dev_rc_branch

[full sonnet]

Sending chat message to player works. But it doesn't show in loading screen :/

[carmine juniper]

Same vibe

[full sonnet]

But I can just enable chat in loading screen, that's actually very easy 🤔
And I'm pretty sure that it once was possible. In Arma 2 probably?
Also I remember having to disconnect to see errors in chat. But that's also not a thing anymore. The server browser display disables chat (and the server browser is still open in the background while you are in the connecting loading screen, so it keeps the chat off

[full sonnet]

There we go

onPlayerJoinAttempt = " \
_pid = _this select 0; _time = _this select 1; \
if (isNil 'CState') then {CState = 1}; \
if (CState == 1 && _time > 2) then {CState = 2; _pid sendChatMessage 'Hello, please wait'}; \
if (CState == 2 && _time > 5) then {CState = 3; _pid sendChatMessage 'Please wait more'}; \
if (CState == 3 && _time > 10) then {CState = 4}; \
if (CState == 4) then {call (missionNamespace getVariable ['PlayerJoinHandler', {'ACCEPT'}])} else {'DELAY'}; \
";

[weary niche]

I'm sure the topic has been discussed previously, but is there any feasibility in a property for CfgWorlds classes that disables the effects skyColor and fogColor have on the horizont object? It seems like over the years the map community has tried a handful of times to find a way around it since the move from Arma 2, and as a result, a lot of maps that aren't islands don't have that larger feel that maps like Takistan had back in the Arma 2 days.

[full sonnet]

I don't know how these effects work. But I guess it should be possible

[ivory ridge]

If something would be fixed with lighting/terrain cfg i would also suggest to fix terrains where the skybox is broken because their creators haven't updated these terrains after the new lighting update deployed in 2016. For example these terrains are effected:
https://steamcommunity.com/sharedfiles/filedetails/?id=543288811&searchtext=aristartis
https://steamcommunity.com/sharedfiles/filedetails/?id=375720350&searchtext=tilos+fs+fr
https://steamcommunity.com/sharedfiles/filedetails/?id=290657458&searchtext=calivigny
I would suggest a kind of "fallback solution" for that: If a terrain refers to the old cfg? the Engine loads the stratis cfg as default (maybe with rpt message).

[full sonnet]

🤡

[carmine juniper]

How it started How its going

[mental fulcrum]

which one it was?

[full sonnet]

New script instruction for nular operators, instead of double-using the get-variable one

[mental fulcrum]

for perf boost?

[full sonnet]

ye, smol one

[full sonnet]

"Hey zeus is freezing when you unsync a unit that was synced to a trigger"
Oh weird, lets look at the code

      for(int j = 0; j < syncedEntities.Size(); ++j) // go through all synced entities
      {
        if(curatorLogic->IsUnitAssigned(syncedEntities[j--])) // if entity is assigned to curator
        {
          ...
        }
      }

https://tenor.com/view/sus-among-us-gif-24132003

Thanks for the helpful comments. But... I thiiiiink we might have something sussy here

[pallid cliff]

j-- = endless loop?

[full sonnet]

This is actually 3 nested for i loops
The second one is a filter/copy_if that was manually implemented as a for loop, and it was done wrong.

The last one is just a find/contains, that was manually implemented as a for loop. And its only there to check if a container contains an object, before removing it from the container.
But the remove operation has its own internal contains check, so that is completely useless.
It actually iterates a hashmap, per-element. To check if the key == object. And if it finds it, it removes it.

Also O(N³) Or is it N^4? 🤔
This is why maintaining RV is so hard

[steel rune]

[pallid cliff]

Spaghetti Code

[mental fulcrum]

How does it even happen, it had to be not tested at all if it went unnoticed xD

[pallid cliff]

I am wondering that too

[mental fulcrum]

This is why maintaining RVshitty legacy code is so hard
FTFY

[full sonnet]

Yeah I assume noone tested this scenario ever https://feedback.bistudio.com/T175342

[mental fulcrum]

write code

not test it

👏

[full sonnet]

But that's not normal code I usually see.
It actually has comments on it explaining what happens.
They are useless comments that litereally just repeat what the code right next to it already says, but it has comments.

[pallid cliff]

Did it say" loops until game crashes"?😄

[mental fulcrum]

They are useless comments that litereally just repeat what the code right next to it already says, but it has comments.
usually a sign of a junior that read somewhere that "comments good" :D

a++; // increment a

[full sonnet]

It forgot to explain why it wants to do a infinite loop
I'm actually helping a guy who's in college starting to learn programming.
He always puts useless comments into his code.
But his teacher gives him full points if he has them, and deducts points if the code has readable variable names and is self explanatory, but has no comments

[mental fulcrum]

college programming is a joke

I'm an expert, I've dropped out of 3

[viscid onyx]

That sounds like it's kind of about getting into the habit. When you've been in deep on a project for a while it can be easy to just think "oh this is readable to me so obviously everyone else will easily understand it" - better to have it ingrained to automatically add comments and not always need them, than to get lazy about it and end up leaving nightmare code for someone else later on

[mental fulcrum]

No, name your things reasonably and leave comments for weird stuff or argumenting why something is done the way it's done if written in a way it does not make sense on the first look (be it a bug in something or some weird behaviour)

good code is self explanatory

[full sonnet]

for(int i = 0; i < syncObjs->_objects.Size(); i++) // go through all synced entities
{
Entity *entity = syncObjs->_objects[i];
...
}

for (auto& entity : syncedObjs)
{
...
}

🥺
(that top piece of code was actually copy pasted 6 times in just one file)

Now the question.
When you switch a unit to a different group in zeus. Should that un-sync the unit from the trigger its synced to? 🤔 Probably not

[viscid onyx]

Yes in principle, but how much do you trust an individual coder to be able to determine when something is actually self-explanatory, versus an arcane hellscript that they only automatically understand because they've been dealing with it so long?
I'm not saying that commenting on every every _var = 1 + 1 is the ideal end state, just that "always comment" is a better habit to start from than "I know what this means so I don't need to explain it"

[mental fulcrum]

Yes in principle, but how much do you trust an individual coder to be able to determine when something is actually self-explanatory
naming conventions, common sense, and most importantly, Code Review!

[viscid onyx]

Probably not if the unit was synced, probably yes if the group was synced.
I think.

[mental fulcrum]

Not a place for this discussion as this can go long, there's fine line between good amount of comments and too much.

tl;dr: git good, write clean code

[mental fulcrum]

groups can't by synchronized to anything

only objects and waypoints

[viscid onyx]

Interesting

I never use syncing but that seems like something people would want to do with it 🤔

[solid crown]

I think a good rule of thumb is that good code should answer to questions "what" and "how", while comments answer to question "why". (I should do that myself too consistently, tend to forget it sometimes 😅)

[halcyon moth]

These days when you put a dumb comment on each line, everyone will think chatGPT wrote it.

[dim bone]

//jump code bellow //this is a comment //this is a comment commenting the comment //...

[radiant lagoon]

Is it normal that uniformType = "neopren"; does destroy the AI movement? When i add this the AI wont following any Waypoints, its doing just some trash

Red just moves to that corner and stops there and does nothing while green moves to blue dest. point and arrives successfully

[trim crane]

Wild guess. Maybe engine expects the value to be case sensitive?

All vanilla values are case sensitive: ["Neopren","Default"]

[radiant lagoon]

Where's this?
Currently ive deleted this uniformType cause it was just a test and ive forgot about it and noticed later that AI is just weird and first i was thinking its my terrain im working on

[trim crane]

Just wrote a code snippet to get all unique uniformType values

All start with captial N so I assumed maybe casing is the problem somehow

I think AI makes units take water routes if they wear swim gear, looks like they tried to get into the water and these rocks aren't blocking path finding properly?

[radiant lagoon]

Same i was thinking too that they search the water route.

Cliffs got properly 1:1 working geometries, spent some hours in them to make them 1:1 like visuals are
I hate it when you cant shoot at what you see

[trim crane]

True, vanilla rocks suck, you can hide inside some of their parts, shoot through some and can't shoot through some other clearly open areas

Perhaps yours aren't blocking navmesh\pathfinding (not sure how its done exactly in A3) properly?

[radiant lagoon]

No they work fine i got no AI going clip trough etc and also vehicles dont have any issues. And i did a lot of such stuff, also some with caves in them & AI workable (but not on this map, here are just basic cliffs without interiors)

When that Neopren is deleted everything works fine like the vanilla soldiers, they go same route and get to dest. point of waypoint with 0 issues

[full sonnet]

It influences the "water penalty" when calculating path.
So neopren units are more likely to cross water, instead of trying to run around it.
The code for that is weird and I would say it's F'ed (with neopren it uses *=, without it uses += in the same calculation..) but no touchy

[full sonnet]

looking at that image. I guess they decided they are faster in water and want to go around north to swim through that pool

[radiant lagoon]

What is the point for neopren in Arma, will it swim faster or why its been added?

Im not really in need of that it was once a test cause found it in the AoW Config and added it and forgot about it and didnt notice that it breaks my soldiers at all

So i had to find the problem first why my soldiers got broken and looked first in my terrain stuff

Is this a "bug" to make a ticket for?

[full sonnet]

Its for when you have a diving suit. With fins on your feet

It is unlikely to be fixed because AI/Pathfinding stuff is too much chaos

[radiant lagoon]

Alright, thank you

[steel rune]

that sounds like an interesting path finding shortcut

[radiant lagoon]

Is it known why the Vanilla AI Tanks are doing weird following Waypoints in Groups? I mean the leader works fine so far but the followers just get off formation, sometimes just stop or stuck somewhere, even turn 360° on same place and get lost the formation

I have reworked my AI in such behaviour, gave them some school (last tank is me not in the group in front of me)

They move in all formations pretty smooth together and also will arrive together the waypoints. Honestly. I dont really know what the real issue is with the vanilla tanks

[radiant lagoon]

Im not sure if it might be the AISteering Component, since i have added my own Sensors and im just using

Components
instead of
Components:Components

so im not inheriting the AISteering Component.. Maybe this disturbs the vanilla tank AI movements in groups?

[scarlet root]

interesting find

A3 broke AI groups with more than 3/4 vehicles (cant tell if since 1.0 or after a later update)

you can switch to the pre A3 AISteering via scripting (has a few other downsides instead and can be done only globally - and not per vehicle/group)

https://community.bistudio.com/wiki/useAISteeringComponent false

your config setup may cause it also use the old system

[radiant lagoon]

Oh nice i got to try this today evening. Thanks a lot!

Im not 100% sure if this makes the vanilla Tanks behaviour that bad in groups, its just what i think since im not inheriting Components for my ground vehicles.

Also a bad thing is i think thats the issue why i cant get smokelaunchers to work.

For my aircraft im inheriting the Components and the Flares work when they get locked by AA

[radiant lagoon]

So this works only with script? Or is this a cfgVehicles entry?

If its only for script, whats about inheriting Components and add the AI steering and leave it empty?

It would be like inactive and the behaviour should go back to the old behaviour?

[scarlet root]

only Dedmen (or maybe also KK/Leopard20) could check now and potentially tell you what exactly it does and how it relates to config definition exactly.

overall reyhard, oukej and the BI programmer probably know most about it
https://forums.bohemia.net/forums/topic/191515-ai-driving-feedback-topic/?page=32&tab=comments#comment-3290606

that aside i am pretty sure you can disable the "A3 type" also via config as you did/or by similar means - we fumbled with the config side a few years ago too

https://community.bistudio.com/wiki/Arma_3:_AICarSteeringComponent
https://community.bistudio.com/wiki/Arma_3:_AI_Config_Reference#AITankSteeringComponent

https://dev.arma3.com/post/ai-path-following-improvements
https://forums.bohemia.net/forums/topic/211039-tanks-tracked-vehicles-driving-and-handling/

The "new driving stuff" is just the PIDs. Set useAISteeringComponent to false or delete the component class iirc (inherited all the way from Car_F base class) and you're back to the "old driving"

(from oukej in the first link)

[radiant lagoon]

Yeah right now the AI works pretty nice with ground vehicles, so wheeled and tracked and amphibious if i dont inherit but looks like the smokelaunchers wont work, ill try now to inherit it and add the AISteering Component but ill keep it empty and ill see what happens with the drive handlings. If it wont work ill keep it the way i already did, the smokelaunchers are not that important for me since smoke does perfomance issues anyway 😅

Thank you sir for looking around, will check these links out

[light lintel]

Very cool stuff! I only know AI tanks to drive okay in column formation.

