perf_prof_branch

[heavy vortex]

It's not generally beneficial for them. Think of the support costs :P

Although there was some openly obstructive behaviour back in the bittorrent boom.

[analog acorn]

no one has support outside of B2B these days, they just have a FAQ that tells you it's your problem and a chatbot that doesn't work

[heavy vortex]

fair

[silk pewter]

added! Cc @uncut linden

[stark falcon]

format ["Items: %1%", []] // note weird extra %

Error in expression <format ["Items: %1%", []]>
 Error position: <format ["Items: %1%", []]>
 Error 1 elements provided, 0 expected

Only happens on recent profiling branch

[heavy vortex]

That never worked, right? It'd just ignore the extra % before.

The error logging is new.

[quaint flame]

It's not a bug but a feature

% symbol can now be added directly without passing it as parameter

[whole cloud]

Thank you. Forgot to filter the check to only print when index > 0. In that case the % is read as %0 which is invalid index too.
Fix tomorrow

Its still an error in the format string though, so maybe we should throw some error on it.

[whole cloud]

It is a bug. To pass % symbol, you need to pass %%

[quaint flame]

Oh well...almost right I guess

[heavy vortex]

That's a new feature, right. Doesn't work in stable?

[quaint flame]

Not yet

[vale shoal]

Dedmen, is there any chance that you will release the crash dump fix on profiling? We have a strange bug which crashes the client (profiling & stable), but no crash dump is created (client will crash with Application Hang).. it might can help to identify what crashes the client. The problem with that specific crash is, that it will crash all players within a specific range. And players are not really helpful in that regard..

[light cargo]

reproduce with a debugger and dump from there...?

[vale shoal]

I cannot reproduce it when I don't know what triggers it

[light cargo]

you could always run a debugger for a while until it happens then.
try enabling windows crash dumps just in case, also does the rpt help at all? additionally the game normally creates dumps when its frozen.
you could try the adplus utility found in the debugger folder of the windows sdk

[vale shoal]

It never crashed for me, I also have no clue where to search. But letting a regular player doing this his brain would freeze. Also rpt isn't helpful. It just quits printing stuff the moment the client crashes. I collected some rpt from the players but there is nothing they have in common in the logs, sadly.

[light cargo]

well you could prepare a script for the player to run that launches the game attached to a debugger that will create dumps on any issue that arises

adplus can do such automated job

[whole cloud]

It comes today
But hang should be a freezedump, not a crashdump

[whole cloud]

2.16.152042 new PROFILING branch with PERFORMANCE binaries, v16, server and client, windows 64-bit, linux server 64-bit
- Fixed: Some cases where the game would not create a crash report on crashing
- Fixed: Vehicle door open states would not sync correctly in MP - https://feedback.bistudio.com/T182082

Known Issue: Doesn't work on Windows 7 or Windows Server 2008.

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh
Note: There are separate Dll files that also need to be placed into Game folder.
🦦

[heavy galleon]

Fuck I can't send images here

insert Bernie I am once again asking meme with "for Linux Addons support"

[whole cloud]

https://tenor.com/view/angry-cat-sour-cat-cat-meme-cat-gif-4305398913986271634

[autumn timber]

Dedmen no angry. Dedmen pliz do

[knotty wraith]

my server is not starting - The entry point to the GetCurrentThreadStackLimits procedure was not found in the DLL KERNEL32.dll .

[whole cloud]

kernel32 is quite essential indeed

If kernel32 is not present, your windows isn't running

Ah. The important info.
What windows version are you on?

Minimum supported client Windows 8 [desktop apps | UWP apps]
Minimum supported server Windows Server 2012 [desktop apps | UWP apps]

Microsoft says that.
Fair, I overlooked that, and we need windows 7 support.
But are you really running a 12+ year old OS?

[knotty wraith]

Windows Server 2008 R2

[autumn timber]

That's vintage 😮

[whole cloud]

Well no profiling branch for you then until next week.
I'll not push a hotfix to make it work on 16 year old OS that is 4 years past its end of extended support

[knotty wraith]

Everything worked before that. There are license issues in our world.

[whole cloud]

Yeah and it'll work again (Because we have to support windows 7)

[knotty wraith]

I just gave the information - no problem.

[heavy galleon]

Thanks I am stealing

[silk pewter]

added(men)

[heavy galleon]

I love you Lou

now nudge him to make the Addons work in linux please

[autumn timber]

I have finally managed to contribute something of value to the Arma community

[silk pewter]

that's why you're a veteran 😎

[silk pewter]

done

[heavy galleon]

[whole cloud]

😠

[silk pewter]

https://tenor.com/view/cuoinotung-pepe-the-frog-explode-gif-17794553

[light cargo]

workaround: revert to previous profiling by replacing files in your copy from google drive

[silk pewter]

https://feedback.bistudio.com/T183112 codePerformance issue known, right?

[spice dirge]

partially matches https://feedback.bistudio.com/T182974 which i've also mentioned in https://feedback.bistudio.com/T182822 meta-ticket, but i don't remember getting any errors in .rpt

[whole cloud]

Its in R3vo's ticket, so known

[opaque ingot]

hello, hello !

veteran linux owner here,
I face a strange issue when I try to use the latest perf build "arma3server_x64_performance_216_152042_v16_linux"

I take file from GDrive, update the server and validate the creatordlc.

I 'just' start the server, no mod, no mission (I remove the server profile too)
$ ./arma3server_x64 > ./server.log 2>&1

then I see that:

14:53:49 Updating base class Mod_Base->, by a3\data_f_destroyer\config.bin/CfgMods/Jets/ (original a3\data_f_jets\config.bin)
14:53:50 Error in expression <0x00050000 + 0>
14:53:50 Error position: <+ 0>
14:53:50 Error +: Type Number, expected
14:53:50 Error in expression <0x00050000 + 1>
14:53:50 Error position: <+ 1>
14:53:50 Error +: Type Number, expected

and few sec later

14:53:50 Error in expression <1.2 * ( ((safezoneW / safezoneH) min 1.2) / 40) + (Safezon>
14:53:50 Error position: </ safezoneH) min 1.2) / 40) + (Safezon>
14:53:50 Error /: Type Number, expected Config entry
14:53:50 Error in expression <17.2 *

there is something I miss, have you any clue ?

reverting to version 2.16.151618 solve the issue
I keep the start log if needed

[whole cloud]

This has happened before, but I fixed it in v10.
Mh I think I found a part of that that I overlooked, maybe its that.

But I tested v13 on my linux server and that one was fine 😕

Nvm that one was apparently not fine either. And noone noticed over the last month 😄
Thank you

[opaque ingot]

huhu !

Thank you for all the great work you do!

[opaque ingot]

let's me know if you want another test

[whole cloud]

Searching for the issue 😄

[opaque ingot]

I know what you mean...

[whole cloud]

Pretty sure I know what it is

[opaque ingot]

a broken hpp ?

[whole cloud]

Initialization order is messed up.
The script command is registered, before the type it uses as argument is initialized

[opaque ingot]

bha ! 🙂

put the cart before the horse!

[whole cloud]

Just had to splatter some init_priority's around. now its working.
I don't know when we'll release it, I would try tomorrow but not sure if we can. Maybe only next wednesday.

I could send you a build tomorrow if you need it for something

[heavy galleon]

Dedmen when I start bothering him about fixing linux: breaks it even more

[rain moth]

next perf build won't support loading mods at all on linux

[whole cloud]

He wants support for "Addons" folder right?
So I'll just change the "addons" string to "Addons" to make it work

[rain moth]

then we can make another ticket to change Addons to AddOns

[whole cloud]

        copyToClipboard format 
        [
            "Result:%1%%2 ms%1%1Cycles:%1%3/%4%1%1Code:%1%5", 
            endl, 
            _result, 
            _actualCycles, 
            _wantedCycles, 
            _testCode
        ];

Wha.

%1%%2 ? Why would you do that. 😄

KK already fixed that one, it should be in dev branch from yesterday I'd think

[void badger]

Were there still FSM issues on the docket? I was mucking about last night and LAMBS (dev) was throwing errors when engaged with explosive rounds

[whole cloud]

No. What kind of errors

[opaque ingot]

no hurry, I'll test when you push it, thanks a lot!

[void badger]

Oh you know what, this might be the enum thing you were talking about

[spice dirge]

can confirm with dev build 152041 on my machine. Thanks.

[void badger]

19:07:35 Error in expression <ambs_main_fnc_doAssaultSpeed; _unit setUnitPosWeak (["MIDDLE", "PRONE"] sele> 19:07:35 Error position: <setUnitPosWeak (["MIDDLE", "PRONE"] sele> 19:07:35 Error Foreign error: Unknown enum value: "PRONE" 19:07:35 File \/z/lambs/addons/danger/functions/fnc_brainForced.sqf..., line 279 19:07:35 ➥ Context: [] L1 () [] L272 (\/z/lambs/addons/danger/functions/fnc_brainForced.sqf) [] L280 (\/z/lambs/addons/danger/functions/fnc_brainForced.sqf)

[heavy vortex]

That looks like a Lambs bug? The documented value is DOWN, not PRONE.

[autumn timber]

You can't have case-sensitivity issues for mods if you drop support for modding

[whole cloud]

It is, they already fixed it internally.
We were not reporting that error before so it was missed all these years

[void badger]

Thanks guys 😁

[orchid mesa]

not sure if it has been mentioned, but the current profiling build has massive lag when you open the player list inside the map

causes big frametime spikes every 1-2 seconds and drops below 10 fps

[tribal pasture]

This is due to format strictness error messages

This is only in KotH btw

I moved my player icons from uiEx to ui and left other arguments intact for when uiEx will be patched (casing issues), but it does massive error spam now due to forced error strictness.

[whole cloud]

On dev branch it's already behind -debug.
But I'm leaving it enabled on Prof because I want to hear about all the issues, especially in vanilla scripts

[heavy galleon]

Question. Would this cause issues for port forwarded servers? Let's say a server hosting set up in someone's basement. Router has the public and is forwarded to the box.

Because we are experiencing this.

Although I've been told that "it's somewhat weird with the public IP on the ISPs side".

[whole cloud]

It shouldn't

your public IP in that case, is normally the same IP as the clients that are trying to connect to you are seeing

I don't know how Steam determines what public IP you have, they determine that somehow. I assume they are pinging their own service and seeing whatever outgoing adress it used

[heavy galleon]

Just got told that if you try to reach whatismyip from the box, you get a different one

And that's the ISP doing something weird af

[whole cloud]

That'll be it then

Same as the other case

[heavy galleon]

Now I am more intrigued to wtf is that iso doing

[whole cloud]

I would've thought that steam already blocked that.
Like how does steam let you say "Hello, I am a server at IP Address A, please add me to your master list" when you send that from IP Address B.

[heavy galleon]

I don't know

I don't even understand how you can recieve responses from different IP

Or how it's setup on ISPs side

Are you planning on removing the feature next week or?

[whole cloud]

Its already off for next build

[heavy galleon]

Appreciated

[gritty wasp]

Speaking about network stuff. Can someone explain why when I host multiplayer mission from Eden and launch another Arma instance on same machine in Servers list at LAN page I have somewhat about 6 server entries and to connect I need to make from one to infinity tries.

[whole cloud]

You probably have multiple network interfaces.
And Arma server listens on all of them

so if you try to see if there is a server, on each of your interfaces, each interface will find one. With a different IP for every interface

[gritty wasp]

Can I mitigate that by firewall rules?

[whole cloud]

probably

[gritty wasp]

Thank you. Will try.

[heavy vortex]

Usually all of them work :P

[empty goblet]

q, did you start the server with -ip= to define the single interface ip ? https://community.bistudio.com/wiki/Arma_3:_Startup_Parameters#Server_Options

[gritty wasp]

I'm taking not about server but full game client and multiplayer preview

[empty goblet]

ok, tho i do wonder if client can be started with the -ip= parameter to do the trick ...

[heavy vortex]

It's a fair question. I don't have enough IPs to be sure :P

[heavy galleon]

You only need one. 127.0.0.1 would be enough

I am willing to borrow it to you for testing

[gritty wasp]

LOL it crash when another client trying to get servers list.

upd: it crashes when another client without -ip= launched. Works when both have -ip=127.0.0.1
upd2: servers list now have one entry but still requires unknown amount of connection attempts to join server. Can it be fixed?

[vale shoal]

I talked with a couple of people using the latest profiling. The specific crash we're having (application hang) doesn't produce a crash or freeze dump file. According to someone, when he tried to create a dump file on the crash dump window it crashed the window too.

