes>& Input, const TNetSimOutput<CharacterMotionStateTypes>& Output

[ember osprey]

I'm doing a Flying game.

CMC doesn't work well with flying.

[meager spade]

this is used in a lot of functions

suprised its not a Templated wrapper

[chrome bay]

🤢

[meager spade]

for both of them

[ember osprey]

yeah lot of that in there.

[meager spade]

as they are the same, and require the same type

just one is input and one is output

so it could have been one variable/param

```using CharacterMotionStateTypes = TNetworkPredictionStateTypes<FCharacterMotionInputCmd, FCharacterMotionSyncState, FCharacterMotionAuxState>;````

hmm

[ember osprey]

Well I get to implement the AI navigation manually using NPP this time. I love doing movement network replication. lol. jk it shouldn't be too hard

@meager spade btw, I had to wrap everything in the Role Authority if, that makes the pawns sync ,but now the since everything is on the server movement is not replicated and I have to implement it myself because i'm using the CMC. Thanks for the help.

[rich ridge]

I m going to explore NPP, so anyone can point to the branch which contains latest NPP commits.

[ember osprey]

https://github.com/EpicGames/UnrealEngine/tree/master/Engine/Plugins/Runtime/NetworkPrediction

[rich ridge]

thanks, I already have master built on my machine.

[ember osprey]

👍

[ivory lintel]

Hey Friends,

i'm trying to test the shortcut to run dedicated server from .ink with -server -log parameters but dont work...the console lock on this image below...

[distant talon]

So, it seems like an initial only replicated variable will act as intended and withold from spamming replications if you call a force net update... unless you set your actor dormant. because of course a dormant actor is what you want spamming your big initial only replicated struct.... UE, why....

[ivory lintel]

trying to use: "C:\Program Files\Epic Games\UE_4.26\Engine\Binaries\Win64\UE4Editor.exe" "C:\Users\Leandro\Documents\Unreal Projects\MyProject\MyProject.uproject" -server -log

[chrome bay]

@distant talon Dormancy closes the actor channel, so when you wake it up again it's treated as a new actor

[potent token]

hi guys,
i have 2 maps (Lobby and InGame)

[distant talon]

That's so confusing, but thank you for the explanation

[potent token]

i set dedicated server with default server map InGame

[ivory lintel]

Any tip why my server is locked?

[potent token]

then i try to open level

but the log says
"LogNet: Browse: 127.0.0.1/Game/Maps/Lobby"

when i open level from client

[ivory lintel]

bind 0.0.0.0 on 7777 port???

[potent token]

i dont know why it browse the wrong map

[potent token]

yes

[ivory lintel]

how to fix?

[potent token]

sorry i respon to wrong message

i have question too regarding dedicated server

[winged badger]

@distant talon Dormancy is tricky to use properly

as Actor becoming relevant for some client won't wake it up from dormancy

meaning that client will have the wrong state

[ivory lintel]

Hey Friends,

Trying to test my project as dedicated server using the shortcut method like "C:\Program Files\Epic Games\UE_4.26\Engine\Binaries\Win64\UE4Editor.exe" "C:\Users\Leandro\Documents\Unreal Projects\MyProject\MyProject.uproject" -server -log

the console open but freeze on [2021.01.19-14.42.24:803][ 0]LogInit: First time updating LLM stats...players cannot connect...like image above...

This does not work anymore? How to fix? Have a new way to do this?

i'm using 4.26.0 UE version from launcher

[twin juniper]

So I just need to spawn a client-side dummy to mimic the real projectile on the server?

[pure frigate]

Yeah: https://en.wikipedia.org/wiki/Client-side_prediction

[jaunty plank]

hello guy i cannot get the defualt online subsystem to connect over the internet im using this code to find the first server for now and join, if there no server then it will just create a new one . it works in edditor but when over the internet it just cant find any sessions

[meager spade]

SubSystemNull is kinda not built for internet

its for LAN

[jaunty plank]

oh is there any way to do it without using steam i want to keep away from steam for this project

[atomic acorn]

How does one fix this issue?

is there a way to get the local player controller?

[winged badger]

don't create widgets for client controllers on listen server

[atomic acorn]

where do I create the widget?

oh

can I do if(!Authority) to stop that?

or whatever it is

[winged badger]

no

you should create widgets only for controller that IsLocalPlayerController()

and you should be creating them from a AHUD derived class

which would eliminate the problem, since HUD is always local only

[atomic acorn]

oh right

I've bollocks'd it up then, cheers

[dark edge]

So is the ideal flow Controller -> HUD -> makes widget?

[winged badger]

Controller doesn't have anything to do with making widgets

except being pushed around as Owner

in my game

its not hard to access the HUD directly and request a widget on screen, from anywhere

[chrome bay]

+1 for using HUD class as a widget manager

Works very well

[woeful ferry]

Hi,

I stumbled upon this in the documentation: "FNames are not generally compressed, so when you are using them as parameters to RPCs, keep in mind that they will generally send the string each call. This can be a lot of overhead."

What should I use instead FString, FText?

[chrome bay]

FName / FString are fine over network

So really it depends what you're sending / using the RPC for

Definitely don't use FText

[woeful ferry]

I have a huge replicated array of available recipes that the players on the server can use, in the game state. It consists of the RowName and a bool to check if that recipe is unlocked.

[marble gazelle]

You could use GameplayTags and enable to use indices to replicate them. this requires that data on all peers is the same

[chrome bay]

If it's a huge array, don't use a data table

[marble gazelle]

or FGuid, 128bit

[chrome bay]

Use a data asset and a big array, access elements by indices instead

Commonly-used FNames such as NAME_None will serialize as an integer, but these names are hardcoded. All other names will serialize as strings.

[woeful ferry]

I'll look into them, never used them.

[chrome bay]

A data table is just a glorified TMap<FName, SomeData>

[marble gazelle]

can't you add such hardcoded names?

[chrome bay]

Yeah you can modify UnrealNames.inl and add more

Less useful for a data table though IMO

[marble gazelle]

yeah

[chrome bay]

GameplayTags FTW though, where it makes sense anyway

[wise bridge]

Hello everyone, I still have an issue with my aim offset replication, so here's a few questions to help me fix this one time for all :
-What value should I use to replicate the pitch accross the network? Control Rotation?
