Pathfinding Lag Fix | Pathfinding Lib | Smart Enemy Pathfinding

[fast crow]

Meow meow!

[crimson jolt]

just curious... how confident are you that this new version is totally stable?

ever since v73 i've been getting a bunch of (seemingly random) inconsistent crashes when landing the ship, and while the crash log is different every time, it seems to have something to do with pathfindinglib/loadstone/dungen

[keen ruin]

not very tbh, I haven't played since the update, only quick tests over the weekend

however, Lunxara has said she hasn't had crashes with it I believe?

someone else has also mentioned crashes like that though, also using PathfindingLib

could be worth testing with only PathfindingLib disabled, I definitely want to know if that goes away from disabling it

[tardy pivot]

i have a feeling that if it was PathfindingLib you would get hard crashes w/o logs🤔

[crimson jolt]

it is super annoying

it is a hard crash, unity crash reporer opens, game closes, etc.

but it is not at all consistent, it is only like a 25-30% reproduction rate for me at this point and the crash logs never have the same stack trace

when the update first released i was crashing constantly (and that was after updating pathfindinglib)

[keen ruin]

nah memory corruption has still caused stack traces to print in the past at least

[keen ruin]

I'm sure

I'd love confirmation that it's related to PathfindingLib, but it honestly wouldn't surprise me, I'll have to open Ghidra soon to make sure all my release offsets match up still

[crimson jolt]

all i can say is that it's definitely happening at the same point each time

it's always after landing the ship before the doors open

in my case it seems like it only ever occurs on dine

but i think lunx was getting crashes that were not specifically related to dine

all of the crash logs have stuff related to loadstone, dungen, and pathfindinglib in the immediate proximity, so they are my strongest contenders for what's going wrong

i no longer use loadstone in multiplayer but i had it on my solo testing profile, havent crashed at all since disabling it but there's no telling if that's just unreliable repro rate

[tardy pivot]

could be a new source of navmesh edits that we still have to hook with a lock

that's the only thing that comes to my mind related to pathfindingLib and the opening sequence

[tardy pivot]

True but in that case the exceptions were not related at all to PathfindingLib 😂 😅

[keen ruin]

gotcha, please keep me posted

the other person encountering that type of crash was using Loadstone, so that would track

[crimson jolt]

yeah

loadstone causes desync on our multiplayer profile and basically makes the game unplayable at a certain point so we dont tend to use it

so if it turns out it is some sort of incompatibility with loadstone it's not really skin off my nose

but obviously it's unideal if that's causing "fairly reliable" crashes, so i'd hope it could be narrowed down definitively and fixed if necessary

[vivid dock]

the only option I disabled was DunGenOptimizations and it never caused desyncs again

[keen ruin]

might be interesting to see some player logs from those crashes actually

it may be a transpiler (or multiple) breaking something, if it's Loadstone that's the problem

[ashen harbor]

@keen ruin There's a crash.dmp file included as well that might help say what it is

[keen ruin]

hmm, interesting, Loadstone is in the stack trace, but it involves navmesh as well

I'll have to try spamming navmesh rebuilds and see if it does anything bad

[ashen harbor]

Oh yeah maybe AsyncNavmesh is doing it?

[keen ruin]

dunno honestly

the stack trace doesn't make it sound async, I forget what that method is called

[crimson jolt]

i have a laundry list of crash.dmp if you want more data

[keen ruin]

player.log is gonna be more helpful if you have those too

[ashen harbor]

Adi said it's AsyncNavmesh

lol

[keen ruin]

idk if that's necessarily guaranteed

[crimson jolt]

i've still got both

[keen ruin]

PathfindingLib has patches to make sure that navmesh modification doesn't happen while async pathfinding is running

[crimson jolt]

i think like, the most recent 3 crash logs all have slightly different stack traces, but they will probably be most helpful

[keen ruin]

yeah sounds good

[crimson jolt]

ok give me a second

here is every single crash i got while i was updating mods to v73

so... 13 total. there were a couple that stood out to me in particular, im gonna see which ones they were

"Crash_2025-10-08_032837303" has some dungen specific functions in the stack trace, related to generating GlobalProps

[keen ruin]

hmm, some involve my ApplyCarveResults patch

also were all of these happening on landing?

[crimson jolt]

a couple of these appear to just be startup crashes, sorry about that - those are probably my bad and not related to the issue

[crimson jolt]

every single crash (at least, the ones that warrant investigation) would happen:

1. after the ship's lever was pulled
2. before the ship doors opened

in my case, specifically on Dine, but i think lunx was experiencing it in general

im gonna try to get a fresh crash log, adi just posted a version with some additional logging

[crimson jolt]

hopefully that will be more helpful than just digging through all of these

[keen ruin]

I wonder if unlocking a door may also cause this?

because that would track with something being wrong with the carving patch

I'll have to run my old tests again though

I'm trying to debug RuntimeIcons rn, something really weird is going on with the compute shader

[keen ruin]

@crimson jolt we can takl more about it here instead of taking over Loadstone lol

unfortunately I'm kinda running out of steam tonight, had a busy day today, so I think I'm gonna stop for now

I went through a bunch of offsets that I hadn't explicitly checked before, everything looks okay

so I'm really not sure what the deal is

if you keep at it, hopefully you can glean some information about what a fairly small pack to repro is

adi may be right about needing to throttle it though, so I might try setting processor affinity next if you don't find something first

[crimson jolt]

i gave it like 30 tries with various settings and kind of burnt myself out

i'll get back to it "soon"

i think im gonna try to replicate it with a slightly large pack first and then try to remove some stuff from there

the PLF + imperium + loadstone theory would be nice to confirm

you're right about the processor affinity, might have some luck there

[keen ruin]

damn this bug is a tricky one

[ashen harbor]

the last time it caused a crash I noticed the console popped up an out of memory error

[keen ruin]

oh wow now that's interesting

assuming it's not data corruption again

[ashen harbor]

Did you or @crimson jolt get any luck on figuring out the crashing problem yet?

