#perf_prof_branch
1 messages ยท Page 30 of 1
And Its super annoying to debug a server remotely.
Local running dedicated server, with latency via Clumsy would be a doable repro scenario
Apparently I forgot how to make Clumsy do anything to A3 local DS.
Could have sworn that udp and (udp.DstPort == 2302 or udp.SrcPort == 2302) used to work.
I used it like that pretty sure
yeah I might have got lucky before and not picked the localhost.
Seems to work on second random attempt. No replication though.
And that's with much worse settings than the real packet loss.
Does anyone have a picture of said placeholder for missing models? Just kinda curious what they look like :)
Ahhhhh hahahaha looks like I reacted to it too- I must've totally forgot. Thank you ๐
Even tried Clumsy on a live server and that doesn't trigger the verification errors. Something else is going on with real packet loss apparently.
What happened yesterday lol
First is some corruption in the vehicle types cache. I have seen that happen before, mainly at game exit.
I don't know how that happens, and it seems to be very rare.
Second one is something with AI, need to look closer at that, but its in a place I fixed previously ๐ค
And what happened here xD That's alot of crashes in a 5 minute span
Crash on startup with auto-restart script?
Not in my case, 1st crash happened after some 5 hours of server being online 2nd after some 3 or so hours of being online.
Oh its that ObfuSQF thing.
Server with messed up files, probably on auto restart yaeh
I accidentally found a repro for the texture loading crash.
A texture is inserted into the cache, even though it already was in the cache.
I wasn't even looking for it 
Does that include clinet crashes or no? @ ping is turned off
yes
I need to see big error signs in arma 3 ๐
ok
Dedmen I don't use ObfuSQF, in regard to server crashes from yesterday
Yeah I know, those crashes aren't yours
And they look totally different
I already wrote what your crashes are, and I didn't mention ObfuSQF
I now know how it can happen.
Its a mistake in new code I "wrote" today, but the problem actually comes from a line I copied from elsewhere and then modified.
The thing I copied was already broken, but was luckily in a situation where the issue would not happen. Probably other places that make the same mistake.
Will take a while to search all places this is done at..
But I have a hotfix for now. Instead of inserting twice and corrupting the cache, if its already in there, just remove it first.
That seems to prevent crashing/freezing ๐ค
Prof tomorrow
As far as I can see the map icon drawing code doesn't make the mistake though.
Quick question is Added: 3DEN Action "CreateTextFile" to create files into mission folder within Eden Editor or is there a way to chose extension of the file instead of beaing .txt to be .sqf or .ext ?
If the mission hasn't been saved yet (so no mission folder) where does the created file go? Just in the base missions folder?
It just fails because it can't find the folder
Sadly do3DENAction doesn't return the result, even though it has a bool for it
I didnโt know you could just execute a .txt file like that. I was always under the impression that script files had to be .sqf. Thatโs actually really cool to know
i heard some stuff about base64 in dev branch, are we getting it anytime soon to profiling or standard?
That was released in 2.20 https://community.bistudio.com/wiki/ctrlWebBrowserAction
oh i totally missed that thanks
thats only in the ctrlWebBrowserAction no?, is it not possible to get a base64 string and apply it as a setObjectTexture?
(This also works with controlNull)
it says on the page
One day I'm gonna try deflate on our saved games :P
Slug Algorithm for rendering fonts on GPU is now public domain.
https://terathon.com/blog/decade-slug.html
Is this useful to you guys?
Maybe for the UI/HUD, but I'm not sure it's worth the cost/development time to implement, especially for a 2013 game 
will there be a full release of perf branch in the future?
Improvements of this branch generally make their way into every next update of the main branch
thank you
just wait for 2.22
no
1.54.134547 profiling & performance v20 server and client, windows/linux
- decreased BE log spam and resolved some issue
https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm
new aprilfool.. lol (because I'm eastern time ppl)
hmm it's upside-down 
psst ๐คซ
tbh it scared me lmao
@whole cloud you are working so hard for us thx so much m8
Finally the correct way to title it, ษษฏษนโ
finally, australian release
Why spoil ๐ฅบ
epic
look out that a Demogorgon does not appear around the corner and attack you.
yup... those that live in the future send spoilers
This is
Just peak
I love all these little easter egg things that show up
Im there's just a few people who actually check this channel ๐
would be funny to keep it that way just for the Perf/Prof binaries
Is there a way to limit FPS with VSYNC off? Arma.cfg refresh setting seems to make no effect.
limitFPS launch param
The video card settings have this option
Thanks
does that work for clients? thought its for servers only
Mm... never tried, just looked it up tho and it does specify server/hc so maybe not
I meant the wiki says it works for server and hc, it doesnt mention regular clients so it probably doesnt work for them
Nevermind its a 2.20 change that says it works on clients too
So.... yes, limitfps works everywhere
ah right
Wierdly, it was only crashing on Altis for us, other maps were fine for a full session
this update is adding this attribute line to units in missin.sqm and gives error when adding new units to mission and opening edit attributes. No mods. Should i make ticket or is this something broken in my end?
class Attribute1
{
property="VehicleCustomization";
expression="if (local _this) then {([_this] + _value + [true]) call (uinamespace getvariable 'BIS_fnc_initVehicle')};";
class Value
{
class data
{
singleType="STRING";
value="[[],[]]";
};
};
};
and it is giving this error
19:59:40 Error in expression <if (local _this) then {([_this] + _value + [true]) call (uinamespace get>
19:59:40 Error position: <+ _value + [true]) call (uinamespace get>
19:59:40 Error Generic error in expression
19:59:40 โฅ Context: [] L1 ()
[] L1 ()
<- VehicleCustomization
Ah so that's where thats from...
Okey so what..
We have a ton of attributes that wrongly have condition=0 and actually mean to be visible.
Except this one attribute, which would rather be always disabled?
yeah ugh
But the attribute also has value=0.
Can I just "if value=0, then accept condition=0 too" ?
Ah the "value" entry doesn't even exist.
What are the chances others have copied using value=0 too, even though it does nothing?
Also weird that this is an Edit control with a string, if you actually use vehicle customization, it sets the attribute to an array..
So why did condition=0 work on this one, but not on any other attribute??
AAAA
Ah F'ing hell.
Of course.
condition="0" DID work for global attributes.
But not for per-object attributes.
So on VehicleCustomization, it should stay 0 and hidden.
But on keyframe timeline attributes that also set it to 0, it should be ignored, because they are object specific.
Schmart
There is one non-object attribute that uses condition=0 
It infects mission files 
Once you open vehicle attributes once, it stays stuck in there.
Going back to non-broken version doesn't remove it
Haha funny April fools, infecting mission files 
Just update the expression to always exit. Next mission save will fix it.
@nocturne obsidian seems like you running w/o date-time timestamps in the logging (if possible use full format next time ๐ )
also if you have crashes, try to use -crashDiag command-line so we can see additional info
It was just too unstable for us to run on main servers sadly
"ShipX simulation would not apply gun stabilization" does this fix artillery computer on boats?
I know nothing about artillery computer
has the same issue. the wave wobbling the ship makes you not able to hit anything. I guess chair might know more about it
just don't wave at the enemy
Was the launcher being able to modify modsets with open game actually implemented at some point? Also, it's still on profiling and not stable yet, right?
I think profiling doesn't have the launcher changes, dev has
And could we? 
Its possible

Dedmen, I reverted from V58 to V55 and no server crashes. Maybe coincidence. I don't see v63 server related fixes so I don't update. Any time frame for the server crash related fixes? Update: got server crash on V55 as well so reverting from V58 was not relevant. Updated to v63. Will see how it goes.
Why can't you try v63? I'll test the same on my linux DS :)
You are right, I should try. It is just problematic to update all servers while no one plays just to do the rollback a few days later.
don't do all servers but a small selection of them and rotate and try doing it in the scheduled maintenance period
yea sorry man
had to drop to v16
cant risk running it too long
lets just say it was worse
no crash for me with v20, but player stuck on "receiving data" after some hours of uptime
Fixing the boat turret stabilization is awesome. The vanilla gmg speedboat was basically unusable at any kind of range, in almost any weather.
whats your maxmsg send set to
@worn obsidian
@empty goblet i guess the only was to fix it then ill turn the logging on but i can only risk a few crashes before i switch back ok ?
Is it an already known bug where your GPS/mini map changes position when using the profile branch? It fixes when getting in a vehicle, moves when you get out.
@nocturne obsidian 64
mission ?
What does changes position mean
The mini map moves from being where it is default to the bottom corner of the screen. I have a video of it happening while playing KOTH but cant post it here for some reason
Do you have a replication case that doesn't require playing KOTH?
I only have a video of it while playing KOTH
All I play really. I could try and make a video elsewhere?
Try creating a mission in the editor and see if it has the same issue there with the same mods.
Alright, the only mod I'd have active is the JSRS.
I had the same issue with the map randomly repositioning. Also using JSRS but other mods too. Maybe something for the JSRS discord
try without jsrs and other mods
jsrs is an audio mod so it's unlikely that it's causing this issue
see above
I'd like to bring into attention a (relatively) new eBPF-based profiling technique that recently became part of opentelemetry: https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation
It requires a symbols-enabled build (so won't work well with "public" builds), but with custom builds on official servers or in cooperation with trusted server owners - this has some interesting capabilities
the game~~ crashes~~ when i alt tab more often now, i thought this issue was fixed not long time ago.
edit: i meant not responding
1gig update
changelog 
nevermind, cdlc 
What the flipity flop
you have beenโฆDedmen'd!!
shortify, dmed or ded ๐
Is today marksman DLC birthday? ๐ค
I totally forgot lol
must be
OBI uses eBPF to automatically inspect application executables and the OS networking layer, and capture trace spans related to web transactions and Rate Errors Duration (RED) metrics for Linux HTTP/S and gRPC services
I don't see how this is relevant
It's tomorrow.
