multiplayer

[winged badger]

and execute only if it is

(its more bandwidth friendly option, and it will also make clients calculate their own linetraces, instead of server doing it for them)

[frank tinsel]

thank you Zlo. I'll give it a try

[frank tinsel]

Worked like a charm

@winged badger Thank you

[graceful cave]

when i enable seamless travel in the lobby game mode, nobody loads the next map including the host

but when its disabled everyone loads in

what might cause this?

[twin juniper]

Is there a way to add input delay to a InputAction?

So that if I am calling an RPC on a input action the player cant just spam

[mild hull]

you could ignore that input until the result of the RPC comes back

[pallid mesa]

you want to... just ignore the input completely

or make the input delayed?

because if your objective is to prevent spamming

you can make something like

[twin juniper]

Yeah

I did a test with an auto clicker on my server spamming E at a rate of 10 every milisecond

and the server

just took a shit

lol

[pallid mesa]

I have a ton of those

let me send you an example

if(LastTimeAttack + 0.5f <= GetWorld()->GetTimeSeconds()){
    Attack();
    LastTimeAttack = GetWorld()->GetTimeSeconds();
}

that .5f is my "anti retarded input" barrier

xD

[twin juniper]

Lol

So you are just cachine the last time, you hit that input

and seeing its been more than .5 seconds

lel

[pallid mesa]

ye xD

[twin juniper]

seems simple enough

[pallid mesa]

.5 could be a variable

[twin juniper]

would be useful though

[mild hull]

you could also count and send them as one rpc, if they is any need

[twin juniper]

if this could be done globally

to ALL inputs

Instead of having to constantly check

[pallid mesa]

you can... make your own input system

xD

[twin juniper]

yea

[pallid mesa]

I... made mine

[twin juniper]

unreal could add a little thing

to enter a number

lel

if 0, no delay

[pallid mesa]

https://vorixo.github.io/devtricks/input-recognition/ <- this

then you treat that array as you want

but for your problem xD I would only do that check in those inputs that are abused

[twin juniper]

oh

[pallid mesa]

/ functions or so

[twin juniper]

InputKey() is called

before the input

goes through

and im assuming if it returns false, it doesnt go through

lol

[pallid mesa]

welp check the parent in the viewport client class

there is not a real way to extend the raw input functionality without editing the engine directly

[twin juniper]

https://i.imgur.com/T3YY4om.png

Lol

that description tho

check a key

lel

[pallid mesa]

hahahahha

dude go to VS

[twin juniper]

😄

[pallid mesa]

don't look at the docs

you can...

certainly override the behaviour of what happens every-time

the input stack gets refreshed on the controller

and maybe, just maybe, you will be able to add manual latency there by erasing the stack content every x seconds or so

[twin juniper]

ive got an idea

@pallid mesa also though

you know your solution would be client sided

im assuming righht?

so someone could just reeset

LastAttack

lol

or change that value

[pallid mesa]

well this solution comes from a SP game, and I use it for a cooldown system

but yes, you are right, it is totally client side

[twin juniper]

though what i can say is that its better than what i currently have

[pallid mesa]

on the function that is getting rpc'ed you can do this same check but server side

[twin juniper]

which is noithing lel

i would rather have a client sided solution for now

than have nothing

lel

[pallid mesa]

but you got a big point there

never trust the client 👍

[twin juniper]

xD

though...

couldnt u just

replicate "TimeSinceLastInput"

and also

the idea is to

"not" use the server for this i feel

as much as possible

we want a client sided check first

then server sided check

before proceeding with the rest of the server rpc's

in my opinion this would be the best solution

What do you think? lol

[pallid mesa]

aye you can combine both solutions, for sure.

[twin juniper]

Basically... Do the check on the client first, if it passes, try it on the server, if it passes on the server, proceed with the rest of the code we need to do.

[pallid mesa]

also don't be afraid of input stashing as long as the function is not very taxing 👍

[twin juniper]

what do u mean?

"stashing"

[pallid mesa]

user input can never be fast enough to saturate a network on a lightweight function

it depends also the situation

and the game and so on...

[twin juniper]

i mean

it can

if ur like

using an auto clicker lol

Spam 10 inputs per 1 milisecond

But that requires the user to use an external script

lol

its "unnatural" input

[pallid mesa]

sure, yeah seeing it that way xD

void yoquese::StartFire()
{
    if (GetOwnerRole() < ROLE_Authority)
    {
        ServerStartFire();
    }

    const float GameTime = GetWorld()->GetTimeSeconds();
    