[ashen harbor]

Someone reported their game crashes with just Pathfinding Lib and PathfindingLagFix present @keen ruin

[keen ruin]

I'm aware, I've been busy so I haven't had a chance to look at it properly again

[karmic geode]

isnt that because of v73 update?

[keen ruin]

I'll have to try with single core affinity tomorrow, if it doesn't repro with that then I won't be able to do much probably

[ashen harbor]

Yeah I'm crashing almost 99.9% of the time when I try to land the ship lol

[keen ruin]

I still haven't crashed once

[ashen harbor]

No

[keen ruin]

and I don't have a small profile to test against still

my only hope at the moment is that running it on a single core makes it more reproducible

[ashen harbor]

I do wonder what mods were in the profile Buttery had since iirc it was small

[keen ruin]

PathfindingLib is updated for v73, but something seems to be finicky about it still to where it'll crash occasionally, and I don't know the cause yet (or it would be fixed)

not sure how much Buttery had narrowed it down, I didn't get the impression they'd been able to repro yet on a minimal one

[ashen harbor]

Hmmm, crashed on a much smaller profile

Gonna see if it repros with just these

AHA

IT DID

I tried to back out to main menu and it crashed

@keen ruin

[ashen harbor]

@keen ruin I tested updating BepInEx and PathfindingLib caused another out of memory crash on ship land

[green gust]

damn, we were really hinging on bepinex update fixing it all

[ashen harbor]

I will say, the updated Bepin build does improve boot times pretty significantly

TS should update it

I imported my profile into R2 to test this so I wouldn't nuke Gale

lmao

[keen ruin]

hmm, if you have a seed and moon and small profile that might help me repro

it's still strange to me that I haven't been able to get it to happen if it's possible to repro with just PathfindingLagFix

I'm assuming your CPU has a decent number of cores?

[next shard]

I have not been able to repro it either

I have a low number of cores (4 💀)

[keen ruin]

I wonder if it's getting triggered by some code in SmartEnemyPathfinding??

[green gust]

i havent been testing that much but i havent had any crashes on my machine yet

[keen ruin]

just to confirm, this was the whole profile?

default configs all around?

[ashen harbor]

Yeah

[ashen harbor]

It does yeah

12900k

[keen ruin]

ah yeah ok

then I have no idea why I haven't gotten it

this is weird

[next shard]

The only thing I can think of is the GlobalRoaming setting, I have it disabled while @ashen harbor has it enabled on da profile I believe

[keen ruin]

only difference is that I haven't really tested with SmartEnemyPathfinding much at all

[hexed moon]

when does this crash happen

[ashen harbor]

Not in the test profile but it is on in my main profile

[keen ruin]

interesting

oh

[next shard]

Hmmm

I mean I landed on a bunch of moons with high Masked spawns and skipped forward in the day

[ashen harbor]

I do wonder why some crashes say Out of memory though

[keen ruin]

then maybe it's just generally having smart pathfinding be used

[next shard]

So they would all try to path to the ship and stuff

[keen ruin]

well the ooms are from it trying to allocate what looks to be a size specified by an overflowed integer

[ashen harbor]

@next shard Why did this throw after a round on Bozo?

Huhhhhh

[next shard]

Whar

[green gust]

nice template paco

[keen ruin]

so nothing is really growing out of control, it's just mad that it wanted to allocate 9 quadrillion bytes or something lol

[ashen harbor]

It throws at the end of every round, perhaps poltergeist lol

[green gust]

i'd be mad too tbh

[next shard]

I'd be mad too to be fair

LMAO

[green gust]

https://tenor.com/view/yeah-dats-what-i-woulda-gif-18253305946678612829

[ashen harbor]

I wonder what is using this tbh

[next shard]

LCMPublishingTemplate

[ashen harbor]

I need to find out what mod is hooking into that rpc

XD

[next shard]

Big mod

[hexed moon]

like going back to orbit or quitting to menu?

[ashen harbor]

It happens when you go into orbit

and the round ends

lolol

[hexed moon]

ok

ive quit to menu during testing a lot but not gone back to orbit

[next shard]

I tried both things while there were like 20 Masked present

[green gust]

lol what i just crashed landing on titan

[ashen harbor]

It can happen on any moon

[tardy pivot]

could you try to reproduce it without AsyncLoggers too?
i Highly doubt it could cause this kind of issues but having a really minimal pack would help a lot zaggy

[ashen harbor]

I crashed several times landing on Experimentation

[ashen harbor]

Funny story

[green gust]

i doubt it'd be smart enemy pathfinding then

thats like, with enemies existing

[ashen harbor]

pacoito has been testing with all the mods I reprod a crash with minus Async Loggers

and he can't repro

XD

[next shard]

💀

[ashen harbor]

But tbf I think his processor also has way less cores

[tardy pivot]

oh ohh 👀

[green gust]

if it helps i dont use async loggers

[next shard]

Downgrade thy CPU

[green gust]

but i also gotta grab my crash log

[ashen harbor]

Yeah I figured it ain't Async Loggers

it was just funny he mentioned to test without it

when pacoito has been

[hexed moon]

i can make a small profile for testing this

a 12700kf might be more susceptible if core count has anything to do for it

[tardy pivot]

what would be great is a reproducible context.
aka smallest modpack + seed
that crashes every time

[green gust]

where was the crash log again

[hexed moon]

so pathfindinglagfix and smartenemypathfinding right

[ashen harbor]

Well you can just grab your player.log it will have it too

[keen ruin]

yea if it happens on high core counts I can't imagine it's related to that, I feel like maybe I need to run smart pathfinding jobs or something to get it to crash

not sure

[hexed moon]

i could add cullfactory for a single seed and disable the other features

[tardy pivot]

i mean smart pathfinding only affects masked atm no?