You must be in Asia ๐คฃ
Maybe I do need to add Timezone checks
Lol
I was already wondering how he got to see it๐
At least put it in a spoiler tag for the others
a marksman birthday celebration should also play the "boom headshot" sound with every tick of the loading bar
Thats actually a neat idea
Cant put a spoiler tag post, posting lol. Im sure the others will live tho
I really do love all the profiling easter eggs ๐ซถ
The idea is that you can get continuous profiling data from live application running in production - at 1-2% performance cost. I.e. if you know that a particular server is struggling under 100 players - you can actually have 100 players on it while profiling, and the act of profiling will not degrade performance. Plus, with backend like Pyroscope you can gather the data continuously and look into issues after they appear - without having to create a test environment and reproduce the issue.
Here is an article: https://grafana.com/blog/continuous-profiling-for-native-code-understanding-the-what-why-and-how/
Tracy already is quite low overhead, and we already have that, and it works on windows too
Tracy is primarily designed to operate in manual sampling mode, from what I read. But if you already use continuous sample profiling on the whole server executable, then yes, this has no advantages. Are profiling builds sampling whole execution flow?
They can if I wanted to
Then the difference is - with eBPF you won't have to enable it, the data would be already there.
Happy to help if you decide to check it out later.
@nocturne obsidian custom life
Tracy generated like 1 GB data / minute from what i recall - and thats for one user/client.
Does eBPF / its tools have some more elaborate data storage/handling?
yeah the april fools one got spoiled
I feel like you need some tracker techno to go with some of these intros
The last featuren I added to my Arma launcher, was a Tracker music player ๐คฃ
bass library in arma 3 files when ๐
Ah those funny crashes...
bet it's a life server
"GTA" ^ thats very ambitious project
Get A Life server, I see
arma can give you life? ๐ฎ where do i sign?
There is one annoying crash with the attachTo thing for handIK and head stuff..
I tried reproducing in vanilla and can't, but it keeps happening. Something is wrong with the face grimace's.
I'll make it show a prompt to create a fulldump, ๐ค that someone actually does and sends it over
Oh I found a way how to replace the crash with a "You dun fucked up mate" error message.
Probably will break all servers that run this bad obfuscation that tends to crash the game. But better that than crashes
"good birds don't obfuscate scripts in game, you have only yourself to blame" PLEASE
- I don't know if obfuscation tool used, or just very badly made mod
- I can't put joke messages for real things

Maybe they'll fade away after that
Judging by the name of the folder its probably Grand Theft Arma, which im fairly sure is a custom obfuscator? Could be wrong, but @sly ivy
They also have "obfusqf" loaded
Hm. ObfuSQF is a fairly common one, maybe they changed.
They have a over 3400 levels deep #include chain.
That's a bit <insert generic insulting slur here>
3400 fuck...
Thats where my log ends, because VisualStudio doesn't want to display more. There is more
I wonder if its somehow recursing the includes? I guess you'd be able tell
I don't understand the point of using obfuscation nowadays.
Also all the crash reports I've seen are from their own server and HC. They're just crashing themselves ๐คท
Don't steal my shitscripts!
you can hide your shame on the server side)
not only for this reason
I've a lot of maps and it is convenient to edit your code from one place (the serverside mod)
Just start adding crashes the moment a obfuscation like character is found.
Stop it at the source
Just call it malformed pbo 
Already did that for another type of obfuscation ๐
But they often use russian characters for it. I can't block russian made mods
why do treadmill work for this?
Just keep updating file headers like that .p3d update to ODOLK last year lol
and block the use of Cyrillic?
Yes, dedicated profile databases have very high compression. I'm mostly familiar with https://github.com/grafana/pyroscope - can't quote the particular compression factor, but the space requirements are low enough that we just run it 24/7 on production.
Tracy also supports zstd compression on tracefile
Also native tracy is more efficient than what ASP does
-# I don't blame you, you've been fighting those crashes for a while haha
FWIW though I think OfbuSQF creates dummy files as part of the obfuscation process. I believe that's probably what's up with the includes chain but not sure 
Honestly should just ban obfuscation... Mostly used by rippers and IP thieves anyway...
Eehhhhhhh that's a slippery slope, I'd just say those who Obfuscate basically have no warranty provided, which is typically how that goes in the world at large anyway
No help offered if you choose to not disclose your sources or content for any reason
Honestly I really don't see the point to banning it. Our servers mission has been being developed and built on for 10+ years. I don't think it's unreasonable we want to take measures ourselves to protect from IP theft when everything outside of the initial framework was made by us.
Unfortunate ObfuSQF appears to cause the issues it does and I'd agree that's their responsibility to fix. Banning tools that help mission and mod authors protect IP just doesn't seem like a great move though tbh.
Honestly not worried about people ripping scripts, more so as a deterrent to cheating in our case.
The higher the bar to rip scripts or see functionality, the better
The game is so far in its life cycle, there's no reason to hide scripts for most reasons, but it still protects us from skids. I understand security by obscurity isnt ideal, but its one of two solutions we have at this point
I post my scripts on discord so others can use it, it only brings people not pushing away from servers.
Personally I keep most of my stuff in server-side mods only where possible. For stuff with assets that can't be, well, obfuscation ain't solving that theft...
But then again, will server side mods be a thing in ARMA 4?
If not, then I'd be looking at good obfuscation lol
I figured for cheating it was actually sorta less relevant. To me seems like it would be easier to get those script during the mission if you have access to memory anyway. I suppose a good way to address low hanging fruit though.
That's totally fair. Only so much you can actually obfuscate on the client and ultimately something they're downloading on their PC anyway if it's not server sided. We just generally tend to run anything not imperative for security on the client for performance.
Always good to run infiSTAR imo...
Started game earlier and this happened - swapped back to main branch and it didn't happen, then I switched back to perf and it didn't happen. So I'm not sure what happened.
Quite an interesting bug 
looks like Dark Souls
bug or creepy trees, who's to say
I think they are creepy trees lol
Had an issue with accessing the UAV terminals on stable, took me 4 restarts of the game and a hop into singleplayer with the same mission to fix it
This game is weird as hell sometimes.
these trees are nothing but witchcraft which the devs should have seen coming
We've called those Treeple since OFP.
Is there even a word for this witchcraft?!?
Marian Quandt vibes
My Fav Birthday ever
"Be not afraid." ahhhh loading screen
So real
Man there are many Anniversaries in April
Ye lol
I wanna see flying dogs on the 14th ๐
looks at table
There is nothing on 14th
But...๐ข
I am in favor of this proposal
you sure you that ain't The day of the triffids mod ? ๐คฃ
I don't know why but because of this screen My Game didn't want to run with some of my Mods ๐
Dedmen, add optional personal birthday form, so anyone can add their own locally.
I wonder if. Windows 11 enforces connecting microsoft account. Microsoft account has your birthday..
Does windows know your birthday? Does windows have API to check if its the users birthday?
I hope not, surely that would be a serious data protection/security issue
Some random app queries that API and gets your name from your account name and hey, we're most of the way to identity fraud
Who signs into Microsoft in first place
Lots of people, it's the default now and not doing it requires an obscure workaround. I have feelings about whether it should be that way, but it is, and many people are just going to use the default.
A lot of people use their name as their account name even if they're not using an MS account, too.
I dont think so
Would be funny
IDK Steam has a birthday thing no?
Aren't all OSes required by law in California or where to ask for BDay and provide API for something? At least checks that you're >18?
Microsoft is now forcing everyone
On 25H2 ?
Whatever windows 11 is doing
For new installs I guess then.
It's jover MS bros
AI tells me windows has it, but no API to query it
Unless you have yourself in the OS calendar/contacts
Well there is https://learn.microsoft.com/en-us/uwp/api/windows.system.user.checkuserageconsentgroupasync?view=winrt-28000 . But I don't consider something an API unless it is win32.
But that just pushes the boundary. Now we need a separate API to figure out which laws the computer and user are bound by....
I think you can still install W11 without Microsoft account with Rufus (https://rufus.ie), but I recall reading from somewhere that Microsoft is planning to make it imposssible to use W11 without Microsoft account
Latest news actually suggests Microsoft is moving away from forcing an account since they got a lot of resistance recently. Time will tell.
with age verification laws coming you might be able to tell soon
grafana/pyroscope
your poll is invalid, you missed one valid option ๐คฃ
where's jets?
above the clouds
in the poll i meant
you'll get used to it
Which DLC's Anniversary is tomorrow?
8
25
4
Helicopters
waits for perf 21 ๐
Waits for @nocturne obsidian to test it first ๐ฉ
thanks dave
btw. you missed Zeus DLC in the poll ๐ same day
Zeus was yesterday
โค
for some reason STEAM considers it today, could be the late night change +1 summer time diff for the availability ๐
Shows yesterday for me...
timezones moment
No comments about the splash screen today
I disable the splash screen ๐คท
BTW, do the anniversary splash screens include CDLC releases?
youuuuuu
not yet
Shame, couple of bangers coming up soon ๐
Dont know why we left them out
like i wrote, i suspect the problem might be the time it got edited/enabled (was 22-23 hour of day) vs summer time and it will show for some at 11th
If only BI had the rights to play Fortunate Son next month and 99 Luftballons this ๐
Yeah, that would make sense that I see Apr 10 in the USA. Weird time to deploy a DLC (even if free) in the Czech Republic ๐
What's the point of a free DLC that's always loaded?
Pings ๐
But all BI DLCs are always loaded anyway
The not platform Contact isn't.
Yeah, apart from that
I wonder if it would be possible to add some parts of the Contact to the main game without the need to load platform. Like alien props and modules
probably because it's not part of the core or first party?
you mean the contact specifics that aren't in platform?
how?
yeah. Since loading the contact as cdlc basically makes most of the game..."unplayable" in a way
contact is split in a base dlc and an optional one, doesn't make it a cdlc since it's first party
yeah, anyway, my point is it would nice to have some stuff from the optional part
as I said, modules and alien stuff
I want to remind you that the alternative to Contact in DayZ became free contentโฆ and yet, after 15 years, we still havenโt been able to enjoy a fully realized Chernogorsk.
you don't like free stuff ?
it's also oriented as SP campaign ...
