perf_prof_branch

[fallow sky]

Just sent it in dm

[smoky ridge]

second

[whole cloud]

A single server having an error in their server.cfg is messing up my graph 😄

[heady plinth]

add backdoor into a3 so you get access to their server and secretly fix it

[silk pewter]

reminds me of these white hat hackers who enter old routers to update the firmware and fix security breaches ^^

[autumn timber]

Reminds me of these white hat hackers who enter old routers to update the firmware and fix security breaches so that only their backdoor runs on that machine.
Fixed that for you.
Wait... I think the color wasn't really white 🤔

[silk pewter]

I was only aware of the good ones 😄

[glass lagoon]

apologies for the kind of question, if there is anything i could rely upon to approximate, preparing comparative analysis, on what engine thinks frametime is?

if i'd try RT kernel for lulz, to see something within context of it

[idle wren]

Can confirm this has fixed connectivity issues on Linux servers.

[fallow sky]

And I can confirm this didn't fix the launching time
Dunno if it was supposed to, but it didn't

[whole cloud]

I still don't know whats up with the launching time

[dusky spire]

Reverting to stable had no effect on the launching time for some reason.
Confusing

[fallow sky]

[hard vault]

I timed how long the game takes to load (huge modpack). My load times went from around 12 minutes between the two loading screens on v23, down to just about 2:17 total when reverting to v21

[static briar]

Is this problem caused by the new Linux compiler?
Reverting to v21 fixes the launching load time issue and everything works fine for me too.

[whole cloud]

If the launch time on windows is caused by the new linux compiler? I doubt it..

[static briar]

That's certainly true. I don't want to believe that a Linux compiler would have any impact on Windows.
Could it be that team accidentally modified the code for compiling on Windows as well?

[static briar]

I think you already notice but I tell you, Stable haven't "fast load patch" yet. so you need to revert to pref/prof v21.

[inland dew]

https://videocardz.com/newz/steam-ends-32-bit-operating-system-support-in-2026

nobody will be able to complain about devs removing a3 32 bit support

[patent sky]

We waited long enough for steam to do it for us

[whole cloud]

Yes. But I went through all the relevant code

I will try to repro it tomorrow

[whole cloud]

Performance binaries, my unit's default modpack

v21 First 58s, Second 46s
v23 First 72s, Second 47s
Stable First 68s, Second 47s

[whole cloud]

More mods
Stable 2m58s, 2m11s
v21 1m35s, 1m17s
v23 2m49s, 2m04s

Ok I see now.
Seems something is slow on the config merge step

[whole cloud]

Huh, I can already see a bad slowdown in a build from 10 days ago. One day before v21.
1m42s though so matches v21.

CPU time is 88s, out of which 18s are spent spinning because the async RPT logging is backed up and needs to increase the buffer size.
This is a bit stupid anyway, I wanted to avoid memory allocations from RPT logging. But in kind of a stupid way. Should just replace it with a pool allocator and get rid of half of that stuff.

This is only bad if loading has lots of RPT logging.
Which you only get if you load large broken modpacks, which I've seen in Jay's logs. And which explains why my units main modpack is barely affected

[silk pewter]

https://tenor.com/view/mujikcboro-seriymujik-gif-24361533

[static briar]

Yeah, it seems that using mods with "malformed" class inheritance config causes an unusually long delay when the game logs the config file changes "Updating base class" to the rpt file.
This long loading time problem did not occur with mods that simply had large file sizes with out not complex configs.

[unborn urchin]

would be nice if people stop ignoring those errors it's the cause of 'when I load x mod y mod stops working'

[whole cloud]

I just don't know why this issue is new..
There aren't any changes in this (except yesterday)

Though I did find a problem.
The RPT thread only checks for new messages every 500ms and sleeps most of the time.
It is supposed to be forcefully woken up if the buffers are full, but that wake up is commented out and has been for years and I don't know why.
So if the buffers are full of RPT spam, it just hangs up for almost half a second until it happens to wake up.
And buffer size is only increased if its been full 64 times.

[heady plinth]

maybe some message could be shown in the main menu if you detect a huge amount of rpt spam?

i dont mean popup. more like a small text somewhere on the side

i guess most modders never look at those logs, so they arent even aware

[fair ore]

big red message that says "please fix your mod lmao"

[whole cloud]

Big modpack with v24

First 1m02s, Second 57s

[silk pewter]

even better than v21

[whole cloud]

The second load time could be 1.5 seconds faster if mods were using sqfc

Another 1 second for non-binarized stringtables
Another 2 seconds for non-binarized configs 🙁

[silk pewter]

can't mod modders ¯_(ツ)_/¯

[whole cloud]

CRISPR disagrees

[ancient smelt]

With the latest version, I started noticing more often that if you're in Zeus, when you're hosting an event for players, the server can "kick" you and not let you in anymore, maybe something is known about this? What kind of data is needed?

[whole cloud]

what kinda kick

what message

[ancient smelt]

"Failed to connect to the server"

[whole cloud]

At the kick i mean

not after the kick

[ancient smelt]

And you can log in only after the server restarts.

One moment

[ancient smelt]

14:42:20 Ref to nonnetwork object 1846356: shell.p3d 3AS_82mm_HE_Mortar
14:42:20 Ref to nonnetwork object 1846359: shell.p3d 3AS_82mm_HE_Mortar

[whole cloud]

nope not it

[ancient smelt]

14:42:03 Client: Network message 190add (update info 1f934677438) is pending
14:42:08 Client: Network message 190fa2 (update info 1f934677438) is pending

[whole cloud]

Sir. The kick message.

Not random log lines

[ancient smelt]

How do I get it? (Just a notification popped up and I closed it...)

[ancient smelt]

14:33:35 Player 1201681629 ([ARF-11] CPL 05-2323) Exceeded anti-flood limit

[whole cloud]

Don't close it

[ancient smelt]

Is it?

[whole cloud]

AntiFlood kick is by default disabled.
If you enabled it in your server.cfg, then turn it back off or increase the limits

[ancient smelt]

Is it about this server-side .cfg?

[whole cloud]

The server cfg yes

[ancient smelt]

And what is this parameter called in the file? (forgive me for the stupid questions, really, not an IT guru)

[whole cloud]

AntiFlood

[ancient smelt]

And if there is no such parameter?

[whole cloud]

Then its off and what you posted is not the kick reason

[ancient smelt]

Perhaps that's when the reason for the kick:
14:43:15 Kicking player name='[ARF-11] CPL 05-2323' ID=1201681629, steamID=76561198103741753. Desync too high (243/ 90 )

[whole cloud]

Yup. That would be it

maxDesync is a server.cfg entry.

[ancient smelt]

The problem is that I have a good Internet connection, like other players, it happens, but rarely and very unpleasant.
Do you propose to bet more? Raise it to 300? (as far as I've read, this can lead to problems on the server)

[whole cloud]

It says it was 243/90 so
I have no proposals

[ancient smelt]

But what about this situation?

[whole cloud]

Yes

[ancient smelt]

I'm using a translator, and he didn't translate it quite correctly. What should I do in this situation? Why can it jump so high?

[whole cloud]

[ancient smelt]

https://tenor.com/view/dog-crying-meme-doggo-crys-megan-soo-crying-dog-gif-5276199764143986284

[heavy vortex]

I don't think we ever had a description of what the desync numbers mean.

[whole cloud]

I did describe it once

but short, magic value.

[heavy vortex]

IME the more obfuscated a mechanic is in Arma, the more likely it's been bugged for two decades.

[whole cloud]

We like bugs

[ancient smelt]

I wanted to express my gratitude for the help, thank you. Also, do you know if there is any way to force Arma 3 to use more memory than it does on its own?

Roughly speaking, I have 32 GB of RAM, but Arma 3 consumes 2.5-3 GB.

[heavy vortex]

A3 will happily use more memory if it has a need for it.

Fly around the map a bit or load a ton of models and you'll see it rise by 2GB or so.

[whole cloud]

-maxFileCacheSize parameter, will make it waste a bit more memory with basically no benefit to you

[autumn timber]

Also, sanity check: you're using the 64bit binary, are you?

[fiery crescent]

Since we have more RAM / CPU to waste these days how about stop AI to see through vegetation foliage 😁

[heavy vortex]

We don't have that much CPU to waste.

[carmine stump]

I bet "NO".. is the answer.

[autumn timber]

If so, then the problem will probaby fix itself around 2.22 🙃

[obsidian condor]

Or Steam removing support for 32 bit systems in general next year depending on which happens first 🙃

[silk summit]

Yeah and fixed collision too, so they don't phase through walls and solid objects

Perhaps an extra check that detects if they are inside another model so they are moved in the correct not occupied position while also disabling their ability to shoot while in that intermediate state

Whatever Enfusion did to fix that issue really

Devs have been so effective and consistent at targeting and fixing issues (sound issues, Ai landing and Ai Driving through rivers were some of the latest most recent remarkable fixes/achievements) that eventually only the bigger legacy issues will remain:

a) Ai Collision
b) Ai visibility
c) Inventory Rewrite

[obsidian condor]

Would be so nice if the AI (especially in my case agents) did not walk through walls and doors

[heavy vortex]

Isn't that a workaround for pathfinding failure cases anyway?

I'd rather it had more of those, not fewer. AIs failing to get from A to B is far more of a problem for me than them occasionally cheating.

[silk summit]

I would rather have them not getting to a location (and recalculating path) rather than cheating and shooting from impossible places or places where players aren't expecting them. Still rather than choosing one why can't both be tackled?

[empty goblet]

erm, AI wasn't seeing thru vegetation since A1 or A2 (aka since the view blocking volume was added to models)

so you end now with 3 different problems

1. badly configured model (for some reason the object is transparent (if the volume is missing the collision / total object volume is used instead), wrong shape (blocking volume != visible volume)

2. AI awareness (information sharing from another AI and memory (already seen you and remembers you position)

3. possible AI bug (shall not see thru correctly configured blocking volume object or similar) or on step 2. (too much info shared / remembered / over reaction to events (ie bug where using detonator instantly revealed your location to all AI nearby)

[heavy vortex]

It's usually just a view model vs rendering mismatch.

AIs don't see the same as players see.

There's a fair amount of variance. CamLaoNam for example feels a lot more player-friendly than Tanoa.

[opal hound]

There's a fairly popular mod that tries to address it, but you're definitely right that a lot of people just don't understand how the AI functions

[empty goblet]

ofc i will repeat my lament of missing better shadows (till horizon) and lighting (to fix lightsource leaking thru objects etc.)

[silk summit]

Is "lightsource leaking thru objects" still a thing? I don't remember having those as much anymore

Dynamic Lights on light sources (other than the sun) would be so cool thou

Extended Multithreading certainly openned up so many possibilities now but I don't think the devs have that much time on their hands

[autumn timber]

Since Forums are up (yay!), may I suggest this excellent post by Gunter Severloh: https://forums.bohemia.net/forums/topic/223493-ai-facts-myths-compilation-list/

See: AI Detection: Vegetation, but honestly, the whole post is worth a read

[obsidian condor]

I don't think an AI just blatantly walking straight through the big doors on the Altis grey warehouse is any form of decent workaround for pathfinding failure cases...

[heavy vortex]

It's preferable to refusing to enter the building.

[obsidian condor]

I disagree tbh... you know if you hear a door open, when they walk through a door without opening it you lose situational awareness

And if you are looking at the time, it is unimmersive

Really is the only thing I really dislike about the AI/agent AI tbh though

[heavy vortex]

..

[obsidian condor]

Apart from them not seeing shadows... ... ... Hurry up. Going to bed soon 😉

[heavy vortex]

Nah, if that's the only thing you really dislike about Arma AI then we simply don't have enough in common to debate :P

Offtopic unless Dedmen decides to fix anything related anyway.

[obsidian condor]

I did say really dislike and did not say that I did not tweak AI 😛 But yeah, OT and Dedmen has said no AI changes - don't think anyone knows how to modify RV engine AI stuff and Enfusion's AI is not good enough for ARMA 3 yet

I'd rather have a new renderer in ARMA 3 than AI changes at this point though

[empty goblet]

maybe add extra event which triggers the door opening when AI just 'ghostwalk' it

[real creek]

https://steamcommunity.com/sharedfiles/filedetails/?id=3542444275
There was this a month or so ago. I have no idea how well implemented it is

[leaden relic]

It's not well implemented, last time I looked at least

[odd granite]

Sooooo my milsim just went to this branch and now all of my xbox controller binds are WACK (I use it to fly helicopters) and idk what to do, arma doesn't even show my controller as being detected, but I can still bind stuff with it. When I try to bind stuff with it, for example, my xbox Y button gets bound to my keyboards F key for whatever reason. Please help

[analog acorn]

You might want to check your controller settings in Steam

[odd granite]

I could kiss you

You're my favorite person

[knotty plinth]

Im also for it, last thing in Arma 3 for me to be fixed

[fiery crescent]

Is this expected behavior? 12 gen i5, dedicated server utilizes 2 cores to about CPU 0 -90%, CPU 2 -100% during heaviest loads. The rest of the cores hover around 5%.

[fiery crescent]

Perhaps it is due to the processor type. Just checked and it has 2 Performance cores:
Intel® Core™ i5-1235U
CPU Specifications
Total Cores
10
of Performance-cores
2
of Efficient-cores
8
Total Threads
12

[heavy vortex]

What are the "heaviest loads"?

[fiery crescent]

200 +/- AI and 10 players.

This is in Escape Missions.

[heavy vortex]

I'd expect a bit more load on the e-cores then. You're not using any weird command-line parameters?

And cpuCount in the RPT is 10?

[fiery crescent]

no weird config. It is default config, MaxMsgSend 128. All network settings default. Tried with HT enabled and disabled seems the same.

let me check RPT

Allocator: C:\Arma 3 Server\Dll\tbb4malloc_bi_x64.dll [2017.0.0.0] [2017.0.0.0]
PhysMem: 32 GiB, VirtMem : 131072 GiB, AvailPhys : 28 GiB, AvailVirt : 131068 GiB, AvailPage : 33 GiB, PageSize : 4.0 KiB/2.0 MiB/HasLockMemory, CPUCount : 12
CPU: 12th Gen Intel(R) Core(TM) i5-1235U
OS: 10.0.19045 (1)

looks like it is seeing 12 cores

I'm still on Version: 2.20.153169

[fiery crescent]

Since CPU has 10 cores how is the count 12? Is it Threads? Perhaps because I left HT enabled or 10 Performance cores + 2 Efficient cores.