[whole cloud]

You can create a dump in task manager.
You mean it crashed task manager?

[vale shoal]

No, I mean when the game crashes it shows the reason for a crash. And you can just click to collect dump information. And then this window crashed too.

[whole cloud]

I don't know which window you mean

We have no window that creates a dump file or collects dump information

[vale shoal]

https://community.bistudio.com/wikidata/images/f/f7/UnusualProcessExit.png

"Show details and prepare report" this button crashes it

[whole cloud]

That doesn't create a dump or collect info. It just puts the files in appdata folder into a zip

Just grab the files out of appdata manually. But I assume there isn't a crashdump in there

[vale shoal]

You are assuming right

[whole cloud]

You can manually create one in taskmanager details panel, while the game is frozen

Before it exits

[vale shoal]

it creates a empty .dmp file

[whole cloud]

funni

BattlEye enabled? Can't do it with BE

[light cargo]

if you're eager about getting your hands dirty you could use adplus utility found in windows debugging tools as part of the windows sdk to record a dump when the game freezes or crashes

that should get the dump in a bit of a more direct manner

[heavy galleon]

not if BE blocks you from reading any memory

[light cargo]

yeah im aware

[light cargo]

um is this normal?

22:11:58 Skipping mods and signatures due to overflow flag being set.
22:11:58 Skipping signatures due to overflow flag being set.```

any clue on this during init?

_namespace setVariable [_objectType, _actions];
>
22:18:38   Error position: <setVariable [_objectType, _actions];
>
22:18:38   Error Reserved variable in expression
22:18:38 File z\ace\addons\interact_menu\functions\fnc_compileMenuSelfAction.sqf..., line 134
22:18:38 Error in expression <sCfg] call _recurseFnc
]
];

[heavy vortex]

Is that where you have too many mods for the steam protocol packet size?

[light cargo]

also this

22:44:37 Error in expression <escriptionParams];};
_descriptionText = format ([_descriptionText] + _descriptio>
22:44:37   Error position: <format ([_descriptionText] + _descriptio>
22:44:37   Error 0 elements provided, 1 expected
22:44:37 File A3\ui_f\scripts\GUI\RscDisplayDebriefing.sqf..., line 45```

[heavy vortex]

Perf/prof branch has some sanity checks on format

The ACE one's something like compileMenuSelfAction being called on a null object. I don't know how that would happen.

[autumn timber]

AFAIR yes, but not you. The server

[empty goblet]

q, was any extra logging added for Can't change owner from 0 to 2 ?

[vale shoal]

Yes, BE is enabled

[vale shoal]

I've got the dump. Where should I upload it? it's 18GB big 😄

[fast hornet]

Sounds like you might be behind CGNAT. Basically your ISP is NATing you and then your Router is NATing you a second time, which makes hosting services basically impossible.
There is no good way around this, IPv4 addresses are exhausted. With some ISPs you can ask them (or pay) to still get a native IPv4 address if they normally use CGNAT.

[heavy galleon]

Well, I am certain that I am not behind NAT.

But I am irelevant, I am talking about the server in someone elses basement. The server hosting works and you can no problem connect to it. The public IP of it works no problem, many arma server, web, ts, others
BUT for some reason it looks like it's "sending answers" from a different IP. Almost like incoming traffic is happening on one and outgoing is being sent from other

[fast hornet]

multiple interfaces within the same l3 network?

[heavy galleon]

doubt, afaik it's barebone, but can send the question to the owner

eh no, there's single internet connection in the house

[light cargo]

@heavy galleon generally that means nat nesting, cgnat or not. check if anything p2p works for you at all. either torrenting or rockstar launcher/grand theft auto online

[light cargo]

thats a consequence of nesting networks, sounds like either the nat'ing is not done correctly or something else.
i have a vpn setup on a server somewhere else that gives me a static ip address and effectively bypasses cgnat entirely. this is done completely with nat techniques

ensuring the ip is the same in the chain is NAT stuff

what im trying to say is that your setup is wrongly configured and having the same ip across the chain of different internal networks is absolutely possible with nat techniques

contact your isp to fix it, nothing that can be really done here

[fast hornet]

as a sidenode, about 15 years ago i was hoping that ipv6 would finally kill NAT within 10 years and v4 would be dead. Instead it's even worse now for many users with that CGNAT crap. 😩

[light cargo]

ikr

cgnat does not even allow p2p unless explicitly taken care of

[fast hornet]

jup

[light cargo]

its the domestic isps dream tho

no hosting for their clients

[fast hornet]

true.. which is another major reason they use it. And most users don't even notice

[light cargo]

and then gamers wonder why they cant play grand theft auto online or use rockstar services at all xD

[fast hornet]

Did anyone experience crashes with v10 to v14 btw.? We have them about once per week rn and i'm starting to wonder if it's even really dedmens fault or the intel 13900 in our dedi box..
Although the system itself is rock stable, just arma server is crashing when it happens.

[vale shoal]

If you have mdmp and dmp files, send it to Dedmen so he can find the issue. I had only once crash on latest perf on the server but it was PhysX. I assume it happened when I connected to RDP.

[fast hornet]

yeah i always send them to him 🙂

last time he looked at them it was some AI going crazy, so probably arma after all.. that was on v11

[feral owl]

Ngrok

[heavy galleon]

Nope

[empty goblet]

{ note: this is new never than today's stable 1.54.133741 hotfix #2 }
1.54.133760 profiling & performance v4 server and client, windows, linux server too

• crash/issues/AI/UI/steam fixes, some tweaks and optimizing (for client too)
  https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[whole cloud]

18GB of crashdump?
I am amazed that Arma uses so much ram, most of the time people complain about it using not enough memory

https://tenor.com/view/oopsies-ohno-gif-25793283

[full nova]

i have 16GB ram, Sytem and Discord + Browser (firefox) ate 60% of it, and welp just barely enough for Arma

[quaint flame]

Ram is cheap. Get more

[full nova]

Downloading more RAM

[feral owl]

I have 32 GB ram

AMD ryzen 7 5700x and 4060 rtx

[light cargo]

windows loves to page things from ram

[whole cloud]

And you are telling me that why? How is this relevant?

[light cargo]

not only that, it got compressed, people complain they dont have enough ram for arma yet they arent getting oom deaths because of os paging/swapping

sorry about that

idk, maybe people expect something different

main people complaints are like arma 3 not using enough cpu/gpu resources lol

[light cargo]

my apologies, it seems that i've misread the message

[empty goblet]

make sure you do some fast lossless compression ontop 😁

[woven loom]

That did happen to me at one point

[whole cloud]

2.16.152058 new PROFILING branch with PERFORMANCE binaries, v17, server and client, windows 64-bit, linux server 64-bit
- Fixed: Crash when providing some command-line parameters with an argument, but leaving the argument empty (-unit=) (thanks @graceful laurel )
- Fixed: Bogus script type errors on Linux servers (thanks @opaque ingot )
- Fixed: Launcher firing effects did not work for remote units - https://feedback.bistudio.com/T182166
- Fixed: Custom missionEnd was broken (thanks @spiral pond )
- Fixed: Game crash when trying to use #vote server command since v7 (thanks @dense veldt )
- Fixed: Compatibility with Windows 7/Windows Server 2008 (thanks @knotty wraith )
- Fixed: Memory corruption and random crashes when a House with Path LOD was outside of terrain bounds (thanks @tall iron )

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh
Note: There are separate Dll files that also need to be placed into Game folder.
🧑‍🍼

[heavy galleon]

insert Bernie I am once again asking meme with "for Linux Addons support"

[tribal pasture]

• Fixed: Memory corruption and random crashes when a House with Path LOD was outside of terrain bounds (thanks @tall iron )
  I was spawning piers outside of the map to put players there during their respawn/revive for years 😳

[whole cloud]

It's only when AI tries to pathfind somewhat near it.
I think

[opaque ingot]

@whole cloud the linux perf version is running fine, thx

[random isle]

we do that still lol

[void badger]

Good old debug island

[empty goblet]

q, anyone noticed that some logs like -mpstatistics which needs to be active in commandline, are auto enabled in profiling (ie not active in commandline yet still producing the logs)

[light cargo]

i think i have noticed that long ago in stable, i had txts about that in my appdata\local folder

[random isle]

on latest pref noticed ram usage super high, not sure good thing or bad lol so far 14gb for server and 16gb for HC
usually they both hover around 6-8gb
i been running mimalloc_v214_lock_pages.dll btw for probably 4 months

[tribal pasture]