I meant why as a DLC ๐
Maybe to showcase the new addition better?
doesn't really matter why since its free and always installed
yeah. But I can imagine a lot stuff from it be used outside the SP.
many are limited to the terrain bounds though
Malden was also free...
Maybe are DLC simply because they have their own addon folders?
some, because not all was made to work in MP
Strange issue popping up only on profiling since a modpack update today
20:43:15 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Companies'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Directions'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Traffic'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Warnings'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Companies'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Directions'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Traffic'
20:43:22 Warning Message: Addon 'A3_Data_F_Enoch_Loadorder' requires addon 'A3_Structures_F_Enoch_Signs_Warnings'
I'm searching through all of the files and I can only see one mod which touches Loadorder which is ACRE and that's only as a requiredAddon, plus it wasn't updated.
Even with debug on I can't get any more info out of this, any ideas?
Is that a fatal or non-fatal error?
(IIRC this one is also a "warning message" when it's fatal)
I think non-fatal, but with a popup
Partially figured this out, it's the albasrah map which seems to be stealing the structures_f_enoch/signs prefix. Apart from the fact that this seems like horrible practice, is it possible it's a load order issue?
*a lot
there you go
some shadow dropped update on profiling?
usually it's CDLC ๐
No also offtopic.
Eden Editor now displays Module Description for legacy (Pre-3DEN) modules
Just when I added a custom module description ๐ข
What I did there is very hacky and I'm not sure if it'll stay. Maybe will only keep it for vanilla modules.
No-one should rely on that working and should rather use proper attribute descriptions
I guess the issue are the vanilla modules anyway, not the custom ones
Something looks off 
What is something
In thermal?
On an empty build, it crashes when firing on any map with the new update
Arma 3?
Is that an affirm?
Me and another friend have this problem now
Firing what? A gun? Which gun?
Did you submit the crash report?
1 minute
so I just started playing with the new version... there have already been two departures on the client. I'll try to play it again.
I can fire guns just fine
Ah...it's my fault, ace didn't update and crashed the game)))) sorry
What
That's not supposed to be possible. A unupdated mod shouldn't be able to crash the game.
Did you submit the crash report?
There's a bug in currentZeroing command, fix ETA ~40 minutes (or 50..)
the third crash of the client, dumped the dump in the personal account
That's right, the game crashes when I look through the scope and try to shoot....
Unless you sit in a car
Apologies if unhelpful;
Getting exact same issue as others.
Confirmed to be ACE causing it as just CBA on its own doesn't proc the error.
Crash report submitted.
Mentions "Status Access Violation; Illegal Read" then some strings.
Maybe something in the new zeroing stuff is causing ACE to cry?
Would the missing error messages for lnb* fail silently or actually cause different behaviour before?
Sadly the Special1 render flag in the Refract shader isn't working for me and nor are the IR laser parameters ;_;
Silent fail
hmm, one of ours is now throwing an error and making a GUI look weird
lnbSetColumnsPos [
108,
[
((ctrlPosition _QS_ctrl_9) # 0),
((ctrlPosition _QS_ctrl_9) # 0) + (((ctrlPosition _QS_ctrl_9) # 2) * 2),
((ctrlPosition _QS_ctrl_9) # 0) + (((ctrlPosition _QS_ctrl_9) # 2) * 2.8)
]
];
It looks reasonably correct to me but it's throwing Error Type Array, expected Number
Some cases had no checks.
Some cases had type checks AFTER using the value. Those would print conversion errors into RPT, and silently have wrong behavior (0 when trying to read a number thats not a number)
I see jup, mistake
another new prof? ๐
yeah but not as critical
such is the life of the prof branch, ty
Is it critical enough to hotfix today. Or to push a new prof on thursday?
I should give dedmen a giant prof button on his desk to do it himself
UI looking weird doesn't sound too critical
I could push another hotfix now (We're already pushing one right now), but I'd rather wait and see if more problems pop up
I can start opening all of the GUIs we have and see what happens, how long do you think before release?
Thursday
I meant the now one but yeah, not critical
The now one 5 minutes ago
I gotta test it but could it be possible that beamColor affects the dot and dotColor affects the beam
incoming changelog for the hotfix?
but yeah the refract is not working with that hotfix either
No. Same as last one
You sure you're one the correct version?
"Profilling - 2.20.153673"
according to the pause screen
I don't know, the code worked when I tested it, and its correctly merged onto profiling branch.
I don't have time to test it again
is it possible there's a typo in the render flag's naming or something?
Special1
looks right to me
I also tested it from the rvmat, the one I copied the snippet from that I sent to you
interesting 
I'll revalidate the files again I suppose, otherwise Im not sure what else might be different
validated, tried running the other .exe's in the root folder, updated drivers, downloaded the files from the google drive, no luck
strange
yeah im officially out of ideas for now lol
There seems to be an issue with disabled channels. I have channels 0 and 2 disabled (via server config), but yet users with profiling branch are able to use these channels.
In the interim you can enableChannel false for those channels via an initplayerlocal in the mission file. Friend confirmed that worked for them
Running those on the client was what seemed to fix it
Thank you!
I refactored it a bit but thought it'd be the same.
Can you send me the line you actually use? Old syntax or new syntax?
The most common crash currently is still the XAudio crash that I can't figure out.
DayZ has the same crash on their end.
I just got notified that DayZ closed their ticket for it, after they found the cause and fixed the crash in a separate ticket that I didn't know about in 2022.
Oh well.. Lets see if this works...
Nvm.
DayZ closed the ticket with "The issue seems to be gone", after several fixes were made, but they were made to code that was only added to DayZ, that A3 never got in the first place.
But, they have a second ticket, for the exact same crash, which is still open, and even after the last fix attempt still accumulated new crash reports and is, till today, unfixed.
The crash tickets are separate because one is crashing in xaudio2_7 and the other is in xaudio2_9.
After upgrading from 2.7 to 2.9, the crashes in 2.7 disappeared! Amazing! But at the same time a new crash appeared in 2.9, in a function with exactly the name, with the same callstack. Oh no, a new issue!
Alas, it shan't be.
could you send me your rvmat you were testing with when you have a spare moment? maybe im missing something in mine
I only have it tomorrow
I came up with some logging that should tell me if the issue is actually on XAudio side like I thought, or if its on our side deleting the buffer sooner than we should.
That depends on people not using -noLogs though 
true is disabled, false is enabled.
In the order Chat, VoN, MapMarkers, MapDrawing.
Which.. looks all correct 
are able to use these channels
What does that mean. They can chat, or use VoN, or both?
Are both disabled, or only one?
disableChannels[] = {0, 2};
They can at least use text channel, VoN no idea. But I have VoN in this channel disabled via infiSTAR so probably not.
Then I wonder if
disableChannels[] = {{0,true,true},{2,true,true}}; Disable von and chat on both
works the same. But as far as I can see it should.
got someone else to also test it (@lethal trench) and bro's got the same results 
This is 2.20.153673
discord shows previews for mp4, you can remux that in obs easily btw
You can just rename the file to .mp4 and it'll work
ah
Found it.
Bug was that the disabledChannels would not be reset when switching servers/missions.
Server always sends the new set of disabledChannels when selecting a mission or joining the server that already has a mission running, which will overwrite them.
But if you leave a server, they stick around. And if you join a server that has not yet selected a mission, your chat might be blocked and you cannot login as admin.
So easy fix, clean it up at mission end, where many things are being cleaned up.
But, the server sending it to you when you join or select a mission, happens BEFORE the mission is loaded.
So it sends the disabledChannels, then the previous mission is cleaned up, and the new mission is loaded.. which deletes the disabledChannels.
We do want to clean up at mission end. But not if the mission that's ending, is a leftover from before the server was joined :/
A bit of a mess, but what isn't.
The commit that broke it was 153666, clearly devils work.
Make -noLogs just not work for few days, months, years, millennia...
in my tests with my new 9070 XT, I noticed that using the -noLogs launch command lowered my FPS by 15%, anyone else noticed this?
maybe only on my specific pc, dont know
retested and now i get a few more fps with noLogs
While testing I found a funky bug, I couldn't join my server, I always got stuck on "connecting".
I just learned how the "custom files" transfer works (We have no documentation of course).
When a player joins the server. The server tells every existing player thats already on the server "Hey, this new player has custom files, do you want them?"
Then it blocks the new player from connecting, until everyone else has replied if they want the files or not.
Then it blocks the new player from connecting, until the server has transferred its custom files to all the other players that want it.
Well in my case, my server has a client connected, that is actually crashed and will never respond, and is waiting for the timeout to kick it.
That one client isn't responding, so now no-one that has custom files can join the server until that one is disconnected.
And these custom files include the Squad logo.
So any player with A3Units or Squad XML logo, will be blocked.
I still don't fully understand this though, I cannot find where the files are sent to JIP players...
Ah nvm I found it.. oh man.. 
I want to redo all of this but there is no time
Okey....
If player joins, we ask every other player already on the server if they want the custom files, and if yes we transfer.
Once the new player is actually joined, he receives every other players files sent to them. We don't ask if they want them, they get them either way, even if they already have the files and don't need to retrieve them.
Man.. I gotta quickly close these files and stop looking at this..
I added a 10 second timeout for when clients don't respond, so that new connecting isn't being blocked..
Quick ducttape fix and walk along..
Huh! This explains why the magical solution of "Hey have you tried removing your unit? it always fixes it for me" worked xD
This bug btw.
Still is present.
But no-one else reported it since my fix. And my fix worked when I tested it.
But today new report of it still being there, and when I test my fix, it indeed doesn't work either.