[heavy vortex]

Count 12 should only happen if you used -enableHT or forced -cpuCount=12

Do not use -enableHT.

[fiery crescent]

Yes, turned the HT ON during troubleshooting.

Without HT:
Allocator: C:\Arma 3 Server\Dll\tbb4malloc_bi_x64.dll [2017.0.0.0] [2017.0.0.0]
PhysMem: 32 GiB, VirtMem : 131072 GiB, AvailPhys : 28 GiB, AvailVirt : 131068 GiB, AvailPage : 33 GiB, PageSize : 4.0 KiB/2.0 MiB/HasLockMemory, CPUCount : 10
CPU: 12th Gen Intel(R) Core(TM) i5-1235U
OS: 10.0.19045 (1)

Well I guess it will only utilize 2 cores.

[void badger]

Bro is running server on a laptop

I was gonna say desktop i5s shouldn't have E-cores but that would explain it

[fiery crescent]

Well, half correct it is small factor CPU used in laptops, but actual box is mini PC. I just switched to 13th Gen i7 also small factor CPU and it blasts 6 cores right now. Something about that i5 12th generation. It must be the 2 performance cores.

[heavy vortex]

In theory that thread director thing might force threads onto the two perf cores. That would be bad though.

[latent socket]

Apologies for bumping an old post. My group is also running a fairly large modpack and v23 severely slowed our launch times from around 10 minutes for the server and 12-15 for the client to now 40 minutes for the client and around the same for the server. Around the same time as the regular branch

[void badger]

I'm just using v21 on my Linux VM dawg

I'm getting ready to stress test it with Zombies and ChemWar this week but it's been rock solid so if it ain't broke

[jagged swan]

Recently, some changes were made to the pylons on the prof branch and after that looks like something start working not as it should be.
Its about getAllPylonsInfo command. On the vanilla Mi-48 Kajman if every pylon is assigned to driver it returns [-1] for Assigned turret value in Singleplayer but in **Dedicated **with the same heli and pylons assigned to driver it returns [0] for Assigned turret value

Executed command:

getAllPylonsInfo objectParent player

Different result of executing command depends on is it singleplayer or dedicated (comman executed locally on player machine)
Singleplayer:

[[1,"PylonLeft1",[-1],"PylonRack_4Rnd_LG_scalpel",4,"0:10000347",[[2.50423,-0.214772,-1.37952],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[2,"PylonLeft2",[-1],"PylonRack_19Rnd_Rocket_Skyfire",19,"0:10000349",[[1.8108,-0.15553,-1.32298],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[3,"PylonRight2",[-1],"PylonRack_19Rnd_Rocket_Skyfire",19,"0:10000350",[[-1.81321,-0.15553,-1.32298],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[4,"PylonRight1",[-1],"PylonRack_4Rnd_LG_scalpel",4,"0:10000351",[[-2.50664,-0.214772,-1.37952],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]]]

Dedicated:

[[1,"PylonLeft1",[0],"PylonRack_4Rnd_LG_scalpel",4,"2:10000347",[[2.50423,-0.214772,-1.37952],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[2,"PylonLeft2",[0],"PylonRack_19Rnd_Rocket_Skyfire",19,"2:10000349",[[1.8108,-0.15553,-1.32298],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[3,"PylonRight2",[0],"PylonRack_19Rnd_Rocket_Skyfire",19,"2:10000350",[[-1.81321,-0.15553,-1.32298],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]],[4,"PylonRight1",[0],"PylonRack_4Rnd_LG_scalpel",4,"2:10000351",[[-2.50664,-0.214772,-1.37952],[-0,-0.997732,-0.0673187],[-0,-0.0673187,0.997732]]]]

If player is pilot in this heli so his turretPath is equal to [-1] and if all pylons is assigned to pilot then **getAllPylonsInfo ** should return [-1] value for Assigned turret. It works well in singleplayer but in dedicated - not

[heavy vortex]

At which locality did you do the pylon assignment?

[jagged swan]

pylons was assigned in 3den editor

[whole cloud]

Lots of RPT spam?

[whole cloud]

"If it ain't broke" doesn't work that well with profiling branch.
Because generally everything goes to stable and then you'll be forced to switch and it's too late to fix it

[whole cloud]

Are you sure it was not broken before the pylon changes?
Or did you only notice the error now, and attribute it to probably be caused by the pylon changes

[magic onyx]

doing the lords work out here

[fallow sky]

Great work. The game starts way faster now! Thanks

[whole cloud]

I can reproduce it on stable, so this is not new.
You can make a feedback tracker ticket.
Important to note that the actual functionality works correctly, the pylon weapon is assigned to the pilot seat, even if the pylon says its on the gunner.
Not sure if we'd fix that, the other pylon commands are generally local effect anyway, and thats just what you're seeing here

[drowsy fern]

Good morning, seems the new prof may have an issue. I have tested with 2 scenarios, and stable branch. Issue didn't appear in stable. After playing for a few minutes on perf the game will freeze with this error:
7:41:57 No alive in 10000 ms, exceeded timeout of 10000 ms

It produced frozen dumps, but I wasn't sure if those got sent automatically

[whole cloud]

Thats the kinda thing that "no noticeable changes expected" means 😄

[whole cloud]

They don't, you can pack them in a zip and send them to me

[eternal kraken]

have the same issue, send you 3 dumps 🙂

[whole cloud]

yeah I see it, on it

The freeze is in custom implementation of a synchronization thing, that Windows actually has natively included, but we couldn't use because its Win8+
Would certainly be nice to switch that over.
But also it can't be the issue because it worked before :harold:
And for extra fun, the thing that broke now was actually not touched by the refactoring

[whole cloud]

Ah yes. Replacing a 1 with a 0 fixed it, two YAAB runs with no problems.
Changing it back, the first YAAB run doesn't even make it half way.

That's good to know, but I don't know why 🤣

Gosh darned be it

I changed 14 thousand lines, and THIS 2/3 character bad copy-paste is the mistake that makes the game freeze.
If only someone would bother to write unit tests

[eternal kraken]

no freeze anymore with the latest update(10 minutes ago)

[whole cloud]

We reverted, new one in ~60m

[uncut linden]

🤔 Should I send it here or in some other channel?

xx:47:57 Error in expression <_pivot","_a","_b"];

private _radius = (_a max _b) * 1.42;
private _objects = []>
xx:47:57   Error position: <_a max _b) * 1.42;
private _objects = []>
xx:47:57   Error Undefined variable in expression: _a
xx:47:57 File A3\Modules_F\Environment\HideTerrainObjects\init.sqf..., line 58

[whole cloud]

run game with -debug to see where it comes from.
Is it actually us?
Is it new on prof? if not then #arma3_feedback_tracker or #dev_rc_branch

[proud cape]

v0?

[whole cloud]

🤐 😄

[void badger]

I'll give v25 a shot 🙂

[jagged swan]

Yeah. Functional works as expected, u r right. But this command helps to determine which pylons assigned to turret. And its working not as it should be. Unfortunately, there is no alternative commands which can do the same.

[whole cloud]

remoteExec on the owner of the vehicle.
But I also think when the vehicle owner changes (player getting into pilot seat) it would still stay broken

[jagged swan]

Yeah. We use it in ACE for missile hud and now i understand why sometimes its not loading proper hud for some pylons...
Anyway thank you for your time

[void badger]

68 seconds to load on v25

[acoustic lynx]

Anyone else experiencing AI tracked amphib vehicles driving squiggly curving paths in water instead of a straight line?

[latent socket]

I turned RPT logging off and even with v25 my launch time is still around 40 minutes

[whole cloud]

Do you know how it was before?

[whole cloud]

Well RPT logging was the only cause I could see.
If there is another different cause, then I cannot reproduce it

Check that your CPU count is set correctly (or not at all)

[empty goblet]

erm, just wondering, what's the drive (HDD, SATA / SATA m.2 or NVME m.2) ?
also could you post your commandline parameters ?

[oblique hearth]

Hey, i did host a dedicated server via the ticking the "server" option in the arma launcher. Then I wanted to use some profiling commands and realized the server commands did not work (after renaming the exe files accordingly).
Interestingly, Task Manager reported two instances of arma3_x64.exe running, so it seems the server doesnt use arma3server_x64.exe. Now my question would be, how to run e.g. #captureFrame on such a server?

[whole cloud]

The Arma Launcher, indeed starts the client executable

If you rename that one though to swap it with the profiling one, profiling would work there too

[oblique hearth]

But from my understanding, there's separate profiling commands for client and server. I tried the client commands and they worked client side, but serverside nothing happened.

Maybe the output is redirected somewhere where i cant see it?

wait, I'm just going to try again.

Also one more question: Are there any negative side-effects of the arma3 executable being used for the server, or is it still a fully functional dedicated server? I love the convenience for some small coop-play, but if it causes issues somewhere else, i could also start the server executable directly. I know hosting a server from the ingame menu can cause issues, that's why i'm asking about the launcher one too.

[oblique hearth]

Ye diag_captureFrame 1;, shows the chart when i press "local exec", but nothing when i press "server exec"

[obtuse bramble]

I run my arma through an NVME m.2, no launch paramaters except show static and skip logos, no pause, no pause audio.
i9 13900K, V21 gave me (identical mods) 22.5 seconds launch time, logging off/on didn't matter.
V25* is still about 2-3 minutes for me, hard to really say if anything* between 22-25.

[whole cloud]

Fully functional

[whole cloud]

It cannot show chart as server has no UI.
Its written to a log file next to RPT
You can also use the capture to file functionalities

[oblique hearth]

I found it. There was a misleading .rpt file in another directory, so i only looked there, but the true RPT files were in appdata. thank you!

[whole cloud]

Decoy logfiles to keep you on your toes!

[unborn urchin]

• Fixed: setDamage 0 would not properly reset the "alive" state of most Entities
  does this actually heal vehicles

like go from destroyed to alive

[opal hound]

When you say report, reports it to you?

[whole cloud]

You find out!

[whole cloud]

Using the crash reporter, but automatic.
It creates a super tiny (50KB) minidump and sends it. Only 4/5 per game run to keep down the spam

[unborn urchin]

i hope not I was using that as a trick to make tanks be dead but not use the full destruct texture, rip

[whole cloud]

I suspected it might be related but I don't know if it is

If it breaks it we'll revert it, just make noise

[opal hound]

Hm not as many of those in my RPTs as I thought, mostly
Client: Object X:YYYY (type Type_ZZZ) not found and Setting invalid pitch 0.0000 for <unit> REMOTE
I can turn it on anyway

[latent socket]

I'm collecting that data from other group members so you have some more data

[empty goblet]

could always go for -1 instead of 0

[knotty wraith]

They were talking about shooting through fir trees - here's an example.
https://youtu.be/EXFTUfLMtDU

[ashen helm]

Do they do so in vanilla?

[unborn urchin]

it does bring them back to life, oof. maybe add a parameter to setdamage?

[whole cloud]

@rose moth can read that tomorrow. If our alternative works we'll do that. Otherwise the -1 idea does sound nice too, but is very annoying to implement.
Or we could add a new command, but I don't like new commands 😄

Mh indeed we already have setDamage alt syntax, yeah we could use that if we need this

[fair ore]

resurrection would be hilarious

[unborn urchin]

its quite clean with it too, gets rid of destruction effects etc

but yeah its gonna badly break spearheads tank destruction stuff

[analog acorn]

Does it work on units or just vehicles?

[unborn urchin]

didnt try it on units

one sec

probably funky on players but ill try it on an AI

uh, seems to work?

[analog acorn]

That might be handy for some purposes

[whole cloud]

What? it can revive dead AI too?

[unborn urchin]

yeah

[whole cloud]

Oh lord

[unborn urchin]

I dont know how functioning they were

as i didnt try very hard i just plopped one down, killed it, brought it back

didnt try giving it waypoints or anything

[whole cloud]

I have seen both Animal and Man types have specific code to handle being revived.
Even though it wasn't possible (something broke it between 2003 and 2006)

[fiery crescent]

I hope it doesn't break revive in Escape missions, remember exactly 1 year (maybe 2) ago we dealt with a revive bug after A3 update.

[analog acorn]

"this can revive things" doesn't mean it's actually touching the scripted revive system.

Also if you look at what the change actually is (setDamage can restore units from 1 damage to life) there's no reason it would break the revive system anyway.

[fiery crescent]

Fingers crossed

It just reminded me hours of trying to fix the revive system. Someone eventually did it or Dedmen fixed it. Don't remember now.

[whole cloud]

Revive system sounds like something KK would've fixed

[analog acorn]

The revive system doesn't operate on dead units at all.
Depending on whether the revive system has handling for it, it might "break" it in the sense that the revive system won't start working again on a unit that has been returned to life, but I don't trust that this works properly on players anyway.
This change doesn't really have any potential to break existing use of the revive system. Especially if it's converted to an alt syntax rather than a default behaviour, as proposed.

[unborn urchin]

yeah all this breaks is doing funky stuff like making a dead unit/vehicle look alive

because now its gonna be actually alive

revive you never die in the first place

[whole cloud]

I do prefer the vehicle being able to revive though. I usually use that when I accidentally crashed the car into a wall and exploded it, trying to fix it before anyone notices

[heavy vortex]

I guess this fixes the issue we have where after doing setDamage 0 on a building it looks fine but the killed event handlers won't fire if it's destroyed again?

Well, Killed EH fires but with objNull. EntityKilled doesn't fire.