-What variable should keep this value? Default camera?
-After that, I just use my Default camera rotation for the AimOffset right?
Here's a picture of my BP_character. And last question : to trace a line, I just have to use my default camera location for the starting point, and my default camera getForwardVector multiplied by a value for the end point right?

[jaunty plank]

i hav emovced to using the steam subsystem and am still having the same issues as befor eothers cannot connect

[jaunty plank]

from my debuging the find sessions is sucseeding

but anything after the for loop dosent get called

at this point im really consused to whats wrong

[dark edge]

That looks like it's gonna join a session AND create one...

[jaunty plank]

im trying to get it so when you click the button if there is no session then it creates one and if there is a session it will join it

[dark edge]

You for sure won't want that loop

[jaunty plank]

would i just use a get

copy

[dark edge]

check if array size is 1 then get if it's size 1

if it's size 0, create

[jaunty plank]

ok

[jaunty plank]

sort of like that

[jaunty plank]

this still dosent work over the internet with the new solution i don't know why still not having it work i have the INI file set up correctly and testing with 2 different PCs on different networks

[dark edge]

Start with 1 button to create and 1 button to find, and see where it's failing

don't start with trying to make 1 button do everything when you haven't tested the individual parts

[jaunty plank]

i have fixed it it was somehting with the shole system

[opaque flicker]

hey we are using UE 4.22.3 , trying to save player EXP in steam leaderboards
using two blueprint nodes "Write Leaderboard Integer" and "Read Leaderboard Integer"
We have a stat on Steam 'PlayerEXP_PlayerEXP" and a leaderboard "PlayerEXP"
After launching the game so that steam kicks in, then create a steam session, "Read leaderboard value" always return 0 and "write leaderboard value" always return true.
however values never save in the leaderboard on steam what so ever
any clue ?!

[wise bridge]

No one ? I'm stuck on this :(

[shadow arch]

hey guys, any idea why the local client running a simple move to would see the character "hopping" or skipping around all glitch like

whereas the non local client would see him moving just fine?

i have a feeling it's a replication issue on the player character but not too sure how to sort that

[dark edge]

@shadow arch where is the move to being called from?

[cosmic trail]

Is there a decent way to sync ragdoll physics between client and server when the ragdoll is cut into pieces and blown up? Or is there just going to be too much variation there between client and server?

[pallid mesa]

Is it essential for your game?

[cosmic trail]

I would really like it to be, basically tanks explode and mechs explode leaving corpses

and the corpses take space and provide cover

[pallid mesa]

interesting, physics replication is challenging but if you do this thing server authorative and you just smooth out server results in the clients you could get around that

[cosmic trail]

I'm not sure how to smooth out the entire pose of the skeleton I guess, it's a lot of info..

[shadow arch]

@dark edge the move was being called on the client instigated by the server. But in the end i solved it. It turned out that there was a setting to enable clientside navMesh. With that setting off the server was moving the character but the client couldn't so that resulted in the client/server synch just teleporting the client character constantly

[pallid mesa]

Usually these limbs are separated objects you can replicate @cosmic trail

[cosmic trail]

@vo oh hmm I guess I've been using the physics assets' constraints breaking ability, which splits the limbs but keeps it as part of the skeleton

[pallid mesa]

let me explain you a real example: Headshots and flying limbs.

take a look at robo recall

usually what you do is you reduce the scale of the bone in the skeleton, and you spawn a limb actor

copying rotation and location of the original arm/head/etc

that way you can apply impulses to separated physics bodies

[shadow arch]

yeah so say you want to rip off a head you shrink the head of the original to 0, create a clone of the original actor and shrink everything except the head

[pallid mesa]

not exactly because it doesn't need to be a skeletal mesh

it can just be a mesh

of just the head, or the arm, or whatever

[cosmic trail]

ah ic, that's the way to replicate it more easily?

[shadow arch]

yeah but the mesh of the arm/head etc needs to be made . I'm assuming the quick and dirty way of i don't have models for those ;P

the good old GTA san andreas method lol

[pallid mesa]

yes and no

that solves the problem of being able to treat limbs separately

however, physics replication is still challenging

[winged badger]

@cosmic trail i think Halo GDC touched that topic

[pallid mesa]

hello zlo!

[winged badger]

hey vorixo

[pallid mesa]

😄

[cosmic trail]

@winged badger recent GDC?

[winged badger]

fairly recent

[cosmic trail]

@vo thanks for the info, seems like I got a bit of work to do

[winged badger]

watched it a year ago or so, all the stuff was still relevant

good watch

[pallid mesa]

if it's a year old

then it's definitively state of the art in limbs replication XD

[winged badger]

i think they replicated the initial state of ragdoll

then let the clients handle the rest on their own

that would be your post-explosion state

[cosmic trail]

ah, I need them to be in sync afterwards too, I got them already in sync pre-ragdoll

[pallid mesa]

yeah... that will bring some errors between clients

yeah you can sync them twice

once initially

[winged badger]

they did talk about it for a couple of minutes

so more then i remember rn

[cosmic trail]

cool, which Halo btw if you still remember?

[winged badger]

https://www.youtube.com/watch?v=h47zZrqjgLc

think it was that one

[cosmic trail]

@winged badger Thanks!

[winged badger]

not 100%

but if its not, its well worth a watch 😄

[cosmic trail]

yeah, looks good. need to keep learning more about networking

Be nice if unreal could supply a deterministic physics setting in the future... 😄

[pallid mesa]

https://youtu.be/h47zZrqjgLc?t=4610

timestamped it @cosmic trail

[winged badger]

watch the whole thing 🙂

[cosmic trail]

haha thanks @pallid mesa

[pallid mesa]

I was looking for it as well haahaha

[meager spade]

In this 2011 GDC session, Bungie's David Aldridge discusses the programming that drove Halo: Reach's online networking.

😄

[wise bridge]

Still no one?

[cosmic trail]

Is there a way to force an RPC to go through to the server from the client when the server is the one that is authoratative on the client actor?