    // If its full auto we don't need to worry about buffered shooting ~ if we hold we want to fire - vorixo
    if (!bWantsToFire && ((GameTime >= LastFireTime + WeaponConfig.TimeBetweenShots) || WeaponConfig.FireMode == EWeaponFireMode::FullAuto))
    {
            ....
         }
}
``` In my shootergame extension as you can see I do this aswell

ServerStartFire() would call StartFire()

so this polling gets done on the server aswell..

as the common SG networking architecture

[twin juniper]

yea

ive seen that be done though

on stuff that doesnt even need it

lol

[pallid mesa]

ikr

[twin juniper]

shooter game is weird imo

lol

[pallid mesa]

yeah you can check my comment about buffered shooting xD

shootergame is just what it says it is

a shooter game with multiplayer support

the networking architecture isn't bad

actually they use networking culling

something I've never seen implemented in any Epic Games game

but in this template, you can notice it by looking through translucent objects

[twin juniper]

what do u mean?

the r.PauseNetworkReplication thing?

[pallid mesa]

they stop the relevancy of the player based on the view frustum of said one

check the template, you'll see what I'm talking about

[twin juniper]

yeah

thats what i was talking about

i tried implementing it

but i dont think it is necessary lol

100% of the time

like i just commented it out in my code lol

[pallid mesa]

well it creates situations that under certain geometry you would be compromising the gameplay

because your clients will stop seeing the other players

so your level design needs to be done accordingly and so on

[twin juniper]

mhm

yea

[pallid mesa]

so just brings more problems than anything

[twin juniper]

thats what had with me

lel

[pallid mesa]

but pretty good though

[twin juniper]

also

i think all it does

is disable the mesh

GetMesh()->SetHiddenInGame(bShouldPause);

is what it calls i think

in their example

[pallid mesa]

I cannot comment on that, didn't check that very closely

[twin juniper]

lol

i dont think it does anymore

than that

[pallid mesa]

idk, tomorrow I'll check to be sure 👍

but yeah, afaik Epic doesn't use it that much

not sure fortnite though but if they use it, it has to be very well done

because I never noticed similar issues to the ones I had with SG

[twin juniper]

yea

i mean fortnite is their most successful game as of yet isnt it?

Am i wrong? lol

[pallid mesa]

yes, well GoW was pretty bumped up back in the days

but speaking about recent ones FNT is the top notch

[zenith wyvern]

i want to store character data 'on log out', so im using EndPlay, but by then the pawn is gone and i cant get the transform

[pallid mesa]

anyways o/

[zenith wyvern]

whats the right place to do this?

[twin juniper]

@zenith wyvern do it in unpossess

in the Player controller

you can override Possess() and UnPossess()

and do not call Super::UnPossess()

until after

u do whaty u need to do

[frank tinsel]

Question: updating location of physics actors already spawned in world. What's the proper way of handling so they update properly for everyone. Right now, my physics actors in properly updated when they spawn in, but if a player bumps them, out of sync

[twin juniper]

dont use physics

3rd person ive heard now who has had trouble replicating physics lel

[frank tinsel]

I've seen it done but there was no proper explanation how

[glad sedge]

Accessing the engine directly maybe?

[frank tinsel]

It was through blueprints

[winged badger]

Repliate movement set to true?

[twin juniper]

@twin juniper you mentioned before that a hacker could just override the clientside input cooldown and spam the input command, but how can they do this? Can they reverse engineer the machine code and set the timeout variable to 0 constantly? Also if they can do this, what is stopping them duplicating the same packet that is sent with an RPC and sending 5000 fake RPCs a second?

[bitter oriole]

Nothing is

They can do that too

[mild hull]

you can force anti cheat tool and you can ignore IPs before they reach the actual server

[bitter oriole]

What does that even mean ? People can just connect to your game and then start sending fake packets with the game suspended with a debugger

The really simple version of things is : nothing that happens on a PC client is safe

On a console, some things are safe, some aren't

[thin stratus]

That's why you have the Server far away and double check

If the cooldown is really already over on server side

CheatEngine allows you to easily search for values in your memory

If they search for the cooldown multiple times right after hitting it, they will probably find a location in the memory that shows that often

If they change that value then they can easily bypass local limitation

[twin juniper]

@bitter oriole so in that case it isn’t even worth having a replicated variable on the clientside check before allowing an RPC to be sent as they can just send as many as they want anyway?

[bitter oriole]

Yeah, it's useless on the client side

Safety checks on the client are generally useless

[twin juniper]

It’s crazy that even this kind of stuff can be cheated, literally anything on the client is unsafe. If game streaming latencies go down to acceptable values in the future, I wouldn’t be surprised if many competitive games won’t be allowed on the client machine and instead you rent a streamed instance

[bitter oriole]

Well, PCs allow for entire control of everything

That's kind of the selling point

Though even consoles can see cheating - mostly by having stuff on the wire between the console and the network

And you don't really need streaming, you need more stuff on the server rather than the client. That's how many games have acceptable levels of cheating

[twin juniper]

MP games would be a hell of a lot easier to dev if the client could be trusted

Also sucks to see so many players blame cheaters when they die even though it could have been just a good player

But yeah you are right that is the reason pcs are so popular

[bitter oriole]

That's why most MP games do server authority + tight verifications + anti-cheat on client + account bans

No single effective measure, but a combination

[twin juniper]

Would love to see some examples of verifications like on Server_Validate functions

Almost all of the ShooterGame ones are just return true

[mild hull]

they only blame cheaters for everything because than they dont have to live with the fact that they suck :D

[bitter oriole]

Validation logic depends on your game, basically put in assertions that can't possibly be wrong in legitimate cases

What's harder is defining what's legitimate

[twin juniper]

Yeah like if you check the frequency of keys being pressed, it could be an aimbot doing 10 things at once or a caffeinated player getting pumped up

Hard to tell which

False bans would suck

[bitter oriole]

Or it could be lag, and all packets incoming at once

[twin juniper]

Yeah!!

[bitter oriole]

_Validate doesn't ban though, it kicks

Banning is up to your licensing, matchmaking system

[twin juniper]

So if it returns false you are kicked to the menu?

Will try it out thanks!

[bitter oriole]

Yes

[twin juniper]

For a typical multiplayer shooter game, would you recommend using GameMode/GameState or GameModeBase / GameStateBase? I'd like to use the leaner option but not sure if I'm missing out on essential features (saw posts a while back saying some things weren't working when using GameModeBase)

[bitter oriole]

I use GameModeBase myself, since I don't have any concept of "match" or stuff like that

GameMode is probably best for something like a MP shooter

[twin juniper]

Perfect thank you

I mean it would be always possible to extend GMbase and copy across any functions, but probably better to just stick with GM because my game is very match based

Would you say GameModeBase was more made to reduce the bloat for non multiplayer / unusual multiplayer games and GameMode suits the more typical setup many games use?

[bitter oriole]

Looks like that, yeah

[twin juniper]

how to replicate random integer

i know that i have to do that on server and then replicate to clients

but how

ok nvm

[brittle sinew]

It's best to pass it as a parameter in that situation (if you need it in an event)—you have no guarantee the variable replication will happen before the multicast is called

If you don't need it in an event immediately, just replicate it like any other variable.

[twin juniper]

i have a trigger with print string of what room type the room is and it shows on both screens, from both players the same value (the one that server sees)

is that good?

[brittle sinew]

Right now it's like that, but there's no guarantee it'll stay like that.

Especially over the Internet, you need to make sure your systems are robust enough to handle ping, packet loss, etc.

Variable replication isn't instant, and you can't treat it like it is, or things will start to go south.

[lilac lotus]

I'm super rookie when it comes to multiplayer / Replication so I would appreciate if someone could help me out with a problem.
I'm trying to make a C++ OnRep variable functionality.
This is my Header

private:
    UFUNCTION()
    virtual void OnRep_WorldMesh();
    UPROPERTY(ReplicatedUsing = OnRep_WorldMesh)
    class UStaticMesh* WorldMesh;```
And this is my cpp
```void ASubClassWorldPlaced::OnRep_WorldMesh()
{
    StaticMesh->SetStaticMesh(WorldMesh);
}
void ASubClassWorldPlaced::GetLifetimeReplicatedProps(TArray<FLifetimeProperty>& OutLifetimeProps) const
{
    Super::GetLifetimeReplicatedProps(OutLifetimeProps);

    // Tell the engine to call OnRep Functions each time a variable changes
    DOREPLIFETIME(ASubClassWorldPlaced, WorldMesh);
}```
Am I doing something wrong? The OnRep_WorldMesh() function never triggers

[ripe raptor]

if I set a variable on both client and server, the client will have its variable until the server one replicates, at which point it will replace it with that one, yes?

[brittle sinew]

Correct

@lilac lotus how are you setting the variable?

[lilac lotus]

@brittle sinew Like this // Check if item has a world mesh if (InventoryItem.WorldMesh) { // Set Specified World Mesh WorldMesh = InventoryItem.WorldMesh; } else { WorldMesh = (UStaticMesh*)StaticMesh; // Set to Null Mesh } I have checked and it goes into the correct statement 😪

[gleaming bobcat]

Can I set Owning Client as authority of rotation somehow ? So the server will not override my rotation ?

[ripe raptor]

Thanks @brittle sinew

[brittle sinew]

And what context is this in @lilac lotus? It needs to be at runtime (I don't think things like PostInitProperties or things like that work), and it needs to be on the server

[lilac lotus]

I'm spawning it at runtime with auto PlacedActor = Cast<ASubClassWorldPlaced>(UGameplayStatics::BeginDeferredActorSpawnFromClass(GetOwner(), WorldPlacedReference, RandomizeDropLocation(), ESpawnActorCollisionHandlingMethod::AlwaysSpawn)); , It's for a Drop Item Function. (It has the finishspawningActor as well). It reads everything loud and clear, it just doesn't set the mesh at this moment. I'm making the On_rep function to Server, Reliable, WithValidation atm and hoping that makes it work : )

[brittle sinew]

Making the OnRep function a server RPC...?

That doesn't really make sense; OnReps are intended to be called on clients.

[lilac lotus]

Hehe, yeah total Multiplayer scrub. I appreciate you point that out, would have learned through trial and error otherwise

[frank tinsel]

Question again. I'll try to clear it up a bit.
My goal is to have items on the ground be synced in location across server/client. As of right now, when they initially spawn in the world or dynamically spawn from players dropped, everything appears to be nice and synced. However, if they are in any other way interacted with once in the world(i.e player running into them), they can get horribly desynced. I have seen both blueprints and c++ solutions but not real information. Does anyone have an idea of counter advice?

or counter advice*

[cinder brook]

@frank tinsel sounds like replication problems.

[frank tinsel]

the players can pick up the items, but the item location is not replicating if the player runs into the item

it will move on both sides

but it won't be synced

[gleaming bobcat]

Anybody knows where CharacterMovementComponent replicates (correcting) owning client rotation which was set only on Owning Client ?

[slim holly]

🤔 I don't think it corrects rotation at all

[gleaming bobcat]

Well. If I will remove replication of movement component then the owning client is not overrided at all and it works

but once I will use movement replication then it's usually overidded

[slim holly]

I honestly don't really know, I've always used custom actor rotation logic

without issues

[brittle sinew]

Look at UCharacterMovementComponent::ServerMove

That's just what I saw quickly looking through the source

[gleaming bobcat]

I assume that

is CharacterMovementComponent is it ?

[slim holly]

no no, movementcomponent has it's own thing

[gleaming bobcat]

Ah damn

and where is sitting ?

[slim holly]

CharacterMovementComponent?

it is character class is it?