Is the new CfgExtensions replace the Battleye whitelisting process ? Can we use dll's on clientside with this new verification ?
see the ticket and related recent discussion in #dev_rc_branch
its validation not whitelisting replacement
Gotcha. Ty. It's a little sad that it's not a replacement. The Battleye whitelisting process still takes a long time
you know, you don't have to use battleye ยฏ_(ใ)_/ยฏ
could we get something like
class CfgExtensions {
onlyAllowWhitelisted = 1; // Only allow extensions which are whitelisted in one of the config classes below
or I guess it could be a server setting, ie, clients know they have only loaded allowed extensions, and server requires that client flag to be set
I was thinking letting you set a load eventhandler, that runs on all extensions.
Then the script can "validate" others too
I don't know how that would be viable. A mod cannot decide that all other mods you have are updated and use the new system.
Servers can use allExtensions command to see whats loaded. And server can read CfgExtensions, and see if the loaded ones are in there.
Hey, does anyone know if there's any reason that this might be causing hitches? I keep getting hitches in frametime, and upon running diag_captureSlowFrame ["total", "20ms", 500, false, 0]; in the profiling build several times, the only thing I can find is steamMngUStatsSt eating exactly 33ms. Is this something that might need to be investigated on my Steam client's end? Also, I only just noticed this now, but I recently surpassed the 7,000 hour mark in Arma; perhaps this is some kind of integer overflow or other bug triggered by the hours? For my test environment it's basically just an ammo box behind the Altis main airport, no mods and no complex mission scripts running.
Should I post the .json file for Perfetto or would this be better discussed in #arma3_troubleshooting? I get the hitches on main branch, so not a profiling branch bug.
Its your steam client or steam API. But I don't know what to do about it
The profiling data itself will not give any more details.
Would you happen to know what API call is being made that I could ask Valve about or look into further? "steamMngUStatsSt" renders zero results on Google, and only the year old message from JonBons when I search this Discord.
ISteamUserStats::GetStat
a couple hundred times, for all our stats
and ISteamUserStats::GetAchievement a hundred times for all our achivements
There is a subscoope called "staEval". Is that small?
Alright, appreciate it. And I'm assuming these should return in microseconds or even nanoseconds?
yes
One sec, lemme check.
977 microseconds.
steamMngUStatsSt first calls GetStat/GetAchievement for all the things.
And then goes into staEval right after, running some scripts for achivement triggers
So all the steam stuff is between those
wonder if this (achievements / steam ui integration) could be disabled/fully skipped via some commandline startup switch to see if it solve some of performance / hitchup / stutter w/e
Probably
if only there's a nosteam parameter that disables workshop but still allows for multiplayer to work
currently if steam is not available you can only play singleplayer
mostly turns into a paperweight game
Aha. Interesting.
i'm confused. do you mean if steam is offline? how do you play online without steam in the first place
or do you mean if by magic steam is not installed but arma is?
how could that possibly happen. hmm
if you're on a steam client older than the steam api support included in the game it will function but without a steam connection
Maybe update your steam client then
But he's running windows 98
I'm glad arma 3 will die alongside steam and won't outlive it
wat
Take the bullshit talk to some other Discord.
what "wat" ? If steam ever goes offline, you won't be able to play the online part of the game.
Yes you will, lan still exists, and so do services like zerotier and hamachi - infact arma 3 is one of the only few games that you can do so with
Also if steam goes down we've got plenty of other issues, so this is just a nothingburger buddy
not necessarily true
direct connect exists ... you can do lan play over tunnel/vpn etc.
also it shows how young you are, Arma already went thru death of MS DirectPlay and GameSpy ๐คฃ and ironically you can still play ACWA, A1 and A2, A2OA in MP
18 years of Steam, I would not call that young, no offense ๐
well, he act like that ... maybe memory loss ... still younger acc than mine
i found if steam client connection is not possible that part gets messed up
cant join servers nor host
Been having a lot of crashing issues with our unit who are using the profiling branch (no crashing with the default public version when we tested). Crash is caused by reloading with ammo in the gun (no crash when loading from empty). Spearhead M1 Garand & Carbine were the two guns that we tested that consistently crashed.
Spearhead does some weird shit with fake magazines to change reload anims, IIRC. Might be related.
We are running mods so it might just be a specific mod having issues with profiling although we werenโt having any issues with it last week.
can confirm it on my end too
I'm gonna guess it's caused by us 'breaking' the reload prior to running the new reload animation, this caused crashes in the past but was fixed
ok tried with V63, no crash, V64, crash and v66 crash
this is wrong it seems to be currentZeroing breakign it again
im double checking but the game seems to crash when the script runs currentZeroing
yeah - no crash when I remove the zeroing check
that's odd
just player currentZeroing [primaryweapon player]; causes the game to crash in vanilla
Nice that you found the cause. Hopefully an easy fix 
I was in commander seat, every time my gunner changed the magazine (reloading after shot or via action menu) my zeroing was reset to "default". I got this in mp koth, haven't noticed this before.
When I use zeroing on the commander's weapon using the rangefinder, after my gunner reloads, my zeroing gets reset. This happens when crew is player, can't get it with the AI
But I thought I fixed currentZeroing in ...... Oh wait thats an alt syntax is it? 
Don't tell me I forgot to copy-paste the fix
I wanted to unify the two functions because they are almost the same code, but I didn't actually do it.. I should've
I was confused too and initially ignored zeroing being the issue till i narrowed it down to it
1.54.134596 profiling & performance v21 server and client, windows/linux
- more logging to track unfixed crashes
https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm
I have a specific technical suggestion that I believe could dramatically improve FPS in dense urban areas, with no downside for modern hardware.
The Problem:
Currently, Arma 3 does not automatically aggregate non-instanced static objects . In a city, every lamp post, trash can, fence segment, and chair is sent to the GPU as a separate draw call. As explained in the BI Wiki, each "section" (material/texture combination) requires its own draw call, and this is multiplied by lighting and shadow passes .
When a player enters a large city (e.g., Kavala or Pyrgos), the CPU becomes bottlenecked issuing 10,000โ20,000 draw calls per frame, not by rendering polygons. Modern GPUs are starved waiting for instructions.
The Solution (Enfusion already has this):
Implement automatic mesh aggregation / batching for static objects at load time or in the background.
Instead of 200 separate draw chair commands, the engine should detect that 200 identical chairs near each other use the same material/section.
It should then merge them into a single, larger mesh (or a single instanced draw call) automatically.
The BI Wiki explicitly states: "Unfortunately Arma doesn't aggregate non-instanced objects automatically like some other engines do" . Enfusion (Arma Reforger/Arma 4) likely already does this.
Why this would help Arma 3 specifically:
CPU bottleneck fix: This would cut CPU render thread overhead in cities by 50-70%.
No content rework: Modders and BI don't need to manually merge objects into single P3Ds. The engine does it.
AI benefits: Freed CPU cycles could be used for AI simulation, which the 2.20 update started optimizing .
Implementation idea:
Use the existing "section" system as a grouping key . At map load, iterate through all static objects within a spatial grid (e.g., 50m cells). For each unique section (material set), merge all meshes in that cell into a single vertex buffer. Update culling to work on the merged bounds. This is standard practice in Unity, Unreal, and even the Frostbite engine.
I understand Arma 3 is a legacy codebase, but the Enfusion job system was successfully backported . Could the batching system be backported as well?
Thank you for your continued support of this game.
TL;DR: Please add automatic mesh batching for static objects in cities to reduce draw calls from 15k to 3k. This would double FPS in urban combat.
I don't think mesh merging within cells would work because a) The objects aren't static enough (damage effects) and b) p3ds are demand-loaded.
Other types of object instancing might work. The cities certainly do have heavy re-use.
Instead of 200 separate draw chair commands, the engine should detect that 200 identical chairs near each other use the same material/section.
It does, if they can be instanced.
This would cut CPU render thread overhead in cities by 50-70%.
But.. we don't really care because the render thread just needs to finish before the simulation thread. And it does.
So it would not improve FPS noticeably
Freed CPU cycles could be used for AI simulation
No.
Could the batching system be backported as well?
I don't think you know how this stuff works. Enfusion uses a scene graph
This would double FPS in urban combat.
lol.
You can't even double FPS by turning all 3D rendering off. Are you a magician?
1 is what decides your framerate (hint, its not the render thread)
2. Is the render thread.
3. Is part where we submit to render thread, some time could be saved there with better instancing
Main thread in this scene is waiting for render thread for 2ms, the 3. is 1.4ms.
At most we could save like 2.5ms
That gets me from 53.4fps up to.. 61.7fps!
A whole +8 fps! Almost double!
If instancing is not active on streetlamps, its a configuration problem.
Maybe they have animations when they shouldn't.
aren't trees and such instanced?
All static objects, with non-changed textures/materials and no animations, should be
yeah. I remember that for example walls with auto-slope to terrain can't be instanced and thus they increase performance drain
basically every sloped object is its own unique object.
Surely there's some diag_enable command that tells you about object rendering and instancing.
yeah, but i'm too lazy to check. all i really remember was that it works with trees :p
frequent = 1 property in geo lod
Got it, thanks for the clarification. What are your plans for optimizing the simulation thread in cities? Or is this already the limit for Arma 3
could be related that you can "shoot out" streetlamps?
The light emitter itself is separate
but when the texture changes yes. But they should be able to instance while their texture is normal. Unless the destroyed texture change is actually an animation
No plans
๐ค ok was just a thought
I haven't looked into any "why is this not instanced"
Because the impact is quite low compared to all the other stuff
yeah all good ๐
the performance improvements we got the last few years is more then i could have imagined ๐
Acting like you tried them
bet you haven't played A3 in years at this point
Regarding the rendering topic, I once tested something in the editor: I placed thousands of a custom object within a 100 meter radius. The static object had an empty LOD (0 polygons) with the frequent flag enabled.
Even though the objects had no geometry, my FPS dropped downtown.