@wise bridge u need to keep the variable in whatever actor is shooting, not default camera, and replicate that actor

[wise bridge]

But it doesn't work for the pitch, or I didn't understand what you said

What should take the value of the control rotation?

[cosmic trail]

@wise bridge are you replicating to the server or to the client?

[wise bridge]

Client

Not replicated > Server > Multicast

And I want the pitch of my actors to replicate correctly

[cosmic trail]

so there's a pitch variable on the server, and you want to replicate that to the client?

[wise bridge]

Not exactly on the server

Evey client got his pitch, and replicate it to everyone

[cosmic trail]

pitch is on one client? and u want it replicated to other clients

[wise bridge]

Yes

And also, there's no specific variable created by myself, just control rotation and the rotation of my default camera

[cosmic trail]

ok do it this way: Update the pitch to the server via the player controller, then on the server, update each actor's pitch via that player controller's update, set the actor's pitch variable to replicate to other clients. Set it to NoOwner replication

you can update the pitch using an RPC from the client to the server, Unreliable Server RPC

then after you update the actual Server's actors, who all have a pitch variable, u replicate that down using NoOwner

so that the original client does not re-update itself, but it's client authoratitive

[wise bridge]

Ok thanks, I'm not on my PC right now, I gonna check this tomorrow, I might ping you again if I got any issue. Thanks a lot again dude :D

[steep flame]

The Cheat manager exists on the client and I want to call an RPC to the server but at the same time, I want to exclude the function from the shipping build.
I tried using the preprocessor !UE_Build_Shipping but I can't wrap RPC's with it. This Error happens when I try it.
Error: UFUNCTION must not be inside preprocessor blocks, except for WITH_EDITORONLY_DATA

What can I do to exclude cheat functions properly?

[acoustic marten]

Anyone know if it is possible to do a server travel in unreal in a way that the player do not notice?

Scenario: Say the player is on one server just doing their thing and then the server decides the player needs to be transported to another server (same map, same position just a different instance with other players) and the new server accepts that without the player in question noticing (as long as there are no npcs or players around ofc), doable? If you have any advice then please do @acoustic marten

[meager spade]

@steep flame put them in CheatManager class

they wont be in shipping builds

[hot robin]

There is no way to push the ccu to more then 100players in one spot even with 10-15k poly Models simple skeletons and changing anything will most likely end up with breaking the replication system right?

[bitter oriole]

The polycount is irrelevant here, problem with 100 users is more the CPU performance, bandwidth, game design, etc

[hot robin]

The Main Problem ist that replication is often used in bp and there is not much when it comes to controll so i have to do it in cpp

unless I would do everything in cpp only then would I do more work than necessary

[bitter oriole]

You can do replication either in BP or C++, same thing really

[hot robin]

yes but without control if I were to run a console application that sends back and forth to the client with tcp and udp

[bitter oriole]

Replication is meant to interact with other UE4 instances

If you're not doing that, don't use replication

[hot robin]

That's what I mean, there is nothing left but to bypass the entire system

[bitter oriole]

There is nothing to bypass really, if you don't want to use UE4 for your server, just run UE4 as single player and ignore all UE4 multiplayer

And then build your own multiplayer system

[hot robin]

I just wonder if it's worth it, do you have any experience with it

[bitter oriole]

Depends what you're trying to do, but for the most part, it's probably a terrible idea

There aren't many scenarios where the UE4 replication model is the weak link of your project

Not to mention a non-UE4 server is incredibly tough since you need to duplicate the entire gameplay code in different frameworks or languages

[plush wave]

Does the steam multiplayer test ID actually work?

Like can you test multiplayer between two devices with it?

[bitter oriole]

Yes

Haven't been using anything else yet myself

[plush wave]

Hmm

Wonder why I can't see the other computer's session

Surely there is more setup required?

Otherwise everyone is using the same id... right?

[bitter oriole]

No more setup required, and no

I guess the UE OSS filters by game name

The only setup needed is enabling the Steam network layer and setting your appid in defaultengine

And having two Steam accounts running their own UE instance of course

[plush wave]

Right right

Did all of that

Must be missing something else

Thanks for the help

Do I need to add more than this?

`
[OnlineSubsystem]
DefaultPlatformService=Steam

[OnlineSubsystemSteam]
bEnabled=true
SteamDevAppId=480
Achievement_0_Id=""

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

[/Script/Engine.GameEngine]
!NetDriverDefinitions=ClearArray
+NetDriverDefinitions=(DefName="GameNetDriver",DriverClassName="OnlineSubsystemSteam.SteamNetDriver",DriverClassNameFallback="OnlineSubsystemUtils.IpNetDriver")
`

[bitter oriole]

No, that looks like the usual stuff

[plush wave]

Hmm

No additional config steps?

[bitter oriole]

Nope

[plush wave]

Do I need to have "OnlineSubsystem", "OnlineSubsystemSteam" as public dependancies?

Right now I'm just dynamically loading OnlineSubsystemSteam

[bitter oriole]

If it compiles it's fine

Do you see the Steam overlay in both instances

[random nymph]

I'm currently thinking about safest way to handle player progression in our multiplayer game. We are using playfab and aws services. What I initially thought about using is just the unreal playfab plugin, but in that case the playfab developer secret key is embedded somewhere in the project even though the pak files get encrypted?

What I then thought is to include a text file with a custom secret key only on server build and then use that key to authenticate http calls to aws lambda functions to alter player data. Would this be a fine solution?

[bitter oriole]

Clearly you never want a secret key on clients

Only dedicated servers you control should have it

[random nymph]

Yeah but in playfab plugin the secret key is setted in the project settings

[bitter oriole]

Just don't use the plugin on the client then

[random nymph]

Hmm maybe I can just leave it out from the settings and read it to the requests from a file on server.

[chrome bay]