[keen ruin]

testing the seed that lunxara crashed on is probably worth

I would be looking into it rn myself but I need to sleep soon

[next shard]

Just keep landing over and over until you get the same seed

[hexed moon]

[keen ruin]

I'll try tomorrow after I fix the cullfactory error

[ashen harbor]

I HAVE NO IDEA WHAT MOD IS TOUCHING THIS

LMAO

[green gust]

[next shard]

It's LCMPublishingTemplate it says it right there

[ashen harbor]

Okay stop trolling

lol

[next shard]

💀

I have no idea

Lmao

Do you have like a test build of some mod still in your profile mayhaps

[ashen harbor]

No

Lol

[tardy pivot]

someone did not even bother to change the project name when using the template 💀

[ashen harbor]

LMAO it's FixPlayerName

@tiny ledge

[next shard]

Thy player name is fixed

[tiny ledge]

💀

I remember I made the changes, but why didn't they take effect?

[ashen harbor]

Maybe you accidentally uploaded the wrong dll?

lol

[next shard]

No the dll's right, the FixPatches class do just be in the LCMPublishingTemplate.Patches namespace

Or wait you mean wrong dll as in da update

[ashen harbor]

Nah by wrong dll I mean the Field Exception error

[next shard]

Nevermind

o

I dunno

[ashen harbor]

I don't think that error threw with the previous release

so it's funny

XD

[tiny ledge]

It seems that BepInEx.AssemblyPublicizer cannot be used as a replacement for reflection.

💀

[ashen harbor]

What? I've seen people use that as a replacement for reflection in a lot of things

[tiny ledge]

Hmm, I did the same thing in version 1.1.0, which has caused an error now...

[ashen harbor]

Cus the method you're using to reference it is inaccessible

[green gust]

lunxara the all knowing

[green gust]

show your csproj

[ashen harbor]

I mean the error is something I've never seen before lol, but it does sound like a pretty straightforward one

[tiny ledge]

        [HarmonyPostfix]
        [HarmonyPatch(typeof(StartOfRound), "SetShipReadyToLand")]
        private static void ResetHUDManager()
        {
            HUDManager instance = HUDManager.Instance;
            if (!((Object)(object)instance == (Object)null))
            {
                instance.spectatingPlayerBoxes = new Dictionary<Animator, PlayerControllerB>();
                instance.boxesAdded = 0;
            }
        }

old version

        [HarmonyPostfix]
        [HarmonyPatch(typeof(StartOfRound), "SetShipReadyToLand")]
        private static void ResetHUDManager()
        {
            var hudManager = HUDManager.Instance;
            if (hudManager == null) return;
            AccessTools.DeclaredField(typeof(HUDManager), "spectatingPlayerBoxes")?
                .SetValue(hudManager, new Dictionary<Animator, PlayerControllerB>());
            AccessTools.DeclaredField(typeof(HUDManager), "boxesAdded")?
                .SetValue(hudManager, 0);
        }

[green gust]

da csproj

[tiny ledge]

Hmm, I didn't modify the template's csproj, but it actually allows me to access private fields when writing code. I'll add Publicize="true" and try again.

[next shard]

Oh yeah that'll fix it most likely

Although wait aren't the game assemblies already publicized

The ones on NuGet I mean

Or does the template not use em

[tiny ledge]

After I manually added BepInEx.AssemblyPublicizer.MSBuild in NuGet and set LethalCompany.GameLibs.Steam to Publicize="true" in the csproj file, it worked normally. It seems that the designer had some issues; when I hadn't installed BepInEx.AssemblyPublicizer.MSBuild, accessing private fields didn't throw an error either...

[next shard]

Hmm weird

[tiny ledge]

ye,template does not have BepInEx.AssemblyPublicizer.MSBuild .

[green gust]

no as in, does the template use the NuGet package for lethal company

cuz the stuff inside that are publicised by default

or atleast they should be

[tiny ledge]

https://github.com/LethalCompanyModding/LCM-Template-TSPublishing/blob/main/src/LCMPublishingTemplate/LCMPublishingTemplate.csproj

It could also be that something like the cache caused a failure in the designer? After all, in some of my projects, I referenced BepInEx.AssemblyPublicizer.MSBuild

[tardy pivot]

assembly publicizer adds some special attributes on the mod dll that tell c# to shut up and allow publicized access. if you do not have said attributes and still try to access private fields ( eg through an already publicized reference ) the runtime will block you

[tiny ledge]

designer didn't report an error, which made me mistakenly think that the template came with an assembly publicizer.

[tardy pivot]

there is a reason why that template is still WIP 😅

that reminds me i should update my own templates 🙃 i still reference v69

[tiny ledge]

Robyn thought that the template wasn't useful, so gave it up.

However, for me, using templates is better than creating projects manually.

❤️

[ashen harbor]

0199d2a4-9194-3f49-6093-42e7b0582d1c

@keen ruin

[next shard]

look inside

[ashen harbor]

I input the seed that crashed into the CullFactory config for ya

@tardy pivot If you're available you could maybe look into it potentially

[ashen harbor]

If it doesn't repro I would just clear the config and try a few lands considering it crashed so quickly

[crimson jolt]

just to get the information out there, i dont use smartenemypathfinding

it might be "one way of several" to trigger a problem but the profile i was replicating the crash on only uses pathfindinglib and pathfindinglagfix

• the other 31 mods

[next shard]

I have a few clues but I'm just confirming it do be this

[crimson jolt]

i dont think any other mods im using implement pathfindinglib off the top of my head

i can try the cpu affinity thing, i've got some time to burn

[next shard]

I got it to repro on my computer

Don't think it's CPU core count

[crimson jolt]

maybe not

i have 6 cores

you said you have 4 yeah?

[next shard]

Ye

[crimson jolt]

wow this game loads terribly with single core affinity

[tiny ledge]

Does providing a dmp file not help in fixing this issue?