[gleaming bobcat]

I have character class yes

If I will just call on Role==ROLE_AutonomusProxy SetActorRotation it's usually overrided back to server pos

server rot

[slim holly]

is it using controller rotation?

or orienting to movement

[slim holly]

orient to movement uses velocity as direction, so it might differ from client and server

since client gets velocity overrides

[gleaming bobcat]

But I m just standing not walking press a button which set the direction

There should be no velocity

I mean rotation not direction

[slim holly]

well, I can say there's always velocity for sure, just very tiny bits of it

I just don't know what is the sleep value for it

I recommend doing custom logic, even if it's a simple one

[gleaming bobcat]

Ok so for the test I will set bOrientRotationToMovement to false just for the test

one sec

Hmm it's still same

Sometimes it works sometimes not

So am I right that CharacterMovementComponent overrides it ?

@brittle sinew I check it several times I did not find any client side call

[brittle sinew]

https://github.com/EpicGames/UnrealEngine/blob/dbced2dd59f9f5dfef1d7786fd67ad2970adf95f/Engine/Source/Runtime/Engine/Private/Components/CharacterMovementComponent.cpp#L7665

Within UCharacterMovementComponent::ReplicateMoveToServer :p

[gleaming bobcat]

But this is called on server

[brittle sinew]

No?

[gleaming bobcat]

I mean from Client To Server

[brittle sinew]

Correct.

But this part is called on the client

[gleaming bobcat]

But I have opposite problem

my Owning client is corrected by server

in terms of rotation

So basically I need to set new rotation on player side straight. Don't wait for any server delays. But server corrects me mostly before he will get know that this rotation is What I want

It;s maybe similar to teleport. You want to teleport player somewhere straight on client side (owner) but server will fix the location straight

for your client (cheat protection and so)

Then it should be implemented same like DoJump what Epic already made on ACharacter

and movementcomponent where you have to use some free slot

But In case of rotation I just need to find the place where the character rotation is overwritten by server

And I have no success so far

and if the place where it's happening is virtual method then I can change the behavior that Owning client is Authority of rotation

because nothing in my game could change my character rotation. Just me (the player)

[twin juniper]

In ShooterGame, when a pawn gets killed, he calls a function on the gamemode that does nothing but call functions on the killer and victim player states. I had thought this kind of thing was meant to be done on the GameState rather than the GameMode. Does it not really matter?

(or casting directly to the killer playerstate from the pawn that was killed as he has the info from the damage event)

[gleaming bobcat]

Just bear in mind that GameMode lives only on Server side. The GameStat on Server and Clients

[twin juniper]

For sure, it's just that I thought the GM was more of a static class handling the rules and major match events like starting or finishing, and the GameState was responsible for the second to second things like monitoring and setting global score, keeping time, etc

[gleaming bobcat]

Basically the GM is ruler and GS is basically the state of the game

But your issue. is it implemented by Epic ? If yes then it's correct :0

[twin juniper]

That's what I thought, but they have timer logic and playerstate score logic being set in the GM

I guess it's correct then 😄

[gleaming bobcat]

Epic is correct 😉

But time is rule

score is basically rule as well

I mean that reach some score means end of the game

sou this should be definitely handled by GM

[twin juniper]

That is true I just always thought the GS tracked that, and only notified the GM when it needed to

[gleaming bobcat]

GameState should be like a bridge for sharing the info

[twin juniper]

It seems it's all done in the GM leaving the GS barely anything to do

[gleaming bobcat]

something like progress of the game

[twin juniper]

Like I had a simple game where the first to 50 kills won, so I had the GS check after each kill if it hit the limit, but only when it did, then I told the GM to do something

Gotta just remove the static nature of the GM in my mind now and see it more like the Gamestate

But the actual doing force and less the tracking force

Also I just noticed that all of the properties in ShooterGame's header file are formatted with UPROPERTY(config)

This means the GM doesn't have its own blueprint in editor, and is tweaked via an .ini file

Is there any benefit to doing this vs creating a blueprint for the GM?

[gleaming bobcat]

Well. You know you can extend you custom GM by a lot of things which can be then modified in BP by designers

this is somthing what you can't do via ini

[twin juniper]

I'm not worried about that as I'll be keeping all logic in the cpp, but I like BP for tweaking defaults

As long as there is no overhead for having a BP, I'll go with it

[gleaming bobcat]

You should not folow this patern

[twin juniper]

How come?

[gleaming bobcat]

You can have logic in CPP that's fine. But you should open the variables to be modified outside your CPP via BP settings

even for you it's very easy to change somthing without compiling the code

[severe widget]

If you mean config vs BP, that's not really a good question IMO

[twin juniper]

That's why I like BP defaults

You can quickly edit

[severe widget]

Also a config file doesn't require a recompile

[twin juniper]

Without closing the editor

[severe widget]

You can even have the config load on command

[twin juniper]

Yeah but you have to reopen the editor every time

[severe widget]

But

No

But

You should do both BP and config IMO

[gleaming bobcat]

thats correct

both

[twin juniper]

That makes sense! I'll probably stick to just BP

Just because I'll be tweaking the children GM properties in editor until right, but maybe a text version would be nice to have too)

As long as there is no extra overhead. ShooterGame just has no BP GM at all, so you can't tweak without changing the .ini

[severe widget]

Yeah so mark it config

[gleaming bobcat]

this is not the best way

[severe widget]

then you can change it later on without having to repackage

[gleaming bobcat]

what about if you want to change the game time

[severe widget]

you don't have to do one or the other @gleaming bobcat

[gleaming bobcat]

or win score cap

[twin juniper]

Oh that is pretty cool (the package thing)

[severe widget]

you can literally have both

[twin juniper]

I can easily change that in BP @gleaming bobcat

They are just EditDefaultsOnly properties

That's the point of the child BP

[severe widget]

EditWhatever, BlueprintWhatever, Config

is valid

[gleaming bobcat]

I agree there is not only one way

[severe widget]

(╯°□°）╯︵ ┻━┻

w/e

[gleaming bobcat]

But I will prefer BP way but it's up to you

[twin juniper]

Thanks for clearing it up! Just was wondering why ShooterGame took that approach, and it's probably because Epic were most comfortable changing ini values

I do prefer the BP way 😄

[gleaming bobcat]

maybe the text file is much more easy to Diff 😃 form the version system point of view 😃

[twin juniper]

Potentially but there aren't really that many defaults, only about 6! It's definitley a good idea to make text versions of BP defaults now that I think of it

Considering it's possible for BP children to get reset

[gleaming bobcat]

I think this is completely up to you I don't see atm any adv/disadv over each other

But from the consistence point of view imho the BP is the way in my project 😃

How many ppl are working on your game ? 😃

[twin juniper]

Just me 😃

[gleaming bobcat]

wow 😃

[twin juniper]

So I don't need designers or anyone else to know what is going on

[gleaming bobcat]

👍🏿

[twin juniper]

Just need easy tweaks

And BP is easiest to tweak defaults fast

[gleaming bobcat]

Agree

In my project. the core and logic is in cpp and the high level things are made on BP level. Which means designers can change the behavior but not a logic

[twin juniper]

That makes good sense

So would they use any nodes at all or just stick with being able to edit defaults?

Or do you let them play with cpp logic nodes you have made?

[gleaming bobcat]

You should make High level functions (methods) visible in BP

then they can for example create struction of questlines

and easily can change which quest is needed before this quest and so on

I mean the parameters should be always visible in BP harcoded values are always bad

[twin juniper]

Nice that makes sense

And yeah I never use hardcoded values

great to be able to change things at a higher level

[gleaming bobcat]

👍🏿

Also UMG is another thing

What I understood it;s faster to use BP for handle the clicks

rollovers etc

faster then have some dynamic_binds

in this case you can have logic in CPP but you will just make visible function called "Resume Game"

but the clicks itself will be handled in BP

something like this

All the fucntions are cpp visible in BP

So power in UE I see in combination of C++ + BP

[twin juniper]

@gleaming bobcat extremely interesting! I was actually recently wondering about UI as usually the widget is created in the editor first, instead of the usual cpp being exposed to BP after

So you don’t actually bind anything here and instead just write the logic in cpp and call the logic you write from BP?