[radiant lagoon]

with just inheriting Components: Components and adding class AITankSteeringComponent{}; the whole AI gets broken, 2 tanks stay there and are like lost and 2 tanks are going to the waypoint

ill do it my way like i did, with the RTS Build menu mode i need the AI working as Commander and to play c&c also vs AI and AI is also needed in player squads, smokelaunchers arent important for me so i can live without that feature, working AI is for me a higher priority

[harsh orbit]

the vehicle must have an AITankSteeringComponent or AICarSteeringComponent in its config. that command can globally enable/disable it
also it seems that command only affects tanks, not cars

[umbral orchid]

I think the smoke launchers are only used when theyre locked too (ie. with a missile)

[light lintel]

There is a way to use the smoke via script, I'm fuzzy on the command now, let me check

I believe I had to use action to get the AI to use the smoke launcher https://community.bistudio.com/wiki/action

Possibly in combination with https://community.bistudio.com/wiki/Arma_3:_Actions#UseWeapon

Alternatively maybe https://community.bistudio.com/wiki/forceWeaponFire might work too?

KK has some examples at the bottom, in the comments:

vehicle player fire "SmokeLauncher";

Or tank action ["UseWeapon", tank, commander tank, 0];

[radiant lagoon]

Somehow i got it to work, got it on the top right and when i press C the sound of it works and the count lowers too but no smokenades show up. Also did the config entries like degrees, in turret etc.. dont know why the nades dont appear

[umbral orchid]

are you using a custom weapon/ammo class? the smoke grenade for vehicles is a scripted effect

[radiant lagoon]

No im using SmokeLauncher and SmokeLauncherMag the vanilla ones

Added the TransportMeasureComponent aswell and these 3 or 4 cfgVehicles entries, degree, how many nades, in turret and one more

It doesnt need any Memory Point in model?

It shows up on the top right amd when i press C it activates so i can listen the deploy sound but no smoke nades show up and the ammo count decreases aswell

My vehicles use KK's oneman script and my vehicles dont have commander slots

[trim crane]

_num=GetNumber (configFile >> "CfgVehicles" >> typeof _v >> "smokeLauncherGrenadeCount");
_vel=GetNumber (configFile >> "CfgVehicles" >> typeof _v >> "smokeLauncherVelocity");
_useTDir=GetNumber (configFile >> "CfgVehicles" >> typeof _v >> "smokeLauncherOnTurret");
_angle=GetNumber (configFile >> "CfgVehicles" >> typeof _v >> "smokeLauncherAngle");

from the BIS_fnc_effectFiredSmokeLauncher script

[radiant lagoon]

Yeah these are the values you can set in cfgVehicles, i got them.

The problem is i can activate it with C and i can listen the sound of it and my ammocount decreases and it does reload (you can do it 2x with 1Mag) but no smoke appears and also no grenades to see, even not at pos 0,0,0 so its like it shoots it but its not there.

[trim crane]

    class SmokeLauncherAmmo: BulletBase
    {
        muzzleEffect = "BIS_fnc_effectFiredSmokeLauncher";

Check what yours has

Maybe you replaced it

[radiant lagoon]

Im using the vanilla smokelauncher, i havent done a new one.

Or does the vehicle need anything to make it work in the turret?

[trim crane]

Maybe the mod replaces muzzle effects?

Not sure if it needs anything to be honest

[radiant lagoon]

No because the vanilla and other mod vehicles still work with smokelaunchers

[radiant lagoon]

Do you have insight of the script itself? I mean where the spawn gets executed, must be something with createVehicle or modelToWorld something

Probably there is something defined. I know the smokeLauncher gets inherited of MGun thats what i was able to see yesterday in the AoW config

[meager sage]

Uses code from VBS2 Smoke launcher by Philipp Pilhofer (raedor) & Andrew Barron
*/``` ![notlikemeow](https://cdn.discordapp.com/emojis/700311897937018890.webp?size=128 "notlikemeow")

[umbral orchid]

that's strange that's all that should be needed, memorypoints arent needed iirc

and yeah the vehicle countermeasures being scripted for some reason is strange i figured it would have been something updated for A3 ( a lot of the older scripted effects were) but it never was

[radiant lagoon]

Probably you are right and i did something wrong in my config, i have to check that all. So it 80% works even with my Components way, just the grenades dont appear. Sounds for me 100% it has to do with the muzzle itself on my side for sure. Ill check today evening, thank you

[trim crane]

The worst thing about is that muzzleEffect doesn't get called for remote entities away from the camera

So if a tank dropped a smoke screen far away and then camera flew closer to it, you'd see no smoke screen

    if(positionCameraToWorld [0,0,0] distance _vehicle > 500) then {
        [_vehicle] call (uiNamespace getVariable getText(_cfg >> "muzzleEffect"));
    };

its fairly low distance too, I'm having to script-call the script just to make sure its there for far away vehicles

But its a bad thing either way, because remote vehicles positions can be way off

And by the time you fly by their position corrects and its far from where smokes dropped

All in all this system sucks and should be rewritten

RHS has custom smokes script that creates global smokes

[radiant lagoon]

Problem is also if people use low graphic options and you got some destroyed vehicles closely the particle effects are lowered and the smoke disappear

[trim crane]

Smokes are still OFP-tier 20 years later

[radiant lagoon]

Wrecks cost a lot of particle effects and for a to long time

Ye i know, im getting rid of them as much as i can

[meager sage]

Inb4 lowpoly mesh smokes

[radiant lagoon]

I have redid the wreck smoke/fire effects and did the duration a lot shorter

[trim crane]

Would still be better tbh

Smoke screens are borderline useless in PvP, they mostly mark your own position and provide a good reference point for targeting

Even if they do cover you, they usually just make your character stand out against smoke more

[umbral orchid]

i mean thats not how you should be using smoke anyway if you're dropping it on your own positon all your doing is communicating 'hey theres someone here shoot here' to anyone with a brain (aka not AI)

you use it to conceal your movement or put it ontop of enemies

[umbral orchid]

ah what a pain, some particles despawn when theyre out of view too

[trim crane]

Do you have BIS_fnc_fire modified?

Or removed from vehicle event handlers

_this call (uinamespace getvariable (gettext (configFile >> "CfgAmmo" >> (_this select 4) >> "muzzleEffect")));

Basically its Fired in EventHandlers that calls projectile muzzleEffect script

Not even engine-driven

[mental fulcrum]

Don't get me on counter measures weapon disappearing in MP

That makes them useless too as you can't use them.

[radiant lagoon]

I got on my vehicles just

class Eventhandlers
{
Init blabla own one
Killed blabla own one

Do i have to add something else there?

So in the Init is the onemanscript by KK and my random number script

[trim crane]

Most BI vehicles have _this call (uinamespace getvariable 'BIS_fnc_effectFired'); in fired EH

This script eventually calls smoke screen script

[radiant lagoon]

Might be worth a try yeah, since i just added Eventhandlers instead of Eventhandlers: Eventhandlers its ofcourse inactive

Ill try this also and will give you answer

[radiant lagoon]

For Aircraft they work pretty well also against player or AI, had no issues with these with my aircrafts and sam sites

In MP

[mental fulcrum]

the issue is in land vehicles/tanks. Basically the weapon disappears for commander and can not be used anymore.

Reyhard was explaining it somewhere ages ago.

[radiant lagoon]

If you removeWeapon and addWeapon + the Magazine it also wont work anymore?

[meager sage]

_this call (uinamespace getvariable (gettext (configFile >> "CfgAmmo" >> (_this select 4) >> "muzzleEffect")));``` ![notlikemeow](https://cdn.discordapp.com/emojis/700311897937018890.webp?size=128 "notlikemeow")

literally the entire function

[radiant lagoon]

love you, it works

thank you a lot

also the formations still keep well in waypoint movements, so everything is prepared now. very cool

just dont inherit Components and they work fine, add the formation values too in your cfgVehicle config to define the distances for your vehicles and they do a great job, probably BI should just remove that stuff again and go back to the old variant before that get added

[light lintel]

Yeah please if we could have working tank/tracked vehicle formations again that would be amazing

[harsh orbit]

if it's a tank then doing useAISteeringComponent false would've given you the same thing

[radiant lagoon]

Does that need a init EH or is this possible to add into the cfgVehicles class?

[harsh orbit]

well it's a global thing

it disables it for all other tanks in the game

[radiant lagoon]

Could you also improve your anti flip / fly tank script? That works pretty good but sometimes it doesnt

[radiant lagoon]

So its only once needed in a mission as example?

[harsh orbit]

yeah. I'll see if it's possible to disable a component per object

[radiant lagoon]

Would be useful for others aswell so not only for my own

[harsh orbit]

yeah it's possible. I'll probably add it

[radiant lagoon]

Would be nice and when this works again for all others it would make it a lot better for many missions etc.

Not sure if you could add your anti flip too somehow into the game so its standard added instead of a subscribe, bit sad tanks and other vehicles are still able to go to orbit or get flipped. Probably a small bug but makes big sadness sometimes

[harsh orbit]

no I won't add it

[pallid cliff]

Have a trauma now after reading through the engine code?😄

[harsh orbit]

I meant the anti flip thing 😅

[pallid cliff]

Ups

[radiant lagoon]

Im allowed to edit yours?

[harsh orbit]

yes

[radiant lagoon]

Thank you

[radiant lagoon]

is it possible that the visual sensor component has been changed or is broken now? before i wasnt able to see vehicles behind structures or walls and now im able to lock and see vehicles behind tanoa office buildings

only if the terrain is not flat so im behind a hill or im more down or up i cant see the vehicle on the radar

[radiant lagoon]

behind structure i can now lock and see enemy on radar

behind "hill" i cant see on radar

[light lintel]

Is this visual or radar?

[radiant lagoon]

VisualSensor

[scarlet root]

@radiant lagoon probably best to move this to #arma3_config - your problems are really not related to this channel

[radiant lagoon]

why not? "my" "problems" are or can be improvments for everyone in any kind of next arma patches if someone has time to look over it / or even fix it

visual sensor is meant to be visual and should not be able to look trough objects / houses

it worked before in previous versions of arma

[mental fulcrum]

make a FT ticket then if it's a bug

[weary niche]

Any news on whether this might be a possibility?

[full sonnet]

Nop, I forgot about it or didn't find the time

[wet citrus]

Hey there, Just wondering if the DoMove command for ai will get fixed eventually?

[halcyon moth]

Fixed how?

[wet citrus]

its inconsistant. The a few devs of AI mods are stopping their work due to this (and other issues with the ai)

[weary niche]

Well here is your reminder then if you do find the time 🙂

[viscid onyx]

• inconsistent in what way?
• is there an FT ticket about this with details of the problem?
• has it broken recently or has it always been like this?
• is it only broken on dev branch, or on all branches? (If all, this is not the correct channel)

[quiet kernel]

Just a pro-tip. There is only a small team of post Arma 3 working on things. Filing tickets, with reproducable steps (yes, even if it takes you several hours to isolate). Gives you far higher chance of getting things fixed then just expecting the devs to do all your troubleshooting.

[halcyon moth]

I gave up on my last AI movement repro because the bugs kept interfering with each other.

[full sonnet]

And fixes also do the same. Fixing AI is very risky and we'd like not to

[ocean sail]

No water splash (or any particles) on bullet hit in v2.14 + prof and 2.15 dev (in any Particles Quality setting selected)
Cannot repro in legacy Arma 3 2.12, water splash particles works fine

• launch any world e.g. Stratis
• spawn player with gun near water, e.g.player setPos [4782.44,2775.76,0.00177091]
• shoot at the water

[formal cipher]

Also on stable 2.14 (without prof)?

[ocean sail]

no all ok on stable

[full sonnet]

When did you test? Last profiling did "Fixed: shotShell projectiles were creating many tiny explosions when passing through foliage"
That should only affect foliage, but maybe?

[trim crane]

Related ticket: https://feedback.bistudio.com/T176022

[trim crane]

Shooting water VS shooting ground that produces water splashes

[full sonnet]

So Blame leopard? 😄

[trim crane]

The issue with HE shots exploding 20 times when shot through trees is fixed though

[mental fulcrum]

the good? it works
the bad? something else doesn't

[tepid dock]

So this is basically him right meow?
https://preview.redd.it/foqe2bvzbi8a1.gif?format=mp4&s=67473f90ef840ba35fdae54430d98a4a3ccd97a5

[unreal arrow]

all of us

[harsh orbit]

I couldn't even imagine something like this being related to water splashes 😅

[lament crow]

Newly onboarded team members faced with the daunting challenge of deciphering and modifying legacy code that's more than a decade old shouldn't shoulder the blame in any organization, particularly within the realm of software-related companies

In this regard, it's evident that Leo has consistently demonstrated his commitment and dedication by giving his best effort, as he always does. I commend the courage of both current and especially new developers who are striving to disrupt the prevailing cycle of minor fixes in favor of more substantial, albeit riskier, improvements

To be honest, I'd prefer to prioritize comprehensive solutions that have been carefully refined, even if it means investing more time, over hastily implementing quick, small fixes

[wise dragon]

so definitely blame leopard

[formal cipher]

Always blame the new guy

[tepid dock]

What they both said!

[wise cargo]

https://tenor.com/view/sweats-gif-25346666

** knowing that leopard isn't the newest guy **

[dim bone]

I allways blame the president.