20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
20:50:52 Ref to nonnetwork object 1780898: <no shape>
```1000s of these lines on client

Wish it also showed a class name so its easier to see what it is

https://feedback.bistudio.com/T171682

Maybe this?

[spiral pond]

can be FX/particles for example

-debug gives you somewhat of an idea usually where its coming from

see also: #arma3_scripting message

[empty goblet]

q, anyone has any crashes for me on 133760 ?

[tulip nimbus]

format ["a %1 b %%2",true,false]
Returns "a true b %2"

Did https://feedback.bistudio.com/T81725 get put on perf? main branch seems fine (returning "a true b false")

["Arma 3","Arma3",216,152058,"Stable",true,"Windows","x64"]

[opal hound]

That seems correct, doesn't the double % escape?

[tulip nimbus]

well is performance prof branch meant to have unreleased stuff not on main stable? the change says 2.18

[analog acorn]

Yes. If you look at the pinned messages in this channel, you can see the branch change notes.

This change was pushed to perf/prof on the 10th of July (#perf_prof_branch message)

[tulip nimbus]

On main the following works
format["scriptName '%3'; scopeName '%3'; private _thisScriptName = '%3'; %4%1line 1 ""%2 [%3]""%4%5", "#", "_filePath", "_functionName", endl] + "_filePath";

but on perf prof it brakes. Provided 4 expected 5. Am i missing something?

[opal hound]

Yeah, there's no 5th parameter

[analog acorn]

You're using %1 to %5, but you're only providing 4 arguments (i.e. %1 to %4). On main the command doesn't report this error, on perf it does.

[tulip nimbus]

hmm interesting thanks

[heavy vortex]

It frequently does, if it doesn't need data changes and maintains network compatibility with stable.

[whole cloud]

The code that prints it doesn't know that the object even has a type, but looks like I can make it print a level higher, next week

[feral harness]

It's a template you will need to add the new method to every object type of modify GetDebugName which you probably shouldn't

[whole cloud]

I already did it, just needed to add it to one

[whole cloud]

2.16.152100 new PROFILING branch with PERFORMANCE binaries, v18, server and client, windows 64-bit, linux server 64-bit
- Tweaked: missionProfileNamespace is now saved on mission end (only if isMissionProfileNamespaceLoaded) - https://feedback.bistudio.com/T183037
- Tweaked: Chat channels which are not available will not appear in the UI channel selection - https://feedback.bistudio.com/T179942
- Tweaked: copyFromClipboard is now available in Eden Editor multiplayer mode
- Fixed: Non-leaders could talk in command chat - https://feedback.bistudio.com/T179942
- Fixed: Crash if a weapon magazine is removed during the "Reloaded" Event Handler (thanks @fast hornet )
- Fixed: "HandleHeal" EH ignored its returned boolean - https://feedback.bistudio.com/T182892
- Fixed: "HandleHeal" EH did not trigger on heal at a medical vehicle - https://feedback.bistudio.com/T183117
- Fixed: "HandleHeal" EH isMedic parameter was always false - https://feedback.bistudio.com/T71433
- Fixed: Tactical Ping was not placed correctly on a rotating mini map - https://feedback.bistudio.com/T165080#2645515
- Fixed: Unable to use compositions or unit/squad logos when the profile folder contains non-ASCII characters - https://feedback.bistudio.com/T148574
- Fixed: "InventoryOpened" event did not receive the ground weapon holder - https://feedback.bistudio.com/T170711
- Fixed: Extensions could not be loaded if the mod folder or game folder path contained non-ASCII characters
- Fixed: Explosive charges could cause very low or no damage at all, when detonated close to the center of mass of a building - https://feedback.bistudio.com/T183118

- Fixed: Performance issues and out-of-memory crash trying to draw a map on extremely large (>128km²) or broken (NaNs in heightmap) terrains - https://feedback.bistudio.com/T183162
- Fixed: sleep command would not work in persistent JIP scripts
- Fixed: Spamming map open/close could result in a permanent loss of sound (thanks @plucky glade )
- Fixed: "Put" Event Handler did not fire on containers - https://feedback.bistudio.com/T175428
- Fixed: Zeus map did not show polyline markers - https://feedback.bistudio.com/T175116

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh
Note: There are separate Dll files that also need to be placed into Game folder.
🩹

[charred holly]

I had just switched to dev to check out the HandleHeal changes 🤣
Nice to see they are included in profiling too!

[light cargo]

that's a hefty update!

what are the additional dlls used for?

[silk pewter]

crime

and cryptomining 😄

[light cargo]

oh come on

[indigo anvil]

you need to integrate crowdstrike endpoint protection as well 😄

[light cargo]

it was a genuine question

[silk pewter]

hence the joke's effect, works better on non-joke questions 🙂 but the question remains, ofc

[light cargo]

also, any idea how old is the bug talked in T179942? is it just a user's feature wish or an actual issue? it has the feeling of being an ancient thing

[feral harness]

ancient

there was no check neither in ui nor on server that the message/von is coming from a leader

[light cargo]

is it from a1 days by chance?

[light cargo]

ah so that finally fixes everyone reading the commander channel?

[feral harness]

i think it predates dinosaurs

[light cargo]

ofp???

[feral harness]

actual dinosaurs

[light cargo]

lol nice joke, bruh :P

i actually wanted to know how far it predates arma 3 lol

[heavy galleon]

Dedmen doing god's work

Dedmen not supporting non lower case Addons on Linux

[silk pewter]

see pinned message

[heavy galleon]

Sometimes we all have to face our fears. I have a "good job" sticker ready for the wonderful day that is yet to come

[hybrid cradle]

i will pay dedmen to keep not supporting it

[worldly badge]

I will pay for multiple Dedmens

[indigo anvil]

cloning Dedmen fo you

[rain moth]

he should add that as a tier on patreon

[empty goblet]

you can be sure after something found today there will be new profiling soonish 😁

[rain moth]

nice, finally dx12 and vulkan renderer

[empty goblet]

no, that definitely ain't it ... yet

[heavy galleon]

I love how dwarden would love to have new renderer aaaaand

[empty goblet]

well once upon time i was dreaming about Vulkan renderer for A3/AR/A4/DayZ ... but then someone shattered that dream

[heavy galleon]

Well, what prevents you from making dedmen to implement a vulcan renderer for A3 now?

[empty goblet]

dedmen specialization ain't renderer/graphics related and there is limit of what superprogrammer can do 😁

[heavy galleon]

I bought vulcan course on udemy, let me at it!

[empty goblet]

also he would be one of many i tried to bribe but none of them took the bait

[heavy galleon]

You did not provide enough beer and sausages in your bribe

[autumn timber]

Here: https://careers.bohemia.net/open-application 😉

[heavy galleon]

🙂 gotta finish school first. Is what is what I was told

[restive pilot]

I don't think adding a new renderer has any merit at this point. It's not just a matter of adding the renderer but also many parts of the rendering logic have to be rewritten to take advantage of it.
Also if a new renderer was to be added (which is unlikely) it should be dx12 because it's easier to adapt dx11 to dx12 than dx11 to vulkan
At this point it makes more sense to invest some time to improve the current dx11 implementation than invest a lot of time for a new renderer
e.g. I noticed that the game does some "things" on the CPU but I think they can be moved to the GPU

[spiral pond]

Did DayZ standalone run initially on the TOH renderer?

[quaint flame]

Dx11 dx12 dx9. Who cares. I want light that doesn't shine through objects : 😭

It's actually amazing that that was not a thing back in 2013

[silk pewter]

"remember the 'per square metre' terrain lighting with cars?"

[autumn timber]

I think you meant "per triangle metre"? ;P

[silk pewter]

hahaha truuue

dang, felt so sharp when A3 arrived with per-pixel on terrain

[rain moth]

with better alpha? 😛

[light cargo]

source engine is older and is more realistic than arma lightning XD

[heavy vortex]

source engine is a corridor shooter :P

[light cargo]

lol wdym

[heavy vortex]

Everything in the map is one giant structure, with pre-baked lighting. Open-world games like Arma need different techniques.

[light cargo]

hmm, how does RAGE does it then? or gamebryo

elder's scrolls always used gamebryo, i believe?

whatever, the point in question is a comparison to similar/equivalent engines in terms of lack of implementations

[silk pewter]

that's the point
there are no similar engines

[autumn timber]

Rage was using id Tech 5, AFAIK

Or another id Tech engine, because I remember reading that Carmack wrote that huge texture support for that game, if my memory serves me well

[heavy vortex]

There are modern shadow-mapping techniques that work fine for open world games, but they're all much heavier on the GPU than A3 is.

[light cargo]

wrong, that's not idtech at all

[light cargo]

the author contributing/implementing a feature to a different engine doesnt automatically mean its their signature engine

[light cargo]

im talking about rockstar engine not the game called rage

[autumn timber]

Oh! Okay 👍

Funnily enough, Rage's (the game) release date roughly matched Arma 3's release (2 years) so i thought that you meant the game

[light cargo]

gta 5 is from 2013 as well

that's the closest match to a3 actually

[heavy vortex]

GTA 5 engine is pretty neat tbh

[light cargo]

aka RAGE

rockstar advanced game engine

gta 4 is more realistic tho

[heavy galleon]

but what about (again) native linux support?

[rain moth]

when Metal renderer for native macOS?

[silk pewter]

Workbench for ArmA please

[heavy galleon]

please no

we already made enough tools to make life happy

[silk pewter]

at this point you're better off patching Linux to support case-insensitive paths 🤣

[rain moth]

that's already mentioned in the wiki page 😛

[whole cloud]

Excuse me people. This is #perf_prof_branch please.
You can put your feature dreams into #arma3_feedback_tracker or #offtopic_arma

[light cargo]

gotcha

[sick arrow]

Isnt dx12 lower level then dx11 so it still requires lot of work to implement regards? But i agree for this old game besideds future proofing and benefit to small part of AMD/Linux users there is no real need for new renderer.

[light cargo]

you can use dxvk if you want a vulkan renderer

[empty goblet]

1.54.133782 profiling & performance v5 server and client, windows, linux server too

• crash/issues/AI/UI/steam fixes, some tweaks and optimizing (for client too)
  https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[restive pilot]

Yeah which is why I said you need to rework the rendering code to take advantage of it. Otherwise it's no different than dx11

[empty goblet]

yes, that works even on windows but the frametime gets massive increase

[light cargo]

oof

[empty goblet]

oh the original idea was to just take what's needed from the DX11v2 renderer which was slapped into DZSA ontop of TKOH engine after 2016Q4

[empty goblet]

ironically that was fixed in DZSA too and some shadows

[analog acorn]

Well, we've come full circle. Time to remake Arma as a mod of DayZ

[quaint flame]

Can someone please push that "Merge" button 😄

[light cargo]

lol

[empty goblet]

that would be too much huge boon for Arma 3 players 🤣

[nocturne obsidian]

crashed after 2 hours

[flat schooner]

You might want to send dwarden a server log :3

[worn obsidian]

23:26:17 SteamAPI initialization failed. Steam features wont's be accessible!

[solemn wyvern]

Chat channels which are not available will not appear in the UI channel selection
does this remove Vehicle channel from the UI aswell? i kinda used that for removing telepathic map sharing from players along with Basic Map Sharing, i was wondering why yesterday i couldn't use the Vehicle channel, is it possible to add a option somewhere to actually enable it again?

[quaint flame]

So...you didn't say it's impossible. So who will be the brave person who presses the merge button?😄

[light cargo]

get on a vehicle and you should be able to use it

[solemn wyvern]

I'm aware but the idea was having it at all times to let players plan mid-mission and simulate as if they were first drawing on their map to then share it because when outside a vehicle you wouldn't automatically share to other players, we usually do Fast-rope incerts so we don't have vehicles nearby, but ngl it's easier just to use the default branch at this point, it only affected 2 of our players with me included, but if possible i would like to be able to re-enable at least the vehicle channel at all times, i don't want to be much of a bother to dedmen

[light cargo]

you're taking advantage of a bug, maybe it could be possible to request moddable custom channels instead

[solemn wyvern]

it's a bug? i just thought dedmen made a QOL change, didn't knew it was intentional, but not gonna lie that second option would be interesting, and i think it would be easier to allow direct channel to create markers

[whole cloud]

@ killzone_kid will see your message when he has time

[hybrid cradle]

i thought we had custom map marker channels already?

[whole cloud]

We do

[light cargo]

See the ticket the last perf/prof changelog references to, it been acknowledged as a really old bug

[light cargo]

TIL, how do you implement them? is it simple?

[light cargo]

sorry, as it was said i stand corrected, just look into implementing that which exists with a mod or whatever applicable to your case

[median belfry]

Please, we need a Arma glow up 🙏🏻

[feral harness]

read it twice I cannot grasp the problem. even before tweak you would not be able to communicate on vehicle channel when not in vehicle, the server does the in vehicle check. the only thing that did not make sense why do you have channel in UI if you cant use it?

[light cargo]

you could communicate with vehicle channels but it would be for yourself only unless it follows you wherever you board a vehicle and get off right ?

[plucky field]

I think he means the map markers. Draw stuff on vehicle channel while on foot, which then gets shared when everyone is in the same vehicle. Now they can't access the channel at all when not in a vehicle.

[light cargo]

yes, it's technically a bug according to bi, would the markers on vehicle channel follow the drawer user across vehicles when boarding and leaving? (and disappearing/showing up for people that are on the vehicle with the person)

[plucky field]

Not sure, never tried this myself. Just what I understood they were doing. But yes, it was a bug and I would go the extra channels route others suggested.