[crimson jolt]

nono i've got a gajillion of those

but ideally we are looking for the smallest subset of conditions to reproduce the issue

we still have no idea how to make the crash happen consistently

or what mods are required

if we could get a seed that crashes every time with just pathfindinglib (or maybe pathfindinglagfix) it would help a ton but i dont think we've found that yet

hey hey hey!

i got this to crash running LC on only CPU 0 (technically single logical processor rather than single core, but still)

[next shard]

Here's the thing uhhh

Try disabling patchOffMeshConnectionStutterStepping in PathfindingLib's config and see if you can get it to happen

[crimson jolt]

first im gonna try disabling pathfindinglagfix to see if it's essential

[next shard]

I'm guessing I never saw it in my testing profile because I had instant landing/takeoff enabled in Imperium

[crimson jolt]

but sure

i will give it a go

[next shard]

I've only had it start happening when letting the ship do its full animation

[crimson jolt]

idk i use both of these with imperium and it would replicate for me

[next shard]

And they do be moving OffMeshLinks

Oh

Hmm weird

[crimson jolt]

it might be worth looking more into that though

zaggy thought it had to do with carving at first but couldnt get that to replicate a crash in testing

not sure if he checked offmeshlinks stuff

[next shard]

Though I've been trying to get it to happen again but haven't been able to, hmmm

[next shard]

I got a really quick crash after landing twice with this profile here

But played a whole quota with that one setting disabled

[next shard]

Although trying with it enabled again and I can't get it to crash, hmmmmm

This be weird

[keen ruin]

I didn't check off-mesh links in depth yet no

I'm up now so I can look into it after I patch CullFactory

shouldn't take long if I'm able to repro that issue

[green gust]

I spawned 50 jimothys and never crashed on titan btw so good luck

Didn't test anything with doors etc though

[keen ruin]

yeah that doesn't surprise me too much I suppose

[ashen harbor]

Let me see if I also stop getting crashes with it off if I do I'll disable it for today and host a lobby

[green gust]

I did also crash on titan on loading but I couldn't find my crash logs so I gave up on finding that

[ashen harbor]

Appdata > Local > Temp > ZeekerssRBLX > Lethal Company > Crashes

yw

@keen ruin So far so good, 3 lands in a row with that option off and no crash

And this is on my normal profile, I almost always got a crash the second I pulled the lever the first time

Lmao

Yeah, it's not crashing anymore

Let me grab that seed I reprod with before

@keen ruin Yeah that seed isn't reproing the crash

I think we found the culprit

[keen ruin]

with PathfindingLagFix off?

[green gust]

Wasn't there, yw given back

[ashen harbor]

I never crashed with it off, but turning it back on I also stopped crashing with this option off

So I think it's the off mesh link stuff like pacotio said

[keen ruin]

that's.....strange

PathfindingLagFix has nothing to do with that setting

[ashen harbor]

Does SmartEnemyPathfinding make use of it though? I also assume some enemy mods relying on PathfindingLib like Scopophobia make use of it too

[ashen harbor]

The theory pacoito had btw is that the ship is making use of it btw

And the game does crash after lever pull usually

[crimson jolt]

i got the game to crash using only pathfindinglib, pathfindinglagfix, bepinex, and that setting enabled

[keen ruin]

no, it's just an optimization of the engine, nothing should affect whether it does something

except whether links are moving, which they are when the ship is landing, so that part makes sense to me

so.. time for me to try repro, does setting seed help with this at all? if so, anybody got one?

[vivid dock]

maybe mods that modify the ship?

the elevator from the mineshaft is a link?

[ashen harbor]

Use that profile I sent ya, it's a small profile with a seed I produced a crash on, if you can't get it to crash just reset it and land a few times on experimentation

[keen ruin]

hm, did you try more than 2 landings? it may not be 100%

[ashen harbor]

I did not, but it did crash on the 2nd landing

[keen ruin]

but not on the first? I thought you meant it crashed on two separate landings

a one-off isn't very conclusive

I got the crash, but it's not consistent, which makes it difficult to determine whether a fix is sufficient

I'll do my best though

[vivid dock]

yesterday i had two crashes on gordion

[keen ruin]

oh, the ship's off-mesh links don't have auto update on, I forgot

hmm

I think I see why the crash is so rare

I'm checking something and silently exiting the update method if it doesn't match

I wonder if I can make it noisy instead...

I gotta downgrade to check

[ashen harbor]

Yeah, I haven't had a single crash since turning off that option in PathfindingLib

it's a good fix for now, ty @next shard

[keen ruin]

all right I figured it out I believe

@next shard what does bozoros have in the way of links nowadays? I wanna test and make sure it's stable there before I release

huh, bozoros doesn't wanna finish generating, I wonder if I goofed something in my install

ok it worked after I enabled Loadstone

[next shard]

Whar

[next shard]

OffMeshLinks that enable when the Ferris wheel windows open

[keen ruin]

I see I see

enabled auto updating on them and no crash yet, so that's good

oop it failed dungeon generation 20 times again with loadstone

[next shard]

Huh

Is it just with Circus facility?

[keen ruin]

uhh not sure, lemme see in a moment

it does kinda look like it, but I've only landed like 4 times

or tried to

[next shard]

I do some funny Doorway stuff to have multiple Entrance/Exits for the tiles, but I haven't had it fail to generate

Thonk

[keen ruin]

dunno if it tries to generate any other interiors with default configs

[next shard]

Low chance for regular facility, Mansion/Mineshaft at a weight of 1

[keen ruin]

gotcha

yeah I haven't seen those try to generate yet

ok it workin time to push update

Version 2.4.1

• Fixed a crash that could occur, usually during ship landing, with PatchOffMeshConnectionStutterStepping enabled.

[next shard]

https://tenor.com/view/yes-yes-sir-yayy-kataman-gif-12260883688244422951

[keen ruin]

I've made it so that if it detects an inconsistency it will disable the patch automatically, wish I'd thought to do that before so that it would've informed me about the issue

but if Unity trolls us again I'll know about it this time

[ashen harbor]

@keen ruin Just a heads up that on moons with lots of masked Global Roaming is quite resource intensive, it caused my fans on my pc to crank up to max lmao

I was no longer able to repro it after disabling it

I initially thought it was smth with Mirage but then reprod it with Mirage disabled

XD

[keen ruin]

ah yeah that's not super surprising

I might be able to throttle it based on load but I dunno how much that'll help

global roaming is a very time consuming background task

[ashen harbor]

Ye I wasn't really surprised by it myself lol

[vivid dock]

global roaming don't work in modded interiors or is it a problem with offense fire exit?