[gleaming bobcat]

exatcly

it's also completely transparent to designers (or not programmers in general)

[twin juniper]

Very cool! How exactly do you create the base cpp class for this?

Do you create a BP widget that derives from a Cpp class and not the standard widget?

[gleaming bobcat]

UUserWidget is my base class

And then you will crate BP based on your own class

with all the function visible in BP

[twin juniper]

Perfect that has gotten rid of any doubts I had about cpp / BP widgets thanks so much!

[gleaming bobcat]

np. Just bear in mind : Completely avoid BP implementation is not the ideal even if you are pure programmer which loves C++ (like me)

same from the opposite way. Doing everything in BP is horrible

[twin juniper]

I can especially see that with the UMG editor because it’s all completely visual and heavily using the BP interface to position things

I mean yeah it’s possible to do the full UI in cpp like ShooterGame but that seems like a massive hassle compared to UmG

Gotta play to the strengths of the tools!

[gleaming bobcat]

I was doing that manually in HUD it's jut pointless 😃 to much work with no sense

doing graphics in code I mean it;s pointless

i like UMG 😃 Have a nice day . I m going to bed

[twin juniper]

For sure I agree! Goodnight and thank you again!

[livid ruin]

Hey guys we're having an issue where we want to create a unique name for each player and display it as text over their head. We made our name creation system in a unique playerstate and we were trying to cast to that playerstate then set text as the playernametag variable. But for some reason we cannot cast to the playerstate because we aren't sure what object to use.

[vague dust]

How much RAM would you allocate for a linux dedicated server VM using a sample project?

[ripe raptor]

3

[vague dust]

the frickin binary is like 800MB lol

[ripe raptor]

4

(I'm kidding)

[vague dust]

is it bloated because its compiled from Windows for linux? Windows dedicated server is like 300MB

[harsh lodge]

Have anyone come across integer overflow error in html5

[frank tinsel]

Clarification question

What kind of replication is best for handling this kind of scenario

If a player wants to pick up an item, I mean

or Alternate question: why does this break my client players ability to pick up items when Run On Server is selected

[vague dust]

how can i debug network connection issue? 4.18.3 - dedicated server running on linux - client joining from windows - disconnects immediately

[2018.02.20-06.23.45:556][306]LogNet: Login request: /Game/Maps/entryMap userId: Invalid [2018.02.20-06.23.45:589][307]LogNet: Client netspeed is 10000 [2018.02.20-06.23.45:656][309]LogNet: UChannel::ReceivedSequencedBunch: Bunch.bClose == true. ChIndex == 0. Calling ConditionalCleanUp. [2018.02.20-06.23.45:656][309]LogNet: UChannel::CleanUp: ChIndex == 0. Closing connection. [UChannel] ChIndex: 0, Closing: 0 [UNetConnection] RemoteAddr: <MY IP>:60720, Name: IpConnection_36, Driver: GameNetDriver IpNetDriver_0, IsServer: YES, PC: NULL, Owner: NULL [2018.02.20-06.23.45:656][309]LogNet: UNetConnection::Close: [UNetConnection] RemoteAddr: <MY IP>:60720, Name: IpConnection_36, Driver: GameNetDriver I

[shadow torrent]

Hey guys, anyone here had a problem with demorec on 4.18? 🤔

[winged badger]

@frank tinsel when you are on server execution, you can get only one game instance - servers, as they are not shared

in that blueprint you replicate an event to server, then grab server game instance, and i imagine no matter which player picks up the item the host gets it in the inventory

using an inventory component, with variable-based replication would be my approach

having just the ServerRPC called to handle the pickup on Server, then replicating variables rather then functions to get the client in sync

[zenith yarrow]

I am using NetMulticast function in gamestate, which executes on the server but not the clients, shouldn't the same code be executed on the clients?

[thin stratus]

@twin juniper You can reference widgets in you cpp class

I don't know if someone told you that yet

It's pinned to the #umg channel

E.g.
UPROPERTY(BlueprintReadWrite, meta=(BindWidget))
UTextBlock* MyTextBlock;

If your child widget has a TextBlock with name MyTextBlock, it'll be accessible in cpp

In case you want to easily access the designer of your umg widget in cpp

@zenith yarrow depends on when you execute it

[zenith yarrow]

@thin stratus I execute the rpc in HandleMatchHasStarted()

[thin stratus]

And that is called when?

Cause if that already calls with only the server connected, the clients won't ever get the multicast

[zenith yarrow]

@thin stratus ok tested the RPC in the DefaultTimer() function , which is called every few seconds, and it worked there... I had assumed the HandleMatchHasStarted() was called when everyone was connected to the map... I'll create a custom event that marks the start of a match and fire the RPC then...

Thanks for helping me out.

[thin stratus]

You can hit delayed start in gamemode

And override ReadyToStartMatch

[twin juniper]

@thin stratus is this as well as creating the base UUserWidget class in cpp?

[thin stratus]

That is the same

In your UUserWidget c++ class, you define what I wrote

And in the UMG BP child class, you create a widget with the same type and name in the designer

And it will get referenced via the C++ variable

So if you have a TextBlock (Title) in your UMG Widget and the C++ parent has the BindWidget variable UTextBlock* Title, you can modify it in c++

@twin juniper

[twin juniper]

Got it thank you @thin stratus !

Yesterday someone mentioned that it’s actually faster to use the on clicked / hovered events from BP, so would you recommend just exposing cpp functions to use in BP for those events and do any binding directly in C++ as outlined above?

[thin stratus]

Tbh, you don't have to use c++ at all if you don't need any non-BP stuff

It really depends on what you are doing

[twin juniper]

I just mean in terms of speed, someone said it’s actually faster performance wise to use the BP hooks for on pressed / hovered etc

[thin stratus]

There shouldn't be a difference

It at all BPs are slowed

He probably meant that setting it up i nBP is faster

As it's just a button press

Compared to defining a function and binding it

[twin juniper]

Ok cool! Thanks! I’ll probably make all the functions in c++ and then use the BP hooks for clarity as its very visual

For most menu stuff performance doesn’t matter anyway because it’s not gameplay

[gleaming bobcat]

Actually not. The Epic said somewhere that is faster or better to handle it via BP

[thin stratus]

In the future, please use #umg though

[twin juniper]

Will do!

[thin stratus]

Workflow wise It's probably nicer via BP

I don't see how it should be faster Performance wise though

[dense citrus]

who knows , how to replicate a local savegame to all

i save my costumization inside a savegame , one material instance and one skeleton mesh and want that other Players can see my Layout , right now server can see what the players have chosen but the clients can see only own layout applyed on all players

thats my setting

[thin stratus]

Does someone know how to reset the LastSavedActor for Spawnpoints?

[pallid mesa]

Setting it to nullptr would help?

[thin stratus]

Yeah but I doubt that's possible in BPs

[pallid mesa]

Mnh, for spawnpoints I usually override their behav on BP's and do my own, found ir quiet limiting =\

Managing those vars myself

[thin stratus]

Most of it works totally well, just for respawning it's annoying

Cause I want to spawn them randomly

Not to the same spawn point again

So I guess I use my own respawn function /shrug

[pallid mesa]

Yeah, it's not that bad! But imho there is lack of variables exposed to BP's to operate better :(

[graceful cave]

im still having the issue where clients get weird replication related bugs if my framerate is too high as the host

capping my framerate lower fixes the issue

[OnlineSubsystem]
NetServerMaxTickRate=50

is in my engine ini but it didnt seem to help

there are no RPCs being called on any tick

[pallid mesa]

I believe 30 is the default value ```
[/Script/OnlineSubsystemUtils.IpNetDriver]
NetServerMaxTickRate=30

[dense citrus]

@thin stratus I can send you soon a sceen with random spawn system if you like

[thin stratus]

Na I have that already, don't worry

[dense citrus]

😉 ok

[thin stratus]

Is "ChoosePlayerStart" called before or after SwapPlayerControllers?