[light cargo]

yeah

[worn socket]

would that also solve the dynamic shadows issue?

[feral harness]

Well I just loaded legacy branch and tested.

1. vehicle channel is not only for vehicle communication but for the same vehicle communication
2. the server will not send message to anyone not in the same vehicle
3. the server will not send message from anyone not in vehicle~
4. markers follow the same rule for where they can be placed and seen

so you either do not remember it correctly or have been using it expecting something which it was never meant to do and in fact didnt do. This is why I am confused.

To save time, please make a repro so that I can verify
before: ....
after: ....

The tweak for the vehicle channel only disables UI vehicle channel availability when not in vehicle, the rest of the tweak more extensive was for command channel.

[heavy galleon]

He used it as a "private" sort of channel just for himself.
If I understand that correctly. I sometimes did that too to track my own position on the map.

I know of an event where they don't want people putting markers on the map and they only left the vic channel available. (No idea why).

[feral harness]

Ok, I see. I have tested a little more, and the vehicle channel is weird. It has other problems, so I will revert the usage of the vehicle channel, because if I make it vehicle specific as it was meant to be I am afraid more people would complain. So next prof it will go back to be weird again

[heavy galleon]

Wait what?
I always thought that markers placed in vehicle channel will get broadcast to everyone in the same vehicle.
No "vehicle in progress" queue just who is in when placed, gets the marker.
Truth be told that updating the marker afterwards will get updated on everyone who has this marker, but that's something you have to live with and I don't see as a big problem

[feral harness]

the marker will be shared with whoever was in the same vehicle at the time, new people getting in will not see the marker, people who received it in vehicle will be able to see it after exiting the vehicle. the marker will be connected to the unit, so people leaving server will not destroy the marker and new people joining into the same unit will see it. it is weird.

Alternatively trying to make vehicle channel available on map only

no point having it out if vehicle as chat and voice wont go anywhere, map markers on the other hand would stay as before

[heavy galleon]

Wait. Chat and voice will work in vehicle even outside of map no?

[feral harness]

yeah only no one can see it or hear it

[heavy galleon]

Wat? When you're in a vehicle, and you speak/send message in the vehicle channel and there are other people in the vehicle channel, those should hear and see it, no?

[feral harness]

yeah only those already in the same vehicle

[heavy galleon]

Yeah, still useful for crew

[feral harness]

actually I wonder if someone is talking in vehicle and a new person enters, will they hear the speaker as they enter or will the speaker need to press the talk button again. I bet it is the latter

[heavy galleon]

Test it out

[feral harness]

meh

I will just revert vehicle channel change it is a can of worms

[heavy galleon]

Getting rid of veh channel outside of vehicle is ok in my opinion. But some people use it as a "private" markers option

[feral harness]

Revision: 152113

[light cargo]

@feral harness but the commander changes are good?

[light cargo]

worth changing it so its attached to the vehicle entity instead of linking it to the people that were present in the creation of the marker?

[spiral pond]

16:10:42 SQFC Code block string uses constant index instead of offset encoding. This is only supported with the main code block. Tried to use constant index: 2
16:10:42 Bytecode loading invalid instruction offset, offset 49 inside content that is only 37 long
16:10:42 ➥ Context: [] L147 (WW2\SPE_Missions_p\GameModesBI_p\SPE_Warlords\fn_SPE_WLVarsInit.sqf)

1. seems the logging is now in perf branch

2. not aware we do any SQFC conversion

so far seen it only for CBA functions

16:10:42 SQFC Code block string uses constant index instead of offset encoding. This is only supported with the main code block. Tried to use constant index: 0

without any filter ref

[empty plover]

the new format argument check is giving me a lot of errors, is there anything that can be done about that?

[analog acorn]

I think the usual advice is to fix your code

[feral harness]

yes

[feral harness]

in the other life maybe

[empty plover]

lots of code to "fix". I never expected something like this to become a problem

hmm RscDisplayDebriefing.sqf also reporting error , now I dont know if that's because me giving wrong argument or already known bug perhaps?

[analog acorn]

Known and fixed on dev. https://feedback.bistudio.com/T182822

[empty plover]

cool 👍

[analog acorn]

The code has always been wrong, the change is that the game now tells you it's wrong

[empty plover]

having not to worry about argument count i consider as feature 😉

[restive pilot]

Well if it didn't show errors before it shouldn't show errors now either
It's not always like the case with RscDisplayDebriefing where you have access to the script to fix it yourself. It could pop up in an old mission or mod

imo the error should go (tho I guess it's ok to show it with an option like -debug or something)

[empty plover]

^^^ lots of errors coming up with peoples missions

[analog acorn]

Error: waitUntil returned nil

[restive pilot]

That was 5 years ago iirc

The game was much more alive back then than it is now. (And the game was not EOL yet iirc)

[light cargo]

-showscripterrors brings all that in normal gameplay right? should i have it enabled or only let it be enabled automatically when editing/testing in eden?

[spiral pond]

DidJip(object) not called on server, this function only works correctly on server. Use DidJip() to get info about local client

is there any point to log that in !isMultiplayer ?

[solemn wyvern]

it is not a problem per se, in the last week build i could do it, i can do it right now on the default ArmA 3 build, we use it solely for drawing in map only, and use an other mod to share it to other people, this is mostly good so people don't just draw enemy positions on map when everybody is miles apart and still recieves the info, if it was meant to be like there's no reason for me to bother y'all,

I will try to find some documentation on map markers and channels, i want a channel that doesn't actually share the markers you make so we can use that mod that shares it instead, as it was said earlier we do have a "Custom map marker channel", if possible can you point me in the right direction?

i would link the message but i got mutted for linking the discord message is right above the message i answered

[solemn wyvern]

i sent the previous message but forgot to re-send (after being mutted) with the reply tag, i was replying to this message here

[feral harness]

does it show errors on stable or only dev?

[solemn wyvern]

and if it's too much of a problem i just found a mod (still need to test) that does restrict markers, so you may not need to revert the changes just in case

[feral harness]

its reverted

[solemn wyvern]

oh ok, i just don't want to be a bother to you guys

[feral harness]

too late

[solemn wyvern]

i can't fathom being a game dev so i wouldn't want to create too many problems for y'all

anyways, appreciated

[feral harness]

its ok usually this means more people will be affected so best course of action is revert in this case

[whole cloud]

Fix the errors if you can. Vanilla ones are all fixed next update. Mod errors should be fixed by mod authors

[whole cloud]

Been for quite a while.

[whole cloud]

Well open your PBO and look?
These errors would only happen if you packed it with a bugged version of HEMTT

[empty plover]

I will fix them, its just that not all code is run in my mission immediately at start so these format errors will be popping up during gameplay. but yeah planning to fix them all

[quaint flame]

Only with debug parameter.

[empty plover]

its like that now?

[spiral pond]

What if we dont use HEMTT? (only mikero tools + addon builder for some parts)

[whole cloud]

Mikero wouldn't pack sqfc

[whole cloud]

it is on non-prof

[spiral pond]

no sqfc

[empty plover]

ok so stable (when it comes to that)

[whole cloud]

Maybe filePatching is on and it finds sqfc via that somehow?
L147 (WW2\SPE_Missions_p\GameModesBI_p\SPE_Warlords\fn_SPE_WLVarsInit.sqf)
What does that line do? call compileScript?

[spiral pond]

no compileScript

largely a3 vanilla function. just modded a bit to support "dynamic factions"

we have compileScript only in a few feature systems and two missions:

P:\ww2\SPE_Core_f\System_Artillery_f\Towing\Init.sqf
P:\ww2\SPE_Core_f\System_Common_f\Eventhandlers\fn_postInit.sqf
P:\ww2\SPE_Core_f\System_Common_f\Eventhandlers\fn_preinit.sqf
P:\ww2\SPE_Core_f\System_Functions_f\Fnc_init.sqf
P:\ww2\SPE_Core_f\System_Tank_Interface_f\Display_Interface.sqf
P:\ww2\SPE_Core_f\System_Tank_Interface_f\Display_Interface2.sqf
P:\ww2\SPE_Core_f\System_Tanks_f\ADS_Init.sqf
p:\ww2\SPE_missions_p\Campaigns_p\SPE_Operation_Cobra\Missions\SPE_M07_Liberation_of_Saint_Denis_le_Gast.SPE_Normandy\scripts\animatedBriefing.sqf
p:\ww2\SPE_missions_p\Scenarios_p\SPE_CO6_Panzerkampfwagen.SPE_Mortain\functions\mission\fn_generateFrontline.sqf
p:\ww2\SPE_missions_p\UtilityFunctions_f\params\fn_paramsSingleplayer.sqf

Only sqfc file in SPE:
p:\ww2\SPE_missions_p\Scenarios_p\SPE_CO6_Panzerkampfwagen.SPE_Mortain\scripts\frontline_data.sqfc

as mentioned before i only saw it mentioned in regards to CBA

[whole cloud]

The context doesn't indicate what file was compiled there, which caused the SQFC error.
It indicates which script file location, triggered the compilation of a bad SQFC file.
I don't know which file it is. I don't remember what was built with the broken hemtt version. GRAD Trenches and DUI, maybe CBA.

Maybe something there somehow triggers compile of a script from some of these mods

[quaint flame]

Does it show without any other mods?

[whole cloud]

SQFC loading code can only run
on Savegame load/save
on "compileScript" command

[spiral pond]

was probably on Savegame load/save

[empty goblet]

so, thanks to @tough stirrup we found out that that issue with MaxPacketSize

• if client joins server and gets stuck-hang on map screen initialization (sees join/leave/kick messages of other players even chat fine) ...
  this is caused by low MTU / packet filtering/blocking by size on the client / ISP side / route
  because the server side assume it can send larger packets to the client
  probably lack of handshake fallback, solution could be if certain handshake packet fails to reach client and back then if larger than default used,
  retry with default (1400) and if that fail use some slightly smaller packet size (1380 then 1300 then 1200)
  or simple do -512 then try +256 +128 +64 +32 +16 +8, at size down from default (this will sort even maxmtu 1200 weirdness)
  another headache of this problem, there is no error entry in RPT / UI about what's the problem at either side client / server

[whole cloud]

Then it would be the game itself. Savegame scripts get stored as bytecode.
Would need repro for it.

[spiral pond]

with CBA running or without?

[whole cloud]

doesn't matter

[empty goblet]

note: koth server needs client to use #perf_prof_branch latest build of profling branch / performance binary

[light cargo]

why?

[empty goblet]

because of fix in profiling and what KOTH v17 introduced

[woven loom]

@ dedmen i will test and see if CTD with GM Force Recon on Cham map is resolved, thanks for the fix!

[whole cloud]

2.16.152136 new PROFILING branch with PERFORMANCE binaries, v19, server and client, windows 64-bit, linux server 64-bit
- Added: Server network anti-flood protections
- Changed: "EpeContactStart" event now triggers before collision damage is applied - https://feedback.bistudio.com/T177724
- Fixed: networkDiagInterval did not print stats for recently disconnected players
- Fixed: networkDiagInterval printed total counter numbers for messages, instead of the number since last print
- Fixed: Server could freeze when too many network messages were received
- Fixed: Waypoint, remoteExec, particle, init.sqf and action scheduled scripts were missing _thisScript variable - https://feedback.bistudio.com/T183544
- Fixed: setUnitLoadout would replace the Uniform/Vest/Backpack container even if nil was passed for them - https://feedback.bistudio.com/T180975
- Fixed: Game still considered shadow view distance (and rendered all objects within that distance) even if shadows are disabled - https://feedback.bistudio.com/T175612

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh
Note: There are separate Dll files that also need to be placed into Game folder.
🚰

[empty goblet]

again , new profiling 2.16.152136 , same new build required on client on KOTH server

[autumn timber]

@heavy galleon it's your moment to shine, again 👆

[heavy galleon]

I am once again asking for case insensitive addons folder on Linux

At this point I am sure he's not doing it just to spite me

[whole cloud]

https://tenor.com/view/bongocat-banhammer-ban-hammer-bongo-gif-18219363

[heavy galleon]

[silk pewter]

I wonder if the Linux path thing will get some luv eventually

should ask Dedmen

[heavy galleon]

True

Let me get an internship, I'll fix it and quit right after

[empty goblet]

that would be waste of fix chances ... there is so much to improve 🤣

[heavy galleon]

Priorities man, priorities

[empty goblet]

i still trying to get MinimalPixelSize (client) and keyFrames (simulWeather, server) to customizable via .cfg/description.ext

[whole cloud]

How much C++ do you do?

[empty goblet]

... the Sentinel's gaze locked on the target ...

[heavy galleon]

Yes

[ashen helm]

psst, you can always lock yourself in Edita and tell them you won't get out until your thing is fixed

[silk pewter]

not Edita please

[heavy galleon]

Mníšek is far

[full nova]

to Amsterdam Office where dedmen is

[heavy galleon]

I might be near Amsterdam in two weeks. A small detour to fix it

[empty plover]

it seems that Perf cant read hard links (says script x not found)

[whole cloud]

more details please. hard link to what?
filepatching with sqf file, and sqf file not present in pbo?
Hard link to pbo?

Just found out, dedicated server cannot load missions inside pbo

[empty plover]

-filePatching is on. for some reason stable loads my mission fine but Perf doesnt find the init.sqf thats inside mission folder (hard linked to there)