[green gust]

It needs to be able to roam

Can't on moons like offense where's there's no nodes and no space to roam

[vivid dock]

Even with your pathfinding mods, enemies will get stuck in the car from the garage tile.

[vivid dock]

probably its a problem with how unity generate navmesh
Sometimes the car navmesh is a curve (working)
Sometimes the car navmesh is plane (make enemies getting stucked)

[modern berry]

that step doesnt have proper navmesh from my testing in the past.. so enemies cant step on/off of it. (assuming thats the march fire exit), same can be said with this #1324682579496144978 message , offense fire exit doesnt have proper navmeshing to other parts of the map, its an independent ledge only accessible from the fire exit itself for enemies.

[vivid dock]

Its Adamance

[modern berry]

adamance has a similar step to march fire exit 3 though, i think?

if you turn on imperium and use its navmesh view and spawn a masked on that step (somehow) you will likely see there is no navmesh... ive never really tested on adamance x.x so cant confirm

you dont need to spawn the masked btw.. just uh yeah.. look at the navmesh in imperium.

[vivid dock]

Yes, offense and assurance problems is that masked can't go to the giant pipe because there's no navmesh between.

[modern berry]

yeah, it will be the same on that step (if its like march's step on fire exit 3), i think its that the gap between the step and the other floor, is too high, for the game to consider it to be a connected navmesh.

[vivid dock]

For some reason this time the masked was able to use that fire exit

But I'll probably disable global roaming. A lot of masked spawned on Titan and my CPU usage reached 80%.

[modern berry]

was it moving inside, to outside?

[vivid dock]

Yes

[modern berry]

the "problem" with fire exit 3 is on the outside, so unless there is a check on both sides of every fire exit (something which i do badly in LI), the entity will not realise it will get stuck outside if it goes through, so will still go through, and i assume then get stuck. in LI i check both sides multiple times in multiple ways so that LI knows whether an exit is "problem free" if it isnt problem free, then i dont use it. im guessing this isnt checked in this case for this because entities generally tend to go "i cant go anywhere" and go back through, but in the case of march fire exit 3, they literally cant move afaik, so maybe they get stuck?

[vivid dock]

The masked that was able to use the fire exit glitch a little like that one and start to chase me

[modern berry]

i guess they recognise that as the nearest point to you that they can reach and run on the spot while trying to reach you, normally if it was a locked door and they could see you, they would run around the shortest possible route, but i assume they dont consider the fire exit to be a route (which certainly, purely vanilla masked dont)

[cold peak]

https://www.patreon.com/posts/pest-control-152566021
Looks like performance of the AI will be improved in vanilla too

[keen ruin]

interesting that he only points to allocation as the main focus for the roaming logic from my benchmarks with/without async pathfinding (and supported by profiles), the vast majority of the jank comes from the pathfinding itself, not allocations

but then again, I did avoid allocations in my hooks that replace some of the vanilla logic for that, and I don't think the profiler gives a good enough indication of just how much time is wasted in allocation, so who knows

gonna be a fun time updating my patches though 😅

[crimson jolt]

???

public void AvoidClosestPlayer()
{
    if (this.farthestNodeFromTargetPlayer == null)
    {
        this.gettingFarthestNodeFromPlayerAsync = true;
        return;
    }
    Transform transform = this.farthestNodeFromTargetPlayer;
    this.farthestNodeFromTargetPlayer = null;
-   if (transform != null && this.mostOptimalDistance > 5f && Physics.Linecast(transform.transform.position, this.targetPlayer.gameplayCamera.transform.position, StartOfRound.Instance.collidersAndRoomMaskAndDefault, QueryTriggerInteraction.Ignore))
+   float num = Vector3.Distance(transform.transform.position, base.transform.position);
+   if (transform != null && num > 5f)
    {
        this.targetNode = transform;
        base.SetDestinationToPosition(this.targetNode.position, false);
        return;
    }
    if (this.carryingPlayerBody)
    {
        this.DropPlayerBody();
        this.DropPlayerBodyServerRpc();
    }
+   if (!Physics.Linecast(transform.transform.position, this.targetPlayer.gameplayCamera.transform.position, StartOfRound.Instance.collidersAndRoomMaskAndDefault, QueryTriggerInteraction.Ignore))
+   {
        this.AddToAngerMeter(this.AIIntervalTime);
+   }
    this.agent.speed = 0f;
+   this.destination = base.transform.position;
}

Vector3.Distance(transform.transform.position, base.transform.position)

(from FlowermanAI())

[cold peak]

it's probably mistype, but works as it should be

[crimson jolt]

well i thought in this case mostOptimalDistance was supposed to be replaced with the distance from the node it wants to hide at

oh wait nvm

i see what i was getting tripped up on

outside of the fact im not sure why he's doing .transform on what is already a Transform i just got confused by dnspy's local variable naming

[cold peak]

bracken was like implemented second (or first) enemy, so probably before player component was used (or gameobject) and now we get double transform thing

[keen ruin]

is mostOptimalDistance removed? that would be good

waste of memory to avoid a single distance calc

(which also caused bugs due to being used too much)

[brittle scaffold]

is this mod still relevant in v81, and does it work fine or is it awaiting an update?