[whole cloud]

that sounds like it.
But the fix was for AI entities, I don't think buildings count as that? maybe they do

[heavy vortex]

I'm gonna laugh if you fixed everything except the one issue that we had :P

[real crypt]

is anyone else getting weird sound bugs since the last prof update?

getting sounds playing from places they shouldn't

i.e we hear people opening their packs like 200m away as if it's next to us

Can't find anything else that's changed on my end

[molten berry]

Cant wait to script Jesus to get resurrected after 3 ingame days

[empty goblet]

custom map / custom tree models , could be other issues than just AI bug, also AI detects you even thru small holes (imho that's one of parameters which i wish was tweakable to lower AI attention / speed of ID threat )

[empty goblet]

undead are back

[empty goblet]

you accidentally introduced necromancy magic into A3 😁

changelog needs change :

• Added Reanimate dead spell
• Added Pristine object state spell

[latent socket]

Just launched, 40:26.05
SATA SDD
CPU Count: 2
Extra Threads: All checked
HT Enabled
No Logs
World: VR

[full nova]

Alchemist gamemode from WS:

[whole cloud]

CPU count 2, with all extra threads (which expects CPU count of 4 or above) with enable HT (which is blocked by CPU count) is nonsense.

Also CPU count 2????

[rose moth]

I always enjoy when a low priority fix to Firing Drill steel plate targets, ends up adding resurrection spells to the entire game 🧟
Let's revert the current change, and then pick your next poison:

1. New setAlive command
2. New setDamage alias with optional parameter to also set alive
3. Reintroduce setDammage and make only it resurrect
4. Let me fix the FD issue by script instead

[worldly badge]

obj setDamage [0, false] revives it? (I didn't red the entire context)

[unborn urchin]

it does yeah

[ripe willow]

I had hole D&D missions in Arma 3 and NGL arma could do that all it needs is just good melee system and magic system.
(NOT MY PICTURE)

[quaint flame]

I vote for setDammage, just because I can.😄

[whole cloud]

Its also perfect for backwards compatibility.
Because you can tell anyone who was using it "Well you've been doing it wrong, that's what you get"

There is also a 5th solution as alternative to 3, use the setArmor command

[heady plinth]

ah, setDam-Mage, as in, magical

[whole cloud]

   msg._who = who;
-  msg._dammage = dammage;
-  msg._immidiateEffect = imidateEffect;
+  msg._damage = pars.damage;
+  msg._immediateEffect = pars.immediateEffect;
+  msg._allowResurrection = pars.allowResurrection;

[opal hound]

I turned on Ref to nonnetwork reporting yesterday and triggered a few, did those come through properly?

[whole cloud]

No, I got zero so far 😰

[opal hound]

oof

not sure what happened then

[whole cloud]

There is crashreporter.log in RPT folder

maybe that.... oh wait

[opal hound]

There's no dates in it but the times don't line up anyway

[whole cloud]

Yeah nah should be fine. Until yesterday the crash reporter was missing because we accidentally deleted it

Could be that crashreporter.exe is not in game folder?

[opal hound]

it's there

[whole cloud]

Welp

[whole cloud]

On the slow loading times. I still didn't find THE issue, but I found an issue.

For every class that has a base-class, it iterates through all classes in the parent to find the base-class AND then again to find itself.
It tries to find its own index, even though two levels up we know the index and could just pass it as parameter 😄

If the base class is not in the parent class, but instead higher up, it will not find it.
And if it doesn't find it, then it doesn't even need its own index at all, so all the time searching it is wasted.

Looking at profiling, 99.98% of the time for setting the base class, is the class trying to find its own index.

So just pass the index along and use it directly.

Modset 1
Before: 43s
After: 41s :harold:

But the CPU load!

Mh also the 10s spike moved around, it shouldn't.

After second testrun: 41s still. :U

Duh the scale of the first is different 🤣 I didn't notice that

[empty goblet]

when you mention the steel plate ... i just realized does the setDamage 0 also repair damaged armor vest and other equipment ? 😁

[analog acorn]

I wasn't aware equipment could take damage in A3

[heady plinth]

if you wanted to position weapons etc. in the world as scenery, you'd place it, disable simulation and then setdamage 0, so it cant be interacted with. always has been

[empty goblet]

so it's also 😁

• Temporal shunt spell

[knotty wraith]

#arma3_ai message
Есть ли шанс на лазейку в этой команде скрипта обмануть ИИ?)

#arma3_ai message
is there a chance for a loophole in this script command to deceive AI?)

[latent socket]

I'm sorry, that was a typo. Its CPU count 4

[obsidian condor]

I love that whoever made the original misspelling of immediate, misspelled it two different ways on the same line. Solid work that 😉

[silk pewter]

we strive for regular performance

[heavy galleon]

Why not all of them AND https://community.bistudio.com/wiki/setPlayable ?

[whole cloud]

As far as I can see setPlayable should work, but only in slot selection screen BEFORE players connect in 😄
Where, afaik you cannot run scripts anyway

[heavy galleon]

That is useful in exactly 1 way

(knowing it is not useful)

[unborn urchin]

lol

[whole cloud]

Found a handful of things today for game start time.
But they only reduced the CPU load and didn't improve the actual end result time

[heavy vortex]

Is there any known modset to replicate the issue that isn't 300GB of banned mods?

[whole cloud]

So far I haven't seen one

[ashen helm]

Petition to sacrifice e in favor of another m so we just get setDammag

[heavy galleon]

Ok listen, how about Damnage?

(cuz only old af people play arma 3 at this point)

[cunning cradle]

I keep getting crashes after an hour cosistantly. Always "GPU Hangs", Access violations.

So many people with the same problem and I can't find a fix...

[heavy vortex]

Is that even the right Arma?

[whole cloud]

We have no "GPU Hang"

[knotty plinth]

About that dammage stuff, is it possible to set/make a global value for it and also for armor values? As example building has 1000 armor and cannon has 100 dammage = you need 10 shots to kill the building.. disabled caliber, speed, distance, penetration, component count(FireGeo) etc.. just simple values

[whole cloud]

no

[whole cloud]

Incoming, being able to profile the game start time yourself 🫂

[silk pewter]

feels like returning the issue to the sender: "hey my game takes ages to load - it takes… 30 min to load terribly-coded mods"
yeeeeees 😄

[whole cloud]

Oh right we need to blame mods, time to add per-mod seperation to the config and stringtable loading 😄

[silk pewter]

genuinely, this will help

[whole cloud]

I also figured out why sometimes captures get completely corrupted and show negative times.

We store the time since start of frame, in a signed 32bit int. Even though our input start/end times are unsigned 64bit.
So they just get truncated. It could only handle a couple seconds of time

I'm increasing the memory usage for profiling from 2MB to 64MB. (Might even go higher)
More precision, more detail.

[whole cloud]

Because there, the long pink parts at the left are file loading, they are very short scopes.
To be able to fit all scopes into the buffer, it keeps increasing the minimum scope length, and if too short its all ignored. So it can fit all the stuff in.
But that means we cannot see the disk reads later on

moar data. Still not enough

[fair ore]

guys I'm loading 500 GB of banned mods from shady Google drives why is my game running poorly

[whole cloud]

That moment when floating point precision is so bad, that scopes are merged together even though they should be after one-another

[silk pewter]

just turn the whole engine 64b, ez

[whole cloud]

There we go.
Hid some scopes. Increased timing precision more, int64 instead of 32, double instead of float.
Now you can finally follow almost all disk reads through the startup.

So you can zoom in and be annoyed at why vanilla doesn't use binarized stringtables

[quaint flame]

Time to binarize them😄

[whole cloud]

861ms for stringtable loading total.
If I enable the cache.. 134ms

The world is saved, I shaved off half a second of your game start time.
Rejoice!
(No I'll leave that disabled because I feel bad about writing that cache file to SSDs)
Although thats only ACE+CBA on a 11s game start..

The question should rather be. What are these 800ms doing nothing

[fiery crescent]

I enjoy following your logic flow Dedmen. 👍

[whole cloud]

Okey so while the initial terrain load is happening, loading the CfgWorlds/DefaultWorld, we would have enough time and space to run Vehicle preloading.

But then also the question, why is loading a empty default world (doesn't even have a wrp file) taking so long 🤔

Loading essentially nothing, is taking 1 second out of my 11 seconds game start time

Ahh.. SimulWeather initialization. Its loading its config, initializing/allocating its textures. And we just wait.
I'm not a fan of waiting..

[whole cloud]

Ugh, vehicle preload reads config entries which can execute scripts.
And weather init reads config entries which can execute scripts.

Both happen together and womp womp

Or I just slap a lock around all config script evaluation......
That will make them super-mega-slightly slower, but I can run config loads in parallel then (probably)

Why did I not have this idea before..
That would also allow me to parallelize the parseConfigs step (1.4 seconds of 11).

But its possible that I did have that idea before, but there is some other problem with it
RPT logging would be a bit woozy with different configs throwing errors in between eachother.
But then make your mod without config errors darn it!

[oblique hearth]

But to make a mod without errors, i need to read rpt

also need to check if weather init etc. do not implicitly assume lock around the entire operation, not just config access, so as to not disturb the global state of affairs

[whole cloud]

Almost everything assumes a implicit lock, by being in main thread only

Ugh annoying.
ConfigParsing all references the global script state. So cannot parse files in parallel because they lock themselves a scope for variables. Sad.

Mods with unbinarized configs and __EVAL statements, could be reading variables from other mod configs that were loaded before them (Even though that order is NOT given by requiredAddons, and instead follows the -mod= parameter order so they have no control and cannot rely on that ever working)

[quaint flame]

Looks like a lot of work for a faster startup time

[whole cloud]

Worth it though 🤣

What are the chances any mod used and depends on that feature that I'm sure is not documented anywhere and that anyone with a binarized config can't use anyway

Although, no, nevermind.
Variables are separated anyway.
Why does it use the global state then 😠

[oblique hearth]

If it's a rare occasion, write a parser to parse if the feature is used x)

[whole cloud]

Waste of time. I'll just change it, publish it and see if anyones mod breaks

[elder adder]

Another days problem

[whole cloud]

I think I can fix it by not the whole config file locking a script scope, but instead every __EVAL statement locking its own.
That makes __EVAL (which barely anyone uses, and binarized configs don't have at all) a bit slower, but lets me run all the rest in parallel

[whole cloud]

Mh.
No crashes, not even any asserts.

Vehicle preload now runs parallel to terrain and simulWeather init
Mildly surprising, but neat

[whole cloud]

Config parsing in parallel also doesn't work because it uses the same unsafe pool allocator as scripts variables do.
But luckily, since I last looked at this, I implemented a new threadsafe variant that I can just drop in there.
It doesn't work on 32 bit, but to hell with 32 bit 😄

Yeah well... Uh.
😄

[quaint flame]

Worth it lol

[whole cloud]

I moved the vehicle preload into a separate thread to run in parallel.
And parallelized all config parsing.
And the load time went from 11 seconds down to.. 11 seconds!

[quaint flame]

At lest config parsing is 0.1 seconds faster

[whole cloud]

Gotta see what it does with bigger modpack.
But likely will run into I/O limit

[whole cloud]

The major drag is that all config strings are de-duplicated.
And that has a lock on it. So every string read from there, runs into the lock.
But I can turn it into a read/write lock and hope that the majority of strings are actually duplicates that are already in there

Excuse me sir. Please.

Down to 10 seconds load time

Still most of it is in the string deduplication and the rest is I/O.

If we had enough ram to not worry about duplicate strings though :harold:

If I move the TexHeaders loading later, to run at the merge step where we have no conflicting I/O though 🤔 Mh.

[whole cloud]

lol 🤦‍♂️
If I have VPN enabled, my game starts slower.
Wuht?

Yeah well.
ACE mod loads https://ace3.acemod.org/version.html which freezes the game until the load is done.

TexHeaders parallel with parseConfigs
Configs 779ms, texHeaders 1s65ms

TexHeaders parallel with cfgMerge
Configs 682ms, texHeaders 395ms

👍

Sadly ACE raining in my parade
I have the tools now with the script handle stuff to make htmlLoad suspend in scheduled and run async. But thats unscheduled script

That's maybe a interesting idea. The engine can now internally create a script handle and waitUntil on it, without you ever noticing. To turn some long-time script commands into async over multiple frames.
But I cannot think of much more than htmlLoad to benefit from it. callExtension I guess?

[whole cloud]

yo that works

Big modpack
10 second faster launch
58.5 -> 49.5 (49.0 with stringtable cache but I said I'd leave that off)

parseConfigs 6.1s -> 2.4s
vehicle preload only cut off 1s by running it parallel to the SimulWeather init

Technically I can probably run parts of vehicle preload in parallel because we can now MT config reading.
But there are so many hidden dependencies (I saw only with modded loading that Car types depended on the terrain which I was loading it in parallel to :D)

But the veh preload is 8 seconds so might be worth it.
Right now I also wait for it to finish before continuing, I might be able to run it in parallel with the world init script, but if the script creates a vehicle class, might be bad. No more time today.

Mh... The start part where it loads the PBO..
In parallel to loading PBO, it also hashes all script files, which it needs later for multiplayer.
Maybe I can move that to run parallel to the preStart script, get rid of 4 seconds of hard I/O load while we are I/O bottlenecked for the pbo loading..
Have to try that tomorrow

Thank you for attending todays live dev-diary

[silk pewter]

one map/dictionary to rule them all? 😬

[whole cloud]

So that last idea, yeah it works.

lBankAddon from 4.1s to 2.5s
BankHashes/DataIntegrity threads from 4.5s to 0.8s
TexHeaders from 1.4s to 2.1s (But the cfgMerge takes 1.9s anyway, we only waste 200ms with that)
addonLoad 10.3s to 8.8s

So, by just moving when in the startup sequence one thing happens, everything gets faster.

Now if I can move that big block, over there, I can win 6 more seconds :u
Oh wait I can't because it needs the terrain 🤔
But I'm sure I can load that just before the preStart script and be fine 🤔

Its all just tetris.

[stark falcon]

ace has a setting for this under ACE Uncategorized

[empty goblet]

oh cmon ... this ain't like writing bogus errors at 2GB/s in loop like >certain other game< ... shave down all those ms , make it zappy rapid start !

[empty goblet]

guess the size needs to increase to 5 TBs of wonky addons to get our normal game load in blink of an eye

[empty goblet]

Players being killed and coming back to life
anyone got this with 153281 ?

[analog acorn]

Do you happen to have any scripts running that might be doing setDamage 0 on the players after they die?

[fair ore]

o7 on it sir I will join more shady milsim units to get access to their mods

[uneven bluff]

This started to happen on my test server because whenever I ran setDamage 0 it would revive my player for a brief moment before I respawned. (Not a big deal atm.)