Cause it's def called before PostLogin

[winged badger]

its called from HandlePlayerStart i believe

when the pawns are already possessed

[thin stratus]

Pawns aren't spawned on my end before ready to start is true

And that happens post postlogin

In PostLogin I assigned a value to the Controller

In ChosePlayerStart I use that value

And it's not changed

So I believe it calls quite early then

But doesn'T matter much as I only use that for Editor testing

Actual game using seamlesstravel

[winged badger]

if you override ChoosePlayerStart

you will break the "PlayFromHere" in editor btw

unless you make a separate check and return w/e the Parent function would return if its in PIE

[dull tapir]

So, starting to get the whole multiplayer stuff I think, though I've come across some replication issues. I want to spawn a camera actor for each player that joins and then set the view target of each player to their own spawned camera. I use the Post Login event to spawn the camera on the server and then I set a reference to that camera as a variable in the PlayerController (though I've also tried PlayerState). The variable is set to replicate and so are the actors, however whenever I test this the variable is successfully set on the server, but on the clients it's just "none", so that when I try to set view target it doesn't work as it doesn't have an input. Anyone know why the variable is not replicating to the clients?

[winged badger]

not sure, but alternate approach: on BeginPlay of the PlayerController, if the PC is the LocalController, spawn the camera actor

no extra cameras on the server that way

[twin juniper]

Is ULocalPlayer relevant only for splitscreen or is it used in dedicated server settings too?

There's some logic in Shotoergame for setting up nicknames, controller IDs and loading persistent user data in a ULocalPlayer class

Not sure if that is splitscreen only or not

[dull tapir]

@winged badger Tried running it on begin play, same result. 🤔

[winged badger]

replication is not involved the way i described

you only spawn a camera actor locally for each player

[dull tapir]

Right, but I thought a server has to spawn actors and clients can't?

[winged badger]

they can, but those actors can't leave their local scope in any way

which for a player camera, just might not matter

[dull tapir]

Oh yeah just read up about it again. Hmmm...

Well in my case it may not matter, but let's say I wanted a spectator that can see what the player is seeing, in that case I'd need access to the camera on the server side, so it's not ideal even if it would work.

[winged badger]

true, but the approach holding a reference in the PC wouldn't work either there

[dull tapir]

Surely there must be a way to fix this replication? I just don't understand how it can be on the server but not on the clients when it's set to replicate.

Yeah maybe, though I originally had the reference variable in a player state

just moved it over to PC for testing

been going at this for about 5 hours lol

[winged badger]

replicate it with RepNotify

print some text in OnRep

see if the replication actually happens

[dull tapir]

Apparently so?

[chrome bay]

@twin juniper LocalPlayers are just that, local 😃

And you do indeed have one for each local player, so in split-screen ,you'd have two.

[twin juniper]

Is it still fine to be used even if there'll never be split screen?

Thanks! @chrome bay

[magic helm]

When is there gonna be split screen in the same VR headset?!

[twin juniper]

Just regarding something like this in ShooterGame ``` GEngine->GetAllLocalPlayerControllers(PlayerList);

    for (auto It = PlayerList.CreateIterator(); It; ++It)
    {
        APlayerController* PC = *It;
        if (!PC || !PC->Player || !PC->PlayerInput)
        {
            continue;
        }

        // Update key bindings for the current user only
        UShooterLocalPlayer* LocalPlayer = Cast<UShooterLocalPlayer>(PC->Player);
        if(!LocalPlayer || LocalPlayer->GetPersistentUser() != this)
        {
            continue;
        } ```

I'm not sure if this is to set up keybindings for each local user and is unnecessary or if it is for the sole local player playing online and is necessary

[chrome bay]

Yeah you can use them, they always exist for the local player!

Not on the server though of course, unless it's a listen server

[twin juniper]

Ok cool thanks!

[twin juniper]

Client doesnt see a newly added component if using Add Static Mesh Component

oh now works

[hidden thorn]

is there a way to get the max amount of players a server can have?

I used this: SessionSettings->NumPublicConnections

But everytime a player joins 1 is deducted from that variable

[jolly siren]

1 shouldn't be decremented from NumPublicConnections. NumOpenPublicConnections keeps track of how many are open.

[hidden thorn]

Yes my bad, I didn't know there were two of them, I was using OpenPublic... I didn't even realize

Thank you

[jolly siren]

no problem

[hidden thorn]

when making an array that holds a list of all the players connected is there a reason to use APlayerController rather than A<ProjectName>Character ??

[jolly siren]

Characters normally don't persist death

There is already an array of all PlayerStates in the game. Why not use that?

[twin juniper]

In ShooterGame when loading and saving data, it seems to treat the game as though there are multiple local players ``` TArray<APlayerController*> PlayerList;
GEngine->GetAllLocalPlayerControllers(PlayerList);

for (auto It = PlayerList.CreateIterator(); It; ++It) ``` which I believe is to do with the fact that they support splitscreen. If I am certain that I won't have more than one local player per machine (i.e. a typical setup for modern online multiplayer games) do you think I can instead write this as  ``` APlayerController* LocalController;

GEngine->GetFirstLocalPlayerController(); ``` ?

And even better if not using GEngine as you seem to have to include the huge Engine.h for that to be recognised...

[chrome bay]

yeah you can safely do that

Or get local players from the game instance

[twin juniper]

Oh getting it from the game instance could be way nicer, thank you very much!

[hidden thorn]

@jolly siren Oh I didn't know there was one

[twin juniper]

Just trying to get my head around what kind of local they are referring to here, since this isn't called on the server, would it also possibly be safe to call GetWorld()->GetFirstPlayerController(), or would that still potentially return the first player connected to a sever, and not the local machine player?

[jolly siren]

@hidden thorn yeah GameState->PlayerArray

john, yeah they are talking about splitscreen

[twin juniper]

Ok cool thanks, so I can basically replace any for loops and iterators for local players with a single first local player reference!

[jolly siren]

yep

[hidden thorn]

Is there a benefit over using GameState->PlayerArray? I was following a doc by by Cedric_eXi and he uses PostLogin(APlayerController* NewPlayer) and then PlayersConnected.Add(NewPlayer) etc

I did similar thing but in my case PlayersConnected is in GameState rather than GameMode

[thin stratus]

PlayerArray is of PlayerState

@hidden thorn

PostLogin is a default event by gamemode that calls when a player is fully accepted

It gives you a ref to its playercontroller

That's different to the PlayerArray that gets filled with PlayerState references

And the one in GameMode is server only

While PlayerArray is somewhat replicated

Well at least gets filled on each client

[hidden thorn]

Is there a reason to have a server only player list??

And yeah I get the difference between PostLogin and PlayerArray

[thin stratus]

Well there is no reason to have that list on clients

The playercontrollers in there are only replicated for local clients

A client doesn't have a ref to other playercontrollers

[lilac lotus]

Is there a special way of handling OnRep in C++ if the variable is of type USkeletalMesh ? Feels like I've looked everywhere I can for the answer.
The event that changes the variable fires & the setup looks good but no changes visually.. 😪

[brittle sinew]

You're wanting the OnRep event to fire on the client, correct?

I believe BP and C++ have a discrepancy when it comes to OnReps—IIRC, in BP OnReps are automatically executed on the server, while in C++ they're not

[lilac lotus]

Yeah I read that, not sure it fires on either server or client though

I'm currently checking references & pointers even though I don't see that being the problem

[limber venture]

hey guys, i am trying to connect to my dedicated server and my client disconnects immediately. This seems to be the error

[2018.02.20-22.37.38:683][138]LogNet: Warning: Travel Failure: [PackageMissing]: [2018.02.20-22.37.38:683][138]LogNet: TravelFailure: PackageMissing, Reason for Failure: ''

Full stack of messages

https://pastebin.com/1qekzkaf

prob should have pastebin that, sry

that is client

this is server, but it seems to only generate client side error, all i see is connection close immediately on server side - https://pastebin.com/6Xzmek2h

[hidden thorn]

if I have the players list as APlayerController is there a way to get the A<Project>Character so I can access variables from that file?

[brittle sinew]

Yes, that's the function of casting.

You can't cast a whole array at once, however; you need to cast each element individually, either when you use it or sometime before that.

[hidden thorn]

I know that what I am asking is how would I go about getting the Character since the list stores the Controller and they are 2 different objects.

[brittle sinew]

AController::GetPawn()

Sorry, I had originally glanced over it and thought you were simply trying to cast the controller to your custom type.