Pretty much anything you put on a client machine, someone will find a way to access. Safest way is to ensure they never get it

[bitter oriole]

Assume anything ever stored, downloaded, streamed, seen by a client, is now public forever

[pallid mesa]

client authorative models are easy at the begining, not that easy when you need to implement an a.c. 😄

[random nymph]

Yeah this is what i thought

[steep flame]

I put them in the CheatManager class. But I call an Server RPC on the PlayerController that I want to exclude from shipping

[meager spade]

CheatManager can call RPC's also btw

have RPC's sorry

so why not move that RPC to cheat manager?

[marble gazelle]

the cheat manager is replicated?

[meager spade]

does not need to be replicated for RPC's

[marble gazelle]

so just needs a network name stable path? I see

[silent valley]

but it needs to be owned by Client to call server RPC right?

[marble gazelle]

since the PlayerController is "owned" by the client this is given I'd say

[silent valley]

yeah, I was talking about the CheatManager

[meager spade]

yeah you have to implement a function, ProccessRemoteFunction

err CallRemoteFunction

in the cheat manager

{
    if (GetOuterAPlayerController())
    {
        UNetDriver* NetDriver = GetOuterAPlayerController()->GetNetDriver();
        if (NetDriver)
        {
            NetDriver->ProcessRemoteFunction(GetOuterAPlayerController(), Function, Parameters, OutParams, Stack, this);
            return true;
        }
    }
    return false;
}

int32 UKaosCheatManager::GetFunctionCallspace(UFunction* Function, FFrame* Stack)
{
    return GetOuterAPlayerController() ? GetOuterAPlayerController()->GetFunctionCallspace(Function, Stack) : FunctionCallspace::Local;
}```

this can also be applied to any UObject to allow RPC's, as long as you can get an actor that ends up with the owner as PlayerController

[twin juniper]

Hi, I have a question related to client correction. When it is replaying saved movements (from UCharacterMovementComponent::ClientUpdatePositionAfterServerUpdate), apparently it is not using the stored movement mode, so UCharacterMovementComponent::PerformMovement will replay saved movement using current movement mode (that can not be the same as the one when saved move was done initially). Is that correct?

[meager spade]

it should pass in the movement mode no?

if not that seems like an oversight

[twin juniper]

apparently not, or at least I didn't see it. I've seen that the client is replying saved movements with the wrong movement mode, so I guess it is not being passed

[jaunty plank]

hello, what would be the best way to reload a level in multiplayer to get the level back to how it was originally, but still keeping all the same players connected

[bitter oriole]

ServerTravel + seamless travel

[jaunty plank]

would that reload everything i tried to just spawn all the players in reposses them again but some of them they wont become posessed

[bitter oriole]

It would fully reload the entire level and restart

[jaunty plank]

ok thanks for your help mate new to multiplayer trying to learn it

[ivory lintel]

i trying to do a replication of collision response on only players

how i replicate this to players that connect on game later?

how i can replicate this modify state to another player that connect in the future?

using variables?

[bitter oriole]

By replicating the state through UPROPERTY

[ivory lintel]

if i change a variable on the server...when future player connect this will sync automatic to all future players?

need to use repnotify?

[bitter oriole]

Yes

No need to use repnotify

This is regular replication

[pallid mesa]

would it be possible to pin this? @thin stratus

[ivory lintel]

ok will try

[ivory lintel]

if not need to use repnotify, i need to make the changes on construct or where?

beginplay?

[ivory lintel]

when i can check if IsPlayerController and not a AI?

[winged badger]

not 100% controller has a function for it, Pawn does

but it comes to return PlayerState && !PlayerState->bIsBot;

Pawn function is called IsPlayerControlled

[ivory lintel]

ok ty

[winged badger]

for PC it should be obvious its not AI

[ivory lintel]

trying to check if get player state is valid

[winged badger]

AI can have a PS

ours do

its not there by default though

[ivory lintel]

on my project thay have because attributes are on PS

need to check if is a bot

not work...IA reach the set collision response node and player not

[acoustic marten]

But I assume it allows for some kind of connection to other services as well, databases and login servers and such?

[meager spade]

beginplay is to early @ivory lintel

would need a small delay

[winged badger]

that should go into override of OnRep_PlayerState

[ivory lintel]

ok

[meager spade]

that is not possible in BP

[bitter oriole]

@acoustic marten Replication won't work with anything but an UE4 instance , no

You would typically use an HTTP connection for that

Or a plain TCP socket

Replication is a real-time multiplayer system meant for gameplay

[ivory lintel]

replication in ue4 is so confuse

when u learn something you dont know anything^^

[bitter oriole]

Replication is a very simple mechanism tbh

It just copies the server state to clients

There are settings and options but that's basically it

[acoustic marten]

Ah yeah, i wasn't intending to use replication for non gameplay stuff, thanks

[jaunty plank]

i have added server travel it works however the client who is the server will stay in spectator mode and wont be assigned a character im using event handle stating new player but it seems to skip whoever is playing as the server but all other connected clients will be spawned

is there some sort of event whcih fires on the server

[rich ridge]

@jaunty plank Does President of UK do game development?

[jaunty plank]

sadly not would be some intresting games lol @rich ridge

[bitter oriole]

Traveling simply restarts the entire game, so the logic should work out the same as when the host player originally launched

[jaunty plank]

yea that what i read doing research into it however the server always stays in spectator mode dont know why it the server

[atomic acorn]

How do I make it so every client has a version of an actor, so when they overlap a UI pops up for them and only them

Atm it keeps popping up for all clients

And I don’t really care for the server handling the overlap either since cheating isn’t likely to happen in this case

[cinder quartz]

When the overlap fires, check if it overlapped with player, then check if the player is locallyControlled, then pop ui?

[atomic acorn]

Ahh I knew I was missing an if, cheers

[peak sentinel]