[analog acorn]

setDamage being able to bring back dead units is a known change (it's in the change log) and it will be changed again soon to not be the default behaviour

[deft oak]

We do

[shrewd frost]

Maintaining and modifying an engine alone is incredibly demanding. Seriously, I think you should sit down with whoever is in charge at BI and have a thoughtful discussion about open-sourcing Arma 3—either fully or partially(for example by opening up interfaces that would allow the replacement of legacy systems such as lighting, particle effects, or skeletal animation)—at the right time. Many well-known engines (Unreal, CryEngine, Source Engine, etc.) have opened their code. I don’t believe open-sourcing would significantly impact revenue; after all, their marketing teams aren’t naive. It would be a shame to let such a well‑built game like Arma 3 slowly fade away over time.

[proud cape]

now while unreal cry and source have their engine source available keep in mind that those engines are products of their developers (as in licensed as prebaked engines)

and thats why the developers of those engines have an interest in making their source available

i dont imagine RV4 being specific to arma and still (from what i can understand from lurking this channel for a few months) having code from OFP would make a selling point

while it certainly would be great to have community support for engine improvements i dont think that what you're stating as precedent would make a reason

still i've got a lot of respect for maintaining the game while arma 4 is most certainly being developed

[void badger]

Considering a lot of that code is in VBS sadly it'll probably never happen

[spiral pond]

Partial open sourcing would already have various benefits like:

• understanding the engine better (ie AI, physics, etc)
• better and more precise documentation
• at least to some extent community could debug/narrow down issues better
• people could submit fixes (for somewhat simple things - otherwise more manpower would be needed as Dedmen's time is spent elsewhere in more meaningful ways)

IMHO a full open sourcing is unrealistic for various reasons. Yet even "just" the ability to build a non MP version would allow way more community development possibilities, enhanced dev tools use/integration, etc. I'd imagine it would require to encapsulate 3rd party code to dlls with obfuscation or such to avoid IP/license issues.

[obsidian condor]

Most (if any at all) is no longer in VBS4 (assuming some is but the code has been divergent anyway for over a decade now). But I second the idea of OpenRV4 (Marek even "liked" my tweet about having it a couple of years ago 😉 )

I say "most" because VBS4 has a new engine but still uses SQF afaik?

Mondkalb would know better though

[heavy galleon]

Daddy dedy, have you been touching projectiles locality recently?
I got some people complaining about things stopping working

[heavy vortex]

what things :/

[heavy galleon]

My addons that do shenanigans with projectiles like replacing them with different type as I can't load mixed magazine

[opal hound]

How do reload tracer mags work then?

[heavy vortex]

Special-purpose config parameter.

[opal hound]

Fair

[opal hound]

#dev_rc_branch message

[heavy galleon]

I am aware of that, that's why my code took that into account
It was a day of very headhurting debugging why it wasn't working at first.
But some people are saying that it isn't working anymore

[opal hound]

Yea I think some locality may have been touched though 😆

[heavy vortex]

The impression I got was that projectile locality didn't change but it no longer incorrectly spammed network messages for local projectiles.

[heavy galleon]

I would hate to go back in the spaghetticode I wrote year ago and redo it
But I will if I really have to

[heavy vortex]

But maybe that means that some local projectiles were actually global projectiles that were cloned on every client?

[opal hound]

Ooh yeah I just read that better, hm

[heavy galleon]

in my case the issue is that I need to replace a shot projectile with a different type, as you can create mixed magazines with any combination of he, smoke, ilum and penetration
and based on the mag configuration I replace the projectiles while in flight, as it just shoot HE by default. (and the combination in magazines are done by players)
But it got reported that people see HE even when they loaded smoke

[opal hound]

Globally?

[heavy galleon]

I think I am assuming that the "real" projectile that is cloned by others is the one on the machine of whoever has the vehicle local

[heavy vortex]

or turret local :P

[heavy galleon]

well, it has only one slot

if you are in, you are in the turret

[heavy vortex]

Static weapon?

I'm not sure how those work. Normally vehicle locality is based on the driver, so it's possible that vehicle and turret locality could still be different for single-turret statics.

[heavy galleon]

yeah static, mortar

[heavy vortex]

I checked a couple and the vehicle locality does change with the turret.

[heavy vortex]

Hmm, had a couple of weird in-game freezes on the latest perf.

[heavy vortex]

Are either of the FROZEN mdmps useful in that case or do you need to force a dump in task manager?

[empty goblet]

1.54.133114 profiling & performance v14 server and client, windows, linux

• another startup crash fixed
• additinal logging on that 'anims' crash
  https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm

[whole cloud]

I did not change anything on them.

There was a "ref to non-network object" error when a vehicle fires submunition (Like the civilian ED-1D demining drone does).
I changed it to not even attempt to send the projectile over the network, when it wouldn't be able to send it anyway

[heavy galleon]

And can we get changed that would make it easier?

[whole cloud]

frozen dumps show where it was frozen. But they are small dumps so might not include all the info needed

[whole cloud]

Changing that behavior would open so many backwards compatibility rabbit holes, I'd rather not touch it

[whole cloud]

Actually vehicle preload compiles scripts (eventhandler and sound expressions) which I couldn't do in parallel to script execution... Bummer.

[stark dragon]

Any updates on the significantly slower load times since last update?

[whole cloud]

Tomorrow we'll publish an update that lets people collect profiling data of game start.
Then I can see where the slowness is

[stark dragon]

Thanks bud.
I know a lot of people think it's too many GB of shady mods but my load times have doubled to ~20min even with just ACE/ACE-X and CBA.

[shrewd frost]

Please share your thoughts on open source. We all know that your views do not represent BI.

[analog acorn]

You were already at 10 minute load times with just ACE and CBA? Crikey

[whole cloud]

I don't think fully open source is viable, atleast not in the next few years. Once Arma 3 goes fully EOL we could consider it.
But if we look at older games, like operation flashpoint/Arma 2, we also didn't do open source these yet, even though its fully EOL. I don't know why that is, it could be licensing reasons, but Arma 3 still contains a majority of that old code too, so if its licensing reasons then they would apply to A3 too.

But we already prepared infrastructure for more open external contributions and code access.

[analog acorn]

Could be partially that older RV games couldn't be open-sourced without also compromising an active game (A3). Once A3 goes EOL, that's it, there are no more active RV games, everything that "matters" is on Enfusion. So that could change things.

[whole cloud]

That could apply to Arma 2 OA yes, but not the older stuff than that.
Multiplayer security is massively different. Old bugs that might have been there and could be abused, have long been fixed.

[shrewd frost]

I also believe that partial open-sourcing is the most practical path, even if it’s just a small portion. As a MODer, I’d love to bring dynamic lighting effects into the game’s particle system. Designing a new particle system with OpenGL is straightforward—I’d only need BI to provide the RV4 interface so I can replace the outdated system.

Gradually opening up different interfaces feels like the simplest approach, and it wouldn’t require a massive team to review every line of code. That way, you can keep focusing on what excites you (like your long-standing interest in reducing the granularity of engine multithreading locks 🔧🧵) while others chase their own passions.

Man, the possibilities are endless ✨🚀—just thinking about it gets me fired up! 🔥

[patent sky]

What in the AI did I just read

[shrewd frost]

The formatting was organized by AI, but the ideas are entirely my own thoughts translated.

[quaint flame]

"Designing ... in OpenGL is straightforward"

You need to hire him

[empty goblet]

OpenGL denied, only Vulkan

[silk pewter]

OpenAL used, OpenGL denied… OpenRV wen?

[quaint flame]

Me waiting for open sourced functions library

[whole cloud]

Same 😢

[shrewd frost]

DirectX, Vulkan, or anything else—it doesn’t really matter. As long as I have the drive, I can learn it. Honestly, what are your hiring requirements? Do you expect a detailed understanding of computer graphics? A deep knowledge of engine architecture? (I might not meet those requirements, I’m just curious.)

[obsidian relic]

which ended up with crash "application hang"?

[silk pewter]

https://careers.bohemia.net/position/engine-render-programmer-praguebrno 😊

[shrewd frost]

No, I’m just curious. I actually looked at your job requirements before, but relocating internationally involves too many complications. I really think you should consider remote work. Also, I’m not ready to dive into Reforger yet, since I’ve already invested so much time in Arma 3. And as far as I know, Enfusion’s particle limit is 10,240—roughly half of what Arma 3 allows—which further dampens my enthusiasm.

[silk pewter]

I was only replying to you saying "what are your hiring requirements?" ^^

[quaint flame]

Remote work is possible sometimes. Tbh if you are serious I'd just give them a call. But that's offtopic anyway.

[quaint flame]

I really hope it happens before a4😬

[shrewd frost]

Yes, I’ve gone off-topic. Right now, I just want to know what I can do to move as quickly as possible toward achieving my goal of replacing the particle system.

[shrewd frost]

Can I get access only to the code related to the engine’s particle system, so that I can get started as quickly as possible? I promise I won’t disclose it anywhere.

[shrewd frost]

Never mind, I realize now that this is a legal/IP boundary issue, not a matter of whether I’m serious or not. In the meantime, while you’re preparing to open up the interfaces, do you have any suggestions for what I could work on? For example, would it make sense to start building a simple particle system now, so that it can be more easily integrated with your interfaces later on?

[empty goblet]

OpenGl denied only because it's legacy tech ... it's not evolving anymore

[shrewd frost]

I believe RV4 is based on DirectX 11, right? Then why is it compatible with Linux? Did you rewrite RV4 using Vulkan? I have to use the DirectX 11 API to build the particle system—other libraries shouldn’t work, correct?

[quartz rampart]

called wrappers probably DXVK

same what Steam/Proton uses for the steam deck

[patent sky]

Its not directly compatible with Linux, only with Proton and that uses DXVK to translate DirectX11 to Vulkan

In the meantime, while you’re preparing to open up the interfaces, do you have any suggestions for what I could work on?
There is no "preparing to open up the interfaces" We have prepared some stuff for potentially more open external collaboration in the future like dedmen said.
and even in that case you would need very in-depth knowledge of the DX11 processing stack and engine code

[shrewd frost]

I know that without the source code, it’s very difficult to get started. Is there any legal way that would allow me to read the parts of the code I need to see?

[silk pewter]

join BI, that's about it 🙂

[spiral pond]

From what i recall Marek and Ondra/Suma did contemplate to open source OFP or A1/A2/OA.
Aside from potential licensing reasons, or potential VBS complications, most likely the reason as usual was lack of time/other prios.

Open sourcing such a complex engine is not a simple process by any means. ie you have to check for potential 3rd party license/IP issues, consider impact on the existing products, clean-up the code to some extent, prepare the whole thing from the legal side, documentation/presentation, public communication and probably a bunch more topics.

There have been a few botched open sourcing of game engines/games after all.
Probably a staged approach would help, yet still involves most these topics and bunch of prep time.

Anyhow possible to share any more info on the startup profiling already?

[proud cape]

i would strongly advise to temper any expectations of source availability

[shrewd frost]

Alright, let’s take a step back. What I want is for Arma 3 particles to achieve the effect shown in the image. Can you or Dedman arrange this? This is an Unreal Engine 5 effect, where particles can use an emissive texture so that specific parts of the particle glow, with both emissive and halo effects, and the light can actually illuminate the environment. (I believe Enfusion already has this capability.) This is not the same as Arma 3’s emissive, which is more of a cold glow effect.

[shrewd frost]

Sorry, it seems this channel doesn’t allow posting images.https://drive.google.com/file/d/1QeSQ5Gscx_fawjpMl-n_NH2vKgm9I0Z0/view?usp=sharing

[real creek]

This is getting quite offtop for this channel

[fair ore]

please no we don't need "vibe coding" in BI 😭

[real creek]

Em dashes and emojis like that only mean one thing

[fair ore]

Can we talk about how the account is two weeks old and joined this discord on the first day of existence

I smell ban evasion but have no direct evidence

[minor epoch]

Its John GPT and arma is its first game, we have reached gamer AI before AGI

[silk summit]

I just want to remark that I had nothing to do with any of this!

[silk pewter]

that's precisely what an AI would say… I know, I am one

[whole cloud]

For the people that have slow game start, please use that parameter.
It will dump two capture files into RPT folder, send them to me

[silk pewter]

is that param temporary or planned to stay?

[uneven bluff]

Updated and now the game hangs when I attempt to join a server through the server browser. Starting the profiling exe seems to work and I can host a local server.

(Downgraded my exe to the last version and everything works again.)

[whole cloud]

Reverting Prof due to hang on MP connect.

Google drive will stay, and you can still grab the exe to do that startup profiling

[fallow sky]

Seems like there is an issue where I can't join a server since last update and the game just crashes

[opal hound]

read the messages above yours

Anyway, for some reason that parameter didn't create any files for me. Not sure if startup was under some limit?

[whole cloud]

Profiling exe?

[opal hound]

I am dumb. thanks

Yeah that's also why those non-network ref reports didn't send, I just can't read 🤦‍♂️

[empty goblet]

was the new command name decided yet ? i vote for magicDamage 🙂

[uneven bluff]

setTransientDamage

setMagicDamage

[spiral pond]

@empty goblet @silk pewter seems no longer possible to create threads in the channel - is that (change) intentional?

[weird it works via the top menu - but not via the channel list on the left]

[whole cloud]

They don't need profiling exe 😄
They need the crashreporter.exe to be present in Arma folder, but it should be.
Btw still got zero reports out of that and crash reports are also massively down since the 17th

[opal hound]

• Added: -reportNonNetworkObject command line parameter that automatically reports "Ref to nonnetwork object" errors (Windows profiling binary only)

will run tonight and see what happens

[spiral pond]

profileGameStart observations

[empty goblet]

i didn't change anything

[patent sky]

ah frick, my include folder only added it to client profiling depot
next prof it will be back for server

[knotty wraith]

... Known Issues:
Freeze when joining MP session...
Waiting for a fix? or do we set it up and help fix it?

[whole cloud]

I'll fix it tomorrow

[real crypt]

i've just spent so long trying to figure out which mod in my modlist is fucked from last night just to find dedmen has broken my arma install 😢

[patent sky]

If you are on prof/perf always check here first

[whole cloud]

Freeze reproduced.
I actually saw this noted before. Thought if I should note it in changelog, but I couldn't find any freezes while testing so thought its probably been fixed in the last decade that this feature had been disabled.
Apparently not!