im running perf from command line so may not have exactly same command line arguments than stable does but I guess only -filePatching matters

[whole cloud]

where is the mission folder?
mpMissions/mission.Stratis/init.sqf ?

running dedicated server or game client?

[empty plover]

its in mpmissions\FOW\fow.Malden I also tried with another fow map and same thing. just trying to run it from editor-MP

[whole cloud]

filePatching doesn't matter there

[empty plover]

ok..

but the funny thing is I think it worked before in previous Perf (just updated the exe today)

[whole cloud]

Yeah game updates sometimes break things

I was worried about this build, which is why I specifically test-ran it yesterday. But I only tested my own mission, without a init.sqf

[empty plover]

yeah i mean maybe someone just broke it

[empty goblet]

and 152136 fails to load \addons\ mission pbo with mission files, only \MPmissions\ mission files works, via mission cycle in server.cfg

[empty plover]

you guys want a test with previous Perf?

[whole cloud]

no

[empty plover]

ok

the one i used today is v19 . hope thats correct one?

and v18 is the previous one

[whole cloud]

v19 today got brokered, v18 before was fine

[empty plover]

ok just checking I had right version 😄

[whole cloud]

Missing a !
It thinks all missions that exist, don't exist.

[empty goblet]

😁 that's too funny typo

[whole cloud]

Don't know why that passed my test run yesterday 🤔

[empty plover]

good it was not hard link issue then

[whole cloud]

The init.sqf issue is separate.
That's fucking C++.
Programmer: Okey I'm giving you a reference to another function, that you should take and then execute
C++: Thank you, I'll happily take the boolean you are giving me

[feral harness]

sounds like Microsoft compiler vs GCC

[whole cloud]

Now I wonder if that works fine on linux 😄
But I won't test it

[heavy galleon]

[whole cloud]

(╯°□°)╯︵ ┻━┻

the init.sqf thing is fixed, but loading missions from mod-pbo (including vanilla ones) is broken. Because I forgot to click the checkbox for the file with the ! fix

[heavy galleon]

What about the checkbox for the file that adds support for case insensitive "addons" folder on Linux?

[empty goblet]

oh lol this explains why i'm staring on failed load mission 😁

[whole cloud]

All fixed now 🐈‍⬛

[indigo anvil]

Depends on the flavour. C++11 or newer ?

[whole cloud]

20, but I asked ilbinek

[heavy galleon]

I did say "yes"
Although I did stop keeping up really at 11

[fair ore]

Is it a prof branch fix that setUnitLoadout properly fires the slotItemChanged event?

Some of my players are on main branch and I think that's causing a bug I can't repro on mine

[whole cloud]

I think there was some missing SlotItemChanged, but I'd think that was a while ago.
More details?

[fair ore]

I'm using cba setUnitLoadout in KP lib and I have a hud element that should activate if they have a particular backpack equipped

For me on prof, it works just fine

For some pals not on prof, it doesn't actually fire the event

[whole cloud]

So prof is correct in that case? So the future is good. So I don't need to do anything? 😄

[fair ore]

I guess you don't lmao

And I need to add a new event to cba loadout set

[hybrid cradle]

yeah i also dont get the problem

its just retail ones so will be 2.18 fixed

(as long as you remember to merge it 😠)

[fair ore]

2.18 gonna be a banger

[charred holly]

https://feedback.bistudio.com/T180338 for reference, but already fixed

[whole cloud]

Yesterday we released the fix for all the mission/script loading issues. Steam should've updated and google drive files were replaced.
So everything should be good now?

[empty plover]

my mission works again

[real creek]

It prompted me with an import thing, said it was gonna make a backup, but both mission.sqm files are now a few bytes brb hopping into a time machine

[heavy galleon]

When you're back, ask dedmen to support case insensitive addons folder on Linux. If someone asks him when he was making the pbo names, maybe it will work out

[hybrid cradle]

dont do it its too funny

[ashen helm]

https://tenor.com/view/holdfast-opinion-rejected-gif-25349141

[azure jewel]

https://imgur.com/a/LzEwej4
does BUG like random zeus lightning and flying Leaflets is from the prof branch?
Peoples around me didnt see it so I guess its local
We are using same mod preset so it may not be mod's problem...?

[ashen helm]

I haven't seen any of those

[azure jewel]

well maybe something else is wrong with me

[hybrid cradle]

test in just vanilla/prof

[whole cloud]

Whatever prompted you wasn't Arma, because Arma doesn't make a backup for you. Find the mod and blame them

[whole cloud]

Can you send me RPT of affected client?

[azure jewel]

I'll send you next time cuz now im using -noLogs 🙂

[vale shoal]

Were there any changes regarding slingloading since V17?

[vale shoal]

The reason I ask is, that I see similar cases where the server freezes and the RPT would output stuff like

19:34:11 NetServer::finishDestroyPlayer(1038565839): DESTROY immediately after CREATE, both cancelled 19:34:11 BEServer::finishDestroyPlayer(1038565839): users.get failed

This was the same when the sling loading freeze happened. It first happened in Thursday again (after months) with V18 running. I first thought it was V19 (because there was a fix for it), but I just installed V19 later on Thursday.

[empty goblet]

1.54.133788 profiling & performance v6 server and client, windows, linux server too

• several more crash cases were fixed since v5
  https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[whole cloud]

no slingload related changes

[naive osprey]

Anyone happen to know what the new "RQ" value represents with the #monitor MP server command? I can't see any documentation on it yet, and only see it with Profiling.
https://i.imgur.com/V4qIh3L.png

[empty goblet]

it's recently new and related to changes we may need to improve ...

[hoary jolt]

Does the default launcher actually launch the profile build from Steam? I dun see the profile launcher anymore even tho I have performance prof branch subbed.

[stark falcon]

look in the upper left corner of the launcher, it should be green

[whole cloud]

Request Queue.
Messages that came in to the server, but have not been processed yet. Should usually be zero, otherwise the server is backlogged

[worn obsidian]

hugly perf with perf5, i try v6

[gritty wasp]

Is it possible to get monitor data by extension? Would be nice to have some kind of Prometheus exporter.

[heavy galleon]

By monitor you mean the #monitor command for admin?

If so, you could use #monitords and then parse it from the DS console output

[gritty wasp]

As much as I now there is now way to parse screen output except setting camera and perform OCR. Sounds bizarre. Proper API for stats or even native exporter would be ideal. But not gonna happen 🥲 As well as case insensitive addons path

[whole cloud]

Don't know what you mean. You can make screenshots now in extensions. Still need OCR if you want to read text.

[heavy galleon]

You don't have to parse screen output no? I am not sure about windows, but Linux just goes to stdin. If you redirect or tee it into a file, you're golden. (I've been saying for year now wog should switch to Linux )

As far as case insensitive paths, we are half way there. Just need to finish the rest

[gritty wasp]

Only couple years ago dedmen did fix linux multicore performance. For decade (from A2 era) windows arma server did perform much better than linux. That is why all 170+ PvP servers was on windows.

[heavy galleon]

I am not old enough to remember that

The future is now.

[gritty wasp]

future with case sensitive paths

[autumn timber]

Arma 3 is Futura - I would even say 😏

(I wonder how many people will get it)

[heavy galleon]

It's just me being busy with other stuff so I don't have my Linux server manager ready yet. Symlinks with correct casing are easy to do. I have an old shell script somewhere that does it.

[heavy galleon]

T-100 has entered the chat

[gritty wasp]

I'm just dreaming about fancy Grafana dashboard that could show all useful stats of running arma server if there was an API to get values directly.

[spiral pond]

doesnt arma already pipe some things. i think to recall both logging and some stats. Ondra (Suma) also made a custom pipe in late a2:oa days for mod sync. sadly it wasnt merged to a3 (afaik)

(on windows)

[heavy galleon]

@whole cloud any work done on extensions lately? OCAP exporter does not load in both perf and dev
I see that dev has a "fix non ascii path" in there. It was reported that it worked on previous dev release

[whole cloud]

Yes quite a bit.
Does not load means what, can you specify?

[heavy galleon]

Does not find the extension. Error in RPT

tried stable and works no problem

[whole cloud]

Send me some repro zip please

[heavy galleon]

[spiral pond]

ie you can/could use Sysinternals DebugView to monitor rpt output live

[rain moth]

yes, you can make your own exporter to prometheus etc using extensions, there are already projects for that on github

[gritty wasp]

Avarage fps and players count? Not very interesting

[vale shoal]

maybe a feedback to that issue.. after switching back to V16, it didn't happened anymore. I heard from others who have the same issue

[woven loom]

I'm getting these script errors on mission end:

16:07:28 Error in expression <criptionParams];};
        _descriptionText = format ([_descriptionText] + _descriptio>
16:07:28   Error position: <format ([_descriptionText] + _descriptio>
16:07:28   Error 0 elements provided, 1 expected
16:07:28 File \A3\Ui_f\scripts\GUI\RscDisplayDebriefing.sqf..., line 45
16:07:38 Error in expression <criptionParams];};
        _descriptionText = format ([_descriptionText] + _descriptio>
16:07:38   Error position: <format ([_descriptionText] + _descriptio>
16:07:38   Error 0 elements provided, 1 expected
16:07:38 File \A3\Ui_f\scripts\GUI\RscDisplayDebriefing.sqf..., line 45```

Happens when the debrief window appears

This is from a force recon mission, but it happened in other missions as well, which did not previously show this error

[whole cloud]

known and fixed long ago, profiling just doesn't get new scripts, which is needed to fix this

[woven loom]

ah okay

I also got these errors while looking through some Western Sahara assets in the config viewer

23:20:32 Error in expression < else {
            call compile _lbData;
        };
        _configPath
    };

    case "textPath": {>
23:20:32   Error position: <_configPath
    };

    case "textPath": {>
23:20:32   Error Undefined variable in expression: _configpath
23:20:32 File /temp/bin/A3/Functions_F/Debug/fn_configviewer.sqf..., line 1025
23:20:32 Error in expression <bis_fnc_configviewer;
        _configValue = [_configPath,_lbData,""] call bis_fnc_ret>
23:20:32   Error position: <_configPath,_lbData,""] call bis_fnc_ret>
23:20:32   Error Undefined variable in expression: _configpath
23:20:32 File /temp/bin/A3/Functions_F/Debug/fn_configviewer.sqf..., line 873

[whole cloud]

Don't know what that is

[woven loom]

Hmm i just expanded a class config entry and was greeted with this, iirc

was last night

I haven't gone through config entries in a while though so don't know how new this is

https://imgur.com/MhYhZtB

Right after i double click the parent class

Yeah doesn't seem to happen on stable branch, happens on perf/profiling, untested on dev

[restive pilot]

Why does the path start with temp/bin?

[woven loom]

the only things i had loaded were WS and RF 😅

[restive pilot]

You haven't extracted it have you?

[woven loom]

...extracted what? 😅 CDLCs? no

[whole cloud]

SQFC, a script compiler bug I've been meaning to fix for 3 or so years

[restive pilot]

I meant "extracted the functions pbo" but nvm

[woven loom]

ahh no nothing like that

[quaint flame]

It's just the config viewer being junk.

Just don't use it

Use Leo's or Terra's

[woven loom]

but it's being junk just on profiling branch 😛

[woven loom]

damn advanced dev tools mod looks pretty dope

[quaint flame]

It also throws script errors on stable.

[woven loom]

Huh didn't when I tested, weird

[quaint flame]

It happens every now and then. Perhaps just specific entries idk

Anyway...no one should use it really, same as with the fnc viewer

[heavy vortex]

Yeah I had a lot of trouble convincing vanilla config viewer to display the deeper class trees (turrets, pylons) when Arma was under any load.

[whole cloud]

2.16.152172 new PROFILING branch with PERFORMANCE binaries, v20, server and client, windows 64-bit, linux server 64-bit
- Tweaked: Anti-flood default values (preventing false positives, and ignoring Headless Clients)
- Tweaked: Minor netcode performance improvements
- Fixed: Netcode related crash
- Fixed: Extensions could not be loaded from the game directory (as opposed to a mod directory, broken since v19)

If you don't want to use the Steam branch, the files are also available for alternative download here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh
Note: There are separate Dll files that also need to be placed into Game folder.
🔥 🦊

[autumn timber]

• Fixed: Extensions could not be loaded from the game directory (as opposed to a mod directory, broken since v19)
  What are the current dll location requirements for:

1. Extensions
2. Dlls that those extensions depend on

Just wondering if the information in the wiki is up to date? Maybe the second point could be added to there as well, by the way? (https://community.bistudio.com/wiki/Extensions#Where_to_put_Extensions?)

[whole cloud]

extensions loaded from all loaded mods dir's, and game root
extra Dll's by default only working directory

[autumn timber]

...which could be literally anywhere on the filesystem? 🤔

[whole cloud]

could be

But game also needs some pbo's and dll's inside working directory, like the "dta" folder

[autumn timber]

Oh, so it will effectively usually be the game root, ok

[rain moth]

If you need same folder as loaded dll you can get that path internally and append to load path

[heavy galleon]

I am once again asking for case insensitive addons folder name support on Linux

[autumn timber]

Yup, that's precisely what I have been doing, but because I needed to decide at runtime which dll with which bitness to use (because they used the same name and I had to put them in different directories).

But I asked nevertheless, because there are people coming to #arma3_tools and asking that question from time to time

[indigo anvil]

no such thing. linux is case sensitive. Geather the fact what kind of path(s) you need and create links that relfect the needed path.

[autumn timber]