[limber venture]

can you set higher verbosity of logging for the packaged builds?

[limber venture]

apparently you can in the defaultengine.ini config

[Core.Log] LogNet=verbose LogOnline=verbose

and then repackage your project

hopefully will find out what this bullshit fucking error means LogNet: TravelFailure: PackageMissing, Reason for Failure: ''

Reason For Failure: '' thats some windows shit right there

Windows is more like Unknown Error: 'Error Unknown'

[hidden thorn]

@brittle sinew Thank You

[ripe raptor]

It seems that Steam marks sessions as InProgress straight away when you create the session?

Nevermind, my bad

[analog hatch]

Hi everyone.

We are having problems replicating AimOffsets. It works on the Server (you can even see the aimoffset of the clients into the server) but the clients can't see any AimOffsets that are not their own's. Anyone had a simmilar experience? Or know how to fix it?

Thanks awesome people.

[ripe raptor]

Do Steam presence sessions actually have to be in progress for them to advertise?

I am getting 0 search results even though there is one presence session that's in the "Pending" state

Actually, I can see the session only if "bAllowJoinInProgress" is true, even though the session is Pending, not InProgress

[cyan bane]

Every so often when i update a replicated variable on my server, my client receives it twice. I am using ForceNetUpdate() directly after I update the variable. Is that normal?

[sly basin]

@analog hatch I had a similar problem a while back. I ended up setting the variables that control the offsets locally and then setting them again in a loop that repeated 30 times a second as replicated, skip owner and it cleared it up. I'm no expert, not sure if that's a good way to do it but maybe this gets you looking in the right place.

[hidden thorn]

Out of nowhere I can't find servers anymore.
I am not sure why this is happening, is there anything I should be looking out for?
The way I tested is I right click the project Launch Game and start 2 instances.
https://i.imgur.com/rrjh8p9.png

https://pastebin.com/qWxUmRL9

[gaunt kestrel]

In Multiplayer I want fairly large map sizes, if I use level streaming in multiplayer and only load the tile any players are on, and the nearest tiles to them, would that but a good solution?

The only thing I could think of is it might be hard on the network streaming in and out levels a lot

[thin stratus]

You can load levels on clients when enabling Client-Streaming

And then let clients load the surrounding tiles

Depends on the view distance I guess

And how big a tile is

[gaunt kestrel]

I think there its about 2ish km?

Client-Streaming? Loading them only on client side?

[thin stratus]

Yeah by default, it's Server-Streaming

So Server loads one and everyone loads it

Client would allow you to only load an specific client + server

[gaunt kestrel]

Oh okay, so Client 1 enters that zone and loads it, the server loads it, but client 2 on the other side of the map doesnt load it

[thin stratus]

Yeah

[next falcon]

can i set a widget in the world only see for one guy who is standing right next to it ?

[hidden thorn]

Well you can check the distance between you and the widget and depending on that show it or hide it.

Or you can spawn the widget locally

[next falcon]

jeah but if i stand next to it and set visibility everybody will see it too

[hidden thorn]

Is the widget set to replicate? also are you telling the server to set it to visible?

[next falcon]

the widget is not replicated and its owning client

and triggerd by collision overlap

[hidden thorn]

So any client that overlaps it makes the widget visible?

[next falcon]

jeah

visible to all players even the guys who are not near

[hidden thorn]

When you overlap do you tell the server to make the widget visible?

Cause if not I don't see why it would be visible to everyone if the client makes it visible locally

[next falcon]

i think the overlap is a multicast...

[thin stratus]

In the Widget Actor

Get the Overlapping Actor

Check if it's locally controlled

Or, if you only have Clients (no listen Server)

Just so "SwitchHasAuthority"

And on Remote exec

Get the Owner of the overlapping actor

That should be the PlayerController, if the overlapping actor is the Character

Check if that is valid

If yes, you have the local client

[next falcon]

nice thanks

[hidden thorn]

@thin stratus have you got any ideas why I have that issues (a few posts above)

I've been looking around for an hour or so but can't seem to get why that is happening

[thin stratus]

Is that Subsystem NULL?

[hidden thorn]

This

// Get OnlineSubsystem we want to work with
    IOnlineSubsystem* const OnlineSub = IOnlineSubsystem::Get();

If yes, well it does pass the if statement.

And it worked last night and nothing has been changed since then

[thin stratus]

No I mean

What Subsystem are you using

[hidden thorn]

I am not familiar with what Subsystem means in this context

Not sure if it helps or not but I followed your "How_To_Use_Sessions_In_C++" wiki

Looking online I think I am using the UEs SubSytem which is set to null in DefaultEngine.ini

[jolly siren]

https://docs.unrealengine.com/latest/INT/Programming/Online/index.html

[twin juniper]

anyone mind looking at my network profile

and telling me what i should be looking at

to optimize for performance

not sure what to look at D:

[candid fox]

Hello, anyone knows how to restart level using Steam hosted sessions without kicking players?

[jolly siren]

@twin juniper What is ClientUpdateAllSlots?

@candid fox dedicated or listen server?

You can use RestartGame or do a ServerTravel to the same map

[twin juniper]

@jolly siren literally just updates the inventory slots

passing an array of structs from the server to the client

to update the icons

and such

sometimes... it doesnt always get updated correctly

so i call that whenever u open the inventory

[jolly siren]

Where are the icons stored? Why would you need to pass them over the network?

[twin juniper]

sec

Actually, my mistake

it sends the "Inventory" which is an array of S_Item structs

with 4 things:
item id, item count, item index, and the inventory component

@jolly siren However, we use the item ID to lookup the stuff for the "item slot/icon"

and it looks it up from a DataTable

[jolly siren]

Yeah passing ID and locally looking it up is the correct way to go. The avg size of that rpc seems quite large tho

[twin juniper]

It loops through all the inventory structs

and passes it to "Update Inventory Slot"

which just sets the widget stuff lol

@jolly siren if u dont mind, how were u looking at that though

https://i.imgur.com/nH57HRu.png

i thought these two parts were the problem

[jolly siren]

just looking at the all rpcs tab

but yeah you have a couple spikes in there

[twin juniper]

any clue what that is caused by

O_o

Or how I could find out?

because at that spike, there is literally NO RPCs being called

however, there is a large number of rerplicated actors...

https://i.imgur.com/a3L71y7.png

so im not sure haha

[hidden thorn]

@jolly siren About the Online Subsystem I looked through that and mine is set to NULL -> DefaultPlatformService = NULL

[thin stratus]

@hidden thorn Well subsystem null is fine

That shouldn't cause any errors

[hidden thorn]

Ahh, could the firewall block anything?

[twin juniper]

The NullSubsystem is literally just there

to give u basics lol

its actually really nice

[hidden thorn]

Or have you had this before and I should maybe look out for something blocking it such as anti-virus or idk

[candid fox]

@jolly siren Thanks for tip. When i do ServerTravel Players stay connected but game state will be None. How do i start new match from there? Tried Start Play.

Also Listen.

[vital steeple]

if i get playerstate owner, it should return the controller right?

yes, it does, just tested 😃

[thin stratus]

@vital steeple Yes, but only on Server and LocalClient

(Can't say that often enough :D)

@Ardivaba#1465 That shouldn't happen. If your GameState is NONE, then you have some other issues :/
When are you trying to access it?

@hazy sableowman#9463 SubsystemNULL is basically LAN Sessions and IP Connection.

Even the LAN Sessions are IP Connections as searching for matches is a simply broadcast through your LAN

So technically, if you local PC doesn't block any internal IPs, it should find the sessions

[hidden thorn]

That must be the issue because I went to uni and tested the exact same project and it works.

[jolly siren]

Does anyone know if it is safe to update session settings such as gamemode, map, etc. after players are connected? At the end of the match players vote on the next map/gamemode and I would like to update the session settings so new joiners can see what map/gamemode it is actually on.

[hidden thorn]

How could I check if my local ip is being blocked

I looked at this: https://answers.unrealengine.com/questions/521300/cant-find-session-in-lan.html

But I dont have "VirtualBox NetWork Adapter" installed or listed anywhere

[jolly siren]

that would only be listed if you are using a vm with virtualbox

[hidden thorn]

Anything else that could be blocking my local ip?