[dim bone]

[full sonnet]

From your friends that brought you setDammage:

[unreal arrow]

Consumme ammount

[carmine juniper]

Bohemmian mmoment

[tepid dock]

Your such linguistical purrists! 😿

[digital vessel]

Any chance of a "haze strength" setter/limiter?
I think haze helps infantry blend-in (because grass isn't rendered at distance) and gives them a chance of hiding from aircraft
Having a script command to keep everyone on the same setting would be appreciated

[waxen vigil]

[unreal arrow]

server enforcement is in discussion so do provide feedback/ideas

[tepid dock]

Back when our group still existed and did PvP-only matches, we would have loved a setting for forcing the minimal grass level, so that everyone who turned grass off didn't have an advantage over everyone else.
We used custom terrains whose topography limited view distance anyway so the lack of rendering in the distance would have been a non-issue for us.

[mental fulcrum]

you can force grass with setTerrainGrid IIRC

[ruby edge]

For the DayZ style Exile server we had i set minimum viewdistance and objectviewdistance of 1.200 and max 2.800.
Since people could cheese with setting objectviewdistance (not render base walls but containers rendered etc).
Also forced a minimum terrain grid value of 12.5 and max of 3.125 (Forcing terrain grid made sure no one is turning grass off).

[tepid dock]

Huh? Why didn't our scripter make use of that command, if the biki says that it's been there since v0.50? 🤨
Well, if it works that way, then disregard my request 🤷‍♂️

[trim crane]

Being able to control terrain geometry separately from grass would've been nice too

[digital vessel]

maybe something like

setHaze []
    taking empty array (setting isForcedByScript = false)
    or tripple [baseHeight, baseBeta0, heightDecay] (setting isForcedByScript = true)

getHaze returning [bool, values]
    0: isForcedByScript
    1: tripple of current haze [baseHeight, baseBeta0, heightDecay]

[true narwhal]

is there a channel with notifications when a new update is released? or do we have to check the dev branch changelog forum?

[carmine juniper]

Just subscibe to the changelog forum

[unreal arrow]

I always check this thread https://forums.bohemia.net/forums/topic/140837-development-branch-changelog/?page=61&tab=comments#comment-3481774

[lament crow]

"Added: Video option that allows reducing haze strength (Standard is default and means the world config settings are used)"

☝️ That one is pretty amazing! Previously the reduced haze mod was pretty much mandatory for many players but the mod did not work on all maps so a vanilla solution was very missed until now! Thanks devs

In the vain of these amazing quality of life improvements done to the game, any chances of a FOV slider? Sorry for asking this now since it was requested too many times in the past but with the dev team strengthened by the presence of new devs perhaps there is more room for such request these days 🙏

[wise dragon]

messes with too much ui stuff

[lament crow]

There are tons of players (and most streamers like Karmakut) who "hack" their way into higher FOV values and I've never seen them complaining about UI issues so those should be minimal and/or tolerable...

Perhaps those UI issues are fixable too, don't know... but most players who mod their FOV don't seem to care

One of the most common complains I get from friends who get into Arma 3 for the first time is the old school FOV that trully feels outdated by now... a proof of that is that when Starfield was released it lacked a FOV slider and it was so requested by the players that it was one of the first things they added with the first patches (yes, it is that important nowadays specially with all the different aspect ration / resolution monitors out there now)

[trim crane]

Didn't dedmen fix that recently? I think it was breaking map icons and no longer does

[wise dragon]

possibly, idk i use base game fov iirc

[verbal stag]

I think all BDC reticles (maybe just in 2d scopes?) and mil dot etc are all scaled incorrectly as soon as you change FOV. Not sure it’s possible to fix that if a slider were to be introduced?
Given A3 is meant to be more of an infantry sim as opposed to just another FPS, having a slider that immediately breaks part of that sim/point of difference isn’t a great look!?!

[carmine juniper]

A3 is meant to be more of an infantry sim as opposed to just another FPS
It's not
as soon as you change FOV
Changing FOV is unsupported feature

[umbral orchid]

it breaks https://community.bistudio.com/wiki/ctrlSetAngle re: ui stuff

see: https://feedback.bistudio.com/T136844

but yeah the big one to me is you basically break most sights in the game with it

I very much doubt that will be fixed

as the only 'fix' for it is to undo their FOV change

[mental fulcrum]

Option to change the FOV in internal camera only (not optics) would be neat and wouldn't break much

[full sonnet]

I think you have been told many times. Chanegd FOV is unsupported and causes many issues.

for camera set fov that should be doable and neat tho

[lament crow]

Indeed, the subject won't come from my mouth again. Actually I am not even among the increasing number of players with ultra wide monitors who find the claustrophobic 70° vanilla FOV completely neglecting their investment. Hell, even for regular aspect ratio 27'' modern monitors something like 70° makes any game feel completely outdated. Luckily Arma 3 is still being supported... just not THAT supported for such feature as custom FOVs but other cool features are in the work I am sure of it 😉 Keep up the amazing job

[mental fulcrum]

But wider monitors do see more

[unreal arrow]

We have FOV commands for camera

[grim adder]

I gotta say you need to do a hotfix asap to include the new video option Reduce Haze strength because it is a game changer for flyboys.
i just tested it and wow what a difference. why wasnt this made as an option years ago?

[ocean tundra]

show screen shot

[trim crane]

If this feature gives gameplay advantage there must be a way to server-wide enforce it for PvP scenarios

Same goes for other some other settings too to be honest

[wise dragon]

not a hotfix because it doesnt fix anything.

[unreal arrow]

do you allow client mods on your servers like reduce haze mod?

both low and very low settings are hard set, cannot be overridden with a mod. Standard is whatever world config is, modded or vanilla. Perhaps we need another hard setting that is close to vanilla haze, doubt anyone would want to set it for themselves but as server enforced option it could work

[grim adder]

@ocean tundra try it yourself. use beketov as an example and see

@wise dragon so you didnt try it then. maybe try it and see. use beketov as an example

[wise dragon]

why would i use a modded map to test a vanilla feature

[trim crane]

No we don't. I meant that when new haze setting in the menus. And that it would be great to be able to force everyone to use standard through server settings somehow (I'd prefer a server-side scripting command)

[unreal arrow]

so what are the downsides of being able to see clearer

[mental fulcrum]

Exactly that. Being able to see clearer which gives advantage in PvP scenarios.

[unreal arrow]

There is setFog/params command for that. Haze in the engine implementation is nothing more than glorified fog that has been separated into own simulation

[harsh orbit]

true but you can't normally set fog in MP (unless you have access to console, or the scenario provides you with something to set it)
what they mean is that you can now just set it from settings. everyone can. anywhere, anytime

[trim crane]

There is no reason not to run "Very Low" setting because it gives slight gameplay advantage over other options.

As for scripting command to control it, having it client-side is fine too to avoid new messages and all that stuff. description.ext parameter is okay too.

[unreal arrow]

Fog is fog, haze is haze, both are fog but haze =/= fog

[unreal arrow]

dont think client side is wise or needed, server side command to enforce either option and server option, where serverside command overrides server option?

[viscid onyx]

That is probably the "most sensible" implementation, but I kinda like the idea of being able to clientside it, for example letting Zeus turn haze off but keeping it on for the infantry

[trim crane]

Client side is easier to implement (no messages\syncs needed) and gives flexibility from @viscid onyx example.

[unreal arrow]

this could be curator setting

in fact could be permanent if curator, ignore server override

[trim crane]

I'd let scripters\designers handle it

[unreal arrow]

but not everyone playing zeus is scripter designer

[viscid onyx]

Zeus was just an example of something you might want to be different. There are other potential uses. You might want to give it to pilots but not others, or to dedicated snipers, or ...
That's just my preference though. Any of the options is better than not being able to restrict it at all.

[ocean tundra]

Hi encountered weird bug with arma3 launcher, when subsribed to 2 mods with the same name, the game will load only one of them regardless of whichever one is selected (i.e if i select only MOD 1 in launcher, the game will still load MOD2)
MOD 1: https://steamcommunity.com/sharedfiles/filedetails/?id=1190237082
MOD 2: https://steamcommunity.com/sharedfiles/filedetails/?id=914182380

[unreal arrow]

#arma3_feedback_tracker + ticket

[trim crane]

Did something change about ESC menu recently? I repurposed "Save" button (idc=103) in RscDisplayInterrupt to do my custom action but it seems that pressing this button closes the menu, compared to stable where it doesn't

Looks like game ignores return of ButtonClick. Returning true used to keep the menu open, but not anymore in dev build.

[trim crane]

Made a ticket with repro: https://feedback.bistudio.com/T177038

[unreal arrow]

could be, the engine action might have gone before the handler.

i’ll fix it before next dev

[radiant lagoon]

if people got time please test this if you can also confirm that the visualSensor is broken since 2.14

Before it was not possible to see targets behind any kind of houses, rocks, objects etc

It worked before like when target is behind a hill it did not show up in radar, now the AI is able to target at you all the time if you are behind any kind of object (hill still works fine)

https://feedback.bistudio.com/T176526

[floral arrow]

R.I.P. Fatigue

[trim crane]

Seeing drawLine3d lines became even worse now that I moved to 4k monitor

Can we finally bring 3DEN thick lines to scripting?

[unreal arrow]

example?

[trim crane]

Area Widget

Hmm, looks like these are just stretched images and are drawn as UI

drawLine2D [from, to, color, width] or something

2d in a sense of it not intersecting with world geometry

A modification to ST_LINE instead perhaps?

Unless there is already an engine way to have drawLine3D being drawn with custom thickness, I'd pick that

[full sonnet]

weep woop weep woop

[ruby edge]

Would that also work for objects or is it just for helicopters ?

[full sonnet]

all vehicles

This currently resets the "filled percentage" of the water inside object to zero to make it float back up.
Probably better to make the fill level a separate command

[unreal arrow]

sweet

now make it MP compatible

[full sonnet]

already is

[carmine juniper]

I think it is very local command so MP is not really a big concern?

[full sonnet]

it must me network synced because the vehicle locality will change when a different driver gets in

you would end up with floating, and then you switch seats with copilot and suddenly you sink

[tall pumice]

story of my life

[full sonnet]

[carmine juniper]

Interesting

[ocean tundra]

what happen to the glass

[carmine juniper]

They said sea is not a cup of their tea

[bleak canopy]

Emergency blow on the submarines =D

[viscid onyx]

Is/will the watertightness factor be in config as well? Probably useful for amphibious aircraft (Sea Kings, floatplanes)

[full sonnet]

Don't ask

[full sonnet]

already is
the script command just lets you edit at runtime

[viscid onyx]

Neat

[umbral orchid]

I cant remember if its still a thing but is there a way to fix planes instant dying when touching water?

[full sonnet]

i think helI's also used to do that and we recently fixed it?

I thought it was also fixed for planes

We can probably make it impact speed based

[umbral orchid]

ill quickly test it again but last time I tried to belly land in the ocean it essentialyl setDamage 1'd the plane soon as it touched the water

[full sonnet]

I crashed one into water last week, but I expected to die

[umbral orchid]

yeah tried it just now with the caesar btt

soon as I touched the water dead

was going very slow too

[ocean tundra]

acidic water

[umbral orchid]

thought it might have been touching any surface without landing gear but could belly land and survive on the ground too

plane not in a good way but still alive at least.

[bleak canopy]

btt is just very fragile. all the planes do take time to sink and die if placed over water.

[umbral orchid]

ah damn

I guess the impact damage is just the same as hitting the ground then

it wouldnt be such a big deal if the plane 'dying' didnt instantly kill the pilot

[trim crane]

The biggest bug I'd love to see get fixed is player being stuck in lean left or right state. Its both extremely immersion breaking and affects gameplay (somebody might be peeking through window but remain unseen because they appear to be leaning).

No repro sadly, but it happens all the time on live server

[unreal arrow]

can you log animation state from local and remote client?

[trim crane]

Can you get lean state through scripting at all?

[unreal arrow]

it’s an animation so you should be able to get the state

[trim crane]

Doesn't seems so, getting same animationState regardless of leaning

[unreal arrow]

interesting

was there a command for it?

[trim crane]

Leaning is also gradual with TrackIR and similar stuff

[trim crane]

I don't think so. Might be a good candidate for a command too.

[unreal arrow]

yeah a getter would be useful

[trim crane]

Either way there is 100% disconnect between local and remote clients of how they see each other leaning, iirc you start seeing people leaning in animations where you can't lean at all too.

I even managed to repro it once after hour of trying with local clients, but it happens all the time on the real server, probably involves latency somehow

[unreal arrow]

So they look normal to local but lean on remote?

[trim crane]

Maybe lean state update gets ignored with some condition. This started several years ago and been happening since

[trim crane]

Yes, normal on local, stuck leaning on random remote clients

[unreal arrow]

how does local know he is leaning in order to exploit it

[trim crane]

They don't, its completely random, but it ends up affecting gameplay, if say sniper peeks through window with no leaning but they appear to be leaning to remote player and completely invisible

[unreal arrow]

certainly something to have a look at. ticket?