When I used hideObject FPS immediately returned to normal, which makes sense since the objects are removed from the rendering queues.
What I found interesting was how the Object detail setting affected performance:
- On Low, I regained a lot of FPS
- On Standard, performance dropped again
This made me think that the setting likely controls the distance at which objects are considered for rendering or at least submitted to the render pipeline / queues
Also I assume the frequent property in the P3D might only apply to baked objects.
the input Im giving here is probably not useful but it was a interesting experience
was the game spamming the rpt or something?
So I am trying my luck with conditionScript for context menu item but I can't get them to work
class ENH_ResetDefaultLayer
{
action = QUOTE([false] call ENH_fnc_toggleDefaultLayer);
text = "Reset Default Layer";
conditionShow = "hoverLayer * script";
opensNewWindow = 0;
wikiDescription = "Resets default layer.";
conditionScript = QUOTE(call ENH_fnc_getDefaultLayer != -1);
// conditionScript = "false";
// SHOW_IN_ROOT;
};
Just to make sure, this is the correct configuration?
script1?
is there script2? ๐
I guess this is somehow related to custom connections?
theCarScript.sqs
Dedmen. Can you on monday check what's the difference between the simple expression that start with hover* vs. selected* afaik one cannot open the context menu without selecting an entity first.
Also there are a few expressions that I don't know what they do.
selectedObjectAttached
selectedObjectAgent
hoverObjectAttached
hoverObjectAgent
hoverFolder (There is hoverLayer, so dunno what's the difference)
I think folder is a legacy name
Afaik it is what it should be, under cursor vs selected
Is possible that right click auto selects and its not possible for them to be different
Ok. That's what I thought.
no , not at all
Reloading P08 Luger from Spearhead causes hard crash with profiling branch, issue doesn't exist with stable branch.
How to reproduce - Launch arma with sprearhead loaded, place AI with luger, reload - 100% repeatability
I have used send report thingy but I can also provide full dump
#perf_prof_branch message yeah, they are aware of it
thx, which unfixed crashes?
Simple Expressions don't support any lazy eval right?
hoverObjectVehicle * {script1} would be neat
And if we think about it this way, what is really the main killer of performance?
Bad sqf ๐
(Only half joking)
Id say it depends on context entirely ๐คท
Arma 3 already runs better (when you consider how much it simulates and what it does) than most modern games do, which may not be high bar but hey
old bandaided unoptimized engine
i think that would be something to ask dedmen however i belive Kharos is right in saying that it depends on context
and theres nothing to gain beating the old engine
I also feel like it's a pretty broad question. Are we talking about client, network, server or just Arma as a whole? Lots of different ways to approach that ๐
https://community.bistudio.com/wiki/Code_Optimisation
https://community.bistudio.com/wiki/Mission_Optimisation
https://community.bistudio.com/wiki/Arma_3:_Simple_Objects
https://community.bistudio.com/wiki/Arma_3:_Performance_Optimisation
Hi, mimalloc for Arma 3 v2.3.1 and v3.3.1 has just released.
Any test result and feedback is welcomed!
https://github.com/GoldJohnKing/mimalloc/releases
Any notable differences between 2.3 and 3.3?
the one which needs -crashDiag ๐ (animation/file/data alike corruption)
It's said by microsoft that v3 consumes less memory than v2. My test shows v2.2.3 performs a little bit better than v3.2.3. I have not carefully compared testing v3.3.1 and v2.3.1 yet.
1.56 TMW
with all these crashes ?
do you not care about your servers at all ?
cos it sure does not feel lik eit
@hallow yew dave you tried this yet ?
No bud, not really been at my pc much today so didn't want to test
I had 6 crashes on v20, if that hadn't happened I may of dipped my toe into v21
But if 1.56 is out tomorrow, it doesn't matter right? ยฏ_(ใ)_/ยฏ
yea
they really care about us
crash crash crash oo new patch
tbh sometimes i wish i never started a arma server
i really do
so glad were dedicating all our dev team to UE4 Game
we had 2 epic devs answer our ticket after 2 hrs
with code samples the works
Running an Arma server is a game itself
Will it break?
Is there an alternative command?
Is faster better?
Is slower better?
IS @empty goblet eating all the chocolate? ๐
even offered to run our build on there test system to show us where we can make improvements
Yes
Probably but won't
yeah....

The alt syntax copy paste also has a bug. When using FCS zeroing, it returns a number even though it should return an array.
That's not documented on wiki.
When copy pasting its a simple oversight.
When I change it to do it properly, I can't even miss that, because its a compile error.
The copy paste has another bug
The same validation is done 3 times, and the compiler even flags a warning for it.
In the non-array syntax this was fixed, but not in the copy..
And another.
There was a crashing issue with FCS zeroing.
The main syntax got a check for it to fix it by at least returning -1 instead of crashing.
The alt-syntax still has the bug
80 lines of code gone, 4 bugs fixed..
Less code = less bugs?
I have an idea ๐
less code = fewer bugs*
bugs are a countable thing
(Finally get a chance to use something my Schools mostly useless english lessons taught me)
YOU WRITE "alot" !!!
amma drive 635 km and slap you in the face with a moist, flaccid baguette
Is that how your gf describes your best part?
what is a gf?
Right or left hand, depends
If you down launch any of the ARMA executables, You'll run into very little bugs ๐
these are my side chicks
we just turn arma 3 into one of those text based rpgs
You are in a field.
>look around
You see a man - 100m - front
You see a flash - you are dead.
but the trick is, that man - 100m - front, is actually behind you
lol
weird, was that discord?
Am I the only who can't login with direct connect from multiplayer menu ? I have to direct connect from launcher since last update

Where does it break?
Server does not show in list? Password wrong?
direct connection, i can't login and no server showed despite being online
but works from game launcher
So the server does not show
yeah only since last update
Fucking hell man
huh
The linux build failed because GCC like to be a problem.
So I made a quick fix to just fix the build and didn't test it.
And changed it from "Try to ping the server and list it" to "Don't try to ping the server"
Seems fair (i didn't understand a shit)
Fix in an hour or so (โฏยฐโกยฐ)โฏ๏ธต โปโโป
we can i call the Support?
???
"we can i call the Support"
what do you mean, do you have an issue?
#arma3_troubleshooting / #reforger_troubleshooting
i need Gamesupport
I believe you
I fucked it up again. Another ~40 minutes
classic GCC
Inventory stuff on ground doesn't move around when you pick up other things, right?
dunno about that but that is not it
๐
Published now
Back to normal. Working fine, thanks!
You can pick up all weapons now without directly looking at the launcher to pick it up
thank you
Now that you mentioned it I might have fixed that too
wow hdr video
What was fixed there?
Meow meow, not sure if this is the right place to post this.
I noticed a small problem with the mini-map being placed half-off the screen. Only visible when joining the King of the Hill gamemode, was normal in 3den each time with squad leader as test.
So it could have nothing to do with profiling's code, but it was only apparently when switching to profiling which is why I mentioned it here.
No mods were loaded.
Tested in Warlords official, and it is normal. So it's likely a KOTH specific bit of code
not only in koth, it happens sometimes to me
ye obviously we don't care, why you even here still complaining then ?
laucher is part of the dead body inventory by default
Is it not currently?
nope, not always
Before, you either get rifle or launcher not both, which one depends on where you point when opening inventory. The weapon that you get also twitches when you open inventory. You should be able to test the improvements in the next prof probably.
Nice
The minimap can be moved in the game settings. Maybe you did that?
i think someone also reported this days ago, it's getting unintentionally misplaced by the game
mimalloc 2.3.1 / 3.3.1
Excuse me for my audacity, this is already more than I ever dreamed of..
Thank you very much for the improvements you have already made, they are greatly appreciated, and while you were looking into related mechanics I wanted to mention another situation that might be worth considering, because when multiple bodies lie very close together or overlap it can be difficult to access the correct inventory, so perhaps it would be useful to provide an alternative way to access nearby inventories, for example an additional slot in the gear interface for the closest body or even a merged virtual container that allows interaction with all items on the ground at once, I experimented with this idea some time ago using a script but set it aside, and I thought it might be interesting to share in case you decide to expand this functionality further.
player addEventHandler ["InventoryOpened", {
params ["_unit", "_container"];
player removeAllEventHandlers "InventoryOpened";
ffabox = "CUP_BOX_RU_WpsSpecial_F" createVehiclelocal [0,0,0];
ffabox hideobject true;
clearWeaponCargo ffabox;
clearMagazineCargo ffabox;
clearItemCargo ffabox;
clearBackpackCargo ffabox;
ffabox allowDamage false;
ffabox setPosASL (player modelToWorldVisualWorld [0,1.5,0.5]);
player action ["Gear", ffabox];
systemchat str ((player nearSupplies 5 select {_x != ffabox && _x != player}) apply {typeof _x});
true; // <-- inventory override
}];
player addEventHandler ["InventoryClosed", {
params ["_unit", "_container"];
deleteVehicle ffabox;
ffabox = nil;
player addEventHandler ["InventoryOpened", {
params ["_unit", "_container"];
player removeAllEventHandlers "InventoryOpened";
ffabox = "CUP_BOX_RU_WpsSpecial_F" createVehiclelocal [0,0,0];
ffabox hideobject true;
clearWeaponCargo ffabox;
clearMagazineCargo ffabox;
clearItemCargo ffabox;
clearBackpackCargo ffabox;
ffabox allowDamage false;
ffabox setPosASL (player modelToWorldVisualWorld [0,1.5,0.5]);
player action ["Gear", ffabox];
systemchat str ((player nearSupplies 5 select {_x != ffabox && _x != player}) apply {typeof _x});
true; // <-- inventory override
}];
}];
other people have this and everyone of these was running JSRS, are you running it?
no, it happens running only cba
if i was complaining ill complain
im here to share you crash logs as you asked
having a rant with my mate dave about server crashes is more than just
im sure its not all roses over there
but you can understand the pure fear of you updating arma when were still left with crashes on perf21 ]
Nah no mods running
I did it, but it is offset off screen only sometines
i said many builds ago not all crashes will be fixed in 1.56 and that will be taken care of in 1.58 per/prof builds
this would be new to me. in JSRS we dont touch any that could cause that.