I turned off my antivirus, firewall and nothing

[twin juniper]

how to replicate add static mesh component?

[vital steeple]

@thin stratus sweet! im running it from server anyway (a spawn queue that cycles through an array of playerstates, gets controllers, and spawns a character for that controller to possess)

works great 😃

[winged badger]

@twin juniper you add the component normally, then you do SetReplicates to true (bReplicates = true from c++) for that component

[twin juniper]

But if I don't want it at start?

[winged badger]

that is fine

[twin juniper]

Because there's a bunch of same blueprint actors

[winged badger]

works either way

[twin juniper]

Ok

But if I need to change the mesh?

[winged badger]

StaticMesh pointer is replicated, so you should just need to set a new one

[twin juniper]

Is there anyway to optimize replication of large numbers of actors

Like, I have about 30+ NPC actors, and even though I've done what I think is more than enough to optimize, it still lags my server.

It literally enables and disables all tick, set rreplication on and off, enables/disables all collision, all ticks on components, as well as registes/unregisters navigation invokers on ALL npc's

yet the optimization is still not enough

am i missing something else?

_<

[manic pine]

brain is turned off too?

[twin juniper]

It literally disables tick on all components

from what i can tell

I'll DM you a snippet

of the code i made

[manic pine]

what about the bot controllers?

[twin juniper]

https://pastebin.com/Rf4p0nbq

So I call SetHibernate(true) when we our collider enters the area, and call SetHibernate(false) when we exit..etc

[manic pine]

the other way around i hope

[twin juniper]

yeah

xD

[manic pine]

right, but the performance problem is constant? not just when enter/leaving the area?

is the performance worse when youre in the area?

[twin juniper]

yeah

well

if no player is in the area

it just despawns the npc's

and performance improves lol

[winged badger]

you sure its the network and not AI controller with PTSD?

[manic pine]

hibernate seems to destroy the ai controllers

in fact, it doesnt even recreate ai controllers when turning off hibernation

[winged badger]

if its the network, and i don't think 30 actors can overload it, even with under 1Mbps connection

state based replication is usually the cheapest

replicating a single enum (1 byte) that holds the current state with RepNotify, then doing what you need on clients with OnRep functions

[manic pine]

im a bit confused too, what exactly is the hibernation function for if you just destroy the actors when leaving the area?

is it just a test to see if turning hibernation on while player is there improves performance?

anyone have any experience with lag compensated projectiles(not hitscan)? bigger ones, e.g. rockets rather than bullets
i want the shooter to actually hit what/where he's aiming at, so i figure fake projectile on shooter, then server does a movecomp.tick of shooterping/2 to sync up to shooter's fake projectile
however, i also want other people able to both shoot down(with hitscan) and dodge those projectiles, so when non-shooters receive the projectile they should tick another nonshooterping/2 for their replicated version, so they can see the actual server version...
but, the dodging of or shooting at those projectiles is gonna take another nonshooterping/2 to actually reach the server, which means they'd need to do yet another tick so their view of the projectile is actually ahead of the server, so in total it adds up to shooterping/2 + nonshooterping ms worth of prediction, which at 100 ping for both clients totals at 150ms... a projectile velocity of 3,000 means the spawn location as seen by the nonshooter would be offset by 3000*0.15= 4.5meters, which would have a nasty gameplay impact up close

[twin juniper]

@manic pine thats commented out

the destruction i mean

@manic pine you destroy the actors when youn leave the "biome"

not when u leave the NPC location

[manic pine]

ah right, so hibernate when in biome but not immediate area of npcs

[twin juniper]

yeah

so if u leave biome, ALL npcs in that biome despawn lel

but it keeps a "FVector" of the location

so if a player comes within1 5000 sq units of the locationo

it spawns a random npc

@manic pine

[manic pine]

right, have you confirmed that your hibernation function actually runs? if its really setting updatefrequency to 0.01(once every 100 seconds), then replication shouldnt be causing an issue

also make sure its not running continuously ^___^

[twin juniper]

@manic pine it is being called

HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 0
HibernationModeChanged: 1
HibernationModeChanged: 1
HibernationModeChanged: 1
HibernationModeChanged: 1```

0 = no hibernation, 1 = hibernation

and its changing when i enter/leave

[manic pine]

ah okay, well thats good.. but then whats causing the performance problems

whats the profiler saying?

if the profiler doesnt give you anything useful, might be worth it to put some prints inside tick/tickcomponent functions just to see if anything's running, since something obviously is

[hidden thorn]

Any reason why I am getting this error?
This works fine on the server but when a new player joins this is what happens.

Access violation - code c0000005 (first/second chance not available)

Urpgprojectgamestate.cpp:39]

//Line 39
if(Cast<ARPGProjectCharacter>(connectedPlayers[i]->GetPawn()->) != nullptr)

In GameMode I have this

void ARPGProjectGameMode::PostLogin(APlayerController* NewPlayer)
{
    Super::PostLogin(NewPlayer);

    ARPGProjectGameState* GameState = GetWorld()->GetGameState<ARPGProjectGameState>();

    if (GameState != nullptr)
        GameState->AddConnectedPlayer(NewPlayer);
}

[brittle sinew]

Looks like you have an errant arrow operator on the end, I'll just assume that's a copy mistake ;)

[hidden thorn]

Yes that is

[brittle sinew]

That said, you should check if the array element is valid before trying to call a function on it

You're looking at a classic nullptr access violation there

[manic pine]

connectedplayers could be null yeah

eh sorry

0 elements

[hidden thorn]

But technically what you see there should work

[manic pine]

or it could be a null member

[hidden thorn]

also the array is not empty as I have the i displayed at the beginning and it returns 0 and 1

[manic pine]

could be null pointers inside the array

[hidden thorn]

I will look into it now

[manic pine]

connectedPlayers[i] dereferences the array, the next -> dereferences the array element

first will fail if index is out of bounds, second will fail if null(or invalid) pointer

@twin juniper any luck?

[hidden thorn]

Ok so what I am getting is this

https://i.imgur.com/xh3LVAU.png

for (int i = 0; i < connectedPlayers.Num(); i++)
{
    if(connectedPlayers[i] != nullptr)
            GEngine->AddOnScreenDebugMessage(i, 0.01f, FColor::Green, FString::FromInt(i) + " - " + connectedPlayers[i]->GetName());
        
}

Top Left - Server other two clients. So the controller on the previous clients/server becomes null but it does recognize that there are more players

[twin juniper]

@manic pine no lel

the -messaging parameter on my server

isnt working

so i cant run profiler on server anymore

[manic pine]

ouch, well you could still try printing in the ticks to confirm that nothing's actually running

@hidden thorn are you trying to access other player's playercontrollers from a client?

[hidden thorn]

Yea

[manic pine]

ah

yeah, thats not legal

[hidden thorn]

I assume if they are in a list it would be a problem as they can't edit but read them

[manic pine]

playercontrollers replicate only to the player who owns it

other players playercontrollers dont spawn on a client at all, they dont exist period

[hidden thorn]

ahh

[manic pine]

youre looking for other players' representation on your client

thats usually done with the APlayerState class(and sublcasses)

listenserver and dedicated server has access to all playercontrollers, clients have access only to their own(as well as any split-screen controllers), but everyone has access to all the playerstates

[hidden thorn]

can you get the player state from the controller?

I am just trying to wrap my head around it

[manic pine]

yes, the PlayerController is actually what creates the PlayerState

it does so on the server, then the playerstate is replicated to everyone

you can also find the playerstates in AGameState

which is another object all players have access to

[hidden thorn]

What sort of info you would store in the PlayerState?