[trim crane]

Looks like no ticket yet, gonna make one

Same for bug and getter or separate?

Here it is in action, watching remote player stuck in a lean

They are shooting people without leaning yet appear leaning right shooting below the window

Just to show that its not a cheater shooting through wall, here are two screens showing how much lean moves the reticle without moving the aim

This is said gameplay effect, people appear to be leaning and invisible yet they aren't on their side and can shoot others

[true narwhal]

i use this for check if its leaning ```sqf
inputAction "leanLeft" > 0
inputAction "leanRight" > 0

[trim crane]

You can double press lean to keep leaning though

[true narwhal]

[viscid onyx]

That's also missing the other half of the disagree check. It tells you when they are locally trying to lean, but it can't be used to identify when a remote client sees them as leaning.

[trim crane]

Bug ticket: https://feedback.bistudio.com/T177317

[viscid onyx]

@ Sa-Matra - did you check whether leaning is detected by gestureState rather than animationState?

[trim crane]

Just did, it doesn't

Its probably another layer of animations not exposed to scripts

[viscid onyx]

I wonder if selectionVectorDirAndUp on certain spine bones could semi-reliably identify it. I think leaning is the only way to get that much of a lateral tilt without being ragdolled
* other than custom animations not used in normal gameplay

[trim crane]

Yeah, it might be possible to get it through scripting, but you'll need to precache non-leaning angles first, probably spawn a dummy character and set it to all animations that allow leaning to record the angles, then calculate amount of lean from that on real characters

Or we could have a scripting command to pull that float number from the engine

[viscid onyx]

Well yeah, I'm not saying we shouldn't get an easier way, I was just thinking if you really really wanted to log desync cases during this debug process

[trim crane]

Scripting command ticket: https://feedback.bistudio.com/T177319

[safe vale]

What is your server's MinErrorToSend and MinErrorToSendNear set to?

I've run into this before where if the values are set too high things like incremental leaning will stop networking reliably to everyone

[trim crane]

One server has:

MinErrorToSend=0.004;
MinErrorToSendNear=0.04;

[unreal arrow]

So leaning is considered unimportant in the same way as turn of the head, so it is possible it fails to force update in certain circumstances. I can bump up importance and see what happens. Not sure about the command, since you can check with input action leanLeft/Right/LeftToggle/RightToggle. How are you going to use it?

[green summit]

So he can check is player leaning with this:
inputAction "leanLeft" > 0
But if Player double taps the lean button it returns false.

[unreal arrow]

leanLeftToggle ?

[green summit]

It only triggers when player double taps. It dosent keep track is player leaning or not.

[unreal arrow]

makes sense

[trim crane]

Was it always like this? Can you compare with say 6 years ago?

This became an issue few years ago, never happened before

[unreal arrow]

what happened (assuming it happens 6 years ago) the lean was separated from position update message and added to animation update message

so technically on every position update it would sync lean before

[trim crane]

Yeah, guessed something like this. Not sure how this can be solved apart from reverting it back, will higher importance really solve it or just make it happen less often?

[trim crane]

AI also lean behind corners, remote players might be leaning, you can't check that with inputAction

[trim crane]

As for usages, I don't have a use for it personally right now, but since its a part of unit animation pretty sure somebody will find it useful for something. If we had it I could've setup a script to check how many players have their lean differ from what your client sees for example, but that's a debug use.

One use that came to mind: I know that people use lean to peek through walls and floors, maybe having a scripting command could help prevent that (make unit stand\sit if they're trying to peek through the wall with lean)

[safe vale]

In my case I went from values of MinErrorToSend = 0.01; to MinErrorToSend = 0.001; and changed MinErrorToSendNear = 0.1; to MinErrorToSendNear = 0.01; and it was fixed.

[trim crane]

Shouldn't higher values make the game send updates less often?

Unless I misunderstand what these mean

[safe vale]

That would be my understanding, so I changed it to send updates more often

Now I think I might've since changed those values and pushed them a little higher but regardless if it's something BI can fix that'd be great.

[worn spade]

player enableFatigue false;

[full sonnet]

It's time again for Arma dev stories
I recently got reminded that Enfusion has a nice memory profiling/logging tool and ported it to Arma 3.

First discovery, during config/script load, the game allocates 4.9GB of memory, out of which 4.8GB are also deleted during that time.
We allocate almost 5GB of memory, out of which we just keep about 100MB alive..

2.94GB are allocated and 2.94GB are deallocated by only one specific thing.
Resolving a preprocessor macro parameter. Every macro parameter that the preprocessor parses, allocates a 64KB buffer, it writes the macro parameter text into it, and deletes the buffer again.

My macro parameter I mean this

#define X(y, z)
X(abc, def)

abc allocates a 64KB buffer to write 3 bytes into it, then allocates a 3 byte string, copies the buffer content into it and deletes the buffer. Same again for def.
🤯
During my load that happened 49k times.

Made a handful of adjustments to use a 256B stack memory buffer instead (most macro parameters are under 64 bytes).
Allocation load during the same phase is now down to 1.9GB

This is for Preprocessor, so using binarized configs or bytecode scripts already bypasses this issue.

[radiant lagoon]

Is this now good or bad? 😅

[full sonnet]

Well its better now

[full sonnet]

That actually didn't bring a game start time improvements, because it was only 50k allocations, out of the 19g allocations Arma does during that process.
Looking at the biggest number of allocations by number instead of size.. 5g are... Preprocessor again 😠

When our game HashMap's resize themselves because they got too big and need to rebalance.
They COPY every element in the map 🤯
Because defines go into a hashmap, we have 9 million allocations because preprocessor defines are copied
And as this is issue with all hashmaps, that could also give some improvements elsewhere 🤞

Fixed, number of allocations reduced by 3g, Game start time till main menu from 1m50s down to 1m30s

[waxen vigil]

so... improvements?

[full sonnet]

magic

[waxen vigil]

[full sonnet]

The real target is to find the memory leak and 3fps bug. But I can't stop myself from looking at performance

[tall pumice]

3fps bug is where the game randomly goes to a crawl?

[full sonnet]

yes and tree's unload and units go into low-res T-pose and such

[mental fulcrum]

Game start time till main menu from 1m50s down to 1m30s
is your testing PC intentionaly slow? that's quite long.

and ported it to Arma 3.
port the renderer, kthx bai

|| ||

[full sonnet]

That's with mods

[tall pumice]

damn. and here i thought my pc was dying when the fps started tanking at random 😄

thought it was a gpu problem or something like that 😄

[full sonnet]

I thought it was well known.
Since before 64bit build was introduced we had this problem and it was never fully fixed
Now I just need to find a repro

[tall pumice]

i never had the issue until around the time we were wrapping up development on western sahara for release. then suddenly every once in a while after alt+tabbing out of the game and going back, i kept getting single digit frames

haven't seen it ingame anymore in a long time now tho

[mental fulcrum]

I think I had it maybe once or twice since 64bit

never since I've switched to RTX4090

[tepid dock]

And to think that people were making fun of the Performance Improvement DLC for Arma

[light lintel]

You need to have a 2013-2018 era system to easily trigger the 3fps stuff

I did get it to crash a few more times over the summer, including slow frame capture, but wasn't able to find the time to upload it

[steel rune]

I've had that issue on my previous build with an rtx 3080

[rigid warren]

3 yrs old feedback tracker, but it was constant thing https://feedback.bistudio.com/T152339
it did not happened to me anywhere else, but on this one server it was always happening after some time

[ruby edge]

Before x64 update 1-2 hours in Koth always triggered the 3 FPS bug for me. (i7-4790k, gtx 770, 16gb ddr3-1600 ram, game on ssd)

[harsh orbit]

afaik that's a GPU bug. I've definitely seen many cases of mem leak on the GPU (at least it seemed like so)

[radiant lagoon]

I never had this problem and we play atm a lot MP with mass (C&C Stuff), but im using a 10years old rig, Win7 and GeForce GTX 770, playing on Ultra/Extreme, never had T Poses or 3FPS

[harsh orbit]

yeah I've never had the 3 FPS bug but I've had the object flickering issue many times

[radiant lagoon]

I only get object flickering / disappearing when im under the bridge, so the other side disappears flicker short

[ocean tundra]

Pleas also fix the constant stuttering caused by the steam A3 infinite download-pause loop

[harsh orbit]

just launch the game via a bat instead of launcher. Steam won't try to update anything then

we can't fix Steam

[ocean tundra]

Something from the game must be triggering steam to do it

[harsh orbit]

no. if the game did it then what I said wouldn't work

maybe the launcher does it?

[ocean tundra]

One fact about it is that it might sometimes stop the infinite download randomly, but as soon as i load a steam workshop scenario, the loop it will start again

[calm wasp]

even on an i9-14900K with a RTX 4080 and DDR5 it happens, after some time

just that the badwidth and/or dual-channel on the stick of DDR5 itself, helps a little bit with that

the higher the view distance, the sooner it will start to happen or the more you fly in a jet over as much terrain as possible

FPS at first starts to gradually/progressively go down, then textures start to flicker/disappear, your soldier can look ugly as hell, when running, it will make (almost) no steps visually, mostly just sliding etc

then, if you still continue to play, despite all of that, the game will simply force-close at one moment or crash

lowering the view distance aa much as possible, helps, for some time
but it's only a matter of time
https://youtu.be/rP_12Qx5nPI

[radiant lagoon]

Probably this has to do with newer GPU's or newer DirectX, such ive never had in any kind in all these years.

Are there more people with newer setup having this issue? So we could do the circle smaller for Dedmen to figure it out

[full sonnet]

After playing some amount of time game is starting to freeze, rendering distance is changing on its own, animations freeze.
"After playing some amount of time" is not a repro.

[calm wasp]

it happens to all the people, provided you play long enough at once, without reconnecting/restarting, with great view distance and/or having covered/loaded most of the terrain in a jet or heli

have always been like that

even in like 2016 on official Warlords

[full sonnet]

I don't have enough ram and time to play for "some amount of time" and record and analyze that

[light lintel]

Problem is it depends on how much textures are loaded and unloaded , which means it depends on length of time, asset quality, texture quality selected, available VRAM, and at least at one point, page file/memory.

So if you're able to artificially force those conditions faster via dev magic then you shouldn't have to spend so much time on it. But having a weaker machine will definitely help speed it up.

[radiant lagoon]

I cant update to newer DirectX with Win7 and my 770GTX so i got to keep on the latest one i can get for Arma 3 with my main rig. I can test with my second rig which uses win10 if i got there such issues

[calm wasp]

there are too many conditions to repro that

[radiant lagoon]

Played over 8hours in one way without restart and players left and connected in a mod test phase we had and a lot of AI and never had any issues like this. We also played a lot weekly COOPS while one was Zeusening, no issues, such OP,'s were mostly around 2.5 hours to 6 hours, no issues

[light lintel]

GM assets seem to trigger it relatively quickly too, with very high/ultra textures

[calm wasp]

well, with few hours of time, even if all conditions are present, you won't be able to repro, unless very little vRAM and/or RAM and flying over terrain with a good objects density and textures quality

[radiant lagoon]

End of year ill buy a new Rig like yours just a 4090 not a 4080 then i can try again

[light lintel]

Win 7?

[calm wasp]

that!

[full sonnet]

I basically never had it.
Then two weeks ago I played a mission where many people had it and half of the players had to restart game.

I need to get to such a situtation

[radiant lagoon]

Yeah my current main rig is 10 years old

[light lintel]

So was mine, but I still updated for security's sake

[calm wasp]

what was the terrain, view distance and after how much time +-?

asking for hardware specs of players that had this is too much, I guess

[radiant lagoon]

I got like around 40 days left until steam stops support win7

[light lintel]

Also what will be most useful for us to submit? Crash logs don't seem to have helped, frame captures are good right?

[radiant lagoon]

The problem is he needs to know what is executing that issue. Frame captures doesnt show anything just fps, it needs to be something like a log file

Also is it approved it comes of the client? Maybe the server could also do such? Does it happen only in MP, hows about SP?

[full sonnet]

When every player has issues, and many so hard that they need to restart, then its probably not a hardware issue.
Settings and hardware will only change the time how long it takes till it cannot cope anymore. But the issue is clearly there

[full sonnet]

Nothing you can submit would be useful.
I need a repro

I once had a mod report that said "fly camera over this terrain and after a while everything will bug out"
it didn't specify "a while", I flew around for 10 minutes and noticed absolutely nothing

[light lintel]

I've had that happen within a couple of minutes over Gabreta while in 3DEN, I've also had it never happen :/

One way to do it is forcing higher LODs on objects by flying closer, then switching to a different map and repeating

But generally placing assets and forcing their higher LODs help to slow things down. Then change map, repeat, etc until it crashes.