Am I* missing something or is it normal that GetOwner()->GetActorTransform() not working on OnRep functions (when playing as client)? In tick, or in non-replicated functions giving correct values but whenever I use it on an OnRep function its returning my old state and server teleports me back to that old state instead of updating my location

[meager spade]

its possible the transform has not updated yet

[peak sentinel]

Its getting updated on local, but server doesnt see it

NetUpdateFrequency is set to 0.9 to demonstrate the movement update

This is my logs, printing location on each function I am using GetOwner()->GetLoc -- Tick Location is not called with OnRep but others are Server RPC or OnRep functions, they dont read the correct value

[fathom fog]

hello guys i've been looking for a course for multiplayer production over unreal engine 4 for a long time, does anyone know a good course for it
I prefer to be from youtube

[meager spade]

http://learn.unrealengine.com

@fathom fog

has some free Udemy tutorials

some are multiplayer

[jaunty plank]

dose anyone have any idea why the server travel feature dosent take the person playing as the listen server out of spectator mode?

[keen grove]

anybody tried using GGPO netcode with Unreal before?

[oak hill]

hey guys, I was wondering if is possible to extend the OnlineSubsystemSteam without having to modify the original classes. Like a plugin or similar

really I need to extend the session interface only

[meager spade]

not that i know off

[oak hill]

oh ok

I've added some Steam functionalities outside of the OSS, but I'm not sure it's a good idea

[meager spade]

iirc its not exposed, so you can't derive from it, also i think you would have to change a lot of stuff

if you are on about FOnlineSessionSteam ?

[oak hill]

oh my god XD

yes, the FOnlineSessionSteam

[meager spade]

yeah it seems like that is not possible i could be wrong, but from my quick 2min look, it does not look promising

[oak hill]

I think you're right, all the classes are private

So I think I've no options but leave my addictional functions outside the OSS

and to write another OSS seems like a nightmare

[hollow eagle]

could probably copy the plugin source and modify it if you want to go that route

[oak hill]

it could work, maybe I'll give a try when I've some spare time

thanks for the suggestion

[cosmic trail]

Ended up being able to multiplayer sync my ragdolls, but pretty costly network updates:

about 60 FVectors per sync tick( every .5 seconds)

think that's ok?

[fossil spoke]

Wouldnt you want to just sync its overall location and let the sim handle all the floppy bits?

Way cheaper, would probably end up looking the same.

[cosmic trail]

similar but doesn't end up the same. I add an impulse on death too, the higher the impulse, the more divergent

I need it the same since I use it to generate cover points and path finding

[meager spade]

does the ragdoll need to look the same? i mean the mesh should not really block stuff

[cosmic trail]

I want to make big corpses and be able to take cover on them, not for smaller ragdolls

[fossil spoke]

Id lerp them to their final resting place once it comes to rest honestly, not during its time simulating.

[cosmic trail]

like ATATs going down

[fossil spoke]

Or try something like that to resolve it to the same place.

[meager spade]

yeah that is a good idea

[cosmic trail]

yeah I did try the lerping final resting place, when it's too divergent it looks quite off

[meager spade]

a lot of people don't ragdoll big things tho, they have death animations

[fossil spoke]

Yep

[cosmic trail]

but its so fun looking? well is 60 fvectors a big load? probably 5x60 over 2.5 seconds

[meager spade]

as long as they are unreliable

you only really care for the final resting vector

and interp the rest

[cosmic trail]

yeah unreliable

[meager spade]

i personally would not do it, but that is personal choice

[fossil spoke]

If its a hero piece and your not expecting to do this X many SMs at the same time, might be a reasonable cost.

But yeah i wouldnt do that either

[cosmic trail]

yeah, I don't expect there to be that many big mechs, but the game revolves around cover. The tanks do this but it's a lot easier with tanks because they don't ragdoll, just one box

[meager spade]

i like to treat my network bandwidth as a precious thing 😄

[fossil spoke]

As you should lol

[cosmic trail]

ok, I'll keep an eye out on it then, will change to quantized too

seems at the border of what you would allow

[meager spade]

yeah, but the best optimization is a death animation, could have multiple based on how it died

[fossil spoke]

Yeah, let the impulse on the Server decide which Anim to choose

[cosmic trail]

be nice if I had more animators to help 😄

[fossil spoke]

🤷‍♂️

[hollow eagle]

another option... don't constantly send updates for all bones. Update a few at a time and let client sim handle stuff in-between updates.
Might work well, might not. Probably worth trying.

[cosmic trail]

@hollow eagle yeah, can do, I can change the update frequency too, it smooths to the new positions every sync

[peak sentinel]

I create a collision from bounds of the actor and simulate ragdoll locally on each connection, it has some problems and not giving what you want (exact same simulation) but it works when you need close results

[cosmic trail]

@peak sentinel that's a good idea!

might change to that and see since that's so cheap

thanks

[dark edge]

Have you tried actually looking at how much bandwidth it draws?

[peak sentinel]

Glad to helped 🙂

[cosmic trail]

@dark edge no, but I have the packet size and frequency so just going off that

[meager spade]

@peak sentinel that works fine, but the bounds will be a square

meaning little gaps will not be passable

which might provide cover

[cosmic trail]

yeah I figure I can make a temporary weird bounds box

the cover points have a grid width apart, so they're not 100% accurate

not FPS accurate, it's more for RTS pathing

[dark edge]

I'd just send it unreliable and make sure you settle into the authoritative resting place and call it a day. The stock physics replication error correction can be pretty smooth if you tinker with it.

[grizzled stirrup]

Is it possible to send a TArray<FMyStruct> in an RPC without it being const?

Reason being if I want to do something to that array on the server I have to make a new copy of the struct array there which feels wasteful

[hollow eagle]

yes

just don't put const on it

[grizzled stirrup]

Oh so I can just send TArray<MyStruct> ?

I think the & was my problem

I was trying to send TArray<MyStruct>&

[meager spade]

no you cant

it has to be a const ref in a RPC, it throws an error otherwise

[hollow eagle]

ah does it?

[meager spade]

yeah

[hollow eagle]

been a while since i tried

removing the & would make a copy anyway lol

[grizzled stirrup]