Old JSRS or JSRS 2025?
By the way, I was very interested in whether an unfinished vehicle from some mod could affect performance?
for example i need some crashdumps and logs on the v21 with negative autoarray (like avka has on BIForum, for now last from him was v20)
Ok. The only 2 people that reported it before had JSRS running.
Oh I meant to check if the lnb stuff was fixed, I didn't see a changelog for it
ssilent fixed
Unfortunately this is not currently possible as the inventory is designed to deal with one corpse at a time. Dead unit drops weapons in separate weaponholders, if you point to a weapon it finds the dead body that dropped it and opens their inventory, if you point at dead body it opens their inventory. This is as good as it gets
The GPS thing is interesting.- I did a change to RscCustomInfoAirborneMiniMap (The minimap for aircraft) 3 weeks ago.
But, that is a data change that prof doesn't get, and it only landed on dev yesterday and its aircraft only, not ground vehicles.
I don't remember any other change for minimaps 
First report was on 04.04. The previous prof was on 02.04. That prof was published two hours after I did that minimap change
whut
No relevant changes in v63, v62, v61, v60, v59..
@left eagle @echo marten Do you know when you first observed it?
Jellyfish asked if we "already know" so I assume its been going on for a while before that?
Holy shit I can reproduce the problem in editor.. The heck
While player is in vehicle.
The vehicles display manager renders the map "DefaultVehicleSystemsDisplayManagerLeft"
Which has a position configured in config, of -0.575909138, 0.569090903
If player is on foot, the players manager renders it.
"DefaultVehicleSystemsDisplayManagerLeft" at -0.696750283, 0.569021881
Its the same classname.. but it has a different position?
class DefaultVehicleSystemsDisplayManagerLeft
{
x="(profilenamespace getvariable [""IGUI_GRID_CUSTOMINFOLEFT_X"", (safezoneX + 0.5 * ( ((safezoneW / safezoneH) min 1.2) / 40))])";
y="(profilenamespace getvariable [""IGUI_GRID_CUSTOMINFOLEFT_Y"", (safezoneY + safezoneH - 21 * ( ( ((safezoneW / safezoneH) min 1.2) / 1.2) / 25))])";
};
Okey. My guess.
One of them is loaded VERY early at game start, before profileNamespace is ready. So it uses the default position.
The other is loaded at game start, when profileNamespace is ready.
The explains Jellyfish's video, when he gets into vehicle, the minimap moves to its default position, but keeps its changed width/height.
Also matches Nostromo's images. He manually moved the left display, to the bottom right. But when he gets into vehicle it pops back to the left default position
Also explains why most people don't notice, if they never change it from the default in the game settings, they never notice a change.
That broke quite a while ago.
When I optimized game startup speed, I moved vehicle preloading around, before profileNamespace was loaded.
And we didn't have any checks like "Reading profileNamespace before it's been loaded makes no sense" so I didn't see it ๐
systemChat str speed vehicle player;```
i received report that this returns values between -5 and +5 if sitting still in a vehicle
it was happening before this month for sure. I dont recall exaclty when it started, but it was some months ago.
maybe even almost 1 year.
that optimization was several months ago
I first observed it yesterday, downloaded the performance profiling via Steam
that's fine, it was first reported before that, as dedmen figured out later
Ah okay, all good :3
Around September/October. That video I sent you was from October but I had noticed a while before that, just never bothered to clip it.
ok
ill try to capture some logs tmw when im online
so what can we expect 156 to run like ?
anyone else replicated it?
any vehicle?
Black Wasp, stationary on the ground, I was running the script command from the pilot seat, On the Salt Flats in Altis
Hmm. Parked planes have been hopping around a bit lately. Not sure if new.
We spawn them in hangars and sometimes they nuke themselves when cycling simulation.
Can't replicate that case in SP or MP hosted though
I have found a similar issue with pylons, but I don't know exactly if it's a me problem or a general one, or if it's even a new problem (because I believe it worked weeks ago).
I use a vehicle (Mi-48) and change pylons 1 and 4 to the driver turret and also the magazines via setPylonMagazine (no remoteExec). Vehicle and both [0] and [-1] turrets are local to me (checked with turretLocal), but yet even after well over 15 minutes the server (via getAllPylonsInfo) has no clue that I changed the pylons 1 and 4 to the driver turret. What the server instantly knows are the changed magazines on both pylons.
Magazines are replicated like all normal magazines are. With pylon as visuals tacked onto them.
But I don't know how the turret assignment works ๐ค
Can't wait for next build
Hm, but are you possibly able to identify if there is an issue with pylon turret? Maybe as an additional info. Vehicles spawned in with different turret assigned weapons are equal on both client on server (with and without the vehicle being local to a player).
why? a build just dropped and you're already looking forward for the next build? what's missing here for you?
6๏ธโฃ 9๏ธโฃ
Another of those questions ๐
mimalloc for Arma 3 v2.3.2 & v3.3.2 is released. Please have a try.
https://github.com/GoldJohnKing/mimalloc/releases/tag/Arma-3-v2.3.2-20260430
https://github.com/GoldJohnKing/mimalloc/releases/tag/Arma-3-v3.3.2-20260430
Changelog
Merge upstream changes v2.3.2, see https://github.com/microsoft/mimalloc#releases for details.
Description
This is a port of Microsoft's mimalloc memory allocator for Arma 3, which ...
Changelog
Merge upstream changes v3.3.2, see https://github.com/microsoft/mimalloc#releases for details.
Description
This is a port of Microsoft's mimalloc memory allocator for Arma 3, which ...
100 to 114?
seems that fixed the small performance drop from previous build and in some cases increased the performance by same amount extra on top
yup, they seem to be more stable but i dont see any performance improvement compared to the previous ones, but im gpu bound anyways.
1.56.134636 profiling & performance v0 server and client, windows/linux
- the usual sync with stable
- additional crash fixes , yet the hunt continue (-crashDiag still has use)
- Fixed: Possible crash in Voice-Over-Net technology (http://feedback.arma...ew.php?id=27974)
https://www.dropbox.com/sh/582opsto4mmr8d8/3BSy9PdRGm
via setPylonMagazine
That command doesn't exist.
setPylonLoadout probably.
I can't see sync for turret. I don't think this ever worked.
So should also be broken in 2.20..
The new non-weapon pylons, do have turret sync. I think this might also be applied to weapon pods.
But you can only get that on 2.21 dev-branch. Did you try it there?
Why do you even reply to Rook's post about ammo fill assignment.
How is that even at all related to your turret issue. Doesn't seem similar at all
setPylonLoadout, yes. Was a bit too early that day.
Ah no, I tried my testing only on prof (client & server).
After reading Rooks text again, I probably was referring myself to getAllPylonsInfo might having an issue.
Whoevers headless client is constantly crashing in their "A3SExtension_x64.dll", please stop. Thank you
"Arma 3 Survival" something. HttpRequestException.
Can you blacklist their IP somehow?
I will notify the owner if he hasn't seen it yet
Hahahaha I realized it earlier and was like hopefully it dosent bother dedmen too much ๐
Has anyone experienced any issues when standing underneath / inside bushes and zooming in?
Strange square shadows are popping up and flickering for me occasionally
Unable to open RPG accessories if open body's inventory, works ok if open body inventory via aim right at RPG
And it seemed to me that you can't put a grenade launcher in your slot without ammunition...
bummer
Dedmen, for past several perf versions I get server crashes, I asked you before and you stated that this is difficult to solve, not sure if recent crash is the same or new issue. This is on latest perf: Crash GUID: 5acefa6e-1da4-46f2-879f-d717dd25a29f
It only happens on one of my servers and the main difference between this server and other is the Vcom mod. I may have to remove it to see if that's infact the cause of the crashes. Can you share your opinion regarding this crash if you happen to have time to look at it.
is there any way to replace the broken model paper-car placeholder via config
No
that would defeat the purpose of the model
purpose of the model is to prevent crashes as far as im aware
empty p3d still accomplishes this
the same as the paper car does
no it doesn't because the p3d being empty can cause crashes
could just add the bare minimum so that it's effectively invisible (or at least less visible than what we have now..)
it doesnt really matter if I cant even change the p3d via config so theres no point in thinking about it
just make a mod that has an 'empty' p3d for the path of missing models
Well
If you can override CfgCoreData the model is set in there, I'm not sure if that affects the empty model replacement
I'm also not sure if it will break certain commands that use it
already tried iirc
you're talking about where it defines like papercarblabla = papauto.p3d
Yes
messing with that will probably just break positioning script commands/engine stuff
as that's what papercar is used for
Probably, unless you can get an "empty" p3d with the same dimensions
it's most likely a hard coded path change, if can't open x path then use papauto.p3d
far better off putting effort into either
A) making it so the models actually exist/the path is correct
B) failing A, make a mod that puts a dummy in the path the game is looking for
can you even do b without duplicating the entirety of a3data_f
I don't mean like that
Yeah that seems to be the solution, it's just a lot more effort tracking them all down
I just mean if you have a mod that has a broken/missing model you replicate the mods path and put a dummy p3d
ah my bad
Yea both of those are way too much since id have to do it for every mod I run that has issues compared to just replacing the paper car
Appreciate it though
does -debug log the missing p3d in question (with config class) ?
it might even do it in non-debug but I forget what the message actually is
oh it's "Cannot open object"
it always displays an on screen error and records it in RPT, regardless of debug
at least for the first time it tries to load it
I know, that discussion is over, it cant be replaced anyway
hey i'm having issue opening sqm, game instantly crashes when trying to open sqm. Tried on regular version and it works fine
It might be useful to send it.