Which is why I think a script that spawns units and moves the camera across a lot of the map very close to the ground should help speed things up

[radiant lagoon]

Easy way to bring it into knees is lets drag a truck into a structure/building and it needs to be stuck. It spawns sooo many little parts flying around and stays on ground and if you do this with 10 trucks you bring every rig into knees

[steel rune]

GM and Weferlingen used to be a problem for me but haven't seen the issue since switching to a 4090

[viscid onyx]

High view distance helps too, I think. I've had it happen in the Editor (max VD) but never in MP (server limited VD).
My feeling is that sometimes under heavy load (high settings and many mods) the game goes into a reduced FPS background mode when you tab out, then fails to go back to normal when you come back. It's still not consistent though (un/fortunately?)

[tall pumice]

well, for me personally, whenever the issue happened, it was after i alt+tabbed back into the game.

didnt need to play for long (i usually dont at all, due to constant restart from testing etc)

it also happened on our ws terrain, which is not exactly hardware dependent

[radiant lagoon]

Are you using one or two screens?
Im using 2 screens and even with alt tab and return i never had an issue. And 80% of my 11k hours arma 3 are also only testing times

[tall pumice]

my way of fixing it was to go into the graphics menu and setting super sampling to a very low value, like less than 50%. then going back into the game a bit, then resetting it back to 100%

i have 3 screens

to me it always felt like my ram was full and the game had issues with reloading whatever stuff it needed.

reducing the graphic settings temporarily allowed it to reload whatever it needed

but again, this is all just theory and no repro

[radiant lagoon]

Ill try this weekend a stress test, doing a 8K texture with the TexView2 SMDI filter, generating layers in TB of my Terrain and playing Arma 3 with a big amount of AI's and let 3 tib harvesters stuck in refinery and recording the game as video, ill try to make a repro crash

[tall pumice]

i dont think this is what is causing the issue, though

[radiant lagoon]

And using PiP scope on extreme PiP setting (this is the only fps drag issue i got)

[tall pumice]

whenever i got the issue, there was barely any ai or complexity in the scene

[radiant lagoon]

So it just happened randomly?

[tall pumice]

more or less, after alt+tab, yes

at least for me. dunno how it's for other people. that said, i didnt had this issue in a very long time now

[radiant lagoon]

I never had it thats weird

I only get a error when im in Zeus and directly alt F4 close game i get an error to press OK

But no crashes or freezes or 3fps bug

[ivory ridge]

Idk if it is related: If I change to a mission with new terrain the server is occupying more and more RAM. Maybe that is also a factor.

I can successfully crash a linux 32bit server with mem param =>more than 3GB (i know, it is out of spec and this was the reason not to send a message about this).

[calm wasp]

sure. has been like this since as long as I can remember

just that people with lower settings and people playing with not that high view distance and maybe few hours or slightly more, moving across not as much terrain, have not that high chances to see this, if ever

on Global Mobilization terrain, it happens much quicker
most probably due to textures resolution and 3D objects quality
and that's where you've played, most probably, when many players had to restart Arma

[calm wasp]

because with your 4090 you need to play more or over more terrain, best possible LoD and trust me, it will still happen 😏

in Reforger, despite more and more detailed and with better textures 3D objects, much much less RAM and vRAM used/reserved and no matter after how much time and even if with ultra+ settings, FPS doesn't tank and no issues

[light lintel]

Maybe 24GB of VRAM has something to do with it 😛

[calm wasp]

this was with my 8 GB GTX 1070 Ti and 32 GB DDR4 RAM, same as in the previous video

https://youtu.be/iH4D-QlxHhw

[calm wasp]

later, with me refusing to restart the game, FPS was so low that there was no FPS, so the game simply force-closed/crashed

that after me lowering the settings and lowering view distance more and more, trying to avoid the need for game restart as long as it was possible

[ocean sail]

I can make all units as player in new Dev rve151158

1. place 2 or more units in Eden Editor
2. make others unit as player
3. result → you able to make every unit as player
   if copy&paste unit with enabled player attribute the copied unit was not be player

[full sonnet]

@unreal arrow 👋

[calm wasp]

also, the higher server FPS, the sooner server FPS goes down very hard and needs a restart
so we had to limit it to default 50 FPS, not only because of CPU load and power bills

it's like there are tickets, like in a Battlefield round/match, where the more players, the more kills, the faster the tickets are drained

so we don't care much about power bills and wanted highest possible server FPS, but being forced to restart the server much more often, since with higher server FPS, server FPS tanks so much faster/more, we had to forget about higher server FPS than default

[harsh orbit]

That's actually interesting

[calm wasp]

the lower the server FPS, the more time it will stay playable

even with default server FPS, we have to auto-restart the server (PvE no headless client) like every 5 hours, when servee full

when the server is less populated and closer to night time les and less players, we manage to leave it without restarting from like 16:00 until 04:00

but are forced to restart it at 11:00 and at 16:00 next time

so server FPS tanking has something to do with how high server FPS limit is and how many players or how much stuff is happening on the server

[trim crane]

inb4 timer\frame counter somewhere gets really big and delta becomes 0 due to float precision and gets fed into function that always expects delta to be >0

[full sonnet]

150k allocations per second while the game is normally running, in editor just looking at terrain and not doing anything.

currentState = defaultState;
Such a simple line of code...

The state contains 30 arrays, each with about a kilobyte of data inside.
The = was first deleting all arrays in currentState, then allocating new memory for them, and copying the data from defaultState.

But.. currentState already had 30 arrays allocated, all at the correct size.
So instead of deallocate, allocate, copy. We can do just the copy step and get rid of the allocations.

Bam, 150k allocations per second gone. That might actually have a relevant performance impact..

Over 20 seconds, out of 3 million allocations before, there are now 300k left
100k are running scripts (Well not news that scripts are terrible allocation wise, don't think I can fix it much)

[trim crane]

After my recent discovery (or more like being able to reproduce and confirm) that even important messages like damage can get split up to be sent over several frames, I thought that other messages like group creation and joining the group can also end up split over two frames and if say the list that holds the messages doesn't guarantee the order you may end up with messages arriving like this on a client:
Frame 1. Join group
Frame 2. Create group

When usually both arrive at the same frame and game first does creation and the joining

This may explain why sometimes you end up with units\players without a group in multiplayer game

This been an issue since OA, you often see players with no group until they respawn or join some other group

I always thought it might be related to lobby groups and that system being ancient and crappy but not, my mission doesn't use lobby group or lobby units anymore its all created by the client and I still end up with players without a group.

Writing this in hopes somebody can have a peek into the engine to see if its possible

[full sonnet]

it guarantees order

[trim crane]

This is also aligns with why this bug happens only on populated server with lots of action, huge message queue and group creation and join arrive separately

private _group = createGroup _side;
private _unit = _group createUnit [_class, ASLtoAGL _position, [], 0, "CAN_COLLIDE"];
selectPlayer _unit;
```This is literally what I do in my mission and I still see players without group each round a several times

[viscid onyx]

Might that not be related to the large red warning here? https://community.bistudio.com/wiki/selectPlayer

[trim crane]

Wish it was more specific, but I had this issue even when the server did group and unit creation by lobby slot, I guess it also does it on the same frame.

[knotty notch]

We do too! Only happens with JIP players who join into a until then empty group.

[trim crane]

Yep, I also see this with JIP players, they end up with no group the first time and it fixes itself after respawn.

[knotty notch]

ah, sadly doesn't get fixed with respawn for us. Only "solution" for us is to have one player in every group on mission start.

Even got a special debug exe from Dedmen but sadly no luck in tracking down what causes it.
https://cdn.discordapp.com/attachments/904790511099994173/1175917043820806245/image.png?ex=656cf916&is=655a8416&hm=da6114a89663ded667905f35ddeb94fab4ec35a22785211c9e516dc398b9f4cf&

[trim crane]

What did it do?

Curious how you were trying to catch it

If I had a special build I'd have it log all group related messages so when I catch some player bugged I can check in the log the history of their group messages

[full sonnet]

My suspicious was that somehow the group gets deleted.

What happens like in your script.
The group creation message arrives over network first, now everyone has empty group.
The unit creation might be delayed and only arrive next frame. Then the unit is created and added into the group. At THIS point, the group that we created previously was null.

Maybe some overzealous "empty group" deletion script was going ham. Because it would see a empty group for one frame

[trim crane]

Since this happens pretty much each round in koth, I could write a script to see if client that sees a player without a group also has the group itself

[knotty notch]

it doesn't crash players for you? for us the client crashes as soon as they interact with a vehicle or something similar.

[trim crane]

Nope, everything goes smooth, respawned unit gets into their group fine

Gonna double check all this, but this might mean the group is not actually deleted and just empty until respawn

[knotty notch]

Hm, seems like we've got different issues then.

[calm wasp]

in our mission, there is also no lobby and, normally, each player that joins, should automatical be a group, but, somehow, certain players, with low FPS or slow loading onto the server, somehow find themselves in a group of another player, that's already been playing on the server for a while, without this player having invited or accepted any invitation from the player that just connected to the server

so, in such cases, the one that just joined, has to go to groups menu and leave the group he joined somehow, without wanting it or the player that was already on the server, has to kick the one that just joined from his group

[trim crane]

How do you have it setup btw? Groups with several player slots in each? Do you have a respawn active?

[knotty notch]

Yup, about... 15 ish groups, most of them with 10 slots, some with only 2.

[trim crane]

I've also seen this and experienced myself, back when we had 1 group - 1 player lobby setup I rarely randomly ended up in another player's group.

But I had an explanation in mind that this player joined the group of the slot that I was playing before so we ended up in a same group

Basically like this:

1. Player 1 - Group 1
2. Player 2 - Group 2
3. Player 2 joins Group 1
4. Player 1 leaves, now Player 2 owns Group 1
5. Group 1 lobby slot is now empty
6. Player 3 joins slot of Group 1, so you end up with Player 2 and Player 3 in Group 1

This theory can probably be easily tested, but I just decided to ditch engine lobby for good a long time ago

[trim crane]

Joined a server, allGroups => [] with tons of players playing

Insane Client: Object ... not found. spam

Probably will be resolved with rejoining but still weird

90 players in the server:

[{isNull group _x} count allPlayers, count allPlayers, count entities "Man"]
```=> `[36,36,42]`

group player is also null

createGroup blufor => <NULL-group>

0 objects received from the server, players running through the air

Nevermind, only some objects were recieved

Looks like allPlayers only has respawned players since I joined

Never seen things this broken

Wonder if I can detect it reliably to just kick player back to lobby

Some objects arrived, some didn't

After 10 minutes on the server allGroups is still [] and I can't create any groups

Rejoined, loaded better BUT one huge public variable that is sent from server only once during init is nil. Other smaller pvars are fine.

What's interesting is that I have a check for it not being nil during loading

So it got recieved from the server, my init code checked that its not nil, then it became nil

allPlayers select {isNull group _x} apply {[_x, getPlayerUID _x, name _x]}
```=> `[[28c3cd58d00# 1789247: o_soldier_01.p3d REMOTE,"","Zer0"],[28c36aae300# 1782516: basicbody.p3d REMOTE,"","vasya"]]`

{count units _x == 0} count allGroups
```=> `0`