[crimson jolt]

it's still relevant, and it will not work until it is manually updated

[desert jolt]

People are reporting that enemies are going "afk" with pathfindinglagfix so it would need to be updated.
Vanilla has some new performance patches to enemy ai so its not as necessary as before but still would be better

[crimson jolt]

it's "less relevant" because on the whole vanilla is a lot more performant but the specific improvements from this mod can still be made even with v80+

[brittle scaffold]

i see, btw @keen ruin i suggest adding a quality of life or bug fixes tag because when filtering for those specific mods category on r2modman your mod isnt there

[desert jolt]

adding to this, theres a Performance tag that pathfindinglagfix should definitely have

[brittle scaffold]

oh yeah especially that as well

[keen ruin]

huh, I wonder if those just didn't exist when I created it

I never look at the tags after I create mods 😅

and yeah, I suspect PathfindingLagFix will still improve things on moons with high enemy power, but I haven't even started looking at the improvements Zeekerss made yet, so not sure by how much

[crimson jolt]

quality of life and performance are relatively new

bug fixes is older but still relatively new

none of them existed when i first released BF if i recall

[lone bronze]

lag fix is pathfinding

[tribal crescent]

pathfindinglagfix breaks the PumaAI's tree climbing ability

to better describe : when the mod is enabled, the Feiopars (and other creatures using PumaAI) will go to a tree and just keep on running at its root attempting to climb and throw errors in the console nonstop

[keen ruin]

It's not updated yet, so it's a miracle if anything works at all yet

[tribal crescent]

lol makes sense

[tribal crescent]

do we know if it's coming soon or still taking time

[keen ruin]

not sure yet how long it'll take to update, I still haven't gotten to reading the changes to the vanilla pathfinding code

hard to say

[vivid dock]

Idk if its possible, but in the next update, if you could improve how butlers open doors when they are chasing someone. Since when they are chasing someone, they go in a zigzag; it's difficult for them to open doors.

[green gust]

Yeah but isn't that intentional?

[vivid dock]

Idk, but I think Zeekers wanted them to be hard to kill on multiplayer so players will try to stick together instead of trying to kill them.

[crimson jolt]

i really dont think butlers getting stuck on doors is intended

just personally

[desert jolt]

is pathfindinglagfix making them stuck on doors?

[crimson jolt]

im not sure if this is the right place to fix it

but

im pretty sure it's a quirk of vanilla

[desert jolt]

this does sound like a zeekers issue

[vivid dock]

every vanilla bug is a zeekers issue

[brittle scaffold]

duh

[keen ruin]

that sounds like something that isn't really within the scope of the mod

the fixes that I've included in the past have been things that couldn't be fixed by other mods without making separate patches for both vanilla and PathfindingLagFix

which by extension means that they are bugs that have been incidentally or even unintentionally fixed and had to be reintroduced in order to add the options to unfix

[karmic geode]

Sorry to bother you zagster, not trying to rush, just curious if you have checked what changes will this mod need?

Is it still needed? Im guessing yes hehe

But im guessing quite a bit will have to be changed?

[keen ruin]

I haven't had the chance, had to fix some things with CullFactory when I had time

I'm honestly not sure when I'll have time, I have friends in town over the weekend and work until then 😅

I'll see what I can do

as far as whether it's needed, I've never said it is and never will, but it will always give a performance benefit to move pathfinding off the main thread

the question is only how much, and you can determine that yourself if you go to a moon with lots of spawns, wait till a lot of enemies have spawned, note your fps, then despawn them and compare the difference

PathfindingLagFix should give you close to that sort of improvement, except when the AI is misbehaving in other ways

[karmic geode]

Yes, i meant needed as in, its an improvement, so why wouldnt you have it on hahaha

Welp take your time then! Im guessing this one is harder to fix than others so i wish you the best of luck haha

And have a great time with your friends :3

[brittle scaffold]

[14:44:43.1560685] [Warning:PathfindingLagFix] Method passed to ILMatcher.Call() was null at ./Patches/PatchFlowermanAI.cs#84 (DoAIIntervalTranspiler)
[14:44:43.1560685] [Error :PathfindingLagFix] Failed to find call to ChooseFarthestNodeFromPosition in FlowermanAI.DoAIInterval().
[14:44:43.1748239] [Warning:PathfindingLagFix] Method passed to ILMatcher.Call() was null at ./Patches/PatchDoublewingAI.cs#71 (DoAIIntervalTranspiler)
[14:44:43.1748239] [Error :PathfindingLagFix] Failed to find instructions to choose player evasion node in DoublewingAI.DoAIInterval().

[vivid dock]

It wasn't updated to v80

[brittle scaffold]

ah

[brittle scaffold]

@keen ruin we need you 🙏

[feral horizon]

Just play without for now.

[karmic geode]

patience

[steep tartan]

https://tenor.com/view/star-wars-yoda-patience-sw3-gif-21070142

[vivid dock]

Its not even 2 weeks since the beta

[keen ruin]

I'm away from my PC until Tuesday

[brittle scaffold]

its tuesday

[green gust]

💀

holy patience

he'll update it when he gets the chance to, you don't have to backseat 😭

[brittle scaffold]

it's just a gentle reminder calm down

[minor python]

You've been popping into every single thread and mentioning their authors.

[vivid dock]

xD

[karmic geode]

No rest for the wicked

[brittle scaffold]

not even true

[lean inlet]

it's tuesday already, but somehow there are still zero pull requests inbound, huh

[keen ruin]

believe it or not I was aware when I arrived home what the next day is

I said I'd be back, not that it meant I'd be able to update it the same day

(and while the errors printed may look like it just needs a recompile, it's very likely that it'll actually need some patches rewritten, I don't intend to neuter the functionality to get it running)

[keen ruin]

starting on it, but there are indeed some patches that need to handle new features which may take some time

first one I've encountered is the doGroundCast parameter for TargetClosestPlayer, and I'm seeing how possible it is to run those raycasts async as well in order to save some extra time there

[lean inlet]

definitely written by non-human, with no human even being in the loop:

https://github.com/Zaggy1024/LC_PathfindingLagFix/pull/9/changes

[ashen harbor]

What the fuck is that pfp

lol

[sturdy zealot]

Yes yes

[cursive sonnet]

Burd

[keen ruin]

what in the hell..........

that would only address one small issue with v81, and it does it in an unbelievably overengineered way

I already fixed that issue in my working tree, there's a whole lot more that needs work

[lean inlet]

AI slop, that's what

like, even the "author"'s name ends with "Bot"

[keen ruin]

indeed

the issue is that this is too niche a codebase for any agent to do a good job with it without the driver actually knowing something about it

it would've been trivial to fix if they had prompted it to use ILSpy's command line tools to find out the new signature

(not that that solves the rest of the issues I'm working on)

it could have even just omitted the parameters entirely and just found it by name, but I specifically put the parameters in there so that things break when signatures change, since that is very likely to break patches

[vivid dock]

Since you looking at the pathfinding code from v80. Do you know if this error is from vanilla?
"IsStopped" can only be called on an active agent that has been placed on a NavMesh.