[whole cloud]

About the -reportNonNetworkObject parameter.
Do you see "refNNObj" dumps in RPT folder?

If they are there then the engine part is working, just the crash reporter part isn't 🤔

[opal hound]

Didn't manage to catch any yesterday, there weren't many vehicles though which is what I usually see cause it

[whole cloud]

But its been a week now

So there should be some

[opal hound]

I wasn't properly running the profiling binary until yesterday

[whole cloud]

That doesn't need profiling binary

OH

It does

its not supposed to
Too late for v28 though

[opal hound]

yea it's fine I'll keep running it

Will the server binary also report it? I have seen a few on that side too but may have been before the fixes

[whole cloud]

Yes

[latent socket]

I do not have that file in the RPT folder

[patent sky]

#perf_prof_branch message
:D

[whole cloud]

Yeah but the dump creation atleast was there, even if the sending away didn't work

[whole cloud]

(Windows profiling binary only)
Oh I guess its really supposed to only be profiling binary. Mh.
I don't remember why, but if I wanted it that way I probably had a reason, I'll leave that then

[obtuse bramble]

It looks like ARPI objects + ACE cause the game to crash on the latest build*

bin\config.bin/CfgVehicles/ARPI_EasternRadio_On/ACE_Actions/: Cannot find base class 'ACE_Actions'

https://steamcommunity.com/sharedfiles/filedetails/?id=1400734335

[static briar]

That mod has truly "crazy" and "malformed" config. so it's not works anymore and author says "DO NOT REUPLOAD" so it's dead.

[static briar]

Restrictions:

• No reuploads on Steam allowed! No not never!
• Montarization in any means is definately not allowed!
• Use on monetarized servers is not allowed.
• If you want to rip/change/host these files to include them in your local clan-repos, thats ok for us.
• If you want to change or implement something, talk to us, maybe we can make an improvement in the official version together.

maybe we can make an improvement in the official version together.
no he is closed the comment section. never.

[obtuse bramble]

I believe you, yet it only stopped working completely due to the latest branch update so that's all I'm reporting.

Main build still works, previous also. That is all.

[obtuse bramble]

Yet at the same time he says if you want to rip/change it - it's ok. Lol
Arma modders.

[static briar]

oh yeah it's time to upload the mods to googledrive or something

[slender lagoon]

Is it just me or when clicking on scenarios the game freezes

im on v28 btw

[violet knot]

Ai VTOL with "Land" Advanced waypoint not working correctly for Ai controlled VTOL vehicles.
They either try to slow down and crash without Vectoring or Vectoring is not recognized.
Launch without mods/cDLC, Repro in VR with Blackflish and standard helipad with "Land" WP on helipad.
Normal branch, no issues.

[deft oak]

Still getting reports of the resurrection bug today:

"There is currently a bug, where sometimes if you die, you can still walk around, shoot, pick up your gun, yada yada. This all happens while the bleed out countdown is on your screen. It happened to me earlier and I just stood there since I was bugged another player shot me and got 3 kill messages, I fell over and my character immediately stood up again. No way to get out of it unless a teammate kills you and starts the revive before you stand up again. Once you are revived it fixes the bug."

[heavy vortex]

vanilla revive?

[analog acorn]

There's an important distinction between being downed and being dead. It sounds like this is happening when you're downed, not dead.

[deft oak]

Sorry, should've clarified. This is from an Exile server. You get "downed" but it actually means you're dead and available for a revive for 5 minutes.

DB has a video of this on their Exile servers as well but not sure if it ever got to Dedmen.

[heavy vortex]

If you're revivable then you're not dead. It'll be using setUnconscious or similar.

You'd need an Exile person to check what their downed state actually does.

Haven't had any issues reported with Antistasi or ACE revive lately, although if it's specific to perf branch servers (also not specified in the report) then that might not have been tested much.

[analog acorn]

If it's a custom system then it's possible that it "revives" you from true death by respawning you on the same position. I wouldn't do it like that, bit pointless to go to the effort of making it work...but you could. Maybe if the engine kill attribution is super important.

[heavy vortex]

Yeah that's also possible. Makes it even more important to know how Exile is handling it.

[obsidian condor]

Pretty sure you are dead and respawned. Does the same in Epoch if enabled on a server - thus have been watching this conversation

Both mod's systems were implemented over a decade ago now...

[deft oak]

you do indeed die

and then get cloned/replaced if you get revived

[heavy vortex]

How does it block the automatic respawn/death logic?

[latent socket]

Either steam is updating my game properly or they aren't showing in my
C:\Users\computer\AppData\Local\Arma 3 folder with the other RPTs.
Not sure if this is a me problem or not

[trim ether]

We're experiencing severe login issues with the latest perf build when deployed to a server. Clients either hang when connecting and do not get to the roster, or once on the roster are unable to select a role. Occasionally they are able to push through, only to not see any other player, regardless of that player being spawned at the same location. The same mission worked correctly with perf v26

[whole cloud]

Are you sure you switched to the profiling binary?

[whole cloud]

More details on hang, total game freeze? Like in v27?

[whole cloud]

What is a resurrection bug?
setDamage had been changed to allow resurrection (NOT a bug, never has been a bug)
Nothing was changed about how you die, literally only the setDamage command behavior.

[whole cloud]

In main menu?

[trim ether]

Stuck on the grey and black banded screen, no words, no logos, etc. I did not run v27, so no idea if its the same. I can pull RPTs from the server, but will have to wait until tonight, if that would help?

[whole cloud]

That code should've never worked. You cannot set the same class twice.
If it worked before then THAT was the bug. Though I don't remember doing anything to fix that.

[whole cloud]

But no game freeze and game killing itself after 30 seconds?

If you remind me tomorrow or next week, I could look at why that broken mod worked in the past. But for now I won't put a priority on something that is clearly obviously blatantly broken, being broken.

[trim ether]

Correct. No game freeze. We restarted clients and server, just to check it was not a one off. But it failed in the same way. Players were running a combo of the latest perf and vanilla clients. We then relaunched the mission using vanilla build on the server, and it worked perfectly.

[whole cloud]

Reproduced and fixed, thanks.

[whole cloud]

Maybe the server was frozen then. Might be same issue as AirShark's
Even if server main thread is frozen, the networking thread will still be running and allow people to get into connecting state

[whole cloud]

Who would've thought that Read/Write locking could be so hard.
Though tbf it really only is because the old implementation was just so bad.

We need to protect one array. Instead of putting a lock around it, it was decided that every caller should grab a lock on their own, before calling any method in the class that manages the array.
And made it impossible to access it without holding a lock.

Then we also duct-taped callbacks into the class.

So to add a PBO, you need to get a write lock, then call Add. Add internally fires a callback.
If you want to access the PBO list in your callback (which you have to, because the callback only gives you an index into the list, not a reference to the PBO that was added. And you want the PBO out of the list) you need to get a read lock.

Now the Add is wrapped in a write lock, and inside the callback which is inside the Add you want a read lock, which you cannot get because its already locked for writing.
The Add could unlock the write lock before calling the callback (which is dangerous because the index might change and be invalid), but it can't do that because the Add doesn't even see the log as it was done outside by the caller.

I sense another big refactoring coming up.. But fixed for now.. "Trust me, I know I already have a lock, just let me bypass all the safeties and access it anyways, thank you"

[silk pewter]

smeels like an added bool trustMeBro = false param! 😆

[whole cloud]

The autopilot changes https://feedback.bistudio.com/T192405 from v23
Wanted to keep the thrust vectoring mode stable when you lost focus to game, but applying that to AI piloted vehicles too was mistake

[indigo anvil]

As long as a handful of people ever touch that code & has enough self reminding comments around it the workaround can hold.

[whole cloud]

Yeah I smothered it with comments. But the main point is that the problematic stuff is done for a bad reason and anyway is risky/unreliable.
Instead of passing around the PBO itself. We pass around and store the index, and grab it from the list every time we need it.
But if a PBO is removed from the list, the index will change. And suddenly you access the wrong PBO. And all the extra code to access the list every time and try to find the PBO, when you could also just store the PBO directly instead of the index to it. Can make code so much simpler and safer and easier to use..

[whole cloud]

or maybe not. Fixing that doesn't fix the issue, it just flies in circles around the helipad. Doesn't even try to slow down. It even speeds up and climbs a bit when I place it lower down in editor.

[plain trout]

I had an issue yesterday that I wouldn’t see the OK button after I returned to the lobby of the MP game. Dedicated Linux vanilla server and me on latest prof. Logging in/out of admin didn’t fix it. Reconnecting did

[whole cloud]

I can even reproduce it on profiling v0 and 2.20 stable.
The VTOL just flies over the helipad, climbing above it, and then flying endless circles.

Either this was always broken, or I need more proper repro steps. @violet knot

[whole cloud]

What do you mean not see?
Its grayed out, or literally not existent?
There is no code to make the button not be there, only code to disable it so you cannot click it when you didn't select a role first
Huh

Thank you discord for just not accepting my edit and not giving me my text either

If you logged in admin try to force start while not all players have selected role, the button will disappear but you will be readied up, server will wait for all players to be readied before starting

I assume, the display stays open in background, so when you after that, return to the lobby, it'll still be hidden?
If thats it then I need FT ticket with repro steps

[plain trout]

Not existent

The server was already started. Will do later then

[fiery crescent]

V25, in escape mission VTOLS fly over the landing zone, they slow down do circle and do it again, never land. It used to work fine but when? I can recall it OK some 2-3 months ago. Not sure when it broke. Helis seem OK.

[fair ore]

I wonder if it's broken in 2.18

[whole cloud]

Well in June, it had the same behavior still. So it must be a long time ago

There were changes to helipads in December and june @feral harness
#perf_prof_branch message

All of which are also in stable, the June ones were merged last-minute.
June:
"helicopter would not land on a helipad placed on an object"
"land command made compatible with VTOL and waiting on elevated helipad"
#arma3_feedback_tracker message

December:
"Helicopter would not respond to change of helipad during landing"

[uneven bluff]

All this sophisticated autopilot, I'd just want the helo/jet to do nothing and keep its thrust and direction the same when I lose focus or switch seats 😛

[fair ore]

same lol

[silk pewter]

ez make it a gameplay option in settings

[fair ore]

yes very good

[charred pagoda]

how to clear mod cache?
my friend's game doesn't start, just stuck on loading

upd: even without mods game stucks on "loading addons"

works on stable branch

ok, i have same issue, lol

game doesn't start at all on latest profiling 2.20.153303

[whole cloud]

:panik:

Should create a freeze dump in RPT folder?

If not you can create one manually in task manager and send it to me

What the heck, I can reproduce.
But only he profiling branch build, not my own build

Luckily I just finished ripping out all this mess that caused the freezes.
But I didn't test it enough yet

yeah wtf.
Profiling exe from steam == freeze.
Profiling exe self built from same code and in same configuration == starts fine.
🤣

[slender lagoon]

yeah same stuck at loading addons

[charred pagoda]

There is not crash dumps. Task manager generates empty dump

[whole cloud]

Ah battleye

[charred pagoda]

Do you still need it?

[whole cloud]

Probably not

[latent socket]

100%. My launcher has the green profiling banner on the top left and I've been on that branch for well over a year. I'm not sure how it could have switched

[whole cloud]

The profiling banner stands for the branch, not the executable

Profiling exe is a separate executable, not enabled by default.

[latent socket]

Yes, arma3_x64profiling.exe or however it is

[whole cloud]

yep

Yeah I see the freeze.
Onto another try then.
But weird that I don't get it locally, but it seems to be a bit random, and the more pbos (and the slower) the more bad

[latent socket]

I will quadruple check later that's what im running. I see no reason it would have changed on me, but 🤷‍♂️ shit happens

[whole cloud]

The profiling.exe is NOT used when you start via the Arma launcher
The arma launcher starts arma3_x64.exe

[latent socket]

My desktop icon for A3 was for the profiling executable

[uneven bluff]

My profiling executable is hanging too.

[latent socket]

Like I said, I'll check again. I'm almost certain that's what I've been running. Will report back that either I'm a dumbo or it's another issue

[whole cloud]

You can also easily check by using the FRAME cheatcode in main menu.
If UI doesn't pop up, it's the wrong one

Fixed, ETA 50 minutes 🗞️

Someday we'll be back to stable and less chaos

[patent sky]

[tropic thunder]

we sure about that? if arma has no chaos is it really arma?

[leaden relic]

https://tenor.com/view/time-check-check-the-time-im-waiting-time-is-money-gif-5751033985188952699

coming up soon dedmen
-# Thought it was a mod issue since I was doing dev work

[slender lagoon]

https://tenor.com/view/patiently-waiting-waiting-waiting-patiently-still-waiting-gif-3590841176648399679

[tropic thunder]

https://tenor.com/view/kermit-the-frog-kermit-waiting-waiting-patiently-still-waiting-gif-17579453702067938124

[whole cloud]

:harold:

We're reverting back to v26, last build is also bad.
Retrying on monday

[stable wasp]

when i use profiling branch it gets stuck at loading addons and never opens

[whole cloud]

https://cdn.discordapp.com/emojis/982322678046294026.webp?size=56

Revert is done

[stable wasp]

which of the dev branch versions u think i should get? for better performance. also do they work online?

[whole cloud]

None of the dev branch versions. dev-branch doesn't work online on normal servers no

[stable wasp]

oh mb i meant performance branch

[whole cloud]

If you mean profiling branch, then the current one on steam, which is now v26

[stable wasp]

the v26 in pins yeah?

from google drive

[whole cloud]

Its on steam. You don't need google drive

[stable wasp]

so just get the perf branch rn? from betas

[whole cloud]

yes

[stable wasp]

sorry im new tho so kinda confused

i got it few mins ago and it didnt work

[whole cloud]

Yes we're a bit chaotic currently

[opal hound]

What are you looking for in terms of better performance?