Hi,
Iโve noticed that something isnโt right with the EnemyDetected group event handler. I canโt say for sure if itโs always been this way or if it started after an update.
It took me a while to figure it out. Now to the problem:
When youโre in a civilian vehicle as an enemy (e.g., BLUFOR), the enemy groupโs event handler doesnโt trigger when an enemy is spotted. However, the enemy (the group) behaves as it shouldโit recognizes me as an enemy and reacts accordingly. Additionally, findNearestEnemy and targets work correctly; itโs just the event handler that doesnโt trigger. It works fine with military vehicles from non-civilian factions.
Im using the newest perf-branch.
Wasn't there a discussion recently about adding a command to overwrite turret stabilisation?
It's currently on dev branch https://feedback.bistudio.com/T162061
Ahhhh.. it's not on the wiki
And I kept reading the wrong changelogs.
I thought it was going crazy.
Sounds like you have multiple. But I only see one report from that server. And thats a crash I haven't seen before.
What kinda mods are you running.
Have not seen the paper car appear even once in the mods and terrains my unit is running.
Did you send the report?
I recently saw some code that special handles when taking over a vehicle of a different faction. Maybe related.
This code is a mess :/ I don't think there are recent changes, so probably been like that for a long time?
I didn't at first cause I didn't know it was profiling causing it
I just did it now
Yeah, it's in the stable version too. I've noticed from time to time in my AI behavior scripts that something isn't quite right, and now I've just noticed it in my new project.
I'll just submit a feedback ticket, right?
Is that the 3DEN fix script that Brett I think has on his steam workshop? That fixes the attribute load times
yeah it is. Just wish this would get fixed in data
maybe it could be shipped for 2.22
I assigned it to KK but he said it didn't make a difference
๐ข
The ticket is still open maybe I'll take it back and just push it
There are also some really slow attributes for units, sadly I don't see which ones
Recently rebooted my PC and havenโt been on Arma 3 for a year or so. Last I remember was that there came out โhyperthreadingโ on the profiling branch, that did actually make a difference in my case.
Is this now on stable? Is there an up to date performance guide that I should follow, since all my settings are gone?
you remember misinformation, -enableht and -threads arguments are not recommended, profiling branch had gotten some specific multithread related enhancements to certain parts but arma 3 has always been hyperthreading and multithreading compatible and capable.
profiling branch puts its successes into the next stable
most guides ill advise about setting unnecessary command lines like those
i dont have a handy link off the top of my head
@whole cloud I can imagine you have a lot of DMs but I sent you the crash report i was talking about when opening SQM
I saw. And yes you're on the pile
Most of it is on stable.
Don't set any settings.
Appreciate it, so I assume I should just put on the profiling branch and it should be good?
Then fiddle a bit with graphics
Alright!
certainly
Any launcher settings*
In-game graphics settings are still relevant ๐
I have a friend who nearly doubled his framerate by setting the -threads argument and the amount of CPU cores via the launcher, supposedly because he was on some old AMD CPU that Arma 3 thought was a dual core. Anyone know if that's been fixed? I'm not entirely sure what CPU it was, I think it was an AMD FX. I do believe the FX series has one floating-point core per two integer cores or some oddness like that that might have thrown the game engine off.
Other than edge cases like that I do agree that it's generally not a good idea to mess with the launcher settings unless you know what you're doing.
Speaking of launcher settings, I noticed while using the perf/profiling branch that there was an mimalloc DLL added to it. Is this a custom BI fork or is it the one from GoldJohnKing?
its gjk's compiled by bi afaik

I believe we're one version behind GJK now but yeah just his repo compiled ourselves and bundled with prof
We take what Windows tells us.
If windows wrongly thinks the CPU is a dual core, I'd assume he'd have bigger issues.
yeah, using windows 
and he went from 8 to 16fps!
a year ago i was (not sucessfully at all) running arma 3 on an athlon 64 X2 1.9Ghz
And even that runs 64 bit!
Yet still people complain when we talk about deprecating the 32 bit build ๐
I believe it's usually Russian Windows 7 instances? (I meanโฆ I hope for them it's not XP ๐น)
underappreciated cult classic OS versions
last year i was running windows 7 and honestly i miss it
I still have it, it still gets updates
you mean windows 10? or premium assurance program?
uhhhhhh 
Almost as unstable as A3 Units
"TCAGame.vars.Arma3Profile"
The chances of me getting that obviously corrupted file so I can reproduce what the problem is, is probably close to zero.
add the file to the automatic report and next time you'll have it :p
That is a good idea, but so complicated
I think I got it though, their config file root, has a base class with name "\x4". Thats impossible, so clearly corrupted. I'll just make the game kill itself
680 crashes, 180 of them are this one server
That server is/was also running a very old prof version. Does not bring me joy :sad:
yea I was not working on it for some weeks, forgive me master 
I got another crash because entity's model was null. But why, papcar should be there :/
Sad.
Want to check logs to see what thing was broken there.
-noLogs
More sad.
make noLogs do nothing on prof and dev :D
A lot of hosting services do noLogs by default

The egg for arma 3 for panels ik does
Pelican and Pterodactyl eggs have it set by default, no idea who decided not having logs was ever a good idea
Yes
Which means a lot of people run noLogs
It was one of those snake oil performance tricks I think
Went unchallenged for so long its written down everywhere
it was useful for a bit iirc, then it got asynchronised and the trick became useless (but remained applied)
Probably some providers know their basic clients are going to run horrible Life configurations with spaghetti mods that generate huge logs 24/7, and turning off logs is easier than having to clean up several gigabytes of log files every week
literally guy responsible for "tech" side of arma group i play in last week
"I suggest to disable logging -noLogs parameter as logging, especially the extended logging, eats quite a bit of resources." ๐
Yeah it's in our onboarding too. I need to fight them on it
If it would be said by dev(dedmen/anyone else) then it would help fight it, like a lot
Eh I think most people who are listening to dedmen are in here and already aware
People still aren't reading the 2.20 guide which says to stop using cpuCount so
I using cpuCount
I went looking to see if there was already a dedmen message on the subject and, well. mmm. maybe a fresh one would be better :U #server_admins message
why?
almost nothing changed since then :D, jokes aside, info for client/server about those things would help fight "snake oil" solutions that still appear (2.20 notes helped a lot)
I don't remember the reason anymore... most likely, after hundreds of benchmark runs, I left this configuration
https://dev.arma3.com/post/oprep-performance-optimizations-in-220
You do not need to use the cpuCount parameter unless you specifically want to limit the number of cores in use. The game automatically uses as many cores as it can, so cpuCount doesn't make it use more.
I'm not aware of any problems with cpuCount.
PR time!
I don't see any information that this will make the game worse... I remembered the reason! I had long problems with windows 11 and getting the P-cores to work correctly.
Yea, noticed some other things so I'll prob make a PR with some stuff at some point
less i/o from that
apparently console logging to disk can hurt performance especially on slower systems
and it's a lot of storage in the very long run
would not be a bad idea at all
you're running the non production/stable binaries so you're obligated to provide/generate logs
less i/o from that
i/o from logging is a joke, unless you run on a 20 year old harddrive with 100 badly made mods that spam RPT warnings hundreds of times per second
and it's a lot of storage in the very long run
I'm not saying you should keep them all
apparently console logging to disk can hurt performance especially on slower systems
I struggle to believe without that insane log spam i said above that same system will run a server anyway if it cant even handle a log file
unless
Not even then, the disk write is async.
Unless he means slowing down reads because disk is busy writing the log, but Modern Operating Systems are pretty good at prioritizing stuff.
I have server logs going back to 2015 (but they are archived off the server...)
maybe someone made a mod to remove the car ๐
the car script
Steal the Car? ๐ฎ
i have had a measurable but small performance impact on a low end system with source engine console logging spam
but i guess its unlikely to happen here
carry on
Source engine sounds like a different game, really 
i know and understand that
just not too far fetched of a thought based on experience
but fair enough, it was done with much newer code design practices and optimizations but keep in mind the engine here is largely the same foundation from the ofp days
some disk scenarios can potentially be like that, some disk/raid controllers quirks too and in cases of using a single drive for the os and game install
the lot of storage part matters in very tight situations, culling/trimming rpts is manual and most are lazy
any system can be a server if you think hard enough
also probably insignificant too on networked environments albeit rare theres a remote minimal possibility that network performance can be affected, but i understand its insignificant and the experience citation i gave affected more the processor rather than the storage itself
culling/trimming rpts is manual
Its not.
Please stop talking crap. Go elsewhere please. Thanks.
apologies, how many rpts are needed to see the game removing them?
I remember you can clamp it to one by adding a lot of tiny files into the folder :P
Default limit is 10 IIRC
I feel like if hardware limitations are bad enough logging causes issues- the solution would be to upgrade hardware 
There are some worst-cases where the RPT gets super-spammy and disk space becomes an issue. We had one where an object would be continually corrected due to being below ground.
Still not sure that's a performance problem though. The physics stuff is probably more expensive than the logging there.
That's fair. We had an issue once where a dev wrote a waitUntil that errored on every frame. It took up storage, but performance wise no one noticed anything. I wanna say PRTs automatically get purged now but honestly can't say for sure.
Also had a bug where the server was spamming logs in unscheduled until it hit the while loop 10k limit, and no-one noticed until we tried to open the RPT :P
That one was jamming the server for multiple seconds at a time so you might reasonably conclude that our player base is stoned.
We have things to make sure spammy things are not spamming. It then just logs "repeated N times in last N seconds". Just need to know which are spammy
Yeah most of them do that. That one didn't, IIRC.
I'll see if I can dig up the message. We worked around the cause eventually.