So only solution is to make a copy on the server then?

True!

[meager spade]

yep

[grizzled stirrup]

Thanks

[meager spade]

i forgot const once on a RPC param

and i had issues, been a while tho, would like to re-check to make sure i am not crazy

[grizzled stirrup]

Man it feels so wasteful though 😦

Copying a big array

Even though it's already there

[pallid mesa]

@fossil spoke maybe worth pinning this?

[meager spade]

i really should put that on my blog

its really useful

[pallid mesa]

it is

[peak sentinel]

You have a blog? Can you send it?

[meager spade]

yeah but i have nothing on it atm

i moved server, never set it back up

might do that the weekend

[peak sentinel]

👍

[pallid mesa]

appreciate it Matt

[limber lodge]

Anyone here do a good referral? I am looking for somebody to integrate EOS or AWS for my project. Paid gig.

[fossil spoke]

Use the Job Board

[meager spade]

When i get my blog post, ill ask for that pin to be changed to the link, as i can provide more details

[limber lodge]

In here? OK will go there

[pallid mesa]

too much mang

[fossil spoke]

@limber lodge #freelance-jobs

Read the #instructions as well.

[limber lodge]

Yeah cant post there.

[steel vault]

I'm reading through the FVector_NetQuantize stuff and they say it's only 20 bits per component. Is that saying each quantized vector only takes 20 bits of information to communicate across the network? I would think you would need at least 24 for 3 values, no?

[meager spade]

@limber lodge did you follow #instructions ?

[peak sentinel]

When bReplicateMovement disabled I cant use GetOwner()->GetActorTransform() on server RPCs because server is not aware of the updated transform, even while I set bReplicates(true); on my actor and movement component. I am not missing something right, thats how its should be?

[meager spade]

it tells you that you need to message the bot

[grizzled stirrup]

Likely wizards at epic doing some magic to crunch it down, also the point of NetQuantize, if it was 24 bits then you'd send a normal vector no?

[limber lodge]

yeah I cant follow the instructions lol:) It wont let me do the command to create a post.

Well thanks for the help. I will just run over to monster or some job site.

[steel vault]

I don't think so because 24 bits would be assuming a uint8 for each of the x, y, z, but they are floats.

[grizzled stirrup]

Ah true!

Huge savings

[steel vault]

I was just saying that me personally, I would make a uint8 quantize with all 3 parts to cut down. I was thinking about trying to help the guy replicate vectors cheaper but even epic was able to do 20 bits which is less than I could come up with. So I was curious how they did that.

[grizzled stirrup]

Likely by shaving decimals off the ends of the floats and combining that data somehow- magic indeed! You can see how they did it in the source I'm sure

[steel vault]

Yea, I'm going through the source as we speak but it I am indeed failing to see the how they did it. Either way, it seems like using their solution is best. So 60 vectors at 20 bits a piece will give you an idea of your network overhead.

I wanted to help him out by simplifying that even more. I was thinking maybe you convert the bone locations to local space which would limit your float values to maybe the hundreds? Then you could get away with even smaller bit representations.

Then you could combine multiple values into one byte possibly

It would be incredibly complex but might cut down a ton on network traffic

[meager spade]

uint8 is only 1 byte tho?

[steel vault]

Right, so I was thinking if you can limit the range of the float you can split one uint8 in two for two values.

Or 3 even

[meager spade]

you cant really use uint8

cause that would not allow negative

@steel vault did you look at WritePackedVector ?

[steel vault]

Haha yea I'm literally on it right now

[meager spade]

its really clever

but you cant use uint8 even if your doing this

unless you divide by a set number and can guarentee it will be less that 255

so you can multiply on the unpacking

[steel vault]

Yea that's something I was thinking of doing

But also a uint8 is 8 bits, no? How the hell do they pack 3 floats into 20 bits lol

[pallid mesa]

20 bits per component

[meager spade]

splitting the float values into components

^

[steel vault]

Ohhhhhh

[cosmic trail]

Ah packing more because local space is smaller. Maybe able to squeeze a bit there. Does unreal do compression per packet?

[steel vault]

I was thinking one vector was a component

[meager spade]

60bits per vector

err sorry

[steel vault]

Thanks that makes way more sense

[cosmic trail]

I’m wondering if the bigger the packet, the more can be compressed per send

[pallid mesa]

it depends on the entropy

[meager spade]

 *        -Scaling factor (usually 10, 100, etc)
 *        -Max number of bits per component (this is maximum, not a constant)
 *
 *        The format is <num of bits per component> <N bits for X> <N bits for Y> <N bits for Z>```

[pallid mesa]

in fp you just sacrifice precision

[meager spade]

what precision in floating points

[pallid mesa]

kekw

[steel vault]

Sacking precision for floating points for bone locations to get a better ragdoll simulation sounds worth it

I was just saying marv that you could do your own form of quantization and pack more info in instead of using FVector_NetQuantize.

[cosmic trail]

Does the scaling factor allow u to do that in the quantized already maybe?

[steel vault]

They appear to use int32

What I was saying is you could use local bone space locations and try to get away with like int8

[cosmic trail]

Yeah that sounds good

[pallid mesa]

you could even do smarty pants things and do your own fixed point arithmetic

to do even a smarter packing

just take a look on how IEEE 754 defines floating point

and do a minor version of it

[steel vault]

Yea I have yet to enter into the realm of optimizing my data structures like this, I was just spitballing. At some point though, I will probably end up trying to do some tricky stuff like this because I too have a lot of transforms to be replicated and will eventually want to ultra optimize.

[cosmic trail]

A little afraid if bit representation on different cpus at that point. Feel like just losing precision is ok, and probably don’t need accurate rotations or velocities either, everything is pretty well constrained. I had to pack physics body velocities

[pallid mesa]

that's called micro-optimising, and it adds tons of noise to the source code

"what the fuck is this" certified moments

[steel vault]

Yea, I would have them off in their own files to prevent that type of stuff or comment it well, but it can be worth it depending on your scenario.

[cosmic trail]

Lol

[pallid mesa]