[stable wasp]

ill re try. i was confused cuz i saw lower specs than me get better performance

[stable wasp]

idk i felt all graphics were simiar in performance

i did change params from launcher and refresh rate from config

but the fps seems to be unbothered even tho it shows 70-100 fps in settings

wait the profiling branch is working rn lol

[opal hound]

idk if there are really any major performance improvers on this branch currently, mostly just crash fixes and minor features

I would probably focus on settings and your own system first

[latent socket]

@whole cloud Should I still run the profilegamestart parameter?

[stable wasp]

using the prof branch fixed my performance

lets keep it this way gang

[deft oak]

This is the same bug The Tall Man sent you a video of this week if that helps at all. Started with v26. Switching to stable fixes it for the players that have experienced it. It also affects all exile servers (and apparently some koth). I'll also dig through exile to see how they handle death/damage more specifically.

[obsidian condor]

For Epoch, you get a death screen and extended (10 mins) wait before auto-respawn happens if you can be revived. Can just manually select respawn there if you want. Some behind the scenes code deals with the revive from death action by another player and creating the new player object with your inventory and modified health stats then uses selectPlayer to move you into the new player object, with the old player object having been hidden...

I'd assume Exile is similar given many of the devs came from the Epoch community originally

[obsidian condor]

Yeah, was just talking to Dave and he mentioned the Hunger/Thirst system too... Will affect Epoch too, will work on a fix. Probably other systems that do a setdamage to 0 will be affected too...

He shared this video too (believe Dwarden and Dedmen have this)

[deft oak]

yes, that's the video I was referring to

[opal hound]

Some Ref to nonnetwork object reports should have been sent through

Seen them sending in the crashreport.log, if you need the refObjNN mdmps let me know

[analog acorn]

Reminder that the setDamage change is going to be changed again, so that bringing back dead units is no longer the default behaviour, but is available by an alt syntax or separate command instead. This is mentioned in the changelogs. It was already reverted in a prof update, in fact, but that update was temporarily rolled back for unrelated reasons.

[deft oak]

appreciate this, I saw it in the changelog (it getting changed to alt syntax) but didn't realize that one was rolled back so just wanted to make sure it wasn't still enabled by mistake

[deft oak]

Just a heads up, after watching the video again and reviewing how the infistar spectator determines the "dead but still watching message" it seems that this also breaks the alive command once a player is resurrected with setDamage. The condition for that message is alive player equals false so an alive check likely won't fix it. I'll confirm all of this when I have a tester later and report back with something more definitive.

[deft oak]

With a quick test it looks like alive does still work after "reviving" someone with setDamage 0 so should be safe to use as a check.

[warm sedge]

Hello All: can someone give me a link to the Take EventHandler

i can't seem to find it

[naive osprey]

https://community.bistudio.com/wiki/Arma_3:_Event_Handlers#Take

[warm sedge]

oh cool thx man

[empty goblet]

tho that question is better for #arma3_scripting and #community_wiki

[warm sedge]

oh shit sorry dam

what the hell em i doing in here lol

sorry

[whole cloud]

yes, it'll be back working properly on monday

[whole cloud]

setDamage 0 will disappear again anyway. No real need for you to fix it, if you can wait to monday.
Just all the builds that included that were broken

Unless my revert of that was bad, but I don't think so I'll check it again

No the code is correct, once its available

[uneven bluff]

Just curious about the performance of a script command: why do deleteMarker and deleteMarkerLocal perform so slowly compared to creation?
createMarker ["testmarker123", getPosATL player];
Takes 0.00051606 ms.
But

deleteMarker "testmarker123";```
Takes 0.17836 ms.
(The reason for the two-liner is because deleting a non-existent marker is fast too.)

[quaint flame]

Probably because createMarker just appends to the array where deleteMarker needs to update the indices.

[whole cloud]

The more I look at it, the worse it becomes.

1. It tries to define ACE_Actions class twice in the same place, that is not valid and never was.
2. It does inheritance wrong
3. Even if it did inheritance right, it would STILL be wrong, because the ACE_Actions class its trying to inherit, doesn't even exist. That radio inherits from Items_base_F, which doesn't have ACE_Actions, and the radio doesn't either.

In the code that finds base class, we specifically have code for this case for when you inherit from a class named the same.
When that happens, it specifically ignores that it found itself, and goes into the base class to find it.
But as I wrote above, Land_FMRadio_F does not have ACE_Actions to inherit, so it obviously doesn't find it.

You can fix this with a extra mod, if you add ACE_Actions class into CfgVehicles itself (probably bad idea), or into Land_FMRadio_F.
But you need your PBO to be loaded before the items mod. And while requiredAddons most likely will generally sort it that you would be before, there is no guarantee for that. But I think it'd work.

Now why did this work before?
I made a performance optimization for finding the base class, which accidentally fixed a bug.

In code its supposed to find itself, but, it was bugged and instead found the first class with same name. Which due to the broken double-defining of class which shouldn't even be allowed, found the empty first ACE_Actions class.
So while finding itself, it did NOT actually find itself, very broken.

But because of finding the first ACE_Actions, it thinks it inherits from itself, which is not possible.
So to fix that, the code early on considers that invalid, and instead tries to find a different base class in the base/parent. It fails to find it.
This causes the base class to actually be null, so the inheritance is silently deleted.

This is just a huge tub full of shit that you just plonked down at my front door.

Do I undo this performance improvement and engine bug fix, to appease your completely messed up mod? I don't want to.

[patent sky]

we gotta rebuild it when we delete one, adding just adds it to the end

[quaint flame]

yay I was right 😄

[uneven bluff]

Ah that makes sense. Just a little inconvenient for me when I try to delete 45,000 of them at once 😛

[patent sky]

https://tenor.com/view/sxv-wtf-gif-23560064

[uneven bluff]

It actually takes a lot longer in my game than the codePerformance measuring thing implies, probably for that reason you just said.

[whole cloud]

We can optimize it if you delete the last marker from the end. But probably not worth it because you're most likely not doing that

[quaint flame]

I always wondered it if wouldn't be more efficient to have deleteMarker <array> and delete a whole range in engine at once?

[uneven bluff]

I can but it's not a big deal enough for me. It's just for drawing debug spawn locations on the map.

[whole cloud]

it would

[quaint flame]

But deleting a lot of markers is an edge case kinda

[uneven bluff]

Definitely.

Was just curious how it worked behind the scenes.

To my mild surprise, markers in general aren't that much more performant than drawIcons, especially if they're supposed to represent moving icons.

[whole cloud]

Two options. I undo the performance fix, but only in stable, and keep it broken on profiling branch so that you stop using broken mods.
Or, I turn the "missing base class" error, into a warning popup. Which doesn't kill the game.

What the game currently does is basically pretend as if the class didn't inherit from anything, silently, without ever telling you about your mistake.
So I'll just do the same, but loudly and annoy you with popups till you stop using that broken mod.

[quaint flame]

PopUp

[obtuse bramble]

Well, thank you for the tldr (jk) and after discussing this with other members in my community, we had the same sort of idea you proposed but were waiting on what you were going to say. So that being said, we'll see if we can jerry-rig it to prevent anymore work for you in the future.

Lastly, I'm just reporting an issue - I didn't see or get this error before for reasons you explained. At what point are we expected to break open every mod every uploaded to the workshop and break workshop policy, while also being pointed out to be the one using ''broken'' mods. This mod was likely fine for many years as ACE didn't change much either, till it did so perhaps that's what's now being exposed.

My point is. I'm ok with breaking and entering pbos of supposed broken things as opposed to bother a BI* developer to fix it for me, unfortunately the policies that could save you the stress are also the ones that can punish me for trying to help.

[obtuse bramble]

And in this case, I would prefer the 2nd option so that you're not responsible (because that's how ppl will blame BI for it) for breaking idk how many milsim and life servers.

[patent sky]

My point is. I'm ok with breaking and entering pbos of supposed broken things as opposed to bother a developer to fix it for me
has anyone tried contacting the developer? that would be the first step

[obtuse bramble]

I meant BI developer, not the mod dev. Sorry

I know Dedmen does his best to work around issues or reprogram things so they don't remain issues, with modders however.. They tend to completely ignore requests even if you already fixed it for them.

[patent sky]

The issue is that the more broken stuff we "allow/tolerate" to just work the worse the problem becomes because people will look to this mod for examples on how to do stuff > make the same mistake and probably even worse > repeat
A popup to warn of a broken mod sounds fair in this case

[obtuse bramble]

Right but I'm trying to find where that line is. If BI only wants to keep it a popup, then I'm all for it. I'm just trying to understand when and where it'd be applicable to create my own fork of a mod, just to prevent even the work to make the popup happen. If that makes any sense.

[whole cloud]

I'll just make it better for everyone.
"Cannot find base class" errors will now be just a warning.
The game will just throw away the inheritance and load the class as if it didn't inherit anything.

That will make bunch of broken mods usable.

[patent sky]

when and where it'd be applicable to create my own fork of a mod
but nobody has even tried contacting the dev yet, and the mod clearly states no workshop reuploads. so that should be the first step

[whole cloud]

There is a screenshot under your first post that shows the line.
Remove the inheritance from ACE_Actions and it'll be fine

[obtuse bramble]

Except he also states:
-# + If you want to rip/change/host these files to include them in your local clan-repos, thats ok for us.

So which one is it? Lol. But yes, I'll do what @whole cloud suggested and see what happens.

Thank you both regardless.

[normal kettle]

Actually his suggestion doesn't even need to reupload the file.

[whole cloud]

If you have a local repo anyway (not workshop) then just edit the pbo and be done with it. Thats the fastest and easiest solution

[obtuse bramble]

And sorry, I didn't mean where the line in code is, I meant where the line is for the workshop policy. If I fix this and get to tell the modder - if he doesn't reply/react and it does break 9K games.. It's a shame and I know ppl will just blame BI for it.

[patent sky]

where the line is for the workshop policy
The license states no reuploads to workshop = no reuploads to workshop, no matter if its broken.

[obtuse bramble]

Fair enough, thanks guys.

[fiery crescent]

Workshop should prevent broken mods from being uploaded or mark them with some sort of number of errors on the main mod page or have two sections of mods, broken and clean for those who download them to be informed of what they are getting. This would most likely encourage devs to fix the mod in the first place. Just a thought on my part.

[quaint flame]

Just that there is no way to check if a mod is broken except manually

[normal kettle]

Yeah the only option would be that modders need to change like a version number somewhere in the config after every game update, so the Workshop checks that it has been updated to the latest game build and then some tag like [OUTDATED] is removed from the Workshop

[quaint flame]

That makes 0 sense

[fair ore]

What R3vo is getting at is that you usually need a human to decide whether something is actually broken

[normal kettle]

I think some game do/did it like that. Might just be enough that update was made in the Workshop no matter what it's

Doesn't actually check if something is actually broken

[quaint flame]

Just because the game updated doesnt mean that the mod is outdated. On the otherhand if the mod relies on a different mod that had a major update it might be outdated

[obtuse bramble]

I believe this is what's probably happened to ARES, as ACE had a few big ones.

[quaint flame]

Likely

[fiery crescent]

This was a case a year ago, today AI to some degree can do the basic SQF logic

[quaint flame]

No they can't

[worldly badge]

Who/How does someone can tell a Mod is working intentionally

Nobody can. No human can. No author can

[fair ore]

I think an author can tell if it's working intentionally but uhm

who's gonna admit it ain't

[empty goblet]

Maximal UI pain Damage mode enforced

[quaint flame]

Only if it's full screen with a white background 😁

You gotta burn it into the retina to have any effect.

[empty goblet]

it could be always made more punishing, there could be timer in the popup button before OK/close is allowed 🤣

[quaint flame]

Followed by a few "Are you sure?" prompts

[fiery crescent]

And count down to FORMAT C:

[proud cape]

i would be the kinda guy that returns errors to sender so people are force not to use hopelessly broken mods, but thats from an architectural/maintainability standpoint

[fiery crescent]

I'm keep getting weird simulation freezes on server V25. It happened a second time during the past 2 weeks. When it happens the simulation freezes until mission or server restarts. Players can move around. VON is working. Interaction with objects not working. AI stops moving. No server crash. This is in Escape missions. Not sure in which perf ver it started. I jumped from v18 to V25 and don't recall it in V18. Since there is no crash log I don't know how to report this. I sent the RPT to Dedmen after it happened the first time.

[light cargo]

bisect with gdrive folder

[fiery crescent]

? Can you explain, sorry for not getting it.

[light cargo]

test the binaries in between from the google drive folder

[fiery crescent]

Got it. Steam downloaded vs Google drive for V25 in my case. Oh no test different versions from Google drive gosh

[static briar]

official escape mission with no headless and no mods?

did you try v26 (latest prof build on steam) and v21 (I think it better than v26)

[fiery crescent]

Escape with CUP RHS LAMBS.No headless. Latest official GitHub Escape not Workshop. Just previous server
versions than v18 and jumped to v25. Will give it shot. Just don't want this to slip into stable eventually assuming that it is a dev build issue which seems like it to me since nothing except server version changed in my server setup.

[void badger]

I feel like there might be a known issue where if a player has some... network issue connecting, they're blocked from seeing the lobby screen, and block any others from connecting as well until they Alt F4?

Experienced that this evening with a player who normally connects fine, validated files and nothing was broken

[whole cloud]

I've seen it too, but couldn't reproduce it enough to get usable logging out of it

[uncut wagon]

Hey there, why is this chat so chaotic? Is the branch going wild?

Dedmen, have you added textures from extensions yet?

[static briar]

For reference, this happened to me occasionally with a previous stable version (around 2.14).
I determined that the cause was the network configuration and a large number of error logs on that time.
I was also using LAMBS, but I was using a lot of mods and a large scenario like KP Liberation, so I don't think there was a direct connection.
Even then, there were no error logs or dumps that clearly indicated the exact cause, so I wasn't able to completely identify it.

[whole cloud]

yes

[uncut wagon]

On stable?

Question is, setobjecttextureglobal will just broadcast the texture to all clients right?

[whole cloud]

Would you check when stable was last update and not waste time with such stupid questions?

[whole cloud]

No textures are not broadcast, except the Squad logo

[uncut wagon]

I could yeah

[uncut wagon]

Ugh, alright, will it just stay local then? I'd love to make a custom bilboard system or something like that

I guess that for it to be seen by everybody, everybody should have the extension

Broadcasting textures wouldn't be a good idea now that I think about it 😅

Aight sorry for these questions lol I'm not really locked in yet

[rain moth]

what will happen on server? and on linux server which as no gui?