Which would be doing the same thing that ilbinek is asking for, but manually and also over and over again, after each mod update

So it's much better to have Arma treat the addons files case-insensitively, because it's unreasonable to expect all the modders in the world to fix their paths or care about creating their paths correctly, in the first place

[indigo anvil]

yea if you are a modder you should be able to work out what kind of links are needed from gameroot for your own content to work.

[autumn timber]

Or another way to look at it: there is no real use-case to having case-sensitive paths for Arma, in general, let alone to having them case-sensitive in only one OS, but not the other

Some modders will just say "works on my machine", some will not even know there is an issue with their mods, since they never used linux, etc...
There also are regular people who are not modders and just want to setup an Arma server using other people's mods.

It's counter-productive to ask all these people to themselves fix potentially every mod they ever want to use if all of this can simply be made a non-issue.

[stark falcon]

Subskeleton index out of range. There seems to be a SubSkeleton index that is greater than the number of sub skeletons, this should not be happening. Please report it with reproduction steps/files.
Not sure if new? Saw when loading a unbinned p3d

[heavy galleon]

But we already have case insensitive pbo names. Until that was a thing, I would have not dared to ask for it

[heavy galleon]

Sadly from my understanding Arma has it's own FS stuff, as it uses virtual FS internally. And that is made in all lowercase.

[whole cloud]

Uh interesting.
Yeah that message is new, but that was supposed to be in binarize exe

[stark falcon]

ace master, just packed into pbos, put down m47 dragon. all kinds of warning about sum of weights and stuff that isn't unusual, just that one error msg seemed new, can post full rpt if helpful

[whole cloud]

The message was added because there was a crash caused by that, and I didn't see how it could happen or really if that is the actual issue.
Seems like it does happen

[empty plover]

this is just a small editor thing but everytime you click the markers button the editing mode changes from 3D to 2D. With the new 3D editor markers this is undesirable

[whole cloud]

I left it because I didn't want to fully commit to 3D markers.
maybe after we've let it run for a game version and know its working perfectly fine, we can change that

[empty plover]

ok

[knotty wraith]

there is not enough opportunity to set the height with a 3Dmarker in the editor

when copying from another map, they may be at an incomprehensible height and it is unclear how to lower them to the ground

[quaint flame]

3den Enhanced will have z value for marker with the next update.

Have you tried snapping them to the surface?

[knotty wraith]

it is often very high, yes, it is possible to lower them with the mouse somehow, but it is difficult

[ashen helm]

but have you tried using the snap to the surface option? No need for using mouse, you just right click and they go down to the ground level.

[glacial mortar]

does this have any benefits fps wise for players? we changed arma3server64 exe with newest one and pasted dll file into arma 3 folder in our machine. should we expect any fps increases?

[inland dew]

try and see

[whole cloud]

you can just run YAAB and test

improving server performance doesn't always equal better clientside performance though

[lilac hearth]

In my machine the gains are about 5% in synthetic test (YAAB tests | client + server performance files)

[empty goblet]

if anyone is at Windows 11 + new AMD (Zen 3/4/5) + recent Windows 11 update containing Zen predictor fixes KB5041587 , more gains #offtopic_hardware message

[whole cloud]

[light cargo]

huh? :o

[empty goblet]

seeing it's all same version, just iteration on the magical surprise ?

[whole cloud]

Only the best ethereal oils being used here

[restive pilot]

It's easy you just make diag_fps return 10 + actualFPS

[rain moth]

vulkan would fit behind that red rectangle

[silk pewter]

DX12 render too

[empty goblet]

vulkan ... that will allow native nix client 😁

[silk pewter]

is it "only" that that is holding penguins back or is there something else (beside .NET framework for the launcher)

[heavy galleon]

Given server supports extensions, I assume it would be no problem for native client

[autumn timber]

There seems to be an "y" or similar, in the center of that rectangle, though -_-

[rain moth]

vylkan

[autumn timber]

You're right, he must have made a typo and it's definitely Vylcan 😄

[empty goblet]

linux always was lacking set similar to DirectX package ... i know there are several crossplatform libraries which are similar but afaik none done most of it and simple while well documented and just works

now over the years the just works on DX part is questionable and the way MS decided to not fix stuff to push new OSes didn't help either cough Xaudio 2, twice (or flat out crippling hw DX audio into software while Xbox does hardware acceleration all the time)

[full nova]

or the weather is colder than usual which helps lower the system temperature results in less throttle.

[heavy galleon]

Nah, it was a neutrino flipping the correct bit in memory that made it somehow skip number of calculations and made them correct anyway

[woven loom]

Can't tell if serious but the system shouldn't be throttling due to thermal limits if it's been built correctly

[light cargo]

not even togl?

[heavy vortex]

Maybe it's just that win11 "branch prediction" update

Wouldn't even be the worst. Fortnite got >30%.

[full nova]

Laptop says otherwise

[woven loom]

yes true BI devs are definitely doing game dev work on a laptop

maybe if they put their laptop in the freezer, it will score another 10 fps in YAAB

ez optimization

[silk pewter]

I feel targeted

[hybrid cradle]

hey i do game dev work on a laptop and im not even a bi dev

well im bisexual does that count

[light cargo]

lmao

what

[whole cloud]

[somber plank]

https://tenor.com/view/star-wars-more-gif-21856591

[heavy galleon]

2.19... It's not like we are gonna see it in the next year

[inland dew]

720p

[whole cloud]

Resolution did not change between both runs

[inland dew]

but 1st run never counts, because of textures/objects loading, giving different FPS vs. the following runs

and need like at least 3-5-7 runs and average that, because of how random Arma is

few FPS is within margin of error

[whole cloud]

I'll leave the benchmarking to you. I already planned to ping you about it then

[inland dew]

dx12?

[whole cloud]

[inland dew]

ok, it is then

finally

higher FPS is better, but consistency is even better

[fair ore]

What is he COOKING!

[heavy galleon]

hopefully joye

[median belfry]

[light cargo]

o.o

[inland dew]

finally GPUs will perform as they can, like in DayZ

[whole cloud]

Its baffling what you find sometimes.
In 2020 there was a freeze/crash issue on linux servers related to reading files.
The issue was with a lock not working properly.

But I just found out, that fix causes considerable lag spikes. In YAAB I see several 20+ms freezes because its waiting for a file load that is supposed to be async.
The fix should've been linux only but was applied on windows too.

[light cargo]

huh strange

[whole cloud]

Okey the bad "fix" was removing a piece of code that should've been there.
Turns out, the person who added that code in the first place, was me.

It was added because we saw a freeze on linux. But it didn't fix the underlying issue, and I later found a proper fix so I removed it again thinking it wasn't needed.
And today I see, for different reasons, that code was a amazing idea.

it looks like a good change, but I'm not 100% sure about this so to be safe we'll keep it disabled for now.
Ah good old 4 years ago dedmen

Now the lag spikes in YAAB, only happen on scene change when a new model appears on the screen for the first time and it needs to load it right now

Luckily it preloads some before benchmarking starts. But it would've been better if it collected all objects inside the benchmark, and threw them infront of your camera for a short time, so the game force loads all at the start. And won't keep spiking mid-benchmark.
But maybe that's also intentional to include that in the benchmark

Like shortly after benchmark start in the thermal vision, it spikes to load the rocket pods of the A-10
Then it zooms down, and spikes to load the wrecked tank model, and then a flying rocket model

[inland dew]

that's why 1st run shouldn't be concidered

[whole cloud]

YAABv2 when? 😄

[silk pewter]

add comment to said code 👀

[inland dew]

DX12 when?

[whole cloud]

There are now 3 comments on that code.
I had to put year numbers onto them to mark when that stuff happened 😄

[inland dew]

touching the artifacts...

[knotty wraith]

What is to be cannot be avoided

[whole cloud]

How annoying. Looking at the times at the bottom. The minimums and average are higher.
But the spike during the black screen is not as high, so the end result average shows as lower

[inland dew]

after 1st run, fps is more realistic

1st run doesn't count

[whole cloud]

Both of them are 2nd run

[inland dew]

less or not all stuff loaded = higher fps, aka fake

[whole cloud]

Before all the changes - 2nd run
After all the changes - 2nd run

[inland dew]

min fps is what you should look at

min fps is 15% higher

that's a lot

the closer the min to the avg, the more consistent/smooth the game feels

if there is a very big gap between min and avg, and max, this means that fps fluctuates a lot, thus choppy game experience

are you just pressing S in YAAB or you really have selected standard preset in video settings?

[whole cloud]

S in yaab.
My video settings are messy, but the same for all runs

[inland dew]

benching with (some) settings lower than standard and benching at lower than 1080p, results in very very inconsistent results between runs

you really have to do standard preset in video settings (not S) and not lower than 1080p

otherwise it's so so messy that you can't even compare stuff

[knotty wraith]

Groove knows what it's saying

[inland dew]

do lower than standard and lower than 1080p and it will fluctuate 10+ FPS between runs, which is useless then

have discovered this behavior many many years ago

and if you do 1080p standard preset (no S), you have to mostly rely on avg numbers, because min and max still fluctuate a lot

already done more than 1000 runs yaab 🫠

[inland dew]

but still important to pay attention to min fps particularly

[woven loom]

May i also suggest using something like CapFrameX or Intel PresentMon to capture the benchmark data, they'll give you better stats and stuff (e.g. 99th percentile etc)

[inland dew]

look at the sky or desert map and your max fps will be much higher, thus much higher avg, which doesn't mean anything, since min fps can still be crap
thus min fps is the most important

it's like Ryzen CPUs have better max, thus avg fps, but you better don't look at min fps...

[knotty wraith]

YAAB is a very strange tool - here I completely agre

[woven loom]

I think it does matter where the min is happening, if it's a random dip then it's not so noticeable, or it might be during one of the loading sequences

Really dedmen should make a custom benchmark 😛

official A3 benchmark scenario 😄

DedBench

[whole cloud]

I am looking at profiling data to see the performance, not at the yaab numbers.
But its good when YAAB numbers confirm what I see and expect

[inland dew]

do capframex add arma exe there, set capture time to 139 secs and set a button to press to start capture and it will end itsrlf after 139 secs

[woven loom]

BTW there are two other benchmarks @ dedmen that might be worth using as more "typical" loads -- Altis Bench and Stratis Bench
Stratis https://steamcommunity.com/sharedfiles/filedetails/?id=172475381
Altis https://steamcommunity.com/sharedfiles/filedetails/?id=173435011

They haven't been updated since 2013 but both work (stratis one shows an error but i haven't seen that affect anything obvious)

[inland dew]

then you can compare several runs to each other

[whole cloud]

Uh that looks easy. Neat, I'll give it a try

[inland dew]

capframex is german, i think

so you run yaab, but only to bench
you rely on what was captured by capframex

just set runtime to 139 secs and assign a button to press to start 139 secs capture

but you really have to do 1080p standard preset (no S)

otherwise impossible to compare

huge fluctuations between runs otherwise

[analog acorn]

Why do other resolutions cause such large differences between runs of the same resolution?

[inland dew]

Arma

even a 1050 ti or 1060 3 gb is enough to not impact 1080p standard preset (no S)
so be it a 4090 or a 1050 ti, as long as it's 1080p standard preset (no S), fps will be same

[inland dew]

in the run in blue, frametime is lower (each next frame delivered/displayed faster) and less fluctuations, thus more stable/consistent fps = smoother gameplay feel, because of 128 MB cache of the good old i7-5775C, despite only 4/8 cores/threads, only 4.4 GHz and only DDR3 2400 MHz

[inland dew]

also you can easily add like 20+ fps to these results, since fps increased by that since 2019, thanks to perf branch

[knotty wraith]

a very cool theme with HC - it can load 4 processor cores by 90%, but can you do something in principle with the game - parallel processes?

and finally make some kind of event that will track the player's life

where it would be possible to catch any moment of his actions

[whole cloud]

2nd run for each.
Green is "new"
its still 720p, I still didn't check my video settings.
But look how close and similar they are. Meh.

top is new

nice

[inland dew]

well, do 1080p and standard preset (no S)

[inland dew]

green runs much smoother/more consistent/stable

but since 720p and non standard preset, can't really compare, because of fluctuations

what you're seeing, you can't know how much the improvements helped and how much of that were fluctuations, because of resolution and settings

[whole cloud]

But I know what actual improvements I've done. How much better they should in theory be. And it matches that data

[inland dew]

also 1 run per per build, even if it's the 2nd, doesn't paint the whole picture
3 runs is the bare minimum, or rather 4, since the 1st doesn't count, because of loading all the stuff
you can check yourself

ok then

[inland dew]

0.1% very close to min FPS, thus very smooth/consistent/stable

no huge dips

[whole cloud]

Does it have some way to count how many frames were longer than 20ms?
It looks to be like 2-3 but I'd like to see actual number

There it is. 48 below 45fps, and 6 below 30.
In the old one

in the new one. Only 28 below 60

I won't get the average much better, but maybe I can fix the minimums a bit more

[inland dew]

minimums is what counts

who cares about "nice" avg, when min can be dogshit

[whole cloud]

I am too often playing Arma with averages of 30 or below 😄

[inland dew]

when it dips, at the wrong moment or often, you die, because it happend when you were touching down in a jet, or firing at enemy or you or enemy abruptly appeared from behind an obstacle and it dipped at that exact moment

severe dips when pressing the button to lock or cycle through targets in a jet/heli

maybe it has something to do with data link (when data link both ways (emit/receive)?) or number of targets with radards themselves, be it enemy or friendly
no idea

have msi afterburner running with on screen frametime graph and see what happens when you press lock/cycle

[knotty wraith]

the best feedback that reports on the feelings of a simple player

[inland dew]

or is it because of the increasing number of players that the data has to be synced/distributed/updated or the more hours the server is online, the more it dips, when locking/cycling through targets (with radards themselves?)

[heavy vortex]

There was a known issue with radars where they'd have extremely poor performance if they scanned outside the map boundary.

Not sure if that was worked around.

Well, the workaround might be that it dips hard when you turn the radar on but recovers afterwards...

[inland dew]

crap avg is a result of crap min

[heavy vortex]

Not really. Often the frame rate is persistently crap even when nothing is being loaded.

[whole cloud]

well min is always below avg. So if avg is already crap, then min must be crappier 😄

[inland dew]

higher min = higher avg

[heavy vortex]

From a practical perspective I don't actually care much about the loading hitches, although ideally they go away.

But then I mostly play infantry where you're doing a lot less object loading.

[inland dew]

fly a jet 1000+ km/h low altitude high view distance 🫠

[heavy vortex]

yeah if I want to fly jets I don't play Arma :P

[whole cloud]

Annoying thing for hitches is, I need a repro to see it

[heavy vortex]

I did a frametime run on YAAB and it was actually a lot less spiky then I was expecting.

I figured there would be a bunch of ~100ms frames in there but it wasn't close.

[analog acorn]

Yes, but it's not always exceptional bad mins that are driving it. Sometimes the frame rate is a pretty flat line and it's some overall thing affecting all frames that's preventing the average from being higher, not spikes dragging it down

(urban combat)

[inland dew]

render, where gpu is cpu bound = crap frametimes

when cpu is waiting on something, gpu is also

main thread waiting on stuff from other threads
or stuff not done on other threads = queuing on main thread

[woven loom]

Well rate of fire is tied to frame rate so... 😛

But more seriously, having Arma run at >60 fps with an X3D processor is great! The minimums i don't notice unless it's a big drop (sometimes on texture/object loading for example). Stuttering is of course not great but Arma usually either has those big drops or just slows down on average during regular gameplay (in my experience). There is microstuttering too at times though.

FOV zoom can cause stuttering on slower systems for example

[whole cloud]

One of the YAAB dips, is the missile hitting a house, and checking for visibility to calculate explosion damage to surrounding units/objects.
The red part is ONE missile exploding.

But the dip only goes to a frametime of 24ms.

Most of the time of AI processing here, is also spent in visibility checks (the yellow block on the right).
If only that could be better..

[heavy vortex]

Well, if you can improve AI visibility check perf, I feel that'd solve a lot more practical issues than one YAAB frame :P

Is it doing raycasting checks to each explosion though? Because that sounds super bad.

[woven loom]

Also 24ms is a big dip 😅 (i guess it depends on baseline though)

[heavy vortex]

(think 20mm cannon fire)

[whole cloud]

On explosion, it does raycast to every object in explosion radius. To check if the path is obscured, so things behind walls don't get damage

[heavy vortex]

Ah, not for auditory checks though?

[whole cloud]

don't know what that means

[heavy vortex]

I mean whether AIs can hear the explosion.

[whole cloud]

Ah thats something different, I don't think it checks if objects are in the way

[heavy vortex]

hopefully not.

[whole cloud]

I can optimize the visibility checks, 90% of it is checking objects, that are not actually in the way.
But there are the last 10%, where the object actually does go through an object, and it has to load the objects mesh, and animate it to exactly check if at their current animation state it would hit it. And there, IF it actually loads the model from disk because its not ready yet, the optimization would blow up.
Happens only in 0.001% of times. But that messes up the whole thing

[hybrid cradle]

0.001% of frames is still 1 in 1000

[heavy vortex]

Nah, that's one in 100,000

[hybrid cradle]

...

true

idk i dropped out of university bc i couldnt math

[heavy vortex]

I do recall one pretty annoying long-frame issue in Antistasi. Only happens in localhost. If you fire a mortar from the back of a truck for the first time, it'll sometimes nuke itself because a long frame causes excessive rotation from addForce recoil.

Because it only happens in localhost, I figured it's probably due to all the AIs checking if they can hear the mortar shot.

[hybrid cradle]

Oh is that what causes that?

Had that tons

[heavy vortex]

I don't really have the tools to prove it, but I couldn't think of any other explanation that fitted the results.

It's fine if you don't have 100 AIs nearby or don't use the addForce on firing.

(or run DS rather than localhost)

[fickle geyser]

I feel like you could limit amount of objects it raycasts per frame

if the damage would be delayed a frame or 5 it shouldn't matter much but less spikes when stuff explodes

[empty goblet]

repro mission would be useful for that problem

[fickle geyser]

I still wasn't able to repro few missions where we had 16km sensors/radars on a helo/cup ship kill frames every few seconds

pretty sure that was related to raycasting too

[whole cloud]

Splitting stuff over multiple frames is quite some effort. Then only thing that does something like that is AI pathfinding.
maybe. I'll atleast try to find some way to make that better

[hybrid cradle]

ive had some instances where AI being above ground/sea has just killed frames tons but could never repro

[fickle geyser]

move to SQF, easy to do here

[heavy vortex]

I feel like it's gonna be an unfixable though, given that you're not gonna mess with PhysX internals.

Interesting part is "why is the frame so long", but long frames are never entirely avoidable.

We can probably just work around it by delaying the "recoil" for a frame or two.

[woven loom]

Dedmen if you're hunting for places where the FPS drops, maybe it's also good to use the ultra or extreme presets to force more intensive settings, but also use YAAB Standard viewdistance settings (press S) to limit the impact of far away objects and textures?
e.g. https://imgur.com/WeqHd6D

And fwiw i see regular excursions past 30ms, with some really big frametime spikes too
https://imgur.com/eWKDr2b

[analog acorn]

Related history: #perf_prof_branch message

[heavy vortex]

Ok this really is a long frame:

18:05:05 "Frame time 0.016"
18:05:05 "Frame time 0.017"
18:05:05 "Frame time 0.797"
18:05:05 "Frame time 0.025"
18:05:05 "Frame time 0.016"

[silk pewter]

wow

[heavy vortex]

(plonk down ~100 redfor AI in approx 1km radius. Fire mortar)

I guess I'll put it on the FT. Seems kinda excessive :P

[whole cloud]

That number is in seconds right? Oof.
I don't care much for a spike for a mortar gunner when shooting. But over half a second is too much

Doesn't need FT yet, I'll take a quick look tomorrow. If I can't easily fix it, then maybe ft

[heavy vortex]

Didn't figure it'd be that much. You get so much visual effect distraction that it covers it pretty well.

[heavy galleon]

I care, mortars are the best thing in this game.
(If you don't use the vanilla system that is)

[heavy vortex]

Ah wait, might be false report from debug console open/close.

[whole cloud]

Debug console opening also shouldn't freeze for half a second

[heavy vortex]

ADT does.

Actually worse on closing than opening. Not sure what it's doing there.

Yeah, I don't get anything detectable just from the mortar fire. Would need to be something else related.

[heavy vortex]

huh, I thought mortar shells were network objects:
18:39:02 Ref to nonnetwork object 1790554: shell.p3dSh_82mm_AMOS

[restive pilot]

I think it's the profile saving thing

Also if you have a breakpoint it will recompile your file to insert the breakpoint in the code (since it's a SQF breakpoint not hardware/os bp)

[restive pilot]

I think all non-controllable projectiles are local (except for mines I think)

[random isle]

None are it’s all local

[whole cloud]

plonked down ~100 opfor AI, in 50 meter radius around blufor in a mortar, fired mortar. FPS didn't even drop below 50

[silk pewter]

I wonder: was it "fire mortar at them?"

[woven loom]

That's what I thought 🤔

[inland dew]

imagine firing at something with a Cheetah or Tigris AA cannons, so fast and so many particles...
or A-164 Wipeout cannon...
FPS drops easily, if you have HDR standard (not low), anti-aliasing and particles of good quality

or just look at water for FPS to drop

or use Zamak MLR or M5 Sandstorm or M4 Scorcher/2S9 Sochor with cluster ammo
mortar is nothing, compared to that

[heavy vortex]

Yeah, I figured there would need to be a frame rate drop for the addForce to nuke the vehicle, but I can't detect one in practice. Just a false positive on ADT. The original replication case stopped working for me anyway so I don't have anything to boil down.

[random isle]

Vulcan fire usually brings my game to its niece

[hybrid cradle]

i think you meant knees

[random isle]

Err voice to text failed

[cobalt spire]

So, I have a question about the profiling branch. I never have used this before as an fyi, and ran into an issue earlier so, Are the profiling branch and the stable branch supposed to be compatible with one another? Like if server is running the profiling branch, are players who do not run said profiling branch supposed to be able to connect or not?

[heavy vortex]

Yes, although it doesn't always work out.

If you do have a clear compatibility issue then you should report it.

[cobalt spire]

Where do I report such an issue?

Because our server is running the profiling branch I think. (arma 3 version 2.16.152172) and clients with the stable branch (arma 3 version 2.16.151618) are unable to connect. Upon connecting the clients receive the following error: Bad version, server rejected connection. And I can only connect when switching the client to the profiling branch.

[heavy vortex]

Here is usually sufficient.

[cobalt spire]

is the information above sufficient or is there anything else needed for this?

[heavy vortex]

It's sufficient assuming that it's replicable.

[heavy galleon]

Arma 3 Console version 2.16.152172 x64 : port 2302
could connect with stable client, issue must be somewhere else

linux server*

(or steam didn't actually switch me to stable)

[cobalt spire]

standby, I will try to replicate the issue elsewhere.
For the server, I cannot say whether it is linux or windows as my server panel is a browser panel through TCA.

I do however have a windows machine (Windows Server 2019) available to me so I will try and replicate it on that.

[heavy galleon]

linux no issue

[cobalt spire]

Just tried to make sure on my Windows Server 2019, with server version 2.16.152172 and the client on 2.16.151618 I seem to connect. So I am unable to replicate the issue. I will continue to look at what I need to do to fix this.

[heavy vortex]

Do the server logs actually say 2.16.152172?

Because if that's Armahosts/opengroup then they don't update their perf/profiling version and it's probably ancient.

[cobalt spire]

yes the log said 2.16.152172 however we have managed to get our server off the profiling branch and are now able to connect. It only took us 2 days to reach our server host hence why I started to explore different routes for solutions.

[kindred radish]

Remove requiredBuild from server Cfg if it exists.
https://community.bistudio.com/wiki/Arma_3:_Server_Config_File

[ashen helm]

was there something done to the cursorObject return value? Now suddenly I get, for example, 3dae8f4b800# 3474161: vn_sign_fob_08.p3dvn_sign_fob_08 instead of 3dae8f4b800# 3474161: vn_sign_fob_08.p3d. Why does the file name show up twice?

[whole cloud]

Something was fixed for an object time that was missing it's reporting

[empty goblet]

1.54.133792 profiling & performance v7 server and client, windows, (linux server not yet)

• several more crash cases were fixed since v6
  https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[plain trout]

we had an issue last Friday where every person in audible distance crashed while playing affecting only perf players (7 out of 50) multiple times. The issue disappeared once everyone switched to vanilla. Unfortunately I was running with -nologs. Can a memory dump help you?

[whole cloud]

For crashes I always need mdmp, not the rpt. so nologs doesn't matter

[whole cloud]

https://cdn.discordapp.com/emojis/635256471675994132.gif?size=128&quality=lossless

[night mulch]

https://media.discordapp.net/attachments/694394891584667683/885935014058000384/image0.gif

[hallow lantern]

why fps are hidden?

[whole cloud]

To not tell people to expect +N fps, and then be disappointed when they don't get it, that's all preliminary still

[hallow lantern]

quite hard to evaluate the improvement, it's something about multithread improvements I suppose?

[whole cloud]

Multiple different improvements coming together

[autumn timber]

Need to figure out what happens at the 87th second in YAAB, since that's where the biggest change seems to be 😄

[whole cloud]

Black screen 😄

[hallow lantern]

was looking exactly at that time, it's the big difference but without numbers I can be looking at a 2-3 fps difference or 10-20 fps difference

[whole cloud]

Oh no that's 81.
Mh yeah don't actually know what specifically is there

[autumn timber]

2.18 Changelog

Changed:

• Improved empty screen rendering performance
• Improved soundless playback performance
• Improved no AI State Machine processing

[silk summit]

I believe those "heavier" optimizations are set for 2.20, like the 2x performance increase on the VR map, a 25% optimization for objects not placed on the scenario and a huge performance increase for shadows (when shadows are disabled) 😅 🤣

[restive pilot]

I think the benchmark pauses the game at that time (setAccTime 0) so probably a lot of simulation is suddenly skipping or something

Actually no that also makes no sense (the FPS drops again at the end) 😅

[whole cloud]

That could be the helicopter fly scene, where the camera is aimed mostly into the sky

[restive pilot]

Oh when you said black screen I thought it's near the end of the benchmark

[heavy vortex]

87 seconds is when the infantry truck appears. The rendering should be pretty thin because you're looking away from the town.

I guess there's still a fair amount of smoke from a couple of mortar rounds.

[whole cloud]

Some of the optimizations improve culling of objects that are around the camera, but not actually visible.
I assume that is having the big effect there.

[inland dew]

when do you think to do shadows faster like you said?

whatever it means

would also really be a thing

[pallid island]

[whole cloud]

I think part of that is already done with what I have

I think my benchmarks are with shadows enabled

[silk summit]

But the big rocks optimizations still not in there right? That would help more with stutters than overall FPSs anyway but still a great fix to have for 2.20 😉

[woven loom]

[woven loom]

BTW, there's this thing when the object's main body isn't in the frame but some part (like a tank barrel) is, and the part will pop in and out of existence with small camera movements. can get very distracting.

Is there a way to avoid stuff like that?

[inland dew]

that's a very tricky one actually

[north glade]

Hey guys can someone please tell me how do i get to the console logs of arma reforger?

[analog acorn]

This channel is for the Arma 3 Performance/Profiling branch. For Reforger, try channels in the Reforger category, like #reforger_questions or #reforger_troubleshooting . Make sure you have the right roles selected in #✔role-selection to see the channels you need.

[restive pilot]

I actually ran a test with roadway PhysX disabled and there was no significant improvement with that (at least not in YAAB)
It will only improve the stutters around those rocks
Anyway the fix is still not ready yet

[north glade]

oh okay thanks

[indigo anvil]

This can be considered an edge case: wireless earbud ran out of power during a longer game and the audio device disconnectes. Result: arma crash. This happens every time. Any way to gracefully resolve this?

[restive pilot]

It's not really tricky without the shadows
With shadows it does get tricky

[silk pewter]

ye, bigger bounding box but to what extent with shadows…!

[restive pilot]

BTW out of curiosity, does this happen all the time or only when the barrel is pointing to the side (or anywhere except the original, unanimated position)? If the latter it's because of a bug in model binarization where LOD bounding boxes are too small because they don't account for animations (model bounding boxes are correct tho)

[woven loom]

Uhh when I noticed it at least, it was to the side

I've also noticed it on other things, although I can't remember specific examples. Wait, actually I remember one, the lighthouse's light beams also get culled when the source is out of the frame

Which I guess is a bit different because the light beam isn't a part of the lighthouse (?)

But yeah I've noticed it a few times with objects

[restive pilot]

I guess so yeah.

[woven loom]

I guess I'll report when I see it happen again, I just sorta ignored it assuming it was an engine limitation

[autumn timber]

and a huge performance increase for shadows (when shadows are disabled) 😅 🤣
Damn, the irony is: this would have fit my changelog perfectly, would have been an actual, real change, and yet I forgot to include it 🥲

[light cargo]

which is that island? official bi terrain?

[whole cloud]

Big rocks are not a problem, they appear in none of my benchmarks. And in the kinds multiplayer missions I play, they also basically never appear.
So I'm not focusing on that.

I optimize the things I see. The things that affect the majority of situations.

[whole cloud]

Make a proper ViewGeometry LOD in the model.

[whole cloud]

For a couple years now I've been hunting some lag spike caused by AI.
I got closer today. AI takes a long time updating its targets. Well if it has lots of targets thats normal, but that would really need to be hundreds of targets to be as bad as what I'm seeing.. What are these targets?

The targets are 1500 props. Shovels, tables, lampposts, fence's... What?

Turns out, this is used for a keybind (Weapons / Next Empty Target, and Next Empty Target (Vehicle)), that is by default not bound to any key (It has one assigned in the Arma 2 key preset).
If you press that key, it searches through these static-object targets. And if there is one there (Which is also lockable by your current weapon) then it locks onto it.

I can't find how to actually trigger it. I can trigger it to search for objects, but it only accepts an object that was detected by a vehicle sensor which.. is none, because there are no sensors that detect buildings?

A majority of AI processing time, is spent tracking stuff that 99% of players never use?

[hybrid cradle]

How much is a majority?

[whole cloud]

There are only two cases where AI group processing time spikes higher than 0.2ms.
One is when it wants to start pathfinding and the grid is not ready yet.
And the other is this.
A group's 0.8ms update time, spending 0.75ms checking props around it.

Comments in code say this is for buildings that are pre-placed on the terrain.
But my sample has about 50 Land_Shovel_F in there. A shovel is not a building?
Why is every AI group, in this compound, keeping their own list of shovels :U

I can understand that you might want to know buildings nearby, you might want to target them for whatever.
But shovels, food cans, wooden tables, lampposts..

[inland dew]

so that's what could cause a good freeze/fps drop/frametime jump, when pressing the button, to cycle through targets, when in a jet/heli...

[whole cloud]

It would only check this huge list, when you press the special "Empty target" button

There are also a few other places that iterate over all targets. But then filter out all non-vehicles (all these buildings/props are not vehicles).
That could be it, checking many objects only to find out that 90+% of them are not vehicles, is probably not good

[inland dew]

well, i have it set so that it's same button for empty and not empty targets, because if a target is empty, i see no reason to not lock and destroy it or decide whether yes or no myself, but lock/show it anyway

[whole cloud]

It also tracks logics (modules placed in editor) even though you can never see or interact with them

[whole cloud]

That would do it then yeah

[fair ore]

Huh, is this related to how pressing the "next target" keybind causes a big frame drop a lot of the time?

[whole cloud]

Ref #perf_prof_branch message

[hybrid cradle]

groove just said that jenna

[fair ore]

Lfkfdnsjdjdns

me when I'm small

[whole cloud]

I have not seen that frame drop yet. Didn't see a way to reproduce it?
Is there a simple way to repro that?

[fair ore]

I'm cute

[fair ore]

I find it happens a lot if I'm in a jet and there's a good number of targets, say ten or so, and it fails to actually grab one

[hybrid cradle]

I can try find a repro later

[fair ore]

See if I can get a vanilla repro when I get home

[hybrid cradle]

Whilst Jenna sends me that spp 😠

[fair ore]

I'm at WORK

[inland dew]

assign empty to same button as next, maybe?

[hybrid cradle]

LATER IDIOT

[fair ore]

There's also a distinct possibility I bound the empty one

[inland dew]

lol, empty can also be non-vehicle

i guess infantry and buildings

but all sort of small props shouldn't be...

[whole cloud]

"Empty Target" is only "static" objects.
Buildings and "Thing" type objects, and Logics.
Stuff that doesn't move, and doesn't get driven and generally also isn't "enemy" ever

[fair ore]

what in the heck lol

Can anything even lock those

[whole cloud]

I don't know of anything that can actually lock such a thing

But surely there must be something, or else this old feature (from Arma 2 or pre-Arma2) would not be here.

[hybrid cradle]

dont think its even possible with current sensor config

[inland dew]

can lock a building with a cannon of attack heli with ai gunner

[hybrid cradle]

buildings sure

but soda cans nah fuck that

[whole cloud]

Yeah I got it. So "Next Empty Target (Vehicle)" iterates the list, when sitting in a jet.
The normal "Next Target" and "Lock Target" buttons don't

[hybrid cradle]

whilst on the topic of radar stuff
is there any chance to get commands like addRadarTarget and removeRadarTarget or something?

[whole cloud]

dunno

[fair ore]

I suspect I just bound it when going through my stuff

[inland dew]

but to lock a building, you can't simply cycle through dozens/hundreds of them, until finally ai is locked on a specific one
buildings are never in the targets list

you can lock a building only when directly/specifically looking at it and then locking/assigning it that way only

[fair ore]

Pebcak etc

[whole cloud]

And 3000 road cones are not causing a noticeable lag spike for me

[fair ore]

surely you will not regret purchasing 3000 road cones

[inland dew]

in a2, you could lock onto any building by pressing reveal target (right mouse button)
works same in A3 (keyboard T), when looking at specific building

but in a2, when locking onto a building, you also had the square on the building, like on vehicles in A3 and the name of the building. so it was similar to a vehicle logic

[whole cloud]

with 3000 invisible road cones, my game runs at 192fps.
If AI were not trying to process them as targets, I'd be at 230fps. And that is only with my player placed down.
This effect would multiply for multiple AI groups each tracking the objects on their own.

[whole cloud]

Oh..
Well. I cannot lock onto anything, because the thing needs to be "detected by sensor" which is a new feature that came with jets DLC.
And there is no sensor that detects a building..
That may have broken this feature.

[inland dew]

well, spawn an attack heli, with ai gunner, select manual fire and select cannon, then freelook at a building and press T. ai gunner will lock onto the building, then you fire cannon at the building

[whole cloud]

Mh doesn't work for me.
When I spam T, in non-manual-fire, I see the crosshair target show up for a split second on the building, but AI gunner doesn't lock onto it
And in manual-fire mode, it doesn't seem to do anything

[inland dew]

i do this for buildings to destroy a specific one by just shooting, while ai aiming at it

or when i visually see enemy infantry and look at it and press T for ai to keep aim at it and me just shooting

[woven loom]

huh can't remember locking on to a building 🤔
is there a difference between map objects and placed buildings? because map objects are simple objects right?

[whole cloud]

ok I got something but its weird.

With laser marker enabled, I can switch to ASRAAM, and then with T move the laser marker.
And then when switching to cannon, the cannon will aim at the laser marker

But when I switch back to cannon, pressing T again does nothing

[woven loom]

yeah i think that happens with tanks as well (like the NATO tank destroyer, if you use the laser pointer it will aim at it)

[inland dew]

oh, yeah, forgot about laser
but when laser, weapon doesn't matter

[whole cloud]

It reliably puts the laser onto the building.
But that doesn't actually use this target list, it just checks whats under your cursor, it doesn't need to go through known targets.
And the "Next Target" button does nothing.

I think thats not it

[inland dew]

but targeting/revealing solo soldier, while looking at it and pressing T, no need for laser

[whole cloud]

But soldiers are dynamic targets. So thats something different

[inland dew]

well, buildings were never visible/selectable in targets list (press 2)

[whole cloud]

yeah that only shows dynamic targets.

Only engine use I can find is literally the "Next Empty target" keys, and I can't find any sensors that detect these, where these keys would work.

[inland dew]

to make ai aim at a building, one always have to look at a specific one and press T
otherwise no other way

it's similar to like when you command ai and tell it to run to a bush, tree or building by pressing button while looking at this bush, tree or building

so visual sensor only

[median belfry]

Are agents doing the same?

[whole cloud]

if they do targeting, yes

[woven loom]

Could it be for line of sight checks? Like sensors also check for LOS

[whole cloud]

no

[woven loom]

Okay 🤔

Out of ideas then haha

[whole cloud]

Well it does check LOS to these static objects, thats why the perf hit is so shitty.
But it doesn't use these lists for that

[woven loom]

Time to remove it and find out what breaks 😛

[whole cloud]

Can you also do that with the "Next empty target" key in Arma2 ?

I suspect T simply grabs object under cursor, but "Next target" definitely would use that list

[woven loom]

I wonder if there was some planned feature that never got implemented and the sensor rework devs just forgot to remove the check 🤔

[inland dew]

there is no such thing in a2
there is only reveal/lock (RMB), so T in A3
and next target (Tab), so R in A3

[whole cloud]

Yeah I'll probably completely remove these objects on profiling.
And on stable keep it at a "everything bigger than 3meters is allowed"

Actually testing it again now.. The next weird thing.
Buildings, like actual terrain buildings, aren't even in that list.
I have PowerPole's, StreetLamps, other Lamps, Navigation lights (Probably from stratis airfield) and the Airport Tower at stratis airfield.
But I don't have any of the other buildings 🤔

But all the comments in the code here say that this is for buildings. But there are basically no buildings in there.
But this code that adds the objects in the first place was rewritten some time in Arma 3. Maybe they forgot that buildings should be in there 🤔

It is quite literally only lamp posts and props. The Airport tower is probably in there, because it has lamps on it.

[whole cloud]

Ah maybe next target and next empty target was on one button back then

[woven loom]

Could it be to shoot out lights???

[whole cloud]

Ai doesn't do that

[woven loom]

Oh I know, but maybe they wanted it to? Isn't it a thing in Reforger? (Or is that just players)