[manic pine]

anything that other players might need to know about that player that's not stored on his actual character/pawn

by default, you can find e.g. the player's name and score there

[hidden thorn]

So like on the actual character you would store health, armor, kills, deaths, team etc. and on the state name, id, etc

[manic pine]

health and armor sure, but what happens to kills/deaths(and team?) when character dies?

they're gone

much more fitting they stay on the playerstate, assuming you dont want to reset them on death

[hidden thorn]

when the character dies do you have to reset them or do variables reset by themselves ? or is that more like a logical thing to do on the character

[manic pine]

well, UE doesnt really have a concept of 'death' per se

there's Destroy, which removes the actor from the game entirely

[twin juniper]

if u spawn a replicated actor in a multicast function

we will get multiples right?

lel

[manic pine]

multicast, when used on a server, is called by the server itself as well as on all the clients that have a replciated version of the object the multicast is called on

[twin juniper]

but

that would mean

u would get 1, replicated to the clients, and then another on the client

so two on the client

1 on the server

[manic pine]

right

[twin juniper]

heh

[manic pine]

and only 1 of the two spawned on the client would have a net connection and thus be replicated

other would be dummy actor that the client has Authority over

[twin juniper]

mhm

[twin juniper]

https://i.imgur.com/0gBrGam.png

lel

my npcs are setup to "run away" when a player gets them below a certain hp

but instead, if no player is lofgged in

it cant complete the eqs query

[manic pine]

oh my... and 28ms for that?

insanity

how big is the grid

[twin juniper]

@manic pine actually the grid isnt the long part

if un look

its the "Trace"

thats taking 5.71 ms

the grid itself is the first one,

its also taking a bit though

2.27 ms

[manic pine]

hmm, are you sure that entire environment query thing is meant for huge open worlds like yours?

i think its for tiny indoors areas

[twin juniper]

i mean

how else would u make something

run away lel

[manic pine]

pick a random direction

^_______^

away from the scary thing, obviously

[analog hatch]

@sly basin Hey thanks we'll have a look at your method. cheers!

[winged badger]

you could make them run away only if the player is close enough for it to matter @twin juniper

that would solve the no player around issue as well, they could just wait then

checking VectorLengthSquared is relatively cheap, much cheaper then traces

[twin juniper]

@winged badger yeah

[agile crane]

Hey, I have one question that I really need help with Unreal engine 4.
I want to make a deathmatch (not team deathmatch) where everyone kill each other, and they all use the same character, I tried doing that but everytime I shoot or someone shoot we damage ourself, like the bullet is colliding our own character, and is there a way to do that it can collide only on other person than ourself? So like we could kill each other.

Please thanks.

[worn nymph]

you need to set the damage instigator at time off firing and also if you are using traces add the controller to a list of ignored objects array

[agile crane]

I don't really get it, all I'm doing is cast to my own character and do health damage inside my bullet, but when I shoot I get damaged myself

[winged badger]

inflicting damage logic does not belong inside a bullet

[agile crane]

Also line trace doesn't work for me, when I do a line trace and try to add damage to enemy or AI's it doesn't work.

Anyone know why?

Oh wait, I think I find a way to do it, thanks for the help 😄

[twin juniper]

Hi. Can anybody help me? I'm trying to make a main menu with chat and friends list in a multiplayer game (session based), but I'm not sure about what I should use. I think I have to make a new network layer, but I don't know if I should use TCP/IP or if there is a better way to do a bidirectional connection in real time (without TCP sockets). It's there a better way to do it without TCP sockets? (Sorry for the bad english)

[hidden thorn]

I have a variable like health that is replicated over the network and it works sort of. The server is not getting the clients health updated but the client can see the servers value being update. While the clients health is not updated between each other.

Do I need to add that in a function with UFUNCTION(Server, Reliable) or a NetMultiCast?

[gleaming bobcat]

Server is Updating Clients not in opposite way

Health is something what should be replicated without messages

[hidden thorn]

I know but this is just for testing

[gleaming bobcat]

Its up you you how you want to replicate

to all clients

but you can be also more specific

when you will use DOREPLIFETIME_CONDITION

[hidden thorn]

Is this the right way of calling it?

UPROPERTY(Replicated)
float Health;

//.h
UFUNCTION(Server, Reliable, WithValidation)
void UpdateHealth();
void UpdateHealth_Implementation();
bool UpdateHealth_Validate();

//.cpp
void ARPGProjectCharacter::UpdateHealth_Implementation()
{
    Health += 0.01f;
}

bool ARPGProjectCharacter::UpdateHealth_Validate()
{
    return true;
}

I just tested it and I am getting the right values displayed for each player

[gleaming bobcat]

Looks correct!

[hidden thorn]

So when would you use NetMultiCast what sort of sittuation?

[gleaming bobcat]

Multicas should be called from server

basically something like Client ---> Hey I m doing something ---> Server will do it and ----> tell other clients about what are you doing

[manic pine]

multicast is nice for e.g. telling all clients to play a particular visual effect

or maybe turning the visibility of an object on/off

the multicast effect can usually be simulated with a replicated property with repnotify too, though without arguments obviously

[gleaming bobcat]

Basically you need distinguish between what you want to replicate really often via replicated value or one shot messages like Multicasts

[manic pine]

careful with unreliable multicasts though, unlike reliable ones they can end up waiting a long time before being sent

[gleaming bobcat]

I am not sure how it works with unreliable in UE but low priority things like VFX should not be never reliable

[manic pine]

unreliable for everything else is sent instantly

its just unreliable multicast that's slow

its a design decision, though i cant remember the exact reason

[gleaming bobcat]

that's fine. It's not important how fast the other player will see my PFX

[manic pine]

i believe it waits until the next netupdate

[gleaming bobcat]

if there is tons of PFX overlapping each other

[manic pine]

yeah, but depending on how youre using it, you could end up with surprising results... e.g. an explosion effect following 1 second after it should

[gleaming bobcat]

so unreliable means it will be delivered but it has low priority ? or it's a chance that it will be not deliverd ?

[manic pine]

unreliable in and of itself just means it wont resend it, ever

reliable will wait for response from client that it received it

otherwise resend after x amount of time

[gleaming bobcat]

Well it's easy to set all to reliable but it's definitelly not good idea

[manic pine]

the fact that unreliable multicast is slow is because they intentionally put it in queue until next netupdate, which is potentially minnetupdaterate

[gleaming bobcat]

So It should be based on your priorities

[manic pine]

yah definitely

[gleaming bobcat]

For example if I put new gun to my hand

it's definitelly low priority

[manic pine]

yeah, though it would arguably need to be reliable

then again, you'd do that with replicated var anyway, whcih is always reliable

[gleaming bobcat]

but what I have in my hands don't need to be replicated all the time

[manic pine]

its not, server only reps vars when they change on server

[gleaming bobcat]

it's something not much changed frequently

ah I see really ?

he is replicating only changes ?

[manic pine]

hmmmmm

actually dont quote me on that, ive never actually tested it

i kinda assumed

there are several systems in the works though, like dynamic net frequency

[gleaming bobcat]

If it's replication only changes that's is fucking poverfull

But still I don't think so :0

[manic pine]

yeah, i kinda assume the engine makes a VarOldValue, then on netupdate it checks if Var != VarOldValue, replicate

[gleaming bobcat]

Do you have any experience with root motion via network ?

There is some support for that in CharacterComponent does it works nicely ?

I mean CharacterMovementComponent

[manic pine]

never tested it to be honest, but its all over the place, including servermove/replicatemovetoserver

so i'd assume it works given how much work is put into it

its even in clientsmoothing

[gleaming bobcat]

well root motion looks always better then not root motion

the question is how it's replicated according to this

It's very sad

https://www.youtube.com/watch?v=H4HpLnJ3gec

On client side it looks horrible

I mean on other client sides. I think the main problem is that there can't be any movement prediction

[manic pine]

hmm so it interpolates between t and t-1, instead of extrapolating from t to t+1 as usual

[gleaming bobcat]

I will test it tomorrow but there are some animation which does not have linear movement speed

and in this case it's not easy to use non root animations

[manic pine]

i guess it could work fine for a lot of games

though not fast paced shooters

[gleaming bobcat]

you know something like asia style animation you will jump then the movement is slowed down in the air and then you will land very fast and then the movement is almost stoped before the hit and then it's very fast again etc

[manic pine]

ah yeah

[gleaming bobcat]

and in this case doing some manual movement to match the animation it's quiet shitty :0

[manic pine]

must be hard working it into the gameplay either way no?

[gleaming bobcat]

😃

[manic pine]

its a very cool effect though