[whole cloud]

Same as warning messages already do

linux I think they just write log message.
Server probably pop up

[rain moth]

that will be fun for hosting providers 😆

[whole cloud]

Why?

[rain moth]

"my server doesn't start"

[whole cloud]

I turned a game force-killing popup message box, into a non-force-killing popup message box

nothing changed about the popup message box being there

[rain moth]

at least people will still have to stop using broken mods

no objection, just curious 🙂

[heavy vortex]

We've had a lot of perf-branch clients (server is stable) in the last few days with intermittent mission startup issues. Symptoms are apparently the display flickering between black & 3d view after clicking continue on the role selection screen, and no progress.

Most players just switched back to stable once they had problems so I'm not sure whether it's been fixed already.

[fiery crescent]

How do I check if the server crash dump has been auto submitted?
Fault time: 2025/10/05 02:57:24
Fault address: 8000B0A3 01:0000A0A3ll C:\Arma 3 Server\Dll\tbb4malloc_bi_x64.dll

[whole cloud]

CrashReporter.log

[fiery crescent]

No CrashReporter.log found in server directory. I sent you dmp directly.

[violet knot]

Sorry I was away for the weekend and couldn't respond right off.

Observed the problem as we use MV-22 on our server's scenario as a mobile respawn and as transit between points of interest on the map (Altis). Players started observing the MV-22 was crashing or getting hung up.

Later tested using no mods and with Blackfish and no mods or Creator DLC and followed the steps below 👇
Using profiling version from Steam, used VR map in editor, put down basic helipad, place Blackfish to flying height about 500m+ away from helipad, set all crew to non-playable/non-player, gave group waypoint to "Land" on helipad. Play.
Blackfish tried slowing down to land but never Vectors from watching it

Inversely, reverted back to standard binaries on Steam, repeat steps above in editor, Blackfish lands gracefully.

I run the profiling/performance binaries (off Steam) on our servers and personally use it on my client.

Reverted Server binaries to standard and now MV-22 operates normally with AI crew.

If this has been fixed, I have not checked as I was away for the weekend and will be on my PC tomorrow evening.

Thanks for your attention to this.

It might have been broken for a while and no one noticed on the server till recently. Not sure when it stopped working accurately. It had to be a few weeks instead of months though, as I was riding in it during some testing and landed as expected. Think either late August to early September and was pretty diligent on checking for updates via Steam update for profiling.

[whole cloud]

Using profiling version from Steam, used VR map in editor, put down basic helipad, place Blackfish to flying height about 500m+ away from helipad, set all crew to non-playable/non-player, gave group waypoint to "Land" on helipad. Play.
Blackfish tried slowing down to land but never Vectors from watching it
Yeah I tried same repro. Not sure what "Basic helipad" means, I used the square one but shouldn't matter.
Mine never really slowed down much, it just ended up flying circles, also never vectoring.
I also tried it starting out as flying or starting out as landed and taking off, same result.
But exact same behavior in 2.20 stable. I even retested it again right now

I suspect the autopilot changes. There was one that could activate while AI piloted, I fixed that in v30 coming today.
Please check then if that fixes it.

if thats not it, then the only other thing I can think of are the Helipad changes, they caused issues before and were fixed multiple times already. But still that would mean the issue is in 2.20 stable.

[whole cloud]

I think with 2.22 I'll stop forwarding the changelogs to here

[eternal kraken]

hey dedmen, the third link in your post direct me to another ticket(one of mine T193138) instead of the autopilot ticket.

[whole cloud]

Ah oops. Yeah

[whole cloud]

v26 released two weeks ago. And the problems only appear now?

[heavy vortex]

First batch of problem reports was ~6pm October 1st, so probably v28? I can't guarantee that they were running the latest perf branch version though.

[stark dragon]

Does this mean we're good to run the command line prompt again to try and diagnose the long start times?

[whole cloud]

yes, hopefully

[opal hound]

Received by automatic reporter or do you want these?

[whole cloud]

I got them already, it finally works

[opal hound]

wicked

[whole cloud]

Actually it did work last thursday and friday. I got two batches.
maybe it still works 🤞

[opal hound]

yea haven't run a session since

[violet knot]

Thanks dedmen!!

[heavy galleon]

This is the Dedmen praise channel, where is Crumble praise channel?

[whole cloud]

Crumbling

[heavy galleon]

Only took over a year for the task to be assigned
Crumble deserves to be praised

[patent sky]

direct it to dedmen instead, this is just copy pasting 10 lines of code 😂

[heavy galleon]

But you deserve some too!
(That is if you actually get one of those changes I proposed in game )

[silk pewter]

long live Crumble! (but not as much as Dedmen because he just copy-pastes) *runs away tailorswiftly*

[fiery crescent]

@whole cloud Have you had a chance to look at the DMP I sent you on Sunday? I don't know why the auto reporter did not work.

[whole cloud]

No. The reporter didn't work because it was deleted, probably

[fiery crescent]

Interesting, I see reporter exe there

[whole cloud]

yes now because we updated it yesterday

[whole cloud]

We'll get back to stable eventually

[eternal kraken]

"Cleaning up MPMissionsCache folder (If above 1GiB, deleting oldest files that are over 50MiB except the 5 newest)" i dont like and dont want this

[acoustic lynx]

Can we disable that?

[whole cloud]

How many files larger than 50MiB do you have?

[eternal kraken]

biggest is domi by xeno, none pbo with 50mb

[whole cloud]

Okey so then it is disabled. You have no files to be deleted

https://discord.com/channels/105462288051380224/1423686204620017685 small survey results say the same.
Basically no-one affected, except the worst case

[knotty wraith]

мне кажется, есть повод переживать, что будут удалены уникальные миссии, которые через некоторое время можно будет найти только в кеше...

it seems to me that there is a reason to worry that unique missions will be deleted, which after a while can only be found in the cache...

[whole cloud]

Yup I had the same worry. Which is why the restrictions are as they are.
Only big files are deleted, which basically no-one would do before the HTTP mission file downloads was introduced.

Its a cache, not a backup. If you want a backup, make a backup

[empty goblet]

@whole cloud this is wrong, i used often the mpmission cache often to check old missions / variants etc., i suggest to move or rename files to not be used

[whole cloud]

How many files are larger than 50MiB?

[empty goblet]

quite some were

[whole cloud]

Then if you want a backup, make a backup

MPMissionsCache is a cache, not a backup

[empty goblet]

it was always persistent cache, so give control over the cache, in era of multi TB SSD with 5GB/s speeds that problem is less relevant than 20 years ago

if the problem is size or version, you can link key it to version of game ie \2.22\ subfolder in the mpmission cache

[whole cloud]

The problem is files accumulating there and filling up space.
Making subfolders or moving files around doesn't delete files

[empty goblet]

then expose it into RPT file and into UI ... make it warning , e.g.```
%drive%:\Users%username%\AppData\Local\Arma 3\MPMissionsCache
exceeds 1GB of space, please clean up old cached missions to prevent game slowdowns
or / and (for the other case of problems)
exceeds 50% of free space on %drive%, please make sure you clean ol missions
or / and
offer the cleanup as backup (the files will be moved into /old/ sub-directory

another thing to consider is option to define the total threshold ... some may want it to be 100MB another one it to be 1TB and the 'large mission' size too (some may consider it 25MB another one 100 etc.)

tons of old missions were already >20MB and that was like ... pre 2020

and lot of mission makers don't listen to 'optimize the size' argument anyway

[uneven bluff]

Can put a button in the launcher, maybe?
It would be nice to have a way to manually trigger clean up. We've had problems in the past where the cached mission was the wrong version and that causes the player to be disconnected from our test server.

[empty goblet]

if we even can update the launcher that easily, would be faster to do it ingame ui 🤣

[uneven bluff]

Or yeah an ingame button.

[empty goblet]

plus it needs to be in both ... because not everyone uses launcher for those things, so in the end it needs to be at least in the ingame

[uneven bluff]

A launch parameter called -makeMyGameFaster. Once all the youtubers make videos breathlessly explaining how to speed up your game catch onto it, it's game over.

[empty goblet]

and i would prefer the ingame to slap me into face that i have cache bloat to clean up (be it manually or by the game but no files 'deleted' suddenly)

but then maybe it's just me who finds useful to be able look at cache folder and find there obscure mission from 2016 or older which got lost and the author is full of joy someone has backup 😉

today's world is more about 'have it now, fast, instant and forget it before midnight'

[uneven bluff]

I wonder what's the average large mission pbo size. People used to complain that Warlords Redux used to be 6-8mb and I had to cut down all the big textures.

[whole cloud]

"to prevent game slowdowns" it doesn't slow down the game

[uneven bluff]

Like Altis Life etc those should be at least 10-20 mb right?

[patent sky]

~50-100MB

[empty goblet]

i have even some koth and wasteland missions in the history cache which were over 25mB

[patent sky]

not 50 tho

[empty goblet]

i have some life missions with 300mb+ size

[whole cloud]

Well the limit is 50. Not 25

[uneven bluff]

Jeeeeez. That's a lot of download on the server.

[patent sky]

http mission file download 😎

[empty goblet]

don't ask, someone though it's great to slap in some pictures with 4k resolution w/o compression and then show them in 50 or 25% size and pack that all into mission file

[uneven bluff]

That was our logo before compression 😄

Cut like 2MB from that alone

[knotty wraith]

v.30 -no one can log in to the server

[whole cloud]

I just tested v30 and v31 on my own server. Mine works fine

[empty goblet]

another extreme cases were sound files, some missions had thousands of them and some with huge sizes (like 1M> sized oggs)

[uneven bluff]

Yeah those can get big. I had an audacity setting that compressed everything down to <10kb each file that made everything sound like they were over the radio so that was a nice plus.

[empty goblet]

so the problem is purely size of the cache for some ? still applies to the fact some may need to make that n times smaller and some would be fine if it's n times larger

[knotty wraith]

18:42:40 glen uses modified data file
18:42:43 Player glen connecting.
18:43:52 Player glen disconnected.
18:51:23 glen uses modified data file
18:51:25 Player glen connecting.
18:51:57 [ANVIL] Fanat uses modified data file
18:51:57 Player [ANVIL] Fanat connecting.
18:52:20 Player glen disconnected.
18:53:19 Player [ANVIL] Fanat disconnected.
18:53:26 Negrbl uses modified data file
18:53:27 Player Negrbl connecting.

[whole cloud]

Yes the size is the problem.
If you want to use it as a backup, then back up the files

[empty goblet]

even level 5 or level 6 are still way smaller than the not needed level 10

[whole cloud]

I had that once, but then I restarted the server and it didn't happen again

[empty goblet]

again i have no problem with that, but you changing how the mpmissionfolder cache works w/o giving chance to actually backup anything, until now it was manual to clean persisting cache

also what happens if someone plays on rp server which has 1.1 GB sized mission ?

yes, absurd example, so what about 4 servers with 255MB sized missions each ...

it defy the point of persisting cache folder because you picked arbitrary value

[whole cloud]

Nothing because it keeps the 5 newest files.

[whole cloud]

Also absurd example.
And also, nothing because 5 count limit.

[empty goblet]

so it needs to be 6 files over 50mb which are in total over 1GB limit ? then he loses the oldest ?

[whole cloud]

yes

[uneven bluff]

Prompt: Your mission cache folder is getting large and slowing your game down, would you like to clean up all but your last 5 mission pbos?
Yes, No, Don't Ask Again.

[whole cloud]

Maybe for the "I constantly rotate between 6 servers that all have above 50MiB mission files) scenario I also need to add a minimum file age limit. Only files older than a week

[empty goblet]

also the slowdown were there in past (unless the code got cleanup since), ie on slooow 5400rpm drive with lot of missions in the cache

[whole cloud]

We don't read the missions in the cache, only the one we need (matching name, of which there can only be one)

[knotty wraith]

Rpt ...
SQF 17:19:15 ["warrespawn1"] 17:19:15 CUP_I_WILDCAT_GREEN_AAF: M681_ROT - Неизвестный источник анимации M681_ROT 17:19:15 cup_i_wildcat_green_aaf: m681_hide - Неизвестный источник анимации m681_hide 17:19:15 ["warrespawn1"] 17:19:15 ["warrespawn1"] ???? 19:02:05 Сервер: SetServerstate никто 19:02:05 Релиз менеджера EPE (0 | 11 | 0) ``
RPT ничего не печатал в течение 2 часов ...

RPT...
```sqf
17:19:15 ["vehRespawn1"]
17:19:15 CUP_I_Wildcat_Green_AAF: m681_rot - unknown animation source m681_rot
17:19:15 CUP_I_Wildcat_Green_AAF: m681_hide - unknown animation source m681_hide
17:19:15 ["vehRespawn1"]
17:19:15 ["vehRespawn1"]
*???? *
19:02:05 Server: SetServerState to NONE
19:02:05 EPE manager release (0|11|0)
````
rpt didn't print anything for 2 hours...

[whole cloud]

Okey nothing needed to be logged for 2 hours, probably because server was empty.
Why is that a problem?

[empty goblet]

ye i think that solves the rotation problem ... i still wish there would be way to define the thresholds tho but depends how cumber is it to implement

[whole cloud]

Anything UI is too much effort for this

[empty goblet]

just setting arma3.cfg would do ... ie tripleBuffering isn't exposed in in-game UI too (tho it should be in the launcher)

[whole cloud]

Average player should be protected from their disk being bloated with endless mission files they won't ever look at again.
Yes there are people who want to keep old missions as back up. If you want to keep a backup, make a backup.

[knotty wraith]

Просто наблюдение и ничего более. Это первый раз, когда я увижу это ... даже если игроков нет, сервер всегда что -то печатает в RPT

just an observation and nothing more. This is the first time I've seen this... even if there are no players, the server always prints something in the rpt

[empty goblet]

that's the thing, lot of people discover the treasure trove of old mp missions in old mpmission because of that not-cleaned 😉

so maybe turn it into one time event 'in the future the mpmissioncache' gets cleaned, make sure you backup what you need or it's gone

[whole cloud]

I have had that bug a couple times now where a player would get stuck in "connecting". Even multiple players getting stuck there together.
Then some player disconnects or timeouts, and suddenly all players that were stuck make it through all at once.