So perhaps the assumption that groups didn't arrive at all or were deleted could be true

also no uids

Rejoined again, pvars arrived properly, no broken players so far

Oh, setTerrainHeight data didn't arrive (lots of changes)

allPlayers select {isNull group _x} apply {[_x, getPlayerUID _x, name _x, getPos _x]}
```=> `[[28a31a8eb80# 1795886: o_soldier_01.p3d REMOTE,"","Home",[19412.6,15970.2,0.00143814]],[28d1dc85480# 1795382: i_fullghillie_f.p3d REMOTE,"","Andriy",[20520.9,10896.6,0.0014267]]]`
```sqf
{count units _x == 0} count allGroups
```=>`0`
Same picture again, no empty groups, 2 players without UIDs and no groups.

Was fine for a bit

Nice players list, same players repeated many times

lol

Rejoined once again, didn't help. Number of times stats are repeated doesn't seem to match number of times same client is repeated in the list

[trim crane]

Somehow this list has players that left long time ago, several times

[unreal arrow]

i know might revert my fix completely as the problem is deeper than it seems, might choose lesser of 2 evils

[trim crane]

Lobby, same picture with lots of repeated clients

[full sonnet]

Congratulations your server is F'ed

[trim crane]

[unreal arrow]

time to restart

[full sonnet]

I tried the terrain switching and model/texture loading thing. And couldn't see anything there.
But I think I found a bug in textures indeed. the FLUSH cheat, does a deep memory clean and should free system memory and VRAM and unload ALL textures.
But... it isn't freeing any memory for me. or atleast nothing that shows up in the memory profiler.. That's probably not good

[calm wasp]

#veterans message

[full sonnet]

Its supposed to do stuff.
Maybe that's the issue. More on that next week

[calm wasp]

only when you manually use it?

[analog fractal]

Oh, so I wasn't going crazy back in 2019, good to know

That's been a thing for a while

[full sonnet]

I just implemented a Superflush cheatcode, that not only flushes the graphics memory, but all memory that can be released (Everything in caches and stuff that can just be re-loaded from disk when needed)
And it basically did nothing

[pallid cliff]

I had tried to use the FLUSH cheat to remove cached textures but that didn't do anything. Not sure though if it was supposed to do that.

[calm wasp]

but also why flush, if I have half of RAM and like 8 GB of vRAM free, yet have the problem with textures, FPS etc, for some reason

shouldn't it be just fine, when you still have a lot of free RAM and vRAM?

[viscid onyx]

I don't think flush is meant to be a long term fix, it's more the fact that flush doesn't do anything could be a symptom of a deeper problem with memory management

[trim crane]

Round end seem to have helped, all players normalized

Just guessed that my recent issues with #shutdown not really shutting server down might have to do with these ghost clients somehow

[scarlet root]

you can try this to trigger the 3 fps bug (rename to the respective terrain) - has fast automated camera flyover basically

[ivory ridge]

I loaded terrains with different terrain assets. First Altis, then Tanoa (maybe a different pbo?) the a terrain with rhs terra core, then one with cup terrains core... I've seen that the memory need is rising from map to map. That was my observation (at least at 32bit linux).

I cannot say that it is connected to the assets itself (i have no profiler) but something is eating memory if the map is changed.

[thick zephyr]

I've noted the same (changing maps will fill up memory), but I've never seen a crash from it, so I just assume that the game leaves some map data in memory intentionally, which just gets cleaned up later if memory is needed

Edit: after a quick test switching multiple times between Stratis and Altis, I can see from task manager that memory does accumulate, but I am pretty sure that it's just assets, and nothing to do with the map in particular. So intended behaviour

[raw flax]

latest dev branch crashes when hitting 0 on the keyboard to bring up the radio menu

[void magnet]

👀

[light lintel]

Huh yeah I think I vaguely remember flush "not working" and didn't use it much

[calm wasp]

last time it worked, it was in Arma 2

[pallid cliff]

I ran this test multiple times on Tanoa, Altis and Livonia without triggering it. FPS get low as VRAM gets filled up to about 98% but that's about it.

[calm wasp]

not enough time then

after certain time, FPS is really really low and objects start to flicker, then, later it will force-close/crash

[keen sundial]

What usually helps me is to load different lods of textures in areas.

A map particularly prone to this is armavir on the workshop which sometimes triggers it after 5-10 minutes

[rigid warren]

weferlingen - no issues with kju testscan mission (4-5 repeats, each ~3:40)
armavir - everytime at the end i get 4fps/freeze, this map is really cursed 😄

[ivory ridge]

armavir - it runs over the island with 120 to 40 fps. After that over water it litterally freezed by under 1 fps

[scarlet root]

are these known for the problem? or rather CDLC and community terrains?

[pallid cliff]

I never had the bug so no clue where it happens

[light lintel]

I've had it on Tanoa fairly reliably, especially after iterating through other maps. Maybe Livonia but not sure (Livonia has other optimisation problems anyway).
I've had slow texture loading on altis but can't remember an actual 3fps crash.

Altis and status usually run the best and with fewest problems, malden 2035 as well

[pallid rover]

Surprisingly nobody mentioned Chernarus 2020 yet. I always had this bug there

[calm wasp]

now I can confidently say, that lower and lower client FPS, with flickering/disappearing objects/textures and/or T-pose and lastly game crash is due to vRAM only

because with a lot of RAM filled - nothing
with more and more vRAM filled - it happens

[calm wasp]

also, when I can start to observe some/any of this, I still have like 8-10 GB free of 24 GB of my 3090 Ti

[calm wasp]

but from the other point of view, even when 8 GB of my 1070 Ti were full, everything was still fine, until a certain time later

[halcyon moth]

But then I'm running the thing on 2GB and it's fine :P

Maybe something correlated with the amount used though.

[keen sundial]

It’s basically only happening with the big forests and it looks like it’s jamming between the different LODs and „can’t decide“ which one to use

[light lintel]

Yup, initially I would get page file related crashes but after a point it was always "DX11 out of memory" issues

6GB VRAM

Increasing page file helped alleviate the page file related crashes with 16GB RAM.

Oh, the out of memory stuff would happen more quickly if I had other things like Chrome open in the background with lots of tabs, apparently that would eat up VRAM too.

[unreal arrow]

Revision: 151167

[calm wasp]

in your case, it's because you make the error to limit the paging file size to a certain size, while it should be managed automatically by Windows, depending on needs

there is no need to manually limit paging file size, because it anyway can't grow bigger than free space on the disk

and even if so, it won't stay and will decrease, if you close Arma and/or the browser

so please, don't add to the mix the errors on your side

[calm wasp]

still had this FPS and flickering problem once, when I simply stayed AFK at the respawn for many many hours
so not moving across the map

[light lintel]

Hmm interesting 🤔

[calm wasp]

since you have only 6 GB vRAM, one could understand such situation

but when one has 24 GB vRAM and like 10-8 GB still free, there is no explanation

[light lintel]

Yeah that's unexpected

[calm wasp]

this is also the reason I've bought a GPU with 24 GB vRAM, thinking back then that 8 GB vRAM of my 1070 Ti were the problem
but now I know it wasn't

[light lintel]

In my case my GPU was pegged at ~5.7GB or more with very high/ultra textures with GM (which would eventually crash) but only around 3-4GB with high textures and GM assets, and would usually not crash within the time I would play. With vanilla assets and ultra it would be fine too.

Last time i tried forcing a crash, I did see some differences between the memory allocators, some would crash, some would end up at 1-5 fps and not crash

[keen sundial]

System memory allocator helps a lots to improve stability

[shrewd flame]

Per case only. There's a good bunch of people using tbb or johnking's allocators because the system one does not bring any performance benefits

[halcyon moth]

Sanity check: There are people who can generate the 3fps bug who aren't using custom allocators, right?

[calm wasp]

yes

[keen sundial]

System doesn’t give best performance but best stability

That includes the default allocator

[shrewd flame]

As John wrote, default allocator can produce the 3 FPS bug so it's still per case. My game crashes less when I run it on tbb.

[calm wasp]

doesn't depend on malloc

[full sonnet]

Can you guess how much vram Arma might be using there?
Maybe 3-4gb?

[calm wasp]

you're talking about actual vRAM usage, not reserved?
because I've never seen actual vRAM usage, but total only, because not so long ago, such an option wasn't even there in MSI Afterburner
so I can't tell you anything about acrual vRAM usage size at all

[keen sundial]

Arma will use any vram it can get. I saw it easily fill up 20GB with the right maps

[calm wasp]

reserved or used for real is not same

[light lintel]

Does perfmon expose these metrics?

[waxen vigil]

marek spanel posted something about fog stuff on dev branch on twitter, what is it about?

[trim crane]

Distant haze removal option in video settings

[waxen vigil]

kinda wonder how it will affect fps and stuff

[trim crane]

Probably wont

[waxen vigil]

well at least more option to make arma screenshot ✨

[trim crane]

Arma 3 Screenshot Simulator

[waxen vigil]

"Oh wow this screenshot simulator has a milsim sanbox stuff attached to it"

[tall pumice]

i thought the distant fog was to allow a better transition of the horizon to the sky

[winter crescent]

The default is still the regular terrain-defined haze, but many people just find the reduced haze more appealing. We're still debating what to make the default in the next update. And there will be some kind of override for specific scenarios / competitive disadvantages.

[tall pumice]

what happens if the terrain distance is set to low? it'll just stop without transition to the horizon?

i mean, with high terrain distance it doesn't really matter i guess, but some scenarios are limiting it to 2000 for performance reasons, etc etc

[light lintel]

Server defaults to 1600 right?

[weary niche]

Any chance that with this new haze setting we could also get someone to look into whether this might be possible? Seems like atleast a few people supported the idea.

#dev_rc_branch message

[winter crescent]

Do you know if it exists as FT ticket, or can you make one please for investigation?

[weary niche]

I don't believe it does currently, but I'll go ahead and set one up.

[unreal arrow]

it’s on DEV could you please test it and report back?

[tall pumice]

i checked yesterday, couldnt find the option to disable the fog

oh wait, is that the haze option?

ok, there is still a fog transition. it's just not blown up so much

1800 view distance, low vs standard haze

looks good to me. almost makes me question why the haze was added in the first place

ok, i see why. looks kinda terrible with low sun 😄

even with 6800 view distance it looks kinda .. cheap?

meanwhile with haze on standard:

so yeah.. i'll keep the haze on. :>

[unreal arrow]

it’s an option

[unreal arrow]

will have a server override as well

[unreal arrow]

is this very low or low?

[tall pumice]

very low

[weary niche]

I have gone ahead and made the ticket. When I get a chance over the next couple days, I'll attempt to include more information about the specifics I learned in my tests from September.

If anyone else has any information on how this works or is curious to follow what becomes of this, please feel free to add your thoughts in the ticket as well.

https://feedback.bistudio.com/T177593

[full sonnet]

Huh I just had 3fps bug. And I fixed it. Flush didn't work. But I set texture quality to very low, my VRAM usage went down to a quarter and now everything is fine again

[harsh orbit]

wow that's awesome! btw how did you repro?! 😅

[calm wasp]

lowering view distance also helps

but!

it's only a matter of time, before lowering textures quality and/or view distance won't help either

I start playing with 12 km view distance and extreme settings and as the time passes, I have to lower the settings and the distance more and more

and at one point, even that doesn't help anymore

[harsh orbit]

wait, by "fixed it" do you mean in the code, or via that workaround?!

[calm wasp]

iirc

[light lintel]

Yup, there we go 😀

[viscid onyx]

Have any vanilla config classes (vehicles, weapons, mags) been removed [semi]recently? We have some older vanilla-only missions that previously worked but are now causing config errors and failing to load; seems to be mission-specific but having some trouble determining exactly what is doing it

[harsh orbit]

Did you check the rpt?

As for changes, afaik there wasn't any

[viscid onyx]

Yes, unfortunately the RPT does not specify the classname. However, I've been doing further troubleshooting and I'm pretty sure (to confirm soon) it's a change to a command

[viscid onyx]

Issue identified: previously, when given an unacceptable classname, commands such as addMagazineCargoGlobal and addBackpackCargoGlobal would fail silently. Now, they fail...loudly, generating a config error and preventing the mission from loading on DS.
To clarify, not a devbranch issue, just asked here originally for recent changes and describing the final result to not leave it hanging.
https://feedback.bistudio.com/T177653

[unreal arrow]

what do you mean not a dev branch issue, does it or does it not fail on dev as described?

[unreal arrow]

22:08:08 Warning: "dsfdsf" is not a valid magazine name! I need all the commands that will benefit from class check in that ticket please (only those that cause config error)

[viscid onyx]

Happens on stable, don't have dev.
* I asked here for dev branch-level information from a dev, and only mentioned the issue for context. I didn't plan to discuss it further here until Leopard asked.

[viscid onyx]

I will check and report inventory-type commands. I don't know every command in the game that uses classnames and can't check them all.

[unreal arrow]

thanks, I can add the check to stop config errors but if I gonna do it i’d rather add it to all problematic commands at the same time

[full sonnet]

Playing multiplayer for hours

So my guess was that Arma doesn't properly know how much VRAM the GPU has.
But it says my 8gb 1070 has 7819mb of vram. So thats correct including the reserves.
But Arma would certainly just use up that memory, it just keeps old stuff cached, no matter how long its not needed, until it needs to free it because it needs more free space.

And if that freeing space is doing the same as flush (which is not working at all) then that will inevitably end in disaster. As long as ALL the textures you see during your gameplay, are large enough to fill your vram.

[full sonnet]

Oh wow. I just exited the game, and noticed that I had a breakpoint set in texture unloading. It was never called during me playing.
And even running flush didn't call it 🤔

Altis memory usage gets up to 2gb-ish, after a while of flying around.
Weferlingen almost instantly bumps to 4gb. So that explains why in GM that bug is more common, it just gets closer to high usage, earlier.

The fun I find.
"Hey move this data please"
"Okey I will copy it and delete the original 🫂"

[full sonnet]

omfg..
Why is flush not working. I clearly see the Engine::FlushMemory method here that frees all texture memory.
Oh... DX11 engine overrides it with this..

Tadaa, flush works

[full sonnet]

And the new Superflush is even crazier

13:20:10 Physical memory free 947 MiB (992866304 B)
13:20:10 Page file free 8.6 GiB (9210191872 B)
13:20:10 Process working set 8.8 GiB (9402130432 B)
13:20:10 Process page file used 14 GiB (14932733952 B)
13:20:10 Longest free VM region: 1398407168 B
13:20:10 VM busy 507846656 B (reserved 2448875520 B, committed 2353938432 B, mapped 149323776 B), free 3786989568 B
13:20:10 Small mapped regions: 36, size 151552 B
13:20:10 VID: dedicated: 8480882688, shared 17151193088, system: 0, max: 8166309888, used: 3915452416
13:20:14 Deinitialized shape [Class: "C_Soldier_VR_F"; Shape: "a3\characters_f_bootcamp\common\vr_soldier_f.p3d";]
13:20:17 Superflush freed 5330778KB
13:20:17 Physical memory free 3.1 GiB (3306594304 B)
13:20:17 Page file free 11 GiB (11872477184 B)
13:20:17 Process working set 6.1 GiB (6513299456 B)
13:20:17 Process page file used 11 GiB (12043939840 B)
13:20:17 Longest free VM region: 1398407168 B
13:20:17 VM busy 288272384 B (reserved 818065408 B, committed 3765174272 B, mapped 149323776 B), free 4006563840 B
13:20:17 Small mapped regions: 36, size 151552 B
13:20:17 VID: dedicated: 8480882688, shared 17151193088, system: 0, max: 8166309888, used: 177881088

I think that should be a workable workaround/fix for 3fps bug

[calm wasp]

so shift + minus and you write superflush?

[full sonnet]

ye. Its not in yet, next week

[calm wasp]

what about server FPS getting worse and worse over time?
faster if more players doing more stuff
even after (most) players leave, FPS doesn't recover and that despite bodies/wrecks and inventory stuff being cleaned regularly

[full sonnet]

Likely completely different thing

[calm wasp]

so despite having like 10-8 out of 24 GB vRAM free, I will still have 3 FPS problem and will still have to flush, like now?
just that flush will work?

[full sonnet]

ye, so you don't need to restart game anymore to fix it

[calm wasp]

but the optimization will only help those that have 16 GB RAM or less and no benefits for those that have more than 16 GB RAM?

[full sonnet]

optimization?
Flush benefits everyone

[calm wasp]

less allocations

[full sonnet]

RAM amount doesn't matter, its fewer allocations, not less memory usage

[calm wasp]

fewer allocations means less stuff happening for nothing?

[full sonnet]

ye

[viscid onyx]

It sounds like this also means the game's auto-flush will now work as well, so you'll probably encounter the problem much less / not at all in the first place

[calm wasp]

so higher FPS or something else?

[full sonnet]

Potentially.
Faster loading times, maybe higher fps

[calm wasp]

only if you manually (super)flush, from time to time

but flushing also means everything has to be loaded again, which means uneven and high frametime, so choppy gameplay, until it's all there again
but then you have to flush again...

[full sonnet]

The game does its own flush in some places.
For example after a camPreload

but then you have to flush again...
Yeah.. You might have to flush every half hour or hour. As opposed to having to restart your game every half hour or hour..

[calm wasp]

good option/workaround for us here, but 99% of average players (and not only) won't be able to benefit from it, since they have no idea about shift + minus and different commands

we will for sure use it, of course

largepages in launcher also doesn't work

I would say

[viscid onyx]

Is this the same kind of flush as the one when you alt-tab, or is this a different thing?

[calm wasp]

limiting RAM and vRAM in launcher to lower values than physically available, also doesn't work, because it continues to grow
like if you have 32 GB of RAM and 24 GB of vRAM, and limiting both to 8 GB in the launcher - it won't do that

[full sonnet]

Different

[viscid onyx]

If the underlying problem Groove_C is describing is not fixable for whatever reason, could it make sense to make it do this kind of flush at the same time as that one? It wouldn't be a true solution, but I think it would silently suppress the issue for a lot of people, and "alt-tab to fix it" is more likely to spread than trying to teach people about debug cheats.
*if the underlying problem can be fixed, obviously fixing it would be better

[full sonnet]

No you wouldn't want to do a full flush on tabbing out

Maybe as a emergency solution if it can't be fixed, but we don't know that yet

[calm wasp]

tab back in and wait 🤣

[steel rune]

what if I don't have numpad? will it trigger by itself when running low on RAM?

[full sonnet]

no

[calm wasp]

Shift + Fn + minus and blindly type flush or superflush

[full sonnet]

The water freeze at the end, is actually the game trying to "generate" roads..
There is something really wonky out there. Its generating some ultra complex roads, and trying to merge them into one mesh..
Outside of the actual terrain, in the water.. why would there be roads out there

It looks like its a road with about 300k vertices, or maybe its 300k separate roads. Haven't seen that before.
But that's not the 3fps bug.

Yes its ONE road

_nVertex = 341118
One road with 341118 points.. Something went wrong when encoding that terrain probably.

[tall pumice]

it's a very good looking road, ok

[full sonnet]

I turned off roads with more than 8192 points. And re-ran Armavir through kju's camera test.
No issues. all fine. Even after a second run
Switched to altis and ran it over that, no issues.

[full sonnet]

I told Arma I only have 2gb of vram (not via -maxVram parameter, that one doesn't work)
And, with its 300mb reserve that it keeps free, it holds me at about 1.2gb of usage.
So the memory freeing if it needs space, seems to be working right

[calm wasp]

so -maxMem and -maxVram are there just for them to be there 😏

[full sonnet]

No, just maxVram. Just as the wiki says

[keen sundial]

Anything I can tell the mapmaker of armavir to change? He was part of my milsim.

[light lintel]

Ooh great to see developments on the 3fps/vram/flush issues!

So if i understood correctly:

• game should be better at flushing by itself after the fix is added
• advisable to superflush manually every half hour/between missions
  ?

[lament crow]

Would be great if the flush can be made automatic once the game detects it needs it to prevent issues rather than having players manually doing it. 99% of players won't even know the fix is there if it remains manual only

[full sonnet]

Something probably broke in his roads configuration, make some broken road shape in his roads shape file (dunno how roads are configured)

[full sonnet]

normal flush when 3fps bug is noticed. Should be enough.
Game does minor flushes some rare times, I don't think it influences 3fps bug, it will still happen and you'll have to flush manually if it does

[full sonnet]

"once the game detects it needs it"
Yup would be great to know when it needs it.
But for me to find that out, I will have to repro the 3fps bug which I still cannot.
And if I can repro it, I might be able to fix it properly anyway

[harsh orbit]

BTW did you try the object flickering thing I told you about?

Dunno if it was related to the 3 fps bug tho. probably a hashmap bug?

[light lintel]

Thanks!

[full sonnet]

Ah! no that one not yet

[trim crane]

Would it be possible to have EPEContactStart event fire BEFORE the damage?

Would make a process of connecting collisions to damage so much easier

Right now EPEContactStart is fired exactly after associated damage it does with HandleDamage fires

If its feasible, I'll make a ticket

Also would be great if that _directHit also changed to false to be more useful or something

Or finally introduce some damage kind flag with a string value

PHYSX, FALL, PARTICLE (fire particle damage), etc

[full sonnet]

Yes, its just swapping two code blocks around

[trim crane]

Made a ticket: https://feedback.bistudio.com/T177724 Can make a repro with logged event handlers if somebody will work on it.

If we had EPEContactStart fire before the damage, you could queue two collided vehicles and then check if HandleDamage source is null and you just had EPEContactStart before.

[carmine juniper]

Pretty much I've just wondered and nothing else, and I couldn't thought up a better channel for it so anyways: terms of RAM and/or VRAM, having, let's say 10 PAA textures vs 10 ui2tex textures that uses 1 (same) PAA texture - do they have big, obvious and considerable differences?

#arma3_texture message
Prety much I've wondered after this post I made

I think I want to test it myself but don't really know what I can do

[halcyon moth]

The PAA textures would get mipmapping and (hopefully) compression.

You could maybe test it by going to extremes.

[carmine juniper]

Okay I think I've figured out. 1000 2k textures vs 1000 ui2tex (SplashArma3). Kinda I've expected this result but it's interesting

private _textures = [] ;

{
    private _r = addonFiles [_x#0,".paa"] select {
        private _res = getTextureInfo _x ;
        _res#0 == 2048 and _res#1 == 2048
    } ;
    _textures append _r ;
    if (count _textures >= 1000) exitWith {} ;
} forEach allAddonsInfo ;

_textures resize 1000 ;
{
    private _t = "UserTexture1m_F" createVehicle [0,0,2] ;
    _t setPosWorld [round (_forEachIndex/100),(_forEachIndex mod 100)*0.5,6] ;
    _t setObjectTexture [0,_x] ;
} forEach _textures ;```
```sqf
for "_i" from 1 to 1000 do {
    private _t = "UserTexture1m_F" createVehicle [0,0,2] ;
    _t setPosWorld [floor (_i/100),(_i mod 100)*0.5,6] ;
    _t setObjectTexture [0,format ["#(rgb,2048,2048,8)ui(SplashArma3,%1)",_i]] ;
} ;```If you want my stupid code