[green gust]

It's not possible to say, can be vanilla can be modded

[tardy pivot]

you can check the stacktrace in AsyncLoggers db, but nothing usefull will be shown if it comes from native land

[keen ruin]

I haven't seen IsStopped around yet, but I'll keep an eye out

kinda goofy that that doesn't just return true if it's not on the navmesh

unity devs

[wanton salmon]

I use it for spider position fix insted of disabling agent when spider is on the wall.

[keen ruin]

oh? you're able to move the spider off the navmesh without disabling the agent?

on another note, only a few things needed involved changes after the TargetClosestPosition patch, so the update should be ready to go sometime tonight

updated some things to match vanilla in the latest version, and a few small things became unnecessary, but pretty much every patch is still relevant

[vivid dock]

Nice

[keen ruin]

Version 2.3.0

• Updated all patches to support v81. This update will not work with prior versions.
• The new ground raycast option in TargetClosestPlayer runs asynchronously.
• The vanilla optimization to avoid short physics linecasts in the path-LoS intersection code is now included in all patches.
• Fixed some edge case handling for the maneater AI sneaking patch.
• Removed the AsyncDistancePathfindingMostOptimalDistanceBehavior option. The bracken stuck bug no longer exists, leaving no other vanilla enemies affected by this bug. The default is now to leave it unset for enemies that use the vanilla sliced pathfinding.

just uploaded, stay tuned in your local mod manager

[vivid dock]

when someone killed the spider start to spam that

[wanton salmon]

ah

[wanton salmon]

agent and mesh aren't actually connected in spider's case

they move independently of each other

[lean inlet]

without PathFindingLagFix a snareflea spams a periodic error (at least not every frame), after I jumped into a pit with it on my head.

I wonder if this mod fixes it?

[18:14:08.7323433] [Error  : Unity Log] Agent 'Centipede' #8: Agent not on nav mesh when trying to set destination

[next shard]

This just be due to a lack of NavMesh down the pit more so than the pathfinding itself

There do be this but I dunno if it's updated for v80: https://thunderstore.io/c/lethal-company/p/JacobG5/LostEnemyFix

I think I remember someone mentioning it wasn't workin right

[vivid dock]

yes, centipede was spam logging this and nothing happened

[ashen harbor]

Yeah this doesn't work on V81 and Jacob said he doesn't plan to update it

[next shard]

Dam

Fair enough

[ashen harbor]

He did just say he doesn't plan to cus it was quite niche lol

[copper cosmos]

I would disagree, it was very helpful lol

But I don't have the money to pay him to update it, so

[next shard]

[copper cosmos]

Jacob is telling me he didn't say that so erm

[ashen harbor]

He said it somewhere in here

LOL

Hold on lemme find how he worded it my OSDD might be fucking with me cus Kylie was fronting when he said it

[copper cosmos]

yeah fair idk

[ashen harbor]

I did just stop existing for like 2 or 3 weeks

☠️

Ah yes message searching is broken rn

Thanks Discord

Oh okay my bad

He basically said he may not update it cus he was busy not that he wouldn't XD

[next shard]

[ashen harbor]

Sorry @swift token Lmao I have no idea how I got such a wrong memory of how you worded that and why Kylie didn't correct me

Kylie: Cus I wasn't paying attention Woopsie

Kylie: I am just very unobservant when not fronting most of the time so if I'm not paying attention I am not gonna help 🤣

[next shard]

But what if laser pointer

[ashen harbor]

Kylie: Where?

[next shard]

[ashen harbor]

Kylie: I will bonk Lacy for Jacob

[oak timber]

Lostenemyfix has been updated just now, issue is fixed it works again

zeekerss broke it for no reason

[coarse hawk]

https://tenor.com/view/project-hail-mary-rocky-ryan-gosling-absolute-cinema-absolute-cinema-meme-gif-10719095557049564481

[oak timber]

ik it was being discussed here earlier so figured good to say

[oak timber]

Okay maybe not literally no reason, implementing the error log that started that discussion was what broke it which i guess is cool because it means the log's more descriptive but it still gets spammed in the use case lostenemyfix was made for which is not ideal

[keen ruin]

I'm curious to see if this is a general issue with all enemies

if so, I might patch it out in PathfindingLagFix, there's no real reason not to since it's purely cosmetic

[ashen harbor]

Yeah would be a good idea

[lean inlet]

yet another day of modding QA o7

[lean inlet]

thanks!

[keen ruin]

you mean for you? I likely won't get to this today, it's very much a backburner thing unless someone else confirms that it's coming from the general destination assignment for every EnemyAI

[ashen harbor]

It is

https://thunderstore.io/c/lethal-company/p/JacobG5/LostEnemyFix/ You can actually look at how LostEnemyFix handles it

[keen ruin]

I'm not talking about including that type of fix

I have no intention of changing game behavior

[green gust]

i think it got put into the setdestinationtoposition function when there's no path or whatever

[ashen harbor]

Yeah but I mean every enemy do just throw that error I mean

[next shard]

It don't necessarily kill them always, teleporting to nearby NavMesh is preferred iirc

[green gust]

so it's the same issue that's always existed, there's just a proper non unity log for it

[ashen harbor]

Ye

[next shard]

But yeah it ain't really somethin that should probably be in PathfindingLagFix probably

[green gust]

i've never seen it happen for vanilla personally except maybe a few times, and only manticoils

[keen ruin]

oh, I didn't realize this was a managed log

[green gust]

unity's log would be a lot more useless

[green gust]

this would be the unity equivalent actually

nothing that tells you which agent is causing the error

[keen ruin]

that's honestly worse, adding an unstuck that the base game doesn't have

well, worse for being faithful to vanilla

it's something zeekerss could potentially stand to do

[green gust]

i think it's technically IsResuming that it would say instead of IsStopping, but either way unity's navmesh logging sucks

[next shard]

Maneater has it technically if you drop it into a pit

[keen ruin]

indeed

[green gust]

i thought maneater's pit thing was fixed ages ago

[keen ruin]

but he could absolutely generalize it if he wanted

[green gust]

unless it still happens if you also dive bomb into a pit

[next shard]

Ye

[keen ruin]

nono paco means the unstuck exists in vanilla for maneater

[next shard]

Ye it doesn't spam

[green gust]

ohh

yeah true, but that one's a lil different i think

it goes from navmesh to navmesh

not non navmesh to navmesh, i think?

[keen ruin]

I think it has two different unstucks, sounds like you're thinking of the grown maneater one

[crimson jolt]

i dont know how the baby maneater's unstuck works

but the adult just gets warped to the nearest AI node if it's stuck on navmesh with no accessible AI nodes for 10s+

so you can't trap it on valid navmesh where it's unable to path

[keen ruin]

yee

[crimson jolt]

i am pretty sure the baby's unstuck works a lot differently though yeah

[keen ruin]

it has an additional condition that it is "stuck" only if no players are in LoS too, iirc

something like that anyway, and that's why the kill spots work

because I believe it also does those checks on a fixed period, rather than a timer

[knotty leaf]

any plans on this?

[ashen harbor]

@keen ruin You did say you were gonna fix this right?

[keen ruin]

ah yes

I knew I was forgetting something

PathfindingLagFix was a bigger priority

[ashen harbor]

Ye it was for sure

lol

[keen ruin]

SmartEnemyPathfinding update for v81 is processing

[vivid dock]

I like global roaming a lot. Unfortunately, it increases the CPU from 30% to 90%.

[drowsy galleon]

I was just thinking of this lol, this mod makes my cores scream

[hexed moon]

Is it really that bad

Ive never noticed my cpu fan going higher than usual with that setting

[vivid dock]

it needs to spawn a lot of masked

[drowsy galleon]

spawn one at a time and watch your core usage

[hexed moon]

Thats probably why I havent seen it

I stopped getting mask spawns some time after I turned it on

I dont think its this mod but something messed up masks

Wait I wonder if this is why rampart was running so badly

Rampart spawns its own masks manually

[keen ruin]

yeah it's unfortunately costly

maybe I should look at making it limit the concurrent jobs, but it might make them look less responsive

I'd have to see

[fiery salmon]

Since V81, On modded planets and modded interiors, it seems like the host eventually freezes after a Masked person spawns. it takes a while and there's not really any logs except for spams of:
[Info : Unity Log] Start coroutine B!!!
Usually freezes close to when the ship is supposed to leave.

I have pinpointed the issue to this Smart Enemy Pathfinding since when I remove it the issue doesnt happen.

My most consistant way to reproduce is with the moon Atlas Abyss and then just hangout in the ship until it crashes (10-15min), but I can reproduce with moons from Moons_of_otherwordly_oddities and interiors from generic_interiors.

[ashen harbor]

I haven't had this problem but good to know

And we did have a ton of masked spawn once yesterday

[lean inlet]

that's in EnemyAI -> CurrentSearchCoroutine

[tardy pivot]

as far as i'm aware SmartEnemyPathfinding has not been updated to v81 yet

[knotty leaf]

it did lol

[fiery salmon]

Might be an incompatibility I'm not sure, It might be only on certain moons/interiors

Don't think this happens on base game planets/interiors. I can reproduce fairly consistently tho

[fiery salmon]

Yeah that spam happens whether I have the mod or not, there's just no other logs I can point too, there's no error

[lone bronze]

i think i had the same issue but in any moon, also with and without the mod still cant figure it out what causes it since i get no error logs

[fiery salmon]

Interesting, I'm fairly sure removing the mod fixed my issue, but I haven't pinpointed the issue, maybe it's something generic.

[swift grove]

https://tenor.com/view/myao53-manul-head-turn-surprised-notice-gif-27549558

something... generic?...

[copper cosmos]

I blame Generic Moons

[fiery salmon]

I did add generic interiors to my modpack recently 🤭

[copper cosmos]

@acoustic furnace must be a Liminal Facility issue /j

[acoustic furnace]

you will disappear faster than monty

[keen ruin]

hmmm that's tricky business

theoretically with a full dump of the game when it's frozen like that, I could find the issue, but I'd have to figure out how to get a managed stack trace from a dump again 😅

and the dump would be the size of the memory that the game is using, so tens of gigabytes potentially

[fiery salmon]

I could maybe get you that

[keen ruin]

you know how to do that? it's in the right click menu (probably in details if win11) of the process

[fiery salmon]

Yes I do, but it's been a while since I did that haha. I'll figure it out tomorrow

[steep tartan]

Noticed that Forest Giant sometimes stand in the same place looking at nothing for eternity and I think this error might be related

[Error : Unity Log] NullReferenceException: Object reference not set to an instance of an object
Stack trace:
PathfindingLagFix.Patches.PatchEnemyAI.GetPathStatus (EnemyAI enemy, System.Int32 index) (at ./Patches/PatchEnemyAI.cs:54)
(wrapper dynamic-method) EnemyAI+<ChooseNextNodeInSearchRoutine>d__111.DMD<EnemyAI+<ChooseNextNodeInSearchRoutine>d__111::MoveNext>(EnemyAI/