There is some very weird timeout code in player connection that doesn't seem to make sense.
Something about time outing the first connection, and then rotating the queue so that "others can connect too".
That sounds like the first player could block others if they are slow. But every player in the queue is checked every frame. The first one cannot block anyone else, I don't know why that is there.
Either leftover from old times and I can delete it, or it actually does serve a purpose and I need to delete it and implement something better.

I'm adding server.cfg option for logging of the connect process so we can see where they get stuck.
But if v31 is stable then only next week

[empty goblet]

there was issue with the socket timeout, where the IP was unable to reconnect until like 600s later ... if i remember right it happens since around the change where BE traffic was merged into game data

[whole cloud]

BattlEye. Notnthis

[violet knot]

Where can we see logging on this? In the connect logs?
I've witnessed this before too and thought it was related to "deform" terrain data. Regenerated the mission PBO and it went away.
Would this cause this too?

AI VTOL Issue
Also as far as the testing you asked for with the Ai pilot in VTOL craft landing, still haven't seen it working after switching back to profiling on Steam. (Should I manually update binaries?)

Still works in normal release (standard, non-profiling) despite circle and square civil helipad. All versions of Blackfish as well with no player control of the vehicle.

If you still want videos or repro steps let me know.

[whole cloud]

Nowhere yet

[whole cloud]

yeah I need better repro steps.
Maybe a VR mission file that does it?

[violet knot]

Where do you want that posted? Unable to post here.

[latent socket]

One of the mission makers in my group says It now takes over 5 minutes to save a mission file. The game freezes for around 4-5 minutes. Doesn't matter how you save the mission, it will freeze every time

[whole cloud]

I just tested saving for that map screen fix and didn't observe any slowness.
Where are you saving?

[latent socket]

Missions folder

[empty goblet]

also another reason to not cull MPmissionsCache, what about people with metered download who play still on same server same large mission for months or years (plus w/e reason it goes over the 1GB and over 5x50MB )
at max would bother with MPmissionCache culling based on % of free space (when it's VERY low like 5% or 3% and less) on the drive (and before doing that made check if the MPmissionCache isn't junction at different drive)

[knotty wraith]

Я думаю, что это было из -за проблем с Steam по всему миру.
https://downdetector.com/status/steam/

I think it was because of the problems with Steam all over the world.
https://downdetector.com/status/steam/

[foggy vine]

Scheduled Steam maintenance every Tuesday

[analog acorn]

There was a substantial unscheduled outage before the regular maintenance, possibly the result of a DDoS attack. Several other services also had outages.

[vagrant otter]

does perf branch ever causes crash when the game is overloaded by AI or assets?

[patent sky]

it shouldnt, if you have crashes submit them using the crash reporter that pops up

or automatically on prof server

[vagrant otter]

i stopped using perf because of that and it's been a long time, happens often when the game is hella laggy from assets and such, switching to vanilla seems to fix the problem entirely tho :v

[patent sky]

well without crash reports we dont know about it

[vagrant otter]

its fine tho, just wanted to share that experience just incase someone is having the same issues, could also be my potato laptop so no worries

[patent sky]

well most prof changes are coming to stable when the next update releases, so it might start crashing then

[whole cloud]

I mean in which mission?
Some big campaign mission, in editor, multiplayer?

[whole cloud]

That's a lot of "ifs" you had to chain together there.

[whole cloud]

Well if you switch back to Prof and get us the crashes, we might be able to fix it.
If you don't, all the changes will probably go to 2.22 stable anyway.
And then you're stuck with crashing, when it's too late to fix it and you have no other alternative to switch over to.

[vagrant otter]

im not complaining dawg, i just wanted to give insight, i can run stable just fine, recreating itself is difficult enough, as sometimes it can tank it and it crashes at the most unfortunate times

its a matter of gamble, but i'll try to recreate it tho

[whole cloud]

Didn't say you're complaining.
I'm just sad that someone has issues and that I don't know what it is and can't fix it.
And worried at what other issues were missing out on.

[vagrant otter]

ah okay sorry, it was a miss interpretation by me then lol, dw i'll try to get the crash report, though sometimes it would crash WITHOUT a crash report..? idk about that issue

[static briar]

Loading time issue since v22 hasn't been fixed yet for me.
"Updating base class" is too long. (also creating cache are not faster)
My friends and I have been using v21 for a while, but because we use a tons of mods, the load times are almost 10x faster than v31.

(start time is rpt title time, ends to "MovesType CfgMovesMaleSdr load time" shows in log, with 100+mods)
v31: 16.8min (launching section are 10min)
stable: 16.3min (8.9min)
v21: 2.2min (0.6min, wow!)

I check the rpt, stable and v31 has almost same "launching" section load time...

Will this loading time eventually go back to like v21?
Or will it become stable with these slow load/launch times?

(yeah I know those mods are bad but v21 is gold...)

[patent sky]

Did you profile the startup with the -profileGameStart argument?

[static briar]

Not yet.
wait, does that change the loading speed?
I thought it was just for displaying logs for investigation.

[patent sky]

I thought it was just for displaying logs for investigation.
yes to investigate it we need to know why yours is so slow

[whole cloud]

Yes I didn't make any fixes.
I added the parameter to find out what's wrong

[patent sky]

is the argument prof only or also perf?

[whole cloud]

Its profiling the start, so prof only

[static briar]

Um, this is a stupid question, but is the prof beta version in the steam version prof or perf? if it's prof, I need download v31 from gdrive.

[whole cloud]

Steam has both files in game folder.
arma3_x64.exe (the default) is perf. But there is also arma3profiling_x64.exe that is prof

you can swap them out by renaming

[static briar]

Ah, thanks. So the result from before is perf. I'll switch to prof and take a log. Thanks.

[whole cloud]

prof should have about the same timing. But with the start parameter it will write a log that says exactly which step took how long

[static briar]

Ok... Thank you for your kindness. I'll measure it later when I have time.

[whole cloud]

Jup got them thanks! Going through now..

allowCrewInImmobile script command, on a local-only vehicle is missing safety check.

But.. That seems very suspicious.
Ref to nonnetwork object 23c0cb35800# 3369452: uk3cb_mtlb_pkt.p3d UK3CB_AAF_O_MTLB_PKT REMOTE
So the log knows its a REMOTE vehicle, thus it inherently must be a network object.
But its still saying its a non-network one...

Oh no
It IS a remote object, the network message to spawn it just came in, but the init script of the vehicle, runs and tries to run that command (Which says its local arg, so it should never be called for remote vehicles. That's a bug in 3CB script, assuming its them setting the init eventhandler)

Or actually, is it ACE maybe?
https://github.com/acemod/ACE3/blob/ffb80c148178724619660354496d2c995be8efa9/addons/vehicle_damage/functions/fnc_addEventHandler.sqf#L29

It is ACE!
https://github.com/acemod/ACE3/blob/ffb80c148178724619660354496d2c995be8efa9/addons/vehicle_damage/XEH_postInit.sqf#L56

ACE is running allowCrewInImmobile (a local-arg) command, on remote vehicles.
And its doing it so early, before the network ID is even assigned properly (Thats a engine bug, I'll take a look at that).
If ACE adds the proper local check to not run that command, that problem goes away.
(Can someone make a ACE PR? You can use it for Hacktoberfest points)

Any init script that would send a network message will fail here.
Because the network ID is only assigned after the Init script.

"local _vehicle" check will work though, that is set before the init script.

Uhm actually allowCrewInImmobile is not local arg.
Specifically has code that if vehicle is remote, it'll send a message and tell the owner to do it.

Still bad to be doing that in an init script that runs on every remote player though

[whole cloud]

Ref to nonnetwork object 43712: tracer_white.p3d CUP_B_9x19_Ball

setShotParents, called on a bullet. Bullets are local-only.
I remember having a conversation about this before somewhere. That someone needed/wanted it...

I'll add a -debug warning for it
Also actually, the script context printing should work for this
The report I received though doesn't have the full RPT, and seems to also not be running -debug on server.

Or no warning.
setShotParents will still have local effect 🤔
But also only on the server side, because the command is server-only.
But the bullet hitting someone will probably not be happening on the server?

[analog acorn]

https://feedback.bistudio.com/T178981
https://feedback.bistudio.com/T180763

[whole cloud]

Ah! Indeed 😢

[eternal kraken]

i have a new observation, when i am playing on the server "HUD Arma3 DK #1 Domination" with perf build i will get the message "You were kicked of the game" or "Session lost" within 30 minutes every time i join (sometimes 5 min sometimes 15 sometimes 25) but when i go back to stable i get no kick/lost. i can reproduce that behaviour. anything i can do to gather information what could be the reason?

[spiral pond]

would it be feasible to cache the merged config as file?
and load if the same hash of the whole modset gets loaded again (dunno when the engine computes that - probably before going by the rpt header tho)

[whole cloud]

Yes, I already have code for it. But the merge step really doesn't take long enough to be worth it
And it also looses configSourceMod information... So, that would be annoying

[whole cloud]

Do you know what the server says, for why you are kicked?

[eternal kraken]

i have found no info about that in my rpt so i can not tell

[whole cloud]

The server would know 🤷

[eternal kraken]

when i try to rejoin instant the server tells me how it is configured i need to wait 90 sec

but after that i can rejoin and play again but with perf build i get kicked within 30 min

with stable i dont have this issue

[whole cloud]

Its up for the server to decide to kick you. It could literally be anything

Maybe they check that you are on profiling branch, and script kick you after 30 seconds 🤷

[eternal kraken]

i dont hope so 😂

that would be evil

no the kicks are random

maybe something like an network message send overload

but i only speculate i dont know

atm i play on stable to be sure and proof this is an perf build problem at my machine

[whole cloud]

But the engine kicks, afaik all have kick messages

[eternal kraken]

last rpt entries

i will send you rpt with PM

look yourself you know better

[whole cloud]

The unhandled message and data is the key. Some netcode change slipped through.
v27 contained a one-byte netcode change that wasn't supposed to be in there

Explains the session lost, but not the kick message 🤔

[latent socket]

Ah, yes. In the editor

[heavy vortex]

Server RPT never had kick messages, did it?

players just... disappear

I think some of them you get in the server console instead but I don't usually get to see that.

[whole cloud]

Is it server console log?

If kicks are not properly logged we gotta fix that

[heavy vortex]

would certainly help. Networking issues are pretty damn opaque at the moment.

You usually have to ask the client what messages they got. And they're not always very observant.

[empty goblet]

that got my attention 😁

[spiral pond]

for many mods with poor configs it might (unless an underlying problem only in perf branch causes this)

from @static briar - more details in #1422598532237627543 and #1301458792323743817

[whole cloud]

That must be a bad bug

But what is causing it.
So this profiling won't work for it, I need actual profiling like vTune to see what its doing there.
And I cannot repro it myself

[static briar]

Yeah, to be honest, I know the my mod presets are a nightmare, but they still work fine in v21 and the loading speeds are the fastest they've ever been.
I'm just wondering why they're so slow in v23 and later version though...

[severe prairie]

yesterday our community had a very weird incident, upon a cruise missile exploding, all players running the profiling branch were booted back to the lobby screen, the players running stable remained in the server normally

of the players that were booted to lobby, no one crashed, or was disconnected, just booted to lobby with no error message, server console showed nothing of note at the time of the incident, the cruise missile was the vanilla launcher from the destroyer ship

our modlist is quite extensive, so i dont know what mods modify that cruise missile

does anyone know of anything that could cause this, or has had something similar happen?

[whole cloud]

I am wondering the same, and I'd love to know

[whole cloud]

#perf_prof_branch message

[violet knot]

@whole cloud Just tested v32 on the mission file I sent you. Ai piloted VTOL still flies over "Land" WP, but able to go back to standard (non-profiling) release 2.20.152984 and it works.
Not sure if you saw my PM

[whole cloud]

PM is still on unread, for probably tomorrow

[violet knot]

No rush, working for now on std, posting for awareness

[whole cloud]

Also seem to have found the game loading slow issue. But weird that it was reported since v23, when the code that I undid to fix the issue was only introduced in v27

[stark dragon]

Thanks for all you do in here.
As someone with a lot of live-service games experience, I know how stressful it must be. 😂

[knotty wraith]

- Добавлено: server.cfg/AdvancedOptions/logPlayerConnectionStates для регистрации прогресса "подключения" игрока
что именно мне следует прописать в server.cfg?

- Added: server.cfg/AdvancedOptions/logPlayerConnectionStates to log player "connecting" progress
what exactly should I register in the server.cfg ?

[opal hound]

  logPlayerConnectionStates = true;
};

Probably

[uneven bluff]

Changes to thermals seem to not play well with the spectrum device secondary. Other handguns work fine.

[whole cloud]

The lag spikes? That's not changes to thermals, that's something else

[whole cloud]

Why does my spectrum device look different than yours?

What am I missing in the repro

[weak panther]

need run game with Contact for UI work

[whole cloud]

I do. If I switch over to contact campaign, it works

[quaint flame]

Seems like he's running the scripts to make it work in the editor

[static briar]

I don't think so. His editor doesn't show folder names (see top right), so I don't think it's loading description.ext or sqf.

[quartz rampart]

a script does not need to be in the mission but it could also be a mod.

[static briar]

yeah I know but that is looks not with any script. I did and can repro it.

wait a min I'll upload the vid

[whole cloud]

Oh the game isn't freezing, just the image/rendering is stuttering but the game is otherwise running high fps?

[static briar]

yeah it's not freezing. unit still can operate with key. only happening while moving forward

for ref: vid setting is ultra

stop moving forward, bugged screen is return to normal

update: it's now happening with combat pace

looks only happening while sprint/jogging

also, stamina bar is still looks fine even screen is bugged

Yes, that only happens at a certain speed...

[static briar]

but not with car/ffv

[whole cloud]

It looks like rendering is not writing to backbuffer. It flips back and forth between the last 2 frames, they are not overwritten.
UI is later put on top thats why that isn't affected

[knotty wraith]

in v.32, addEventHandler "HandleDamage" is broken (ACE mod is present)