[halcyon moth]

Which result is which?

[carmine juniper]

PAA is top, ui2Tex is bottom

[halcyon moth]

I guess the PAAs do compress pretty well then :P

was expecting more like 3:1

[carmine juniper]

Well can consider that PAA is lossy compressed texture

[halcyon moth]

Although I guess it might not even have loaded the full-res textures. Not sure how Arma works there.

[full sonnet]

Ah yeah PAA's are DXT compressed usually. Ui2Texture... not sure actually. They are rendered directly to gpu memory so I don't know if they use compression.
If not I can probably enable DXT on them 🤔

True, displaying a 2K paa texture, doesn't mean its displaying 2K. It might only load a lower mip

[unreal arrow]

So when animation messages were separated from position update they lost some priority but most importantly a force timely update push after no updates timeout. I am adding that back so no matter what the lean should update after max 5 s if stuck.

[trim crane]

Thanks, gonna help multiplayer a lot!

[trim crane]

Since you were looking at character movement, maybe you can have a peek at unit speed issue @ #arma3_animation ? Its another multiplayer gameplay ruining issue.

[unreal arrow]

is there a ticket to save me browsing through channel?

[trim crane]

No ticket yet, will make one tomorrow

[trim crane]

Execute the code in code snippet, observe speed in hint, watch the videos to see how bad interpolation gets because of huge speeds

Otherwise I'll file a ticket with it all tomorrow

[unreal arrow]

what code

[trim crane]

#arma3_animation message

[unreal arrow]

yeah I will need a detailed ticket. from first glance it connected to prediction

[trim crane]

https://feedback.bistudio.com/T177832
Made a ticket

[trim crane]

We have a server running 2.14.151210, there has been a recent streak of issues where players got killed seemingly by vehicle collision with no actual collision happening

https://youtu.be/nSm_0QYXN4c?t=162

Pictured here, player gets out, runs away for few meters and dies, in other cases death gets attributed to helicopter pilot, so it is a roadkill

[carmine juniper]

Do you mean you're suspecting that, the “turning player” issue is causing this?

[trim crane]

No, unrelated

This feels like collision damage somehow started to apply on remote units

As in some client saw remote helicopter collide with remote unit and broadcasted the damage to them?

I did some tests with vehicle collision recently, unless unit owner decides that collision happened, no damage will be done, no matter what remote clients see and do to them

Thus my assumption that remote clients started doing vehicle collision damage to units

Since people are running stable builds, maybe its server doing that?

Server weirdly extrapolated helicopter movement, saw helicopter collide with unit, then broadcasted the damage?

Doing guesswork here, maybe somebody can see if any changes to this logic were done recently?

[formal cipher]

Do you mean 2.15.151210?

[trim crane]

Perf build, aren't they 2.14?

[formal cipher]

That would go to #perf_prof_branch

[vernal pollen]

DPI glitching finally getting fixed in 2023 no way

[trim crane]

It creates a lot of mess even with normal DPI and as for fixed, I still don't know how it can be properly fixed at all 🤔

[unreal arrow]

I have noticed you can kill player by just piloting a small drone into them. Do you know if this is new?

[trim crane]

I think it was always like this, but in this case no collision happens yet player still dies

I tested this on latest perf and couldn't reproduce it

Unless local unit sees the collision, no damage happens

[unreal arrow]

there was an issue with ghost unit because it was killed remotely but not locally. I fixed that but that was strictly rotor collision related

[tall pumice]

you could always kill soldiers with piloting a drone into them

[unreal arrow]

should this be looked at like maybe deal damage depending on speed rather than instant kill?

[unreal arrow]

what killed this one, collision with the soldier?

[trim crane]

I think it was the helicopter, in other cases they get killed by the pilot so its a roadkill

https://youtu.be/gH0xdTAPimE?t=135

Here is another example of this recently

You can see both are attributed as teamkills with a helicopter in top right

In this case helicopter pilot doesn't see the collision either, I don't think soldiers had it either

Thus my assumption that maybe some other client saw remote heli collide with remote unit and passed broadcasted the damage?

I also tried to see if I could get killed like this on a live server but I couldn't

[formal cipher]

Are the server and the clients that it happened to both running perf branch?