don't worry about this stuff, as you are defining your own FP minimal interpretation

so you would choose your endianness

the problem would be reading the content of types, but if you are writing directly on bits there is nothing to worry about

[cosmic trail]

Ah ic, but doesn’t sometimes low bit mean high bit when u are using instructions to access?

[pallid mesa]

yes, but your most significative bit will be always in the same place in your interpretation

you aren't re-implementing int

mainly because it is "your" interpretation

you decide where to write the bits

however the standard output will print differently depending the endianness of your final system - based on the container type

but since you are reading bits within positions you are fine

[cosmic trail]

I think I get it. Going to make that a later resort though or maybe find an implementation of ‘small float’ somewhere

[pallid mesa]

k, been working recently in some compression algorithms for the CCSDS (Consultative Committee for Space Data Systems) standards, so I know this stuff 😄

[steel vault]

If your local bone space floats go above 255 try int16

Yea you are going to be the person I come to when I have to work on this stuff myself in the future 😉

[pallid mesa]

NOW... for the context of your game marvg

I would just recommed you replicating the first and the last stance within your ragdoll simulation

using net-quantize

[cosmic trail]

I did try that vo, it doesn’t work well, interpolating looks funky when it diverges too much in the middle

[pallid mesa]

then 3 snaps

[cosmic trail]

3 is probably ok, I’m doing 5 right now

[pallid mesa]

I wouldn't expect that much divergence when replicating the first stance :(

but it happens... reeeee

[cosmic trail]

First thing I did was set it up exactly the same so that’s the first. Then I apply the same impulse to all the clients, then sync in between to keep the divergence in check

I think it’s not a lot most the time, but sometimes it’s very odd

Like maybe something got touched at a particular angle and caused a huge angular velocity

[steel vault]

Yea that's why I wanted to cut down on his vector traffic so he could have the same amount of precision with less traffic.

[pallid mesa]

have you tried different interpolation methods?

[cosmic trail]

I can also clamp angular velocities of the body instances to help

[pallid mesa]

I suppose you did

[cosmic trail]

Yeah I tried a bunch even using blending with animation to the sync poses

It felt very non-simulated when u do that. I wanted the physics feel of things falling and bouncing

[pallid mesa]

👍

you could also skip certain joints if you think replicating those is just unnecesary

[peak sentinel]

Just a random question, are you replicating all bones or some of them? @cosmic trail

[cosmic trail]

Gonna try clamping my velocities too, I feel that’s when things can really diverge when it hits a joint in the right way to cause it to float off far

@peak sentinel all right now since joints can break off

[peak sentinel]

👍

[steel vault]

I assume you are also replicating rotations?

[cosmic trail]

Yes

[steel vault]

Yea so I do think you really need to optimize those structures so you can try to get as many updates as possible in.

[cosmic trail]

Yeah will see what I can squeeze out. Thanks

There's a netquantizeNormal, 16 bits per component

[meager spade]

only for normals

😄

-1 to 1

[cosmic trail]

yeah true, rotations only

[meager spade]

no things like impact normals, normalized vectors

[cosmic trail]

yeah but they mostly describe an orientation?

[cosmic trail]

nvm, realize this as I'm actually trying to convert it

it can describe an orientation but not a directional rotation, since u can have many ways to arrive at that orientation

I wonder if it's enough for bone transforms tho, will have to see

[scenic horizon]

Hi all, probably a nooby question but learning unreal (and how networking is done in unreal) right now, not new to networking in general. Question that's confusing me as to whether I have something set up incorrectly - is the "Print String" screen log shared by all connected clients? I'm playing in the editor as a listen server with one additional client. Have a "Switch Has Auth" on BeginPlay in the ThirdPersonCharacter blueprint (testing in the demo project) that just prints out "We have auth of objectname" or "We don't have auth of objectname" for each spawned player but instead I'm getting more entries in the log than I'd expect.

I'd assume that the game starts up, listen server client spawns in, prints out "We have auth of [character he just spawned] (167)", then the second client is launched, spawns in and the server should print out "We have auth of [client character (C_0)]", and THATS it as we have two characters, so EventBeginPlay should be fired (and prints out) twice. But the server is also displaying TWO additional logs, the print strings from Client 1? If this is how it works and logs are "replicated" (or just writing to the same log file on disk maybe?), is there a way to turn it off to just display the logs called locally?

[meager spade]

@scenic horizon those logs are correct

Server and client both run beginplay, so client is going to print it twice (one for server client and one for himself), server will do the same.

so 4 prints there is 100 percent correct.

[scenic horizon]

@meager spade Thanks for the reply. I'm a bit confused then, because what you said seems contradictory in my mind. If the client is printing it twice, why does the client show 4 logs?

[meager spade]

it doesn't

it shows 2 in that screenshot

and server shows 2

[scenic horizon]

Why does the clients window show logs printed from the server?

[meager spade]

cause printscreen prints to all screens

its just the way PIE is

single process

[scenic horizon]

Interesting. I guess that clears it up, just I wasn't expecting the client to show logs that were exectued from a server. Do you know if this is just a debug thing in the editor?

[meager spade]

yes it is

[scenic horizon]

Like if I was to test on an actual network, it would only show the logs executed from that launched instance of the game (and not from all connected clients?)

[meager spade]

correct

[scenic horizon]

Cool cool, thanks so much. For a moment there I thought I somehow had twice as many players being spawned and didn't know why! 😆

[meager spade]

np, happy developing 😄

[thin jacinth]

I've been looking at aws gamelift. I've got few questions: should I use spot or on-demand instances.

Should I use Windows or Linux pricing? What is the main difference between them. (MAIN QUESTION)

thanks and please make sure to @ me if you reply or help me. 😊

[hollow eagle]

1. https://aws.amazon.com/ec2/spot/
2. depends if you want to run on windows or linux...