If its the falling under ground one I know where to find it
Yeah that one. Can't find an example atm.
paints memory of tens of GB rpt / log files ๐
I'm not saying you should keep them all
the history is full of digital horrors
especially if you run public zeus server with script compositions
then you wish you had more than TB
where bad scripting is involved, you don't want to write logfiles ... let alone to see them ๐
and now when you realize lot of hosting barely used 120 GB SSDs (tho i remember even smaller)
I removed VCOM mod, seemed ok for few days and got another crash: 03:08:27:968 CRASH REPORT : Report Result: - SUCCESS - Crash GUID: 9280bfc0-ff63-478c-af90-c3ba7364e103
PhysX objects cleanup ๐คท
Seen that a couple times, but its one of those corrupted memory things with no way to see what caused it
I've finally managed to report this bug.
Since this perf branch with the number 2.20.153671 (i.e., 3-4 versions before the current one), there has been a bug with dynamic groups.
The bug occurs in multiplayer when there is at least one player on the server.
I think it has to do with โFixed: Many lnb* script commands were missing error messages for invalid argumentsโ
Here is an excerpt of the rtp error:
19:27:57 Error in expression <achIndex, 1], [1,1,1,1]];
_listbox lnbSetPictureColor [[_forEachIndex, 0], >
19:27:57 Error position: <lnbSetPictureColor [[_forEachIndex, 0], >
19:27:57 Error Typ Bool, erwartet Array
19:27:57 File \A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf..., line 655
19:27:57 โฅ Context:
19:27:57 [] L1 ()
[] L3 (x\cba\addons\xeh\fnc_initDisplay.inc.sqf)
[] L77 (a3\functions_f\gui\fn_initDisplay.sqf)
[] L128 (a3\functions_f\gui\fn_initDisplay.sqf)
[] L130 (a3\functions_f\gui\fn_initDisplay.sqf)
[] L12 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L115 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L115 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L12 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L172 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L172 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L12 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L371 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L375 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L375 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L12 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L606 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L668 (\A3\Ui_f\scripts\GUI\RscDisplayDynamicGroups.sqf)
[] L648 (\A3\Ui_f\scripts\GUI\RscDispla
19:27:57 <-
19:27:57 [] L1 ()
[] L8 (A3\functions_f_mp_mark\DynamicGroups\fn_dynamicGroups.sqf)
[] L897 (A3\functions_f_mp_mark\DynamicGroups\fn_dynamicGroups.sqf)
[] L899 (A3\functions_f_mp_mark\DynamicGroups\fn_dynamicGroups.sqf)
[] L901 (A3\functions_f_mp_mark\DynamicGroups\fn_dynamicGroups.sqf)
Probably need to move this post but testing with 32GB of RAM 2.3.2 results in a lot of popping up of objects ingame, while 3.3.2 feels smoother but after 2 hours in game, it would CTD. This is with the current profile build.
I will retest with 64GB later.
It subjectively feels like perf branch has gotten less stable recently
I've had a lot more freezing issues in the last 2 weeks compared to previously with varying modpacks
might just be up to something I broke on my end tho
I had microfreezes every 5-10mins few updates ago, but still not sure it was related to arma, or smth else
For me the game seems to freeze up completely for up to a minute more often but not regularly, sometimes I have to restart
are u on the latest perf?
Yep steam branch
hm. I suspected g-sync and/or nvidia shadowplay aka instant replay were the cause.
do you have enabled these things?
I didn't have any issues lately, everything was smooth sailing
since then, updates have been released for both Nvidia, Windows, and perf branch itself, so problem is gone
will rev 153745 be on profiling branch?
Where did you get that number
Ah
I did set up a mission with dynamic groups. With myself on a dedicated server. And I cannot reproduce this issue.
I can see it calling lnbSetPictureColor, with correct arguments, no errors.
What am I missing?
Ah needs two players on server.
Then _color = true, literally.
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-6, n-10, clsync, i-0, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-7, n-10, clsync, i-1, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-8, n-10, clsync, i-2, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-9, n-10, clsync, i-3, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-10, n-10, clsync, i-4, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-11, n-10, clsync, i-5, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-12, n-10, clsync, i-6, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-13, n-10, clsync, i-7, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-14, n-10, clsync, i-8, s-10;
16:53:34 ADDCUSTOMRADIO WARNING: channel not verified; DATA: c-15, n-10, clsync, i-9, s-10;
Is this a profiling message?
The script file has color selection code TWICE.
The first one has a default value, the second one is almost exactly the same code copy pasted, except its missing the default..
Yes @feral harness
A repro would be nice
Something about creating custom radio channels
hmm just some initial creation
(radioChannelCreate [[0.96, 0.34, 0.13, 0.8],'Command channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,1,1],'Aircraft channel','%UNIT_VEH_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[0.4,1,0.4,1],'Primary AO channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,1,0.4,1],'Secondary AO channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Alpha channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Bravo channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Charlie channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Delta channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Echo channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE]),
(radioChannelCreate [[1,0.4,0.4,1],'Foxtrot channel','%UNIT_GRP_NAME ' + '(%UNIT_NAME)',[],TRUE])
I'll see what I can do about repro
Thanks, we would need a repro, this is definitely an anomaly this is why it got logged
Is fixed, but its a script change, which won't land on profiling branch.
is the T198797 link intentionally twice?
Probably. Both related to server sided scripting roles.
ok
Nice
It brought 3 features, the EH's I can easily combine, but combining command and EH's into one is annoying
so theres a joke about 67 but not about 69?..... i hate the internet culture of today
both
I'm lazy today
well, lazy days do happen
Too busy* I mean ๐คฃ
Its perfect ๐ฅน
that's... extra nice
@opal hound just had to ruin it 
or 777777 999999
What do you mean by popping up of objects ingame?
probably means pop in, things showing up instantly from thin air in the user view instead of being already there beforehand
sounds like a bottleneck somewhere else, could be paging
it would be nice to see the compare of those 2 in video ...
When playing, turn around 180, many objects, trees, buildings, POP out. It appears to be some kind of loading in or LOD issue, where it goes from no objects or the lowest quality and starts to load in the proper LOD/Textures.
I noticed this on CUP Chernarus 2020, with Extreme QUALITY Settings, BUT OBJECT and VIEWDISTANCE at STANDARD settings. 1600m limit.
3.3.2 does NOT do this on the same settings, same build (LAST PERFORMACE build before the update from yesterday).
I am back to using 227_20260203 for stability, no issues, same settings. I will run through all the mimalloc at the same location from 227 up to 332 again later and take screenshots of any differences.
This was with Win10LTSC, Steam, A3 launcher and ARMA 3 running. AFAIK, no other task was running. No Reshade, No DXVK.
I am unsure how to screencap video as I run ARMA 3 in a WINDOW ATM.
Hm. Yeah, noticed it slightly when I turn too fast in some spots, Chernarus 2020 too, though not substantially problematic for me as it's just a couple of trees in a distance within a fog
just use obs, either window or game capture, should work, same with other software, playing windowed shouldn't be a problem to capture @hoary jolt
๐ฅณ
2.22 release in the next 5 weeks (+3d) confirmed.
lol
I just got them now, what's going on ๐
Bohemias birthday
Happy Birthday Bohemia
happy existmas bi!!!!!
2 Birthdays in a row, WWW
Intercept does not seem to load properly with dedicated server v69 - 2.20.153752
Seems to load properly with dedicated server v68 - 2.20.153725
No changes other than updating the branch version. Happens with or without any other launch params, and happens whether its the only addon loaded (with CBA as the requirement) or not. Seems to work fine with a client executable.
v69
9:38:09 Initializing Intercept
9:38:09 CallExtension loaded: intercept (D:\SteamLibrary\steamapps\common\Arma 3\!Workshop\@Intercept Minimal Dev\intercept_x64.dll) [] []
9:38:09 Intercept initialization part 1/3: 1
9:38:09 Error in expression <interceptRegisterTypes parsingNamespace>
9:38:09 Error position: <parsingNamespace>
9:38:09 Error Missing ;
9:38:09 Error in expression <interceptRegisterTypes parsingNamespace>
9:38:09 Error position: <parsingNamespace>
9:38:09 Error Missing ;
9:38:09 Intercept initialization part 2/3: any
9:38:09 Intercept Loading Plugin: PLUGIN1
9:38:09 Intercept Loading Plugin: PLUGIN2
9:38:09 Intercept initialization part 3/3: Tried to call before initialization
v68
9:41:11 Initializing Intercept
9:41:11 CallExtension loaded: intercept (D:\SteamLibrary\steamapps\common\Arma 3\!Workshop\@Intercept Minimal Dev\intercept_x64.dll) [] []
9:41:11 Intercept initialization part 1/3: 1
9:41:11 Intercept initialization part 2/3: true
9:41:11 Intercept Loading Plugin: PLUGIN1
9:41:11 "Intercept: Load completed PLUGIN1"
9:41:11 Intercept Loading Plugin: PLUGIN2
9:41:11 "Intercept: Load completed PLUGIN2"
9:41:11 Intercept initialization part 3/3: -1
Tried missionHTTPDownloadBaseURL parameter.
Is it ok that Player-Name header contains some hash or smth instead of player name?
Base64?
#perf_prof_branch message ok found
Revision: 153779
Trying to escape all the characters that we cannot use there is just way too annoying. Base64 is simple way to pass it along without risk of it being missinterpreted.
But then we have the mission file name in there bare... mh.
But filename needs to correspond to filesystem limitations so it's probably already rid of bad characters. Unless you use a weird filesystem that allows quotes and linebreaks in filenames..
- that'd be their own problem so

but players could do whatever
Ah yeah, we just forward the name the client got from the server there โ
Hatchet h60 ๐
Of course, those guys push the engine to the limits.
I fixed the crash in todays prof though
What's the fix? Ignoring the attachment attempt?
Only going a few levels deep instead of infinite until we're out of space
I won't get to it till monday
We can just run the older version, just hope a solution is found before next major update 
By the wayโฆ just between us, he doesnโt work on Mondays ๐