[unreal arrow]

the collision was most probably pilot side, would be interesting to see pilot video

[trim crane]

Stable for clients, perf for server

[trim crane]

https://youtu.be/9_AYHe-LN6Q?t=165

Here is an example from pilot POV

What's interesting is that this is the safe zone, all damage gets ignored through HandleDamage

Yet player somehow dies

https://youtu.be/ZXYC0ckTOtc?t=160
Both pilot and passenger die here

[formal cipher]

@full sonnet Possibly PhysX suspension changes causing some issue here?

[trim crane]

Disregard "killed by GM6 Lynx", its just last enemy player damage being used for scoring

Did it become a thing recently?

[unreal arrow]

could be back rotor kill (sorry replied to the wrong vid before)

[trim crane]

Did this become a thing recently?

[full sonnet]

@harsh orbit 🫂

[unreal arrow]

yeah both rotors kill now but it is suicide unless chopper is moving then the kill is attributed to the pilot

[trim crane]

Considering how weird these kills are and that HandleDamage is ignored, I assume server does the damage without checking locality?

[unreal arrow]

actually it now checks locality and applies damage where it is due whereas before it was applying it remotely hence ghosts

[trim crane]

So it was fixed in one of the recent perfs?

[unreal arrow]

it would have been there on 8th

that guy otherwise would have been ghosting

[trim crane]

Server on the video is exactly 2.14.151210

Nevermind, not exactly

[unreal arrow]

assuming it was prof, the video is 8/12/23

[trim crane]

2.14.151108: 2023/11/04 - 2023/12/09
2.14.151210: 2023/12/09 - Now

[tall pumice]

dunno about that. thing is, getting hit by a drone with it's rotors.. is gonna hurt real bad. those rotors will slice you up if there is no guard around it.

[trim crane]

Rotors are guarded though

[tall pumice]

not necessarily

[unreal arrow]

yeah hurt hence +damage not kill, unless was already hurt

if wounded is finished with drone, thats understandable

[trim crane]

Can we somehow expand damage messages to include more info about damage source so we for example can filter out rotor damage through event handler?

I guess right now you can't tell collision damage vs rotor damage at all

By source I mean kind of damage

I'd suggest to include this info in ammo field, but its gonna affect backwards compatibility so I guess it could be another EH argument

[unreal arrow]

if killed by rotor by moving chopper you should have instigator and vehicle, suicide unit itself i think just like before

[trim crane]

But how can you tell if it was rotor or helicopter body?

Say I want things to be like before with no rotor damage

filter out rotor damage, leave body collision damage

[unreal arrow]

dunno there is no system and looks like body collisions are screwed anyway, for once none of them have instigator

[trim crane]

Player-driven collisions provide the driver at least

[unreal arrow]

try it on dev

driver as killer not instigator right?

[trim crane]

Just tried walking into working rotor, bumped several times into it and finally died after few attempts, feels unreliable

[trim crane]

Been like this since A3 Alpha, if player ran somebody over they would be the killer

Didn't try on dev yet

[unreal arrow]

that maybe didnt touch that there is some half-bothered calculation is happening probably to save on performance

[trim crane]

Unit got bumped away few times (probably that collision shape shifting), but finally died after few attempts

[unreal arrow]

rotor kill shold give you pilot instigator and killer chopper, but only if chopper was actually piloted andmoving

[trim crane]

So its different from vehicle collision kill then?

[unreal arrow]

instigator will be there yes

[trim crane]

Body collision kill:
Killer = Pilot
Instigator = Null

Rotor collision kill:
Killer = Heli
Instigator = Pilot

?

[unreal arrow]

yeah try it

has to be moving or this is suicide by walking into rotor

[trim crane]

Nope, 0 HandleDamage calls, player just dies

23:18:03 ---------------------------------------------------------
23:18:03 "Frame 139252: Killed Both: B_Soldier_F"
23:18:03 ["B Alpha 1-3:1 (Sa-Matra (2)) (B_Soldier_F)","B Alpha 1-3:1 (Sa-Matra (2)) (B_Soldier_F)","<NULL-object> ()",true]

Same for local on local, no HandleDamage events, just death ignoring everything

AND the damage is done to remote units unlike normal collision damage which is calculated where units are local

Test case:

1. Player 1 in heli
2. Player 2 freezes for 15 seconds
3. Player 1 touches Player 2 with back rotor
4. Player 1 flies away
5. Player 2 unfreezes and dies

Looks like Player 1 decides the damage for Player 2, if same test is done with vehicle collision, until Player 2 sees that they've been hit, they won't get any damage

Rotor kill also fires HandleDamage on remote unit

But this might be that bleeding HandleDamage bullshit fire

lol this also works for non-server clients

So I guess if some client sees that a remote heli touched a remote unit with a rotor, that unit will die

Now multiply that by say 50 players

Everyone will die near a helicopter

becuase some client somewhere will extrapolate the movement that it ends up near the rotor

So in total:

• Cool to see a new feature but it would be great if we could filter it out with HandleDamage to make things work like before
• I think unit itself should decide if they get damaged from back rotor, not the other way around, same as its done with vehicle collisions right now

[unreal arrow]

it wasnt rotor kill if instigator is null

it is hardcoded this is how you know

[trim crane]

["Arma 3","Arma3",214,151225,"Stable",false,"Windows","x64"]

Just tested rotor damage in and out

[unreal arrow]

doesnt say dev

[trim crane]

You mean it used to do damage without instigator but does now?

[unreal arrow]

no this is some other damage

there are millions of ways to die

the rotor kill adds instigator

[trim crane]

Gonna try the dev now

But in that version it just kills ignoring HandleDamage and has unit itself in Killed

[unreal arrow]

and by moving i mean flying moving, off the ground moving

are you good at piloting?

I suck but I managed to get it once or twice

[trim crane]

What's the condition in the engine? Velocity check? Touching ground?

[unreal arrow]

yeah

[trim crane]

Either or both?

[unreal arrow]

and presence of pilot

both

player pilot i think

[trim crane]

Ok, gonna do tests on dev

[trim crane]

It won't attribute the damage to AI pilot if they're moving?

[unreal arrow]

no it is to reward skilled pilots

[trim crane]

So you couldn't filter this out if you died from AI helicopter rotos then 🤔

[unreal arrow]

would be suicide no instigator

unless remote contrlolled by player

[trim crane]

Tried on the dev, it seems to be identical to perf

I scripted the helicopter to stay in same position (isTouchingGround = false) and have velocity ([5,5,5])

 0:21:22 ---------------------------------------------------------
 0:21:22 "Frame 115445: Killed Both: B_Soldier_F"
 0:21:22 ["B Alpha 1-3:1 (Sa-Matra (2)) (B_Soldier_F)","B Alpha 1-3:1 (Sa-Matra (2)) (B_Soldier_F)","<NULL-object> ()",true]

[unreal arrow]

you are doing it wrong

[trim crane]

Player was in the helicopter, not touching ground with setVelocity each frame

Another player died from main rotor, no HandleDamage, just Killed

[unreal arrow]

2 players
player in heli called heli [little bird]
another playable bob
hosted server
on server lift up and start moving
on client execute
bob addeventhandler ["killed", { showchat true; systemchat str _this}]; [] spawn {sleep 2; bob setposasl (heli modeltoworldworld [0,4,0.5])};

[trim crane]

Same

productVersion => ["Arma 3","Arma3",215,151218,"Development",false,"Windows","x64"]

[trim crane]

[unreal arrow]

sorry for thé vid broken shoulder

[trim crane]

get well

[unreal arrow]

thanks

[trim crane]

Is my version right?

[unreal arrow]

this is on internal but i can try dev just to be sure gimmie few min

[trim crane]

Engine being on also desynced somehow, doesn't sync over time either

Fuel syncs like each 5 seconds

Turning engine off and on doesn't help, still off on client

Never seen this before

[unreal arrow]

odd mine is fine

verify integrity?

[trim crane]

Client h getHitPointDamage "hithrotor" => 1
Server h getHitPointDamage "hithrotor" => 0

Thats why engine doesn't turn on on the client

Why HitHRotor damage doesn't broadcast is the question

[unreal arrow]

client flying?

[trim crane]

Server flying

[unreal arrow]

it gets cancelled when slicing soldier for big choppers maybe

[harsh orbit]

Man-Vehicle collisions are not PhysX related

[unreal arrow]

try a remote wall does it register at choper locality?

[harsh orbit]

Is this related to choppers only?

[unreal arrow]

yes rotor collisions are custom

[trim crane]

Touch a remote wall with the main rotor?

[unreal arrow]

yeah

[trim crane]

From my tests helicopter owner does the damage with the rotors

[unreal arrow]

my guess it will break rotor

[trim crane]

Unlike normal roadkill damage where unit owner decided if they got hit or not

[unreal arrow]

testing dev

["Arma 3","Arma3",214,150957,"Stable",false,"Windows","x64"]

wait not dev arrrrh

[trim crane]

Broke the rotor, repaired with setDamage 0, its all in sync again 🤔

[unreal arrow]

what gethitpointdamage say on both?

[trim crane]

0 on both now

What an oddity

[unreal arrow]

i mean at damage stage

ok 1 dev is coming up

["Arma 3","Arma3",215,151218,"Development",false,"Windows","x64"]

[trim crane]

Both get it at 1 once it breaks

Figured it out, I had allowDamage false on heli owner side, but not on other client

If you touch the wall with invincible heli, nothing happens for you but other clients have your helicopter's main rotor destroyed

It syncs back after some time now though 🤔

[unreal arrow]

works too

[trim crane]

Nope, I can't get it to work, this time killer was pilot and instigator is null but this is normal heli collision

[unreal arrow]

odd

[tall pumice]

looks like deleteGroupWhenEmpty doesn't work correctly

[unreal arrow]

lemmie make autonomous mission

[trim crane]

Finally did it

 1:09:25 ---------------------------------------------------------
 1:09:25 "Frame 511259: Killed Both: B_Soldier_F"
 1:09:25 ["B Alpha 1-3:1 (Sa-Matra (2)) (B_Soldier_F)","B Alpha 1-2:1 (Sa-Matra) REMOTE (B_Heli_Light_01_armed_F)","B Alpha 1-2:1 (Sa-Matra) REMOTE (B_Soldier_F)",true]

onEachFrame'ed victim unit in the air then carefully touched with main rotor

Still no HandleDamage

That's why players are dying in safe zone on our servers, it works like a scripted damage and bypasses HandleDamage

All of this makes me wish for some kind of HandleDamage extension, have another argument describing source of damage packet.

"SHOT" - direct hit damage
"SPLASH" - splash damage (if possible to distinguish)
"FIRE" - Particle damage
"EPE" - EPE collision
"EPE-CREW" - EPE collision propogated to crew?
"FALL" - Unit fall damage
"ROADKILL" - Unit collided with a moving entity (roadkills too)
"ROTOR" - This new rotor damage
```This would make things so much easier

But considering that the engine is 20 years old, it probably gonna take a whole lot of work

[unreal arrow]

so you dont need repro?

well i made it. start mission with 1 client in multiplayer join slot 1 start another client and join the hosted server...boom you are dead

[harsh orbit]

In today's dev branch update (as well as profiling branch v9), you might've noticed this in the changelog:

Fixed: Various PhysX issues related to the vehicle suspension system
For those who are wondering what these fixes do, see these videos for comparison:

Before:

After:

While the results in synthetic tests look promising, there could be cases of broken "backward compatibility", so we would like to ask you to keep an eye out for these issues and report them immediately if you see them (preferably with repro steps):

1. Tanks not able to climb over obstacles they could before
2. Vehicle wheels getting stuck under terrain after collision/getting run over by another vehicle
3. Vehicles getting Arma'd into space after changing locality in MP
4. Other PhysX issues

Also if you happen to use this mod (and similar mods), please remove it
It doesn't really help much with this issue anyway, and in this case interferes with the results of the fix
https://steamcommunity.com/sharedfiles/filedetails/?id=2191542091

[tall pumice]

how do i get that achievement now

[lament crow]

It might not mean much for some, but for me this sort of work restores my faith in humanity, thank you

[light lintel]

Arma space program has been terminated rip

[hot wave]

Does this change help with ragdoll and vehicle physx interaction?

[craggy geyser]

what a nice surprise

[lament crow]

Early Christmas 🎁

[trim crane]

Can this damage be put through HandleDamage so we can filter it out in scripted safe zones? (no allowDamage used there)

[unreal arrow]

handle damage at what locality?

[trim crane]

Locality of target of course, that's the only place for HandleDamage to be at

But speaking of locality, I think collision check should be moved to victim side, but right now it seems that helicopter decides who it hit

Pretty much all damage with rare exceptions is checked and done on the client side in the game

You know how strange extrapolation and prediction can be, this will end up with people bumping into rotors because of it all the time, such collision must be calculated on victim locality side because that's where the most accurate position is

[unreal arrow]

rotor collision is calculated where vehicle is local, you are asking for it to be calculated on every client