[thin jacinth]

Thanks I'll check the link out soon, is there a down side if using a Linux based server instead of a window? @hollow eagle

[hollow eagle]

it's not an upside or downside, it's dependent on what platform you're familiar with

no one here is going to answer which is better, because the answer depends on you.

windows is more expensive due to licencing costs and higher resource usage, but may be "easier" to work with. Linux is... linux.

[thin jacinth]

Ah ok got ya, thanks for the huge help. @hollow eagle

[chrome quest]

Question please. If I want to store really persistent data about a player beyond each game session. Is there a class I can use?
I know PlayerState allows for data persistence beyond the lifetime of a pawn but is there any that exists beyond the lifetime of the PlayerState. e.g I want to store player level, achievements, skins unlocked e.t.c.
And if I'm using external solution like cloud storage, is there any that can store unreal data types?

[bitter oriole]

You're asking in #multiplayer so I'm guessing this is for a multiplayer game where progress matters and you don't want player levels to be changed by players

If that's the case, you need the UE4 dedicated server (and never the client) to tell a central server (that you will develop) about the player's progress, and I guess that server (using Python or something like that) will store the progress to a database

You won't be able to store Unreal types directly but UE4 can serialize to json, for example, which is easily processed in Python through an HTTP socket

[chrome quest]

@bitter oriole . Thanks. I see what you mean. It's the serialization to json that I was missing

[silent valley]

@chrome quest An easier way that might work for you is to use Dedicated server, but use the built in Unreal savegame code.
I create a savegame per player and load it when they connect, and save it when they leave.
This way the native type serialization works just fine.

[wise bridge]

Someone?

[bitter oriole]

But then you need to always connect to the same dedicated server, which is a bit dumb

[silent valley]

true, but my savegame system also reads/writes to s3 directly

so it's not tied to a server

[chrome quest]

@silent valley, I was considering that also but then I didn't know what class will hold the save game for each player and how to recognize each individual player. Also, you'll need a save game to hold the save games and I don't think save games can store classes

[bitter oriole]

Works too I guess. I usually advise against binary serialization, json is built-in for all UE4 types and is a lot more reliable and debuggable.

[silent valley]

it's not perfect but it works for my game design which is more like Rust - persistent world where player progress IS tied to a specific server instance

fair enough, didn't realise json was an option tbh!

[silent valley]

Yes they can hold classes - any Unreal reflected type really.

we mainly store inventory in there, so the items are all serialized into the savegame

savegame name is tied to SteamID

I'm sure Strangers solution is more robust and better for more gametypes, but we're a small team and it was quick to implement savegames

[chrome quest]

Wow. I didn't know that. Someone, somewhere told me otherwise

That they only hold primitive data types

[silent valley]

not at all, to save entire player for instance:

FMemoryWriter MemoryWriter(PropertyData, true);
FSBSaveGameArchive MyArchive(MemoryWriter);
playerCharacter->Serialize(MyArchive);

there's lots of examples for how to do it

[bitter oriole]

Though this means you serialize the entire player - any update adding UPROPERTY field is likely to break saves entirely

Very much not how I would attempt this

Like, this will also serialize basic replicated Pawn properties, for example

Or the components

[silent valley]

no

only properties tagged with Savegame

[chrome quest]

@bitter oriole , @silent valley , thanks I think I can see a solution forming up. I'll bother you with more questions when they come up. 😁😁

[bitter oriole]

Alright then

[silent valley]

all other properties we recreate if necessary in the Serialize function

works fine, but I get your point about compatibility

[paper lava]

Hey, what are you all using for server browsing? Is there something prebuilt to query the server for player info and such?

[fading smelt]

How to get rid of UGamestatistic:GetPlayerPawn(0) and GetPlayerController in My Map and Minimap manager which lives on the level as Actor?

This manager changes a texture of your map/minimap when you go to another location.

[winged badger]

if you have some c++ involved

just make better accessors

a static function returning local player controller via GetGameInstance()->GetFirstLocalPlayerController()

[floral zephyr]

hi, I just tried setting up a dedicated server for UE4 without having much knowledge about networking in the first place.
whenever I run the test_server.exe and open a game-client and type in "open 127.0.0.1" the server log says: "NotifyAcceptingConnection accepted from: 127.0.0.1:49191" etc.
and if I try to open another game-client and type "open 127.0.0.1" in that I get the same accepted message in the server, but from another port, and I don't see the 2 characters within the same game / server.
did I do something horribly wrong or am I just mistaken?

[fading smelt]

It will work on multiplayer with dedicated server?

[winged badger]

it will return null on dedi

as there is no local PC

[fading smelt]

ok

[winged badger]

APlayerController* USolsticeObjectLibrary::GetLocalPlayerController(const UObject* WorldContextObject)
{
    UGameInstance* GI = UGameplayStatics::GetGameInstance(WorldContextObject);

    return GI ? GI->GetFirstLocalPlayerController() : nullptr;
}

basicaly that

    UFUNCTION(BlueprintPure, Meta = (WorldContext = "WorldContextObject", CompactNodeTitle = "LocalPlayerController"), Category = "Local Player")
    static APlayerController* GetLocalPlayerController(const UObject* WorldContextObject);

[fading smelt]

But can I replicate a location property of local PlayerController to a server so my teammates can see me on the map (I thinking to add this feature in the future)?

[winged badger]

acess pawns via playerstates in playerarray

minimap requires no replication whatsoever

more effective pattern is

make an MinimapComponent

add it to the Actors that need to be shown on minimap

have the component register the Actor with the minimap system on BeginPlay

and then minimap system just displays Actors that registered

that way adding a say MissionObjective on the minimap is as simple as adding that same component on MissionObjectiveActor (just sitting on the objective location) and changing the icon on the component so it doesn't look like Pawns