#perf_prof_branch

1 messages · Page 46 of 1

whole cloud
#

u serious right now?

restive pilot
#

oh that's from the previous changelog meowsweats

#

didn't notice...

#

btw why does allMissionObjects's performance vary from terrain to terrain (no mission objects placed)?
is it because it still has to search all "grids"?

e.g. VR: 0.08 ms; Stratis: 1.4 ms; Altis: 23 ms... meowsweats

whole cloud
#

all grids, the more grids the bigger

#

Also it iterates over ALL objects, and filters out terrain objects

#

more terrain objects, more work

#

More being filtered out also makes the MT stuff more efficient percentage wise

restive pilot
#

so I guess that's out then... notlikemeowcry

if I do entities [[], [], true, false] + allSimpleObjects [] + (agents apply {agent _x}), are there any other important objects left? (those created during mission, excluding projectiles and footprints and flies that are returned by allMissionObjects, since I don't need them anyway)

vivid rune
#

My idea for MT:
MT: many threads collect all of object for each "grid" parallel. In this phase also the sort out of terrain objects is done.
ST: copying together the single "grid lists".

whole cloud
#

That is what the code already does

vivid rune
#

nice 😄

restive pilot
vivid rune
#

Is the copying together performed after all MTs have finished or puts it the data from finished grids already in while some grids aren't finished?

whole cloud
#

already

vivid rune
#

second nice 😄

#

The only thing that i have in mind for maybe speed up this copying is some cascading of the copying.

whole cloud
#

The end step is converting raw data into script values, and that only works in main thread.
The worker threads are shoveling data to mainthread and it eats it up

vivid rune
#

This "copying together the grid lists"-thread is a worker or the main thread?

whole cloud
#

Workers put into a lockless queue and main thread takes out

#

Write faster oof

vivid rune
#

Mh. I do not have all the information that you have, so I only can provide my 2 cent:

  • main thread is doing more than one copying/move to become all data that it needs to create the list of objects, if I understand you right
  • I would say the creation of the whole list can be done by a worker thread and at the end it informs the main thread "here is the whole list (sorted)". Main thread has then only to converting whole data to script values (maybe this can also be done by worker thread).
vivid rune
#

Or another idea: What if the "grid-worker threads" creates already "script value-lists"?

whole cloud
#

maybe this can also be done by worker thread
no, only mainthread can do script stuff
main thread is doing more than one copying/move to become all data that it needs to create the list of objects
no. It gets a object, and converts it into a script object value.

patent dome
#

Used v8, set cpucount to 6 on our 8 core machine and it still froze the server box. Didn't freeze when just one person was on the server, but with 15 players it froze in about 15-20 mins

novel ravine
#

Have not done this yet, but will try tonight. Just experienced another freeze/crash now. However there is something I wanted to point out. After a full machine freeze/crash, when I launch Arma through a .bat file I use after rebooting the machine, within 5-10min or so, the A3 server magically restarts itself without being told to and sits at 0% CPU Usage as a Background Process and never gets going -- basically what I had originally DM'd you about. So I have to kill the process and relaunch it manually again. The machine isn't frozen when this happens, but after using v7 this happens whenever I reboot the whole machine now, there is always a secondary random restart from A3 within a few minutes before things stabilize again.

I doubt this info is necessarily helpful, but it's something to note I think. Just really fucking weird. Going to try messing with the latest hotfixes + perf builds as well as the parameters you mentioned tonight though.

patent dome
#

Hmm... We haven't experienced that behaviour after the hardware restart.

heavy galleon
#

Probably asked already, but wasn't able to find it in the time I allocated into the search.
The steam version, the binaries used by default are perf, or prof?
Is there performance increase for client (for MP), or only noticeable on server side?

hushed seal
heavy galleon
#

Thanks for the info! I'll look around and test it a little, if it suits my usecase

hushed seal
#

The improvements are definitely noticeable on the server side, client side it's harder to tell.

heavy galleon
#

The server part is awesome and :praisedeadman: emote is needed.
But I want to squeeze as much performance as I can from client now.
(Sadly the YAAB has lower score for me) (Not sadly YAAB does not represent my usuall use, will have to wait till night to test it on wog)

wispy shuttle
#

Same for us, we had a 106 people TvT operations yesterday and everything went really fine! (perf v9)

silk summit
#

I am actually happier because most stutters got mitigated (For me). I prefer lower FPS but no stutters rather than high FPS with periodic/constant stutters

#

The only stutter left standing for me is related to the big stones that cause up to a second stutter when driving/shooting near them... but devs won't tackle that at this point even thou it can cause huge headaches when using some workshop compositions and worshop maps that use those stones indiscriminately.

#

Funny thing is that SOG Prairy Fire devs fixed that issue with their optimized stones yet on vanilla the issue remains

#

Damn I wish SOG was vanilla and Arma 3 2035 the CDLC and not the other way around 😆

vagrant zodiac
#

A couple of questions regarding the latest updates:

  • Since the 2.08 hot fix, is there any expected behaviour change with regards to AI?
    A game mode we use a lot to practice skills now seems to have the AI machine-gun-throwing hand grenades at us. (No other mod updates since this change.)
    I can't see how this would have changed with 2.08 hotfix so we're scratching our heads but thought it worth asking.

  • Since around the v7 profiling update, one of our scripts we use for target hit detection now has issues. We have some bespoke targets with names selections for each 'zone' of the target. We then use hitPart to detect what part of the target has been hit so we can then generate a score system etc.
    Since roughly the v7 profiling branch, we seem to have localisation issues with this. So now we seem to have things like hits being counted per hit multiplied by the number of clients connected.
    I think there were some updates to bullet/missile behaviour a few updates back. Could this have caused this?
    Could we also get this kind of issue if some of us are running the profiling branch and others are running the main/stable branch?

silk summit
vagrant zodiac
#

I'll try to get the script writer to post. (As I have been for some time...)
Unfortunately, my scripting skills are very poor. I just make mods.

hexed gorge
#

The AI optimizations to increase clients performance when not using reveal, target, and others like functions remotely, was abandoned?

silk summit
silk summit
#

Most meaningful features and fixes take time but are not dropped. Only my feature requests are dropped, most other just get postponed 😉

hexed gorge
#

😅

whole cloud
whole cloud
charred lava
#

New profiling branch: Heli can crash upside down and be fine; explodes sliding on its skids in standard flight model meowsweats

novel ravine
#

Still need to change -cpuCount (I am assuming when I use this, I should not also be using -enableHT?) But I believe this issue is definitely caused by CPU Usage capping out and getting stuck. We just had a crash now and this is what my infiSTAR metrics show for CPU usage: https://i.imgur.com/K0M91y0.jpg

I will update -cpuCount on my next scheduled restart tonight just to be able to report back. Meanwhile, my launch parameters currently include:

-limitFPS=100 -enableHT -hugePages -bandwidthAlg=2 -autoinit

I was using a malloc parameter until our previous crash and tried removing that to see if it made any difference.

novel ravine
#

Got -cpuCount set now. I assume a crash will happen again within the next 72hrs, but will follow up if/when it does.

warped rain
#

whats the consensus on basic.cfg nowadays?

fickle geyser
#

What's vmVer in proflining trace?

# 
 wSimu; 0.30574; 316.70587;""
  memLo; 0.33362; 0.00002;""
  vmVer; 0.33499; 282.77429;""
#

playing an OP with my unit and having terrible stutter every few moments.

#

it was only me having it, switching to stable seems to have fixed it.

whole cloud
#

Hm never saw vmVer before

boreal wigeon
#

Anyway to test network related issues? We had an operation yesterday running with about 60+ players. After about an hour server was having desync and froze, however server-side fps was still around 30-50. I checked the network with #monitor and in was around 0-1 Kbps. While out was around 1500-3000 kbps.

hushed seal
#

And the way test is pretty much trial and error while graphing the #monitords output

boreal wigeon
# hushed seal What's your basic.cfg?

It's very minimalist right now, as the server (software) should know best, at least that's what I hoped for.

MinBandwidth            = 1000000000;
MaxMsgSend              = 640;
MaxSizeGuaranteed       = 512;        
MaxSizeNonguaranteed    = 256;    
MaxCustomFileSize       = 0;

I used to set MinErrorToSend and MinErrorToSendNear before as well to x2 what is default.

#

However I had fpsLimit to 300, I've dialed it down to 150. As maybe it's doing too much network related tasks in a second when nothing heavy is going on?

edgy comet
#

minBandwidth may be too high

boreal wigeon
#

Well I do have a 1Gbs connection. And it has been said before in this channel that it should be put to a number that it couldn't reach.

gritty wasp
#

that it couldn't reach.
No. You messed MIN desctiption with MAX.
Min said: "Bandwidth the server is guaranteed to have (in bps)."
1Gb plan commonly do not guaranteed 100% of time
also second part
"Increasing it to too optimistic values can increase lag and CPU load"

hushed seal
edgy comet
#

Our server also has a 1 Gbps connection. Had lag issues setting minBandwidth to over 100 Mbps or so. Have it set to 55 Mbps currently.

gritty wasp
#

We have 80Mbit while having 1Gb plan. One day I tried recommended 80% of plan (800Mbit) and everything fall apart for us. So I revert it to 80.
Actual traffic around 200Mbits but varies.

boreal wigeon
#

I'll try lowering the MinBandwidth then to around 80% then. Ill also look at your messages @hushed seal thanks for the information.

whole cloud
whole cloud
boreal wigeon
#

I heard that was used by the BI servers 🤔, BI knows best right? 😅

gritty wasp
whole cloud
#

When you set any realistic minBandwidth for your server, Arma will basically never reach it anyway.
Whether its 100mbit or 1000mbit

whole cloud
gritty wasp
whole cloud
#

If you look on map screen, you'll see that each player has a bandwidth

#

Arma negotiates bandwidth with every player individually

#

MinBandwidth/playercount is just the upper limit per player

kindred radish
#

Couple of weeks after we wiped our "DayZ style" Arma 3 server and it's still running between 80-90 FPS, no matter how many players are online.
Unlike before this major changes the constructions created by players also seem to have no affect on the server FPS. 👍

[2022-03-27 13:53:35] [info] AEOG-ServerMetrics: Build:149115 | Runtime:113min | Players:43 | AI:37 | allMissionObjects:5740 | AllVehicles:372 | ObjSimulated:1882 | AllAliveEnt:885 | activeSQFScripts:14 | FPS:85)
[2022-03-27 13:55:35] [info] AEOG-ServerMetrics: Build:149115 | Runtime:115min | Players:43 | AI:37 | allMissionObjects:5424 | AllVehicles:364 | ObjSimulated:1673 | AllAliveEnt:862 | activeSQFScripts:13 | FPS:86)
[2022-03-27 13:57:35] [info] AEOG-ServerMetrics: Build:149115 | Runtime:117min | Players:42 | AI:39 | allMissionObjects:5634 | AllVehicles:349 | ObjSimulated:1736 | AllAliveEnt:862 | activeSQFScripts:18 | FPS:85)

whole cloud
#

Heh.
While testing some other stuff I just noticed smth.
While my Arma is running at about 100 fps on a empty map, 45% of frametime is used for the engine to check if it can use its spare time to preload some textures/models around the player.
It spends 45% of frametime (at ~100 fps) with checking, and the results are empty every time 😄
I wonder what happens if I just take that out notlikemeow

100fps + ~50% on top. hoo boi

#

10% of frametime is spent for drawing the mouse cursor (crosshair) notlikemeow

#

That 50% thing should theoretically get me from 100fps to 200 :kek: (This thing only shows up when frame times are high and Arma has time to waste, so it doesn't matter at all for low end fps)

silk pewter
#

can't wait next perf!!1!

whole cloud
#

I'll probably not release that

short flame
#

Can I get a quick rundown on what exactly the performance-profiling branch is and what it does? I've been told to try it without any explanation.

whole cloud
#

performance branch is a early-access beta-ish build.
With early bugfixes and improvements, and experimental features that we don't deem stable enough to put on.. stable which are mainly performance tweaks.

short flame
#

I get about the same results in YAAB with both stable and perf-prof (about 2 fps more in stable), except if I have the perf-prof installed and try to run stable I get notably worse results, by like 6 - 8 fps.

#

Or rather got, when I ran it before the 2.08 update.

whole cloud
#

perf doesn't care about high end fps.
Main target is to raise the lower fps.
Bring players from 5 fps to 30.
I don't care if we make players run 110 fps instead of 120. So if your Arma runs well, then yeah it might run ever so slightly worse

short flame
#

Good point, and a good goal! I can't even tell if my fps is higher than 30 without looking at my top-left corner, and I've got a 12600K hitting 36 - 68 fps on Tanoa in a heavy mission so 😄

#

Basically double of what my old i5 3570 got.

gritty wasp
whole cloud
#

yes

gritty wasp
#

then why it will not be released?

whole cloud
#

Because its there for a reason, and it provides no useful benefit besides maybe Groove's lust for more high-end fps

inland dew
# whole cloud Because its there for a reason, and it provides no useful benefit besides maybe ...

You're so so wrong man.
Having 100 FPS with almost no view distance, just for benchmarking purpose, no really ised graphics settings, it doesn't mean anything.

When FPS drops already by 10 or more with only like 1000 m view distance, at 3000-4000 m, graphics settings one uses to really play, FPS is like 2 times lower.
And from 2 times lower FPS that I have with settings I play, remove these >10 FPS lost from 2.04 to 2.08 and it's considerable then

#

So please, stop talking about high end FPS, since it's only standard video settings and only like 1000 m view distance.

FPS seems to be so high, only due to settings used to benchmark.
It isn't high at all when playing at usual 24/7 settings in Warlords or KotH

feral harness
#

can one even see 70 vs 100fps diff on A3

inland dew
#

If I test with 500 m view distance, you will say I have FPS too high. So what if it's only 500 m view distance...

inland dew
feral harness
#

69?

silk pewter
rain moth
#

with a good monitor there's a huge difference between 60 and 120

inland dew
#

Also saying that it's only me losing FPS with 2 last patches is completely not true.
Anybody can test and see yhat FPS is lower.
People with lower end hardware lose not less FPS than me with better hardware

inland dew
#

Take a notebook or a PC with a 5-7 years old hardware and see for yourself that FPS is lower as well on them from patch 2.04 to 2.06 and now 2.08

#

Why should I be happy with 43 FPS at settings I use to play if it was close to 60 before?

edgy comet
#

Depending on the area, 100+- FPS isn't too hard to achieve on non-modded servers without AI, even with a high player count. Not maxed graphics, but still.

restive pilot
#

I think the problem might be that the current performance improvements were tested on the same PC (or maybe LAN) and only one or two clients
so the performance optimizations do not really apply to "real world" cases, e.g. MP games with lots of clients at various framerates, high pings, etc. and the game is bottlenecking in other places that were not visible during internal profiling.

whole cloud
#

When with "current" you mean before I started with my stuff then maybe

#

Most of my stuff comes from real world profiling data

empty goblet
restive pilot
#

do you guys actually do profiling on official servers?

empty goblet
#

nope, that ain't what i do for last 9 years (longer if we go pre A3)... sigh

#

dd isn't the first nor (hopefully) last programmer doing server/mp performance and other improvement over the lifetime of A2/A3 engine

restive pilot
#

isn't it possible to use one of official servers as Dedmen's guinea pig for a few days?! 😛
he might find something interesting...

empty goblet
#

he gets snapshots from me all the time ...

restive pilot
#

but those are from the server alone right? maybe something could be going wrong on the client side?

kindred radish
#

Anyone else noticed more frequent player kicks due to "Wrong signature" ?

silk pewter
#

sign better!

hexed gorge
vale shoal
#

he must be a doctor 😄

random isle
fickle geyser
#

I don't think so.

restive pilot
#

doctors write like that all around the world

random isle
#

there was funny video where it was like just a line that was meaning different things 😄

autumn timber
#

Weird. I was trying to add a 👆 reaction to Leopard's post but Discord keeps removing it

#

...and shakes the whole screen while doing so (a strange easter egg?)

#

Anyway, what he said 😄

whole cloud
autumn timber
#

Oh well ¯_(ツ)_/¯
(I don't actually even remember talking to leopard but if he feels that way towards me that's his choice)

kindred radish
#

Server is running for 49 minutes and there are already 49 kicks due to wrong signature.
Even longtime players who never had this kicks that often are complaining now 🤷‍♂️

feral harness
#

what does rpt / server log say for kicks?

hexed gorge
#

@kindred radish may be back to stable to check if kicks continue?

pale mica
#

I've been noticing users getting kicked because of that, it includes both vanilla .pbos and mod .pbos

#

I can't pinpoint exactly when it started, but it was after 2.06

#

and it also happens with and without the perf branch

#

My hunch is that it's linked with some issue with the users's network

kindred radish
#

3 hours mission runtime and i count 60 kicks due wrong signature in the server logs.

21:02:50 Player Reaper: Wrong signature for file C:\Program Files (x86)\Steam\steamapps\workshop\content\107410\843593391\addons\rhsgref_c_canoe.pbo
21:04:06 Player Seth: Wrong signature for file E:\Games\Steam\steamapps\common\Arma 3\!Workshop\@3CB BAF Vehicles\addons\uk3cb_baf_vehicles_bulldog.pbo
21:04:18 Player .MajorDV: Wrong signature for file a3\vegetation_f_enoch\bush\b_rosacanina_1s.p3d
21:04:57 Player Reaper: Wrong signature for file addons\map_altis_data_layers_00_01.pbo
21:05:03 Player Bigsz: Wrong signature for file addons\soft_f_gamma.pbo
21:06:14 Player Safal: Wrong signature for file a3\supplies_f_heli\cargonets\cargonet_01_ammo_f.p3d
....
whole cloud
#

Which day did that start on?
We didn't touch any of that 🫂

restive pilot
#

I used to get those a lot, but I'm pretty sure that was related to my connection (or maybe server cfg thing's fault)
anyway, after I switched to a better network it went away thonk

pale mica
#

I was looking over this, and the Wiki says that 131072 is the default

#

Is it just me or this whole configuration file is confusing?

whole cloud
#

wiki is wrong

pale mica
#

What is the point of the "MaxBandwidth"? The wiki says: Bandwidth the server is guaranteed to never have (in bps). This value helps the server to estimate bandwidth available.

whole cloud
whole cloud
pale mica
#

why isn't it hardcoded to 99999999999999999999?

#

ah

whole cloud
whole cloud
#

Can someone tell me if loading OldMan is broken on prof?

vivid rune
#

V.09 perf: First try to resume Mission but cannot because save game is incompatible so i restarted campaign.
Loading happened to 50% then fast to 5-10% at the end and halts there. The "Loading wheel" at the left corner is spinning and "Messages of the Day" are still switching with fades to black. ESC don't react.
Windows 10 64bit, RAM occupied: 4075,5MB, CPU 0,4%
Second try with prof: Same. RAM occ.:4153,1MB, CPU around 1%

whole cloud
#

Thanks o7

whole cloud
#

2.08.149174 new PROFILING branch with PERFORMANCE binaries, v10, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Inventory item lists are now sorted - https://feedback.bistudio.com/T163668
  • Fixed: Unable to click the bottom-most item in the inventory if a scrollbar is present - https://feedback.bistudio.com/T163667
  • Tweaked: Inventory UI listboxes now have the class names of the items inside lbData (previously only magazines had them)
  • Fixed: Script compile errors due to missing closing brace were not logged to RPT
  • Fixed: exitWith potentially exiting the whole script (instead of just one layer) if the script also tried to compile a script file with a missing closing brace in it
  • Fixed: AI would calculate a path even if "PATH" is disabled via disableAI - https://feedback.bistudio.com/T160446

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

whole cloud
fickle geyser
#

I guess I should PR change to use lbData

whole cloud
#

it uses lbData.
But if lbData returns "" (which now it doesn't anymore because of that change right under it that you might notice is not ordered correctly because its between the fixed argh) it had a fallback that doesn't work anymore

restive pilot
#

(also I pinged the wrong message...)

#

no nvm, it wasn't a bug meowsweats

whole cloud
#

:popcat:

vagrant zodiac
#

I’m really not very good at scripting but would guess from the changelog that this won’t affect/improve the hitPart issue we’ve been seeing in some of our script localisation?

vivid rune
#

Mh maybe something that i doesn't notice before:
If you put many items in your character that the "overall load indicator" (withe rectangle on the bottom of your inventory is full) than you cannot put more things from e.g. the ground by right click. But you can by drag and drop put more things in the inventory.
I think it should be able to overload because if you do this, the character is slowly.
So the right click thing is buggy or is this a notice for the player: "Hey the limit is reached, from here on it is your responsibility"?
The problem is, some mod (ACE+CBA) switch of this slowly walking and then this behavior has no reason anymore. Idk how they mange this but maybe there is a indicator to switch this right click limit off if ACE+CBA switch off this slowly walking?

  • start VR with soldier with bergen backpack
  • put in bergen many rockets as you can
  • if overloaded, character walks slowly and cannot run
  • try to put more thing by right click in inventory ->doesn't work
  • try to put more thing in inventory by drag and drop -> work
wanton thorn
#

Hey - regarding last update today -waypoints are no longer working in Editor-

Yesterday it was working perfectly, today I just installed last update (small one through steam)
Now if when I set waypoints for units they never move when you run the scenario all units affected , even loaded previous scenarios - Units don't move

#

so if you set a waypoint for tank or vehicle , when you run the scenario you expect it to move but that is not happening anymore

restive pilot
#

probably related to this fix:

  • Fixed: AI would calculate a path even if "PATH" is disabled via disableAI
night mulch
whole cloud
#

I hope, yeah the hitpart stuff is not listed

vivid rune
whole cloud
#

They should only not move if Path is disabled

#

eee

#

fix tomorrow then

#

So in SP editor, if you place unit with WP, and preview it doesn't follow WP?
Or only in MP(but self hosted works)

whole cloud
#

Actually maybe not hitPart

#

So much work for tomorrow 🫂

spiral pond
#

hm units not moving here any more too

#

thought it was something from the mission or mods, but maybe from the perf update

#

any way to verify/confirm than to switch branch?

restive pilot
spiral pond
#

ty

whole cloud
vivid rune
whole cloud
#

Okey besides the AI path change all other things that could be it were tested for dev branch, and they might only happen in remote MP.
So it has to be that AI path somehow.
But PATH should not be disabled for normal units.... Wuht

spiral pond
#

tested in SP and hosted MP

#

v9 works fine. so must be something from today

restive pilot
restive pilot
#

tho it makes no sense since CBA doesn't have anything related to AI pathing thonk

whole cloud
vivid rune
#

Vanilla guy doesn't move in SP

restive pilot
#

maybe some race condition thonk

whole cloud
#

The code is literally
If disabled(path) then return false
Right before AI would plan a path.
Well that's gonna be fun tomorrow

spiral pond
#

no CBA here

empty goblet
whole cloud
#

Worst case I just take the fix out again and leave it unfixed.. it's just minor performance thing anyway

vivid rune
#

I tested "Old man" (it starts) 🙂 And i drived a little around and then a jet fly by. meowhuh
This is not this recorded path for vehicle, right?

restive pilot
whole cloud
#

Why?

#

It's so unimportant

restive pilot
#

I use my own path finding for my AI 😛

autumn timber
hexed gorge
#

if (true) then {true} else {false};

rain moth
#

you forgot the true == true

mellow lark
#

or boolean method() { return true;}

autumn timber
#

The code was supposed to always return false. Mine does but yours always return true 😛
@hexed gorge @mellow lark

mellow lark
autumn timber
feral quiver
#

I noticed something was off when simplex support helicopters stopped working.

fickle geyser
whole cloud
#

remind tomorrow

vivid rune
#

I tried YAAB with v10. The quietest benchmark run ever. But my FPS doesn't go up. -.-

kindred radish
#

Not sure if anyone else has this problems but it looks like since the newest prof build there might be a super weird problem with the inventory.
https://gyazo.com/359fa1322aacfc24c519c9d74ed806b4 (Server v9/Client last prof build)

Vest, Uniform, Backpack seems to be not visible in the inventory while getUnitloadOut shows it properly and the corpse also visually shows the gear properly.

[[],[],["hgun_P07_blk_F","","","",["30Rnd_9x21_Mag",30],[],""],["U_B_GhillieSuit",[["Exile_Item_InstaDoc",2,1],["DDR_Item_Dressing",3,1],["DDR_Item_Firstaid",1,1],["Exile_Item_PlasticBottleFreshWater",1,1],["DDR_Item_Bread",1,1],["DDR_Item_Antibiotic",1,1]]],["UK3CB_BAF_V_Osprey_Holster",[["DDR_Item_Bloodbag",1,1],["DDR_Item_Cereal_Crunchin",1,1],["DDR_Item_Vitamins",1,1],["Exile_Item_MacasCheese_Cooked",1,1],["AeoG_Item_Defibrillator",1,4],["20Rnd_762x51_Mag",4,20],["20Rnd_762x51_Mag",1,3]]],["B_Bergen_dgtl_F",[["DDR_Item_Yummy_Juice",2,1],["SM_Zombz_Item_Heart",1,1]]],"rhsusf_mich_bare_norotos_arc_alt","G_Balaclava_oli",["Rangefinder","","","",[],[],""],["ItemMap","ItemGPS","","ItemCompass","Exile_Item_XM8",""]]
#

Gonna restart the server and try with v10 on the server nootlikethis

kindred radish
#

Edit: Client/Server on v10 and the same problem for me in this test.
https://gyazo.com/f5cea2d6192913a0758a1d85f14dbc1e
You can clearly see the corpse has a uniform but the inventory doesn't show it.

[[],[],["Exile_Weapon_Taurus","","","",["Exile_Magazine_6Rnd_45ACP",6],[],""],["TRYK_U_denim_hood_3c",[["DDR_Item_Redgull",1,1],["Exile_Item_CockONut",1,1],["Exile_Item_Bandage",2,1],["Exile_Magazine_6Rnd_45ACP",1,6],["DDR_Item_Drugs_Cocaine",1,1]]],["rhssaf_vest_md99_md2camo_rifleman_radio",[["rhsusf_100Rnd_762x51_m61_ap",2,100],["rhsusf_mag_15Rnd_9x19_FMJ",1,15]]],["TRYK_B_Carryall_blk",[]],"H_Booniehat_oli","G_Spectacles",[],["ItemMap","","","ItemCompass","Exile_Item_XM8",""]]

Another dead player and same thing -> https://gyazo.com/4f34dd5250f9738e62deda9a04ee5edb
Same thing for dead AI's inventory, so it seems to be consistent for us.

Edit 2: I reverted back to stable (Client) and i can see the gear again inside the inventory UI.

fair swan
#

I am experiencing the issue of all AI units not moving, which I see was already confirmed by others... but I also noticed the enemy AI inside warlords sectors disappearing for me when I choose to attack a sector. The garrison will despawn when opposition is present. This did not happen until after the update. Duplicated with MCC by placing enemy near sector.

https://streamable.com/51uxkm

plain trout
kindred radish
plain trout
#

Nope. Big collection of mods. The uniforms are vsm. We also use ace/rhs and a ~few~ others

vivid rune
plain trout
#

I also couldn’t use my grenades

#

Looks like it’s not really playable until the next release

vivid rune
plain trout
#

Yes. It’s a private modpack that was updated on Saturday and we have two people running performance branch that had the issue

vivid rune
plain trout
#

😄

whole cloud
#

IsAIDisabled vs DisabledAI meowfacepalm

kindred radish
whole cloud
#

yes also got other reports about it

#

will go right next

#

Yep got the inventory thing

#

containers were mixed between items and magazines and they got lost in some refactoring
I'll probably instead just put them right at the top

#

either top or bottom.
I think having containers at top makes more sense than them being between items and magazines at the bottom

feral harness
whole cloud
#

My coloring was broken when I wrote that I didn't care enough to fix it

feral harness
#

the moral of the story… let’s use c++ casts

silk pewter
#

game engines should be in Java *

rain moth
#

allowing non boolean statements as conditions was a mistake 😄

feral harness
#

boolean definition was wrong!

silk pewter
#

boolean definition is innocent!

whole cloud
#

The HitPart firing for remote projectiles will also be fixed

whole cloud
restive pilot
whole cloud
fickle geyser
#

How the heck it is lagging my game then.

#

And only on profiling blobsweats

#

I need to try again on same mission.

#

Could non default memory allocator affect it? (CMA)

whole cloud
#

yes

inland dew
fickle geyser
#

It was literally after starting the game.

#

And goes away after switching to stable as I said earlier.

fickle geyser
#

And ofc I can't repro the issue when being alone on the server blobunamused

vale shoal
#

I just want to go back to this issue. Also limiting the network use of Nextcloud to 250 Mbit (around 1/4 of the servers capacity) will bring the CPU usage of Arma 3 to a maximum.

whole cloud
#

can you maybe run profiling build and do a diag_captureFrame to show where the cpu time is spent at?

#

2.08.149187 new PROFILING branch with PERFORMANCE binaries, v11, server and client, windows 32/64-bit, linux server 32/64-bit

  • Fixed: AI units not following waypoints
  • Fixed: Uniform/Vest/Backpack containers not listed in dead unit inventories
  • Fixed: Entity HitPart event would fire for remote projectiles

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

whole cloud
whole cloud
#

I found a better solution for the upside down helicopters that both prevents random explosion, and keeps explosion when you upside down crash.
for next prof then

vale shoal
#

@whole cloud is diag_captureFrameToFile also work? And how much frames should I set?

whole cloud
#

yes, just one or two

#

captureSlowFrame might work better for you

#

on server all is write to file

empty goblet
fickle geyser
#

captureSlowFrame it works on dedi? I never got it to capture anything.

whole cloud
#

Well i hope it does

hushed seal
#

is there a big performance impact running profiling vs perf exe?
It would be nice to know whats causing low server FPS but don't want to risk having even lower FPS for the whole OP

empty goblet
#

by default profiling branch uses performance binary

#

profiling binary is slightly slower because it uses debug layer

hushed seal
restive pilot
#

depends. 5-10% iirc

heavy vortex
#

I don't know if there's a list of what the codes in the profiling output refer to. I guessed some of them but others are quite opaque.

vivid rune
#

V11: if optic in first slot in vest or suit, it cannot equiped direct to the outfitted weapon (drag and drop)

#

no secon also not possible (first is ACRE 343

whole cloud
#

and thats a new bug?

vivid rune
#

yes i think

#

i doesn't have to put optics on boxes with inventory bevor

#

from inventory it was always possible i think

whole cloud
#

cannot reproduce, I put holo sight into vest, and drag&drop back from vest to gun, all fine

vivid rune
#

it is heavy modded

#

i play something different with not so many mods. stay tuned 🙂

#

For detailing: I wanted drag and drop the optics from vest to equipped weapon. the weapon slot was white (OK) but it doesn't equip the optics after releas of mouse button

  • ACE+CBA
  • ACRE 2
    *simple inventory overhault
  • UAS + Expansion mods
  • CUP weapons
  • niarms aio
  • on Livonia
  • and many many other mods
#

maybe important: Character was overloaded

bleak garden
#

Can you reproduce the bug without mods?

vivid rune
#

yeah iam playing with some friends now. i go to vanilla server and try it again

restive pilot
vivid rune
#

yeah but ammo is working

#

it was the buris optic on the lynx

kindred radish
vivid rune
#

Mh. Arma and only Simple Inventory Overhaul in SP is working
also all mods together in SP
I try it in MP again

vivid rune
#

happended again with scorpion Evo 3 A1. Ammo and optic musst first lay on ground bevore equip is possible. (heavy modded server)

heavy vortex
#

I've had a similar problem with just CUP units/weapons/vehicles, a few weeks back.

#

That would have been pref/prof but several versions back :P

charred holly
#

11:17:44 Destroy waypoint not linked to a target: Near target acquisition is slow and may even select friendly unit.
Getting a lot of those (no mods). Only noticed them today, I don't know when it started but I feel it wasn't there a week ago.

spiral pond
#

this is like this since OFP

random isle
#

on latest pref users reporting high amount of shutter (screen lock up)

vale shoal
#

I didnt managed to get the diag_capute commands working, always the same error:

23:47:57 Error in expression <diag_captureSlowFrame ['total',1];> 23:47:57 Error position: <['total',1];> 23:47:57 Error Missing ;

23:48:51 Error in expression <diag_logSlowFrame ['total', 1];> 23:48:51 Error position: <['total', 1];> 23:48:51 Error Missing ;

23:38:15 Error in expression <diag_captureFrameToFile 1;> 23:38:15 Error position: <1;> 23:38:15 Error Missing ;

whole cloud
#

you are not running profiling

#

that command only exists in profiling

vale shoal
#

I ran that command server side

#

and the server is running profiling

hexed gorge
#

Performance or Profilling? The default EXE is Performance. You need to rename.

vale shoal
#

arma3serverprofiling_x64.exe , right?

kindred radish
random isle
#

v10 perf got this error Error in expression <if (local (_this select 0)) then {[(_this select 0), ", [], false] call bi> Error Missing } Dont have this code in the mission file anywhere

vale shoal
# kindred radish Yes

thanks. I successfully extracted some stuff out of these commands and sent a PM to Dedmen with all logs

whole cloud
feral harness
whole cloud
#

No thats all of it

#

its the quote

vale shoal
#

Dedmen, could you already take a look into the logs I sent you?

whole cloud
#

maybe next week

bright latch
#

I also got the generic script error of doom in eden, couldn't figure out what fnc it called though 🙂

feral harness
feral harness
#

it clearly errors even in the editor why no one reported it?

#

"Land_RepairDepot_01_civ_F" createvehicle position player

edgy comet
#

I don't launch my game every day, but the first time I got that error was yesterday.

rose moth
#

These types of errors only recently started getting reported in Dev / Prof, no matter them existing in the data for much longer.

whole cloud
#

I suggest you don't look at error log while playing old man 🫂

feral harness
#

I dont see why wouldnt it show before, the error is straight forward and there are only 2 type of objects that were affected

whole cloud
#

every missing } error wasn't printed

feral harness
#

you mean it was a bug in error reporting? Oh you!

woven oar
#

Are there bigger implication with fixing this?

whole cloud
woven oar
#

So you created more work for yourself.

whole cloud
#

not if we don't fix them

#

And the work has always been there, we just didn't see it

restive pilot
hexed gorge
#

Fix those errors can summon, more errors!?

vale geode
#

Does the Performance branch have any real benefit or increase to performance for higher end GPU/CPUs?

gritty wasp
#

Nothing noticeable. One man always report degradation.
Server side is more performant. And perf mostly like beta version for experimental features and tweaks

novel ravine
#

Just wanted to follow up:

We are still currently using v7 and I have enabled -cpuCount as you suggested a few days ago with my last post in here. Since then, we have yet to experience a system freeze/crash. I still want to wait another few days/a week before declaring it 'fixed', but this is the longest we have been without a freeze for a while now. Just thought you'd want to know since I guess that means it could have been a multi-threading issue that was causing it. After a week or so, I will probably port us over to newer performance builds, but at least v7 seems stable for us now. Will keep you posted if we have another crash or freeze though.

whole cloud
#

NeilZar is testing a different potential fix currently

buoyant jungle
#

Is running diag_captureSlowFrame on a server supposed to log the output to a file? diag_captureFrameToFile successfully creates a file in the .rpt directory, but I can't find anything created from diag_captureSlowFrame. Running arma3serverprofiling_x64.exe (latest version) on server. No errors in rpt.

kindred radish
#

so i guess no

buoyant jungle
#

How do you then profile a slow frame on a server? Write some logic to detect a slowdown and then use diag_captureFrame to capture the next frames? https://community.bistudio.com/wiki/diag_captureFrame description says it will capture to a file on server instead of UI, so it's not conclusive to me whether or not the slow frame capture is supposed to do the same.

gritty wasp
#

Arma profiler by dedmen. Not the same but great for profiling scripts

buoyant jungle
whole cloud
whole cloud
vale shoal
#

Try this

fickle geyser
hushed seal
#

is there any documentation on what that profiling UI is displaying? and what the numbers in "tree" mode mean.

hexed gorge
#

I remember to use it on server some time ago (1, 2 years?) and it wrote a text file near rpt files.

#

diag_captureFrameToFile

whole cloud
hushed seal
#

😦
does anyone know what the ms stand for in the tree mode?

heavy vortex
#

milliseconds.

silk pewter
#

Microsoft

#

Must Succeed

hushed seal
#

sorry, the numbers preceeding the unit

#

simTr; 1.35760; 0.02157;""

heavy vortex
#

It's kinda standard profiling output. The parent number is the total for everything underneath.

#

Difficulty is that no-one apparentl;y knows or tells what most of the five-letter categories are. You can figure some out by what's in them.

hushed seal
heavy vortex
#

oh, the second number is the total.

#

first number is more like an offset but I'm not sure what the point of it is.

#

Like you can calculate the first number from the second number.

#

I guess it's the time offset of the call, and they're basically in order.

#

But the second number is the useful one.

hushed seal
#

second number is time spent on <insert 5 characters>?

#

thanks!

heavy vortex
#

wSimE is quite fun, has vehicle simulation costs with the classnames.

#

there's another one somewhere with identifiable scripts in it.

random isle
#

i know for fact there big boat does it

heavy vortex
#

If there is then I haven't seen it. Purely personal experimentation.

hushed seal
#

any unofficial collection of findings? Our server sometimes takes a random nosedive FPS wise and I'm trying to figure out why

random isle
#

When we spawn big cup naval ship we go from 45fps to 12fps instatly, ones we delete the boat it slowly recovers. But i feel like there is also a ground vehicle that does it but its hard to find out which one

heavy vortex
#

Spawn them all in editor, capture a frame, look at wSimE and you should see it.

random isle
#

nah that wont do shit, has to be MP loaded server

heavy vortex
#

It's fine locally?

random isle
#

yep

heavy vortex
#

@hushed seal Nosedive where it stays low afterwards (usually) or a short-term blip?

hushed seal
#

stays low afterwards, recovers once everyone disconnects

heavy vortex
#

We had similar issues with the Antistasi community servers when they were set to maxmem=3072

#

Would apparently run fine for a bit, then dive to ~5fps and stick there. Sometimes recovered after a clearup.

whole cloud
hushed seal
#

👍 thanks

hushed seal
#

Has anyone experienced server FPS being severely degraded once more than 4GB of ram usage are reported in #monitor?

random isle
#

are you on 32bit exe ?

#

usally for 64bit ones you over 6/7GB game starts to page your harddrive like crazy and its where FPS goes to 2-5

fickle geyser
#

Also happens if you have a lot of players in same area and bunch of RHS helis in same area too.

restive pilot
#

does it have to be RHS?

fickle geyser
#

Propably not but so far I mostly had it with their helis, might be related to sensor setup.

restive pilot
#

or maybe scripting? do they use any scripting for their vehicles?

fickle geyser
#

No it's not scripting.

#

In the profiler it's clearly AI targeting / sensor.

oak nymph
#

What is the current procedure for sharing crash dumps?

whole cloud
#

sliding into my DM's (during a workday) or making a FT ticket

oak nymph
#

I'll DM you then. Many thanks!

whole cloud
#

workday :harold:

fickle geyser
fickle geyser
#

Basically every heli was taking 30ms for tgTrk on every client, causing stutters.

#

Something similar happens with CUP frigate as the guy above mentioned, in my experience even 2-3 of them can kill the frames.

#

And IIRC it shows the same in profiler, bazilion of tgSee

random isle
#

@celest sparrow good info here

hushed seal
rapid cradle
fickle geyser
#

Chinook issue was on chernarus, frigate one I don't remember but I think it was modded.

gritty wasp
#

Hmm why all clients needs to know what heli seemeowhuh lock detection?

heavy vortex
#

because commands like targets and knowsAbout generate all their data locally.

random isle
#

EntityAI SubSkeleton index was not initialized properly (repeated 13049x in the last 60sec)

#

getting few of this bad boys

oak nymph
#

Is there still a comand line parameter to suppress logging of objects not found?

hushed seal
#

class advanced options in server.cfg @oak nymph

oak nymph
#

Thank you Mildly_interested

rapid cradle
whole cloud
novel ravine
novel ravine
#

I've just now removed -cpuCount. We'll see how the next few days go. Usually, it freezes within 48-72hrs, so I imagine it wouldn't take long for me to get back to you about it.

whole cloud
#

can you let it dry run without players to not harm them? notlikemeow

kindred radish
#

Exile players are strong, they can endure that since they all have PTSD from Arma already aviator
We run the same framework (ExileMod) and also test everything on the live server, its a more realistic test environment tbh.

autumn timber
#

Testing on production. Love it! 😁

hushed seal
autumn timber
#

Just buy 80 Arma copies, what's the big deal? 😉

kindred radish
#

Can't really simulate the stuff players do, so there is no other way to test reliably imho.
And since it's a hobby and we don't earn money with it there is no pressure and most of the players are understanding.

vale shoal
#

I figured out a interesting fact. The problem with Nextcloud sync raising the CPU load of a Arma 3 process is only related to my Exile servers. With A3 Vanilla (but with the same mods excluding Exile) there is no increasing in CPU usage.

hexed gorge
silk pewter
hushed seal
#

I wonder if you could pay one of these captcha clickfarms to do the testing for you...

vale shoal
#

funny enough that I never figured that out (because I generally don't download via the machine). When the CPU load gets 100%, the download speed is reduced to around 1 MB/s. Strangely enough

kindred radish
#

Never had such problems, I download updates while exile server is running and no problems. (steamcmd or windows updates)

vale shoal
#

maybe a bug in the Intel Net Drivers? 😮

vale shoal
#

Also when only connecting via RDP to the machine the A3 processes jump up to 100% usage. This is so annoying. Cannot do anything on the machine....

kindred radish
stray prawn
vale shoal
#

@kindred radish

AMD Ryzen 9 5950X
128GB DDR4 ECC
Asus Pro WS 565-ACE

hushed seal
kindred radish
stray prawn
vale shoal
#

I also talked with our provider (Hetzner), they only gave me the advice to update the drivers. Indeed the Intel software updated the drivers from 12.18.11.1 to 13.0.13.0

#

The mainboard uses Intel I211AT as Controller just fyi

stray prawn
#

i got two dedicated servers running arma 3 servers, a xeon without issues and amd ryzen 7 5800X which has the issue

vale shoal
#

I'll dm you

vale shoal
#

but, well.. this driver messes up with extdb3 - I rolled back to 12.18.11.1, no problems with extdb3..

hushed seal
#

haven't tried it on our 3600 server

hexed gorge
#

Any performance improvements currently being worked on? 👍

whole cloud
#

no

tawdry gazelle
#

(Enfusion is currently being worked on) meowawww

whole cloud
#

new prof probably tomorrow, I'll try

desert olive
#

has anyone reported on strange server freeze issue for V11 yet ?

#

my group hosts a weekly event every Wednesday and we get ~50 people each week

#

last week the server would just freeze when everyone was loading in and downloading the mission

whole cloud
#

windows server 2016? full server freeze?

#

or only arma server freeze

desert olive
#

Linux

#

yellow chain

whole cloud
#

:sad:

desert olive
#

not a hard crash

#

the behavior felt similar to the freeze issue that was happening a few versions back where the server would stop doing anything but remain running and alive

#

yellow chain, no crash, increasing dsync warning in logs

hushed seal
#

do you all spawn in at once or is it staggered?

desert olive
#

we attempted staggered too

#

took longer to die but it happened eventually

#

this issue does not happen on normal days when we get ~20-30 people playing the normal map rotation

rain moth
#

which nic and kernel?

desert olive
#

Intel I210 Gigyabit nic

#

not sure about the kernel, we upgraded the distro like 2 days ago

#

we were running debian 10, so whatever was the latest on that distro

#

basic.cfg in case this is somehow related to how many messages are sent

# 
MaxBandwidth=1000000000;
MaxMsgSend=1024;
MaxSizeGuaranteed=512;
MaxSizeNonguaranteed=256;
MinErrorToSend = 0.010;
MinErrorToSendNear = 0.10;```
# 
    maxPacketSize = 1430;
    MinBandwidth = 16000;
    initBandwidth = 1250000;
}```
#

current kernel is 5.16.0 but like I said, we haven't tested on this one yet

#

does MaxMsgSend somehow affect how the server sends the mission file to joining players ?

#

if MaxMsgSend=1024 is per frame, could we be chugging the network if server fps is high and a lot of people joining ?

rain moth
#

check kernel logs if you're seeing any nic reset entries

#

something like e1000e 0000:00:1f.6 eno1: Detected Hardware Unit Hang:

hushed seal
desert olive
#

isn't 48 awfully low ?

hushed seal
#

Yup, but so far it worked fine for our group and atleast 4 others who I told about that, all in the range of 50 to 100 players

desert olive
#

however one of the guys did say he had trouble SSHing to the box when all this was happening

#

was getting timeouts

rain moth
#

might be the e1000 UDP offloading issue then

#

it resets the NIC which takes a little while

#

I've been able to confirm this happens with Arma servers, both for our own boxes and other communities

#

there's a workaround which disables the offloading, we've had no problems since

desert olive
#

will try thank you

rain moth
#

ethtool -K <interface> tso off gso off tx off rx off or some combo of it

kindred radish
desert olive
#

I reduced it to 256 for the event today, hopefully it will help

hushed seal
#

Oh there's a neat website to plot the output of monitords...

desert olive
#

it would be nice to be able to run #monitor via SQF like you can do with other diag commands

#

unless there is a way that I don't know of

rain moth
#

you can via serverCommand

desert olive
#

ahhh this is useful thank you

hushed seal
#

anyone else having issues running the monitords command? monitor works fine but monitords doesn't show anything in the server console nor the console log file

whole cloud
#

2.08.149255 new PROFILING branch with PERFORMANCE binaries, v12, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Adjusted helicopter upside down damage handling
  • Tweaked: Savegame scheduled scripts are now saved as bytecode
  • Tweaked: Multiple changes in the audio system, potential crash fixes
  • Fixed: High Command was not working - https://feedback.bistudio.com/T163682
  • Removed: AvSetMmThreadCharacteristics due to potential Windows Server freezing

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

whole cloud
quaint flame
#

Custom connections?

whole cloud
#

Not mentioned in changelog because clientside compatibility

quaint flame
#

👍

restive pilot
whole cloud
#

no

#

I wanted all scripts, which with many mods does have a significant impact, but thats disabled now.

vale shoal
#

AvSetMmThreadCharacteristics... sounds interesting. Could this also solve the problem I have with my high cpu usage?

restive pilot
patent dome
#

Prof v8 of 2.06 is when it was introduced.

rain moth
#

Too bad for Alder Lake owners

hexed gorge
rain moth
#

From previous perf changelog,

  • Tweaked: Use AvSetMmThreadCharacteristics to tell windows that we are a Game
#

since alder lake has heterogeneous cores the scheduler needs to figure out how to schedule threads among performance and efficiency cores

hexed gorge
vale shoal
heavy vortex
#

High CPU usage is normally what you're aiming for in Arma :P

rain moth
#

no you want good server and client fps, but that usually leads to high cpu usage 😛

vivid rune
#

Server and client v12:
Edit:

  • killed ai
  • loot something from AI uniform ->work
  • switch uniform by right click on AI uniform ->work
  • loot something from old uniform laying on ground->work
  • switch uniform back by right click on old uniform on ground->work
  • loot something from AI uniform on ground ->doesn't work
    (all by drag and drop)
    After the first switch the uniforms laying as paket in the near of dead body (not part of the dead body anymore). Maybe that has influence of the problem?
    I tried this also with backpack. This works.
    I also tried older versions of profiling branch. This is seen since v02 (00 and 01 i cannot test because it crashes).
vivid rune
#

Another thing:
had a silencer in uniform. Switch unifrm by righjt clikc on uniform on ground. Switch happen but uniform on ground (my old uniform with silencer) has no item in it

#

if i switch back, silencer is in uniform

vivid rune
#

Hi,
I changed server to normal branch, client is prof. Now it seems to work

empty goblet
#

unfortunately from my observation the 149255 seems to end with noticeably worse server-side FPS on WL with lot AI

vale shoal
hushed seal
heavy vortex
#

I would guess that's Arma using blocking sends and UDP having ultra-low priority, but that should only be an issue if you're actually upstream-constrained...

vale shoal
#

the CPU usage of the Arma 3 processes is generally 5-10%, depending on the players. But when this bug occurs, all processes going up to 30% and the general CPU usage is 100%. It's not only RDP. Its downloading via browser, Nextcloud or all kinds of network usage.

#

I cannot kill the processes every time I want to work on the remote

#

I also sent Dedmen dumps, but I didnt get any answer till today sadly. and I have no idea what to look in these dumps

hushed seal
#

that really sounds more like a hardware problem and arma going nuts because its network threads are going mental, does capture frame still work under these circumstances or not?

vale shoal
#

well, the hoster gave me the opportunity to change the hardware. maybe I can take this into account. capute frame worked

#

but without Arma 3 running, the server itself behaves well without any problems (for my view)

hushed seal
vale shoal
#

Dedmen got the results per DM some weeks ago

whole cloud
#

dedmen forgot about it

vale shoal
hexed gorge
vale shoal
plush apex
#

Hi, I ran a large TvT event the last sunday, 90+ players, no AI, using the last perf (v12). Server FPS were very high and fairly stable through the game (it lasted 2hrs 10 min), but we experienced very long time for downloading data. I was monitoring the server via #monitor in chat as well as #monitords and my monitor in chat was suddendly interrupted, while instead the monitords on server was running and fps were high. After a certain time I had a lot of monitor messages in chat, sent all in once. Players took a lot of time to pass from briefing to game, and they took a lot of time to download the data as I said before. But everything was fine in the end and no complains about server performances, just this unusual long time to spread data in the network (at least what appears to me). Any hint? Thanks in advance!

hexed gorge
spiral pond
#

[Feature Request] Add ability to adjust the number of tank tracks and foot steps shown
https://feedback.bistudio.com/T156441

Dedmen:

For foodsteps (humans and animals) the time is
50 seconds for thermal vision or
10 seconds for non-thermal vision.

The time to live for tank tracks is 60 seconds, but it also cleans up if the track is outside of a bounding sphere of 20m.

So max lifetime is 60 seconds, max distance from tank is 20m.
If I understand this right.

So its not as simple as tweaking a number of "max active tracks"
But I don't see a reason not to up the lifetime to like double or triple the current value, and also up the range for tank tracks.

Any chance to just up the values for perf/prof branch and see if people can notice/measure any fps impact?

random isle
#

better yet ability to turn it off lol

fickle geyser
#

They stay on if you disable their simulation.

whole cloud
#

Should be possible if its not config and if I don't forget

silk pewter
#

food steps 😋 yumyum

hexed gorge
plush apex
#

@whole cloud something cooking in your kitchen for this? #perf_prof_branch message
We ran a similar event on March 3 using the stable and we had not this behaviour, less server FPS though
Do you (or others) have something to suggest? Might it be related to MaxMsgSend as suggested in some posts above?

whole cloud
#

¯_(ツ)_/¯

gritty wasp
plush apex
gritty wasp
#

We play with same basic config for over 5 years on a3.
I can’t check config right now. But as much as I remember mms 256. Min bandwidth 80 mbit even on 1 gbit server connection. Also we have modified errors values.

We don’t have freeze issues caused by Microsoft parameter which was removed in v12

Also mission file load (when you go from lobby to briefing) depends on speed defined in socket section. (If defined.)

plush apex
#

Thanks so much @gritty wasp ! I have this in class sockets
class sockets { maxPacketSize=1430; initBandwidth=1250000; };
This is the same for a while now though, I did not change it, but only the last sunday we had such slow downloading data

spiral pond
#

this is not on stable yet, right?

whole cloud
#

Don't know

#

probably not

vale shoal
#

@whole cloud btw thanks for your reply about that high cpu usage problem, but I'm 100% sure that it doesnt have any to do with scripts. There must be something in Arma 3 netcode. I restarted the Arma 3 servers, normal cpu usage. Disconnect from RDP. Reconnect to RDP, cpu usage jumps to 100% instantly

tawdry gazelle
tawdry gazelle
hexed gorge
#

Wow! 👍

vale shoal
#

Anybody of you also noticed that Arma 3 Server processes run WMI Performance Adapter?

tawdry gazelle
#

You mean Windows Multimedia something or Windows Management Instrumentation?

vale shoal
#

this high cpu usage problem is so annoying, again connected via RDP and the usage jumps to 100%...

#

err, just to be sure. I found this error when closing a Arma 3 process. should it be like that?

Name of the corrupted application: a3chernarus.exe, version: 2.8.149.255, timestamp: 0x62601b53 Name of the corrupt module: mscoreei.dll, version: 4.8.4380.0, timestamp: 0x608baf11 Exception Code: 0xc0000005 Fehleroffset: 0x000000000003fb28 ID of the faulty process: 0x33f8 Start time of the faulty application: 0x01d860457bb36fb9 Path of the faulty application: D:\Servers\arma3\chernarus\a3chernarus.exe Path of the corrupted module: C:\Windows\Microsoft.NET\Framework64\v4.0.30319\mscoreei.dll Berichtskennung: 6a16abd3-75b3-4a1d-9816-bb3380b012e2 Full name of the corrupted package: Application ID relative to the corrupted package:

restive pilot
#

That's not a vanilla exe

vale shoal
#

its actually the perf prof exe just renamed

#

I have a server tool running which scans for the process, if its not running the server is restarting automatically

restive pilot
#

Well that's a .NET module. I don't think Arma uses .NET at all thonk

silk pewter
restive pilot
#

Yes

#

But not the game

silk pewter
#

is there such a thing as perf launcher though?

restive pilot
#

Well there's a small bar at the top that shows the branch when you use the perf (or dev) branch thonk

vale shoal
#

well, if Arma 3 don't use any .NET module, how is this error present in the Event Viewer? 😄

spice dirge
#

some custom extension that chainloads it, maybe?

cold vale
#

Have you gone back to basics to ensure vanilla arma shows the same behavior and after a reinstall of windows that the issue still shows up as I've never heard of RDP causing behavior like you're describing

vale shoal
#

Sure, I even changed the hardware because we thought it might be a issue with the Ethernet Controller, but nope. Reinstalling was also done (well, a switch from Win2k19 to Win2k22) and both have the same issue.

whole cloud
#

Can confirm that Arma exe doesn't use .NET

feral harness
#

only .com and .org

silk pewter
#

bohemia**.net**
checkmate

gritty wasp
vale shoal
hexed gorge
#

2.10 has a date? When? 👍

silk pewter
#

in the future!

autumn timber
#

2.10 actually IS a date

#

it means the 2th of October

#

(or 10th of February if you live in the US)

silk pewter
#

ISO-8601 FTW

autumn timber
# silk pewter ISO-8601 FTW

Nah, using common sense is enough 😛
You just need to count the number of countries and how they interpret the string to know which one is "correct" 😛

silk pewter
#

(well, that too)

#

big-medium-smol
smol-medium-big

then there's "that guy"…

hexed gorge
#

Correct date is year-month-day so we can sort it correctly 😀

autumn timber
#

Wow... so that actually means that US date representation really is middle-endian! 😱
I actually googled it and that's the name for that format 😄

whole cloud
#

But thats a sunday :sad:

silk pewter
#

so 2.12 last patch confirmed then 😏

patent dome
#

No, 2.28 will be the last one. Release date: 2031.2.28
If we skip a couple years due to missing the release date, it could be pushed back to 2035.2.28

hexed gorge
#

Great, sooner than I thought 🧘‍♂️

restive pilot
#

well current version is: 2.09.149299
see you in year 149299 guys 💀

hexed gorge
#

😮 I hope Jeff Bezos sells immortality pills before I die.

vivid rune
hexed gorge
#

🙅‍♂️

vale shoal
#

I encounter a strange bug since the latest perf release which cannot be killed in any way

https://abload.de/img/screenshot2022-05-140z8jss.png
-> this process is marked as "suspended"

https://abload.de/img/screenshot2022-05-14048kfw.png
-> there is no data found for that process, but it's running

https://abload.de/img/screenshot2022-05-140z6jl0.png
-> Process Hacker (and any other tools) cannot kill it because PID 1664, the process, is already closed but it's still running

https://abload.de/img/screenshot2022-05-140v5kvs.png
I'm sorry that I only can provide the errors in German, but they literally mean "Cannot kill process, access denied" in any way.

random isle
vale shoal
#

no, but I might think it's related to the -debug parameter

whole cloud
#

that doesn't make sense, debug only adds more RPT logging

whole cloud
#

2.08.149367 new PROFILING branch with PERFORMANCE binaries, v13, server and client, windows 32/64-bit, linux server 32/64-bit

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

whole cloud
restive pilot
#

(abandons us now)

whole cloud
#

Haven't done much profiling branchable stuff, all the TI stuff cannot go because shaders stuff

hexed gorge
#

- Fixed: Floating bodies and PhysX objects after a building collapse - https://feedback.bistudio.com/T63050Nice, bye bye my lovely code 🥲
I use "BuildingChanged" mission event handler to fix that.

#

If the construction is hidden or deleted, will the bodies still float?

whole cloud
#

the new code from KK does the same, just in engine

silk summit
silk summit
feral harness
vale shoal
#

I didnt upgraded to v13 yet, maybe "Fixed: Several game crashes" will fix my issue

whole cloud
vale shoal
#

Yeah thats true

vivid rune
#

Hi,
have found a bug with profiling (remote server):
CIVs can wear all clothes.

  • start mission as civ with a basecap (other huts are too big for suit inventory)
  • as civ go to dead CSAT, open inventory of dead CSAT
  • put your basecap on ground () ->work
  • take CSAT suit and wear it (right click) -> work
  • take basecap and put in CSAT suit ->work
  • take basecap from suit and drag and drop on Head-Slot->doesn't work
    -> if you try last point, server throw: "Server: Object 2:19 not found (message Type_203)"
#

server and client v13

#

In local MP there is no problem, only remote.

whole cloud
#

only on profiling, or also on stable?

#

sounds like that would also happen on stable

vivid rune
#

I think it is only on prof. First I recognize this on playing Dynmic Bulwarks on Prof. Then switched Server to Vanilla and probs are gone.

#

I checked to prov 2.08 v2 i think (v0 and v1 are crashed on my server). all 2.08 prof seems to have this issue

vivid rune
whole cloud
#

yes non-profiling normal stable arma

#

Thanks notlikemeow

rain moth
#

mission on weferlingen dropped to 5 fps with 100-250 ms render latency all of a sudden

#

i.e. good old 3 fps issue

rain moth
#

now another player got it with stable branch

#

weferlingen is unplayable in large scale events, the fps bugs occurs way too often

vivid rune
hushed seal
#

server on profiling.exe, clients usually perf or stable

rain moth
#

we've had participants with the issue on almost every large scale weferlingen event 😦

whole cloud
vivid rune
feral harness
#

Could you add a 100% repro?

vivid rune
hexed gorge
#

Hi, i'm Donnovan, your big friend. Please fix the map crash problem, you already have the Crash Report.
Add Phone verification to avoid bad people. Thanks and bye. 🌻

silk pewter
#

@hexed gorge wrong channel, Walter

whole cloud
empty goblet
empty goblet
#

so anyone else got 0 fps bug on 367 server, while 255 was fine ?

marble mason
#
  • Tweaked: Star visibility in night vision
#

Is this not suppose to work on dedi?

#

I have both server and client on perf

whole cloud
#

its a clientside fix

proud crane
#

Pingus

patent dome
#

This has to be the most useless ping I have ever received...

silk pewter
whole cloud
#

2.08.149459 new PROFILING branch with PERFORMANCE binaries, v14, server and client, windows 32/64-bit, linux server 32/64-bit

  • Added: Ability for mods to distribute Eden Editor Compositions in their data
  • Added: filePatchingExceptions[] server config option
  • Tweaked: Unauthorized Headless Client log messages now also includes the IP address
  • Fixed: Unloading a magazine from a weapon into a container could overfill the container
  • Fixed: Eden Editor script error when the Undo-action is used - https://feedback.bistudio.com/T162775
  • Fixed: Shots from remote players would have aiDispersionCoef applied to them, even if the shooter is not AI - https://feedback.bistudio.com/T163665
  • Fixed: Flag cloth would not animate on vehicles in 1st person view in some cases - https://feedback.bistudio.com/T164798
  • Fixed: The amount of surface dust generated from ammo impacts ignored surface humidity
  • Fixed: Possible crash when the refresh button inside the audio options removes a device, the last device was selected, and the cancel button is pressed

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

whole cloud
hushed seal
#

Is there anything in the profiling build to debug what's using so much RAM during a mission?

whole cloud
#

Eh, yes. There is a memory info dump, I think when you do a capture frame it dumps that to RPT too

#

But it will only roughly tell you what types of script variables use how much, not where they come from

hushed seal
#

👍 We've seem to be getting the good old 3FPS bug with disappearing trees again and have no clue where it's coming from.

#

It correlates with server ram usage so we're hoping to just avoid it by keeping our mission small.

night mulch
#
  • Added: Ability for mods to distribute Eden Editor Compositions in their data
    how to use this feature?
whole cloud
night mulch
#

awesomeness thanks

vivid rune
#

@whole cloud
Thanks you and the other guys for fixing things in A3.
I would ask for this bug in profiling:
#perf_prof_branch message
This blocks me to use perf build on my potato server. And the server needs your perf build. 🙂

analog acorn
#

If you have a reliable repro and have confirmed it on the latest prof, you should make a ticket so it can be properly tracked.

vivid rune
heavy vortex
#

In both 149367 and 149459 I've had several chunks like this in the RPT shortly after starting a localhost mission:

 4:15:21 Unexpected network message, possible hacking attempt id=2, name='********', steamid='*********************', msgType=387
 4:15:21 In last 1000 miliseconds was lost another 4 these messages.
 4:15:21 User tried to set invalid material 
 4:15:21 User tried to set invalid material 
 4:15:21 User tried to set invalid material 
 4:15:21 User tried to set invalid material 
 4:15:21 User tried to set invalid material
#

Main branch doesn't do it. Any idea what this is?

graceful nexus
#

I have been having issues in Zeus with placing down my custom compositions. My custom compositions and steam workshop compositions/steam subscribed content appear on the list, and the steam workshop compositions have no issue being placed, although with my own custom compositions, they aren't placing down at all, or even loading when I select it and attempt to place them down. Is there a fix for this? I have had no problems with it a few days ago, but all of a sudden I'm having this issue and I'm not sure how to fix it.

  • I have tried to restart my game, PC, opening Eden Editor and then closing out and joining a Zeus server, although they still don't want to be placed.
  • Another thing is that my custom compositions work and place just fine in the eden editor, and the problem only occurs as game master in a Zeus server.
    -- @empty goblet @whole cloud I inquired about this issue in another Arma 3 related discord server and I was told to ask here and tag you two. Would there be a fix for this?
  • Another note is that I use the Arma 3 Profiling Branch version along with the companion mod and have had no issues with this before, and this bug appeared to me just recently in these past days with no fix that I've found.
whole cloud
whole cloud
graceful nexus
whole cloud
#

I forgot a ! when I added support for mod compositions :harold:
Is fixed next build

worldly badge
#

You forget it!!!!!!!!!

grand spire
cosmic silo
#

Issues occured after the update, no word as of yet as to wtf is going on or if there will even be a fix.

analog acorn
whole cloud
#

fix build comes tomorrow

whole cloud
cosmic silo
cosmic silo
#

My fault. Should have looked a bit higher up.

whole cloud
cosmic silo
#

Regardless, thank you for correcting me and the quick work on resolving the issue!

#

Is there any particular timeline for when the update will go live?

whole cloud
#

Tomorrow afternoon-ish europe time.
Unless something unexpected goes wrong

whole cloud
#

2.08.149534 new PROFILING branch with PERFORMANCE binaries, v15, server and client, windows 32/64-bit, linux server 32/64-bit

  • Added: "Unlisted" visibility for Scenario and Composition uploads (translation is missing for now, so it shows up as an empty text field) - https://feedback.bistudio.com/T163573
  • Tweaked: Improved performance of nearestObject if the desired object is close to the source position
  • Tweaked: Improved performance of isFlatEmpty if objects blocking the space are close to the center
  • Tweaked: Improved performance of nearestObjects, nearObjects, nearestTerrainObjects, nearSupplies, nearRoads when scanning a large area
  • Fixed: Inventory desync issue - https://feedback.bistudio.com/T165733
  • Fixed: Helicopters 'hopping' when landing due to telling AI to disembark
  • Fixed: Placing custom compositions in Zeus not working since the previous update
  • Fixed: Linux server was autodetecting cpuCount as the number of threads instead of the number of cores
  • Fixed: onUnload was not called when a title effect display was destroyed

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

whole cloud
autumn timber
#

What about the Profiling Branch Companion Mod nowadays? Does it still make sense to load it or has everything been merged into the game already?

patent dome
#

I believe that is just the sqfc files of functions_f and ui_f, so not everything has been merged yet.

autumn timber
#

I've been playing with it enabled for a while without any (visible?) issues at all. Maybe it would be time to consider merging them?

rose moth
#

We reverted it after 2.08 due to issues in loading savegames. That should now be fixed, so we're trying again in 2.10.

whole cloud
#

On savegame load scripts were compiled from string, instead of loading the bytecode.
And as script state gets serialized by index of instruction that was last executed, and bytecode (due to optimizer) has less instructions.
The game re-ran some instructions after savegame load that it shouldn't have.

But since recently we have bytecode serialization for savegames and its been tested on profiling branch for a couple months now, and it should work fine :harold:

vivid rune
whole cloud
#

that should generate a freeze dump? does it?

vivid rune
#

i tried cpucount=1 but not working

#

where can i find this

whole cloud
#

in RPT folder

vivid rune
#

i tried vanilla server this seems to load

vivid rune
whole cloud
#

ooooohh linux server. Thats a quite important information

#

no it does not

vivid rune
#

linux 32bit in docker

whole cloud
#

32bit? notlikemeow

#

why

#

ahh

vivid rune
#

because phenom II

#

mission is loading and does something till it hangs

whole cloud
#

can you check in htop which thread is at 100%?

#

thread names should be enabled I think 🤔
Or mabe it was only enabled on prof

vivid rune
#

htop?

whole cloud
#

yes, or top or whatever process monitor you like that shows threads

vivid rune
#

yeah i will try. i would message in 2 hours

sinful smelt
#

After today's profiling build release on Steam, a couple of our developers are getting repeatable (but not yet tracked down to a small root cause) game crashes that were a regression from the previous build. They're on windows, x64. I wish I had a small reproduction case for you, but I haven't had to time to dig deeply. Our current 100% reproduction method is planting a blasting charge on Olympus - I just haven't had the time to dig down to root cause. I have frozen mdmp from one of them I can provide if that might help. Based on timing, I'm assuming this is the 2.08v15 build, but I'm not in a place to attempt to dig further personally right now, and I'm unsure if it'll even happen on Linux/Proton.

Just let me know who to send the mdmp to or if more details would help, though I doubt I'd be able to offer anything without spending some more time digging down to a simpler reproduction case.

autumn timber
sinful smelt
#

I'd assumed, I just try not to DM people like that w/o hearing something from them first and trying the community for people who've already tracked it down. If he mentions something I'll toss it his way right away.

Given the 100% reproducability for us right now, I was sort-of crossing my fingers for other's cases to help narrow down the scope.

If he mentions something I'll send it over to him straight away; thanks 🙂

whole cloud
sinful smelt
#

I haven't been able to get away from regular work to look into it myself yet, but, looking at the changelog on the forums, one thing that caught my eye is there are a few nearestObject calls in the vicinity of where it was crashing

whole cloud
#

win32u!NtUserMessageCall [fail:0]
user32!RealDefWindowProcWorker [fail:0]
user32!RealDefWindowProcW [fail:0]
uxtheme!DoMsgDefault [fail:0]
uxtheme!OnDwpSysCommand [fail:0]
uxtheme!_ThemeDefWindowProc [fail:0]
uxtheme!ThemeDefWindowProcW [fail:0]
user32!DefWindowProcW [fail:0]
thonk I may need more dumps

whole cloud
sinful smelt
whole cloud
#

Ah thank you. Sad, but much better 😄

unknown.png
#

Well they probably didn't implemented nearestObject as efficiently as possible, because if you do it wrong, it causes freezes 🤣 crap

#

I'm gonna place a bet that something called nearestObject at 0,0,0 coordinates and thus the grid index got negative :stamp:

whole cloud
#

I won my own bet blobcloseenjoy

tawdry gazelle
whole cloud
#

Fix for the freeze should be able to go out today, in maybe 3 hours

heavy vortex
#

Just had an odd case on localhost profiling branch where a bunch of remoteExecs that were pending when I aborted the mission were apparently executed immediately after restarting it (before initServer).

#

Not sure I can replicate but it seemed... interesting :P

#

(saving is disabled, for what that's worth)

whole cloud
#

Eh 🤞 I mean

#

more unexpected bad things 🥐

silk pewter
#

🥐 ← you missed them btw, forcing me into travelling once more to your office

rose moth
patent dome
#

I hate unexpected bad things NotLikeBOYE

silk pewter
#

talking about me?

whole cloud
silk pewter
#

ooooooh 😄

#

since we spent the night together @patent dome is not the same with me

patent dome
#

That one night changed a lot

sinful smelt
quaint flame
whole cloud
#

The freeze fix is ready, but our buildserver decided that yesterday was monday and also he stood up with the wrong leg so he's not been in the mood to work right.
But today is a long day so hopefully it not be too long

whole cloud
#

2.08.149549 new PROFILING branch with PERFORMANCE binaries, v16, server and client, windows 32/64-bit, linux server 32/64-bit

  • Fixed: Freeze when calling nearestObject with a position that was close to the edge of the terrain
  • Tweaked: Voice over Net is now by default handled via the server instead of peer-to-peer (config option exists to switch it back)
  • Tweaked: Vehicles now change locality to the effective commander if no driver or copilot are available

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

whole cloud
#

@unborn urchin meowheart

unborn urchin
#

cool will try it out soonish

#

did it go into the devbranch update too ?

whole cloud
#

I think that was after yesterdays dev branch was built

unborn urchin
#

yep definitely working! works for hasDriver = -1 and the statics

#

was monitoring local vehicle player and it would go from false to true after a moment and the driverless tanks actually could be driven now

whole cloud
#

So I fixed driverless tanks by accident? 🤣

unborn urchin
#

pretty much, I thought for a long time it was a locality issue

whole cloud
#

@summer raft FYI you asked about that two months ago

unborn urchin
#

as they'd work fine in local hosted/sp but not for clients/dedicated

cosmic silo
#

@whole cloud you are now my favorite. Don't tell the others.

rain moth
#

oh no, now I have to resurrect the stridsvagn S / 103 😦

#

does it work with AI this time or still only player?

whole cloud
#

It would also transfer locality to a AI commander pretty sure

rain moth
#

I think previously if you had a tank with hasDriver = -1; the AI would never do anyhing

unborn urchin
#

i only tested with players as that was the only issue I was aware of

summer raft
#

@whole cloud well that's awesome. Thank you sir for remembering me!

gritty wasp
#

I don't know how it works in engine, but maybe it is possible and easy to fix. Countermeasures from driver(or any other slot) available not only to effective commander?

whole cloud
#

Won't do

wooden saddle
#

The Arma 3 dev that does the profiling (performance build), is he here on discord or contactable?

restive pilot
#

was that a joke? thonk

wooden saddle
#

No, I like getpos instead of the rest, I never joke

#

But

#

What, what is funny, Arma official ppl aren't around here or what

restive pilot
wooden saddle
#

Nice I will nag him then, thanks

silk pewter
#

uh-oh…

mellow lark
#

looks at his bio 😬

silk pewter
feral harness
vivid rune
#

i have a hunch

tawdry gazelle
wooden saddle
wooden saddle
#

Also it doesn't involve dragon fire, although that's not a bad idea

worldly badge
#

What do you mean by particle hard limits

whole cloud
#

Hardlimit for particle count, between 16k and 19k in engine. If you have more particles the engine just deletes them

wooden saddle
#

Yup, ironically the first search result about it is a thread by polpox lol

mossy anchor
#

Been testing some comps and found out when applying textures or materials to the falcon drones, while they apply them when leaving a area of around 200-300 metres they will return back to their normal version which doesn't happened to any other aircraft or vehicles I have tested out.
Also discovered that both hurons (unarmed and armed) camos can't be swapped over to the different camo with the armed variant unable to swap to black and the unarmed unable to swap to green once saved as a comp then placing them down. Also testing setobjecttextureglobal on the hurons doesn't work on them as well.

plain trout
#

I wonder if it is possible to fix the keyframe animations working again on dedicated servers

spiral pond
#

@plain trout got a ticket?

plain trout
#

haven't made one yet

#

lost my login data once more

rugged owl
#

What could trigger this in RPT? Only happens on this branch
12:47:53 Tried to AccessTargetList for non-local AI group while non-local targeting was disabled! Engine will now enable non-local targeting which will hurt clientside performance.

heavy vortex
#

In the main branch, all AIs do full visibiliity checks on every client, unless you use disableRemoteSensors. This data is needed so that commands like targets and knowsAbout will return data on any client.

#

Prof has an optimisation where it only does visibility checks for local AIs. targets will work as before if called on local units. If you use it on non-local units, it switches back to doing visibility checks everywhere.

hushed seal
#

aka some AI mods trigger it

#

afaik LAMBS doesn't

hexed verge
#

lambs should only check knowsAbout and the like local to ai. besides that messages usually pops up at the very start and lambs routines only really run upon danger events

heavy vortex
#

Antistasi will trigger it due to undercover-related checks. Need to rewrite some stuff.

silk pewter
#

@tribal pasture ↑

tribal pasture
#

@silk pewter Made disableRemoteSensors true; much earlier, did disableAI "all" on agents, lets see if it helps

#

In general my mission is player-only apart from lobby Logic units and few static agents

silk pewter
#

beware of (AI) mods too

tribal pasture
#

No mods are used, strictly vanilla

silk pewter
#

you, I like you

whole cloud
#

There are also some in engine things that can trigger it

silk pewter
whole cloud
#

no

silk pewter
#

bad Dedmen, bad!

tribal pasture
whole cloud
#

no

tribal pasture
#

We run player vs player with no AI mission and if we do have this triggered, then everyone will

whole cloud
#

yep many people get it

knotty moth
#

mal;loc

wooden saddle
#

We're talking about murlocs right

tawdry gazelle
#

Hi, was the crash issue since V15 fixed? 😅

#

(I guess not since there's still no update published)

whole cloud
#

I'm a professional thing forgetter

whole cloud
# tawdry gazelle Hi, was the crash issue since V15 fixed? 😅

Thats so weird.
Server is crashing on mission switch, while cleaning up cut effects, like cutRsc script command.
For one the crash doesn't make sense to me, and second a server has no UI, there should never be any UI effects there 🤔
Its probably not reproducible because most servers don't have UI scripts running on a server without UI

#

It thinks there are 101 elements in an empty array and crashes when trying to remove 101 elements from a array that has zero.
I cannot see any changes in v15 that would do such a thing 🤣

tawdry gazelle
#

It's a heavily modified and modded escape mission... Neither the mission file nor the mod list was maintained by me so... notlikemeowcry

#

That's the most strange point: on V14 it never crashes like that 🤣

whole cloud
#

nvm I missed one change.

unknown.png
#

Sus

#

A title effect related change, and now we get cutEffect related crashes thonk

tawdry gazelle
#

Ah...! meowawww

whole cloud
#

I cannot reproduce it. The mission must be doing something special.

#

@tawdry gazelle how hard would it be for me to get that mission running?
Do you get crash on every restart/mission switch?
When exactly do you get it? I assume after you ran that mission, and then restarted it or switched to a different mission

whole cloud
#

Well I tried more ideas to repro. And found a different crash in same place 😄
Creating a Rsc on layer 4, which has a onUnload handler which recreates itself on layer4.. bad.

whole cloud
#

Found it.
Game is clearing all cutRsc effects at end of mission.
If you have a onUnload handler in it, that creates a new cutRsc effect while engine is currently deleting all effects... bad.

or more technical for those who care

onunload = "diag_log 'unload config'; if(isNil 'WantCrashStuff') then {WantCrashStuff=true; 50 cutRsc ['myRsc','PLAIN']}";
That finally got the same crash.
Array clear sets size to 0, then destructs all elements (which re-adds one element and re-sets size to 51), then deletes data

So now we have data be nullptr, and size be set to 51.
Next time clear is called, it again tries to set size to 0. It sees it has 51 elements that need to be removed.
Tries to remove the elements from _data (it knows _data cannot be nullptr because size is not 0), but _data is nullptr and poof.

Thats some good detective work here

#

@tawdry gazelle can you send mission file. so we can check in their code if thats really what they're doing

tawdry gazelle
#

I am not sure if it can run without mods... It is a heavily modified and heavily modded escape mission. I will send you some of the files.

#

emm... Although the mission file is heavily modified, I bet the Rsc something / UI things should be left untouched...

whole cloud
#

cannot find it inside the mission file. maybe its a mod.. eh.
But I reproduced the exact same crash and fixed it so should be fine

tawdry gazelle
whole cloud
#

Thanks for finding, thats a tricky one

whole cloud
#

2.08.149598 new PROFILING branch with PERFORMANCE binaries, v17, server and client, windows 32/64-bit, linux server 32/64-bit

  • Added: Publishing a scenario to Steam Workshop now adds tags for used DLCs by default
  • Fixed: Combobox entries were not properly clickable if the list was bigger than its parent control
  • Fixed: Redundant nulls returned by the 'agents' script command - https://feedback.bistudio.com/T75506
  • Fixed: Floating ground weapon holders after building collapses - https://feedback.bistudio.com/T63050

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

whole cloud
#

Crash fix is in there, I'm just too lazy to put it on changelog

rugged owl
#

Is -noPauseAudio planned to become a default branch setting anytime?

whole cloud
#

Are you asking if we are planning to release any more game updates?

#

Thats a weird question to ask

#

no Arma 3 is not abandoned.

autumn timber
analog acorn
rugged owl
#

Except while being on profiling

analog acorn
#

That's interesting. I take it you also have -noPause, since -noPauseAudio requires it?

rugged owl
#

Yes

#

Normally I'm always using this build, but the most recent update broke a horizontal scroll bar in king of the hill. So I have to swap to default now and without that parameter it kinda sucks 😄

analog acorn
#

It's not in the actual 2.08 patch notes. I guess the wiki is incorrect.

whole cloud
rugged owl
#

Didn't have the time yesterday, sorry for not making a report then.
Since the last perf branch update this horizontal scrollbar in particular has stopped functioning. This is in king of the hill https://i.imgur.com/nA1ZIbX.png
Unable to click and drag.

Imgur
#

@whole cloud

#

No script errors or rpt entries about it unfortunately

#

The Select button work just fine, Close button does too. It's just the scroll bar in the bottom

whole cloud
#

Thank you, I'll investigate

hexed verge
#

Anyone's linux server restarting after a mission finishes and you select a new mission recently?

i am on Arma 3 Console version 2.08.149595 x86 right now, so latest profiling version

the logs themselves are not useful as well. They show, game finished, then you see it waiting for next game, list stringtable.xml issues of missions, then you see the file being truncated and server restarted

log: https://pastebin.com/LLB5SF09

#

the No entry error is normal as we do remove the vanilla missions via a mod, iirc probably even a cba optional.

whole cloud
#

Are you on profiling v15 or v16?

#

A crash on mission restart was fixed in v17

hexed verge
#

I updated this morning via steamcmd

#

i would assume it is the lasted

#

notice how you announced version 2.08.149598 but mine says 2.08.149595

#

that is 4 commits behind

patent dome
#

I'm wondering how you got 595. Dev is 590, prof is 598, and our checks reported no version differences during publishing.

hexed verge
#

good question, I have this as steam cmd prompt since forever
steamcmd +login user password +force_install_dir /opt/fparma/arma3/server +app_update "233780 -beta profiling -betapassword CautionSpecialProfilingAndTestingBranchArma3 " validate +quit

#

i ran the update once more and it is still at 595

patent dome
#

Ah, I see we don't check the version of the Linux binaries when publishing...

hexed verge
#

mystery solved?

patent dome
#

As to how you have an older build, yea.

#

Correct versions should be available on steam soon.

patent dome
#

Correct version should now be available on Steam and GDrive.

hexed verge
#

lemme check

#

Arma 3 Console version 2.08.149598
yup that is now the latest

mossy anchor
#

When applying textures or materials to the falcon drone, while they apply them when leaving a area of around 200-300 metres they will reverted back to their normal camo and only comes back once getting close to them which doesn't happened to any other aircraft or vehicles.
Also discovered that both hurons (unarmed and armed) camos can't be swapped over to the different camo with the use of comps as they are reverted back, armed variant unable to swap to black and the unarmed unable to swap to green once saved as a comp then placing them down. Also testing setobjecttextureglobal on the hurons doesn't work on them as well.

restive pilot
#

I just noticed it

whole cloud
restive pilot
#

I'll check. wait one

whole cloud
#

actually, horizontal you mean right? not vertical

#

vertical is what was fixed and what me and QA tested

restive pilot
#

I meant horizontal yes meowsweats

#

is there any part of the game that has a horizontal scroll?

#

I can't think of any 🤔

restive pilot
pallid pebble
#

The company overview maybe? But I think thats just scrolling to zoom

#

as in the one when you click on the arma logo

#

(I think thats where it is at least)

restive pilot
#

no it doesn't have one

pallid pebble
#

hm

restive pilot
#

you just rclick with mouse and drag

hexed verge
#

(unrelated to the current discussion) the server crash between selecting new missions on linux servers is now gone

unborn urchin
boreal wigeon
#

There are vertical and horizontal scrollbars when you click on the DLC's (jets for example) on the main menu.

tribal pasture
#

Speaking of scrollbars, it would've been useful if mouse wheel scrolled horizontal scrollbar when there is no vertical present

tribal pasture
#

Or have a config value for it

#

Probs can script it though

feral harness
#

By default? I dont think changes like that that affect every user would be approved. On the other hand scripted solution is available and can be provided per custom control

plain trout
whole cloud
plain trout
restive pilot
#

yeah both vertical and horizontal scrollbars are broken (can't be dragged with mouse)

whole cloud
restive pilot
#

I guess it wasn't noticed because most people use the mouse scroll wheel

#

I don't think there are any vanilla examples for horizontal ones... meowsweats

rugged owl
whole cloud
#

So I reproduced the horizontal one.
Arsenal vertical works, Eden entity list vertical works, eden attribute doesn't, and even scroll wheel doesn't work for me

#

Yeah I had a stoop moment obviously

#

if (out of bounds) {
if (in scroll bar) processScrollBarClick();
return;
}

I wanted to get rid of the out of bounds check, and just commented out all of it meowfacepalm

#

its only controlgroups so thats why listboxes still work

restive pilot
whole cloud
#

fixed nau :wobcat:

timid goblet
#

Was there any changes to setposasl or setdir by chance?

#

Within the last month in a public update that is?

restive pilot
timid goblet
#

Ah just curious, having some weird issues with it out of nowhere.

#

Specifically the changes not syncing over a server, but I'm thinking maybe it just needs to be executed differently now.

dapper heron
#

@whole cloud when using _ctrlGroup ctrlShow true on a group type control, is it intended behaviour that all child controls are shown, even if they were previously hidden specifically? If it is intended behaviour, any chance of getting an alt syntax to prevent this behaviour please? There are obviously messy workarounds but I don't feel clean using them 🙃

restive pilot
dapper heron
restive pilot
#

ctrlEnable

whole cloud
silk pewter
#

@whole cloud "MapSingleClick" Mission EH doesn't have _ctrl value, worth adding? (even though it is now in a different order than other EHs)

#

(shift/ctrl/alt vs alt/shift)

whole cloud
#

don't we have a "is key pressed" script command?

vagrant zodiac
#

Noticed something odd that's snuck in with the last update or two on the performance branch:
Tested and confirmed when running just ACE3 and CBA
Equip a Kestral 4500.
Show (via hot key or self interact etc.) Kestral and as opposed to it staying on screen until you ask for it to close, it stays open for only a couple of seconds.

Double checked that performance returns to normal on the stable branch.
This feature exists in SP and MP.

We've been experiencing other ACE3 oddities on main Ops starting at around the same time but none of these have been repeatable / consistent so might be completely unrelated. But perhaps hints that the above is more 'bug' than 'feature'?

whole cloud
#

I noticed that too notlikemeow

whole cloud
#

Helo, anyone running Linux server perf/prof on 32bit?

gritty wasp
#

@vivid rune told that they use phenom. Not sure if it is perf

hexed verge
whole cloud
#

We're upgrading our linux compiler with next profiling update.
It shouldn't break anything but just in case you notice something off report it meowheart

whole cloud
#

2.08.149626 new PROFILING branch with PERFORMANCE binaries, v18, server and client, windows 32/64-bit, linux server 32/64-bit

  • Fixed: Game was unable to launch on CPUs from 2008 or earlier
  • Fixed: Clicking on scrollbars in ControlGroups was not working

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

whole cloud
restive pilot
#

Game was unable to launch on CPUs from 2008 or earlier
was it supposed to?! 😅

autumn timber
#

AFAIR according to the minimal requirements, yes it was 😦

#

Processor: Intel Dual-Core 2,4 GHz or AMD Dual-Core Athlon 2,5 GHz

patent dome
whole cloud
#

aa

silk pewter
#

my still working i7 930 begs to differ, Arma 3 works "fine" (on Windows)

quaint flame
autumn timber
quaint flame
#

I was, until I got an arrow in my knee.

pallid island
mossy anchor
#

Seems to be an issue when applying textures or materials to the falcon drone, while they apply them with the comp however when leaving an area of around 200-300 metres they will reverted to their normal camo and only comes back once getting close to them which this issue doesn't appear to happen to any other aircraft or vehicles.
Also discovered that both hurons variants can't be swapped over to their other camouflage which also implies to swapping their texture or material when placing them down as comps.

gritty wasp
#

Do you really think that writing same message 3 times helps? Maybe you should try fourth?

silk pewter
gritty wasp
silk pewter
#

oh, got it indeed

mossy anchor
#

Welp it a issue that still hasn't been resolved and I haven't gotten anything back for it so I been waiting a few days for anything as it usually appears to be flooded by other issues happening at that time and with them being fixed so I just been patient and ensuring that it doesn't get forgotten by other messages.

restive pilot
silk pewter
mossy anchor
#

Understood will do for now on with bump messages, also will attempt the sleep on the drone though I not used a sleep script before

restive pilot
#

in other words, if you do it in object init, use spawn

mossy anchor
#

I believe I get what to do, I attempt to do it on the Drone to see if it resolves the issue

mossy anchor
# restive pilot try some `sleep` since a few updates ago attempting to set textures and material...

Turns out I have no idea on working code involving sleep or spawn xD, also to give a description on the Falcon Drone issue while the material/texture does apply to it and can be seen the view distance of it has a range of around 200-300 metres before it shows the texture of the normal nato falcon which with the texture should being a csat one and hostile to nato means it will be misidentified as friendly.

night mulch
#

Its most likely a problem with the model p3d, the faraway resolution lods probably dont have a hidden selection assigned

mossy anchor
#

That could be the issue

whole cloud
vagrant zodiac
# whole cloud https://github.com/acemod/ACE3/issues/8954

Thanks. We think (but need to retest to confirm) that one of the other bugs (stamina system breaking) is related to assisting digging trenches.
Will attempt to test this with profiling and stable version some time this week if possible

rain moth
whole cloud
#

2.08.149650 new PROFILING branch with PERFORMANCE binaries, v19, server and client, windows 32/64-bit, linux server 32/64-bit

  • Tweaked: Lights with dayTime and flare now automatically increase their brightness during daytime in order to stay visible - https://feedback.bistudio.com/T161893
  • Fixed: Possible crash on savegame loading

If you don't want to use the Steam branch, the files are also available for alternativedownload here:
https://drive.google.com/drive/folders/15p9j7C2nHUt6NoVfChX4YFuqzFXzblJh

whole cloud
night mulch
#

Yesssss

fickle spade
#

just started today morning on previous version of profiling branch and got a reverted user's IP in logs

#

returned to master branch

whole cloud
brave barn
#

IP connected players is reversed
My IP on localhost 127.0.0.1
On connect to local server rcon ip is 1.0.0.127

Friend IP 134.19.135.139 but on server is 139.135.19.134

Started few profiling versions ago

fickle spade
#

octets are swapped

whole cloud
#

where in server log? Where is it displayed? what example log message

fickle spade
#

few minutes

brave barn
fickle spade
#

Linux server. In console log got messages

12:25:27 BattlEye Server: Player #15 ХАКЕР (100.189.128.45:9) connected

Normal output

12:54:36 BattlEye Server: Player #3 ХАКЕР (45.128.189.100:2304) connected
brave barn
#

Broken since version 208_149459_v14
In version 208_149367_v13 IP works fine

whole cloud
#

ok so battleye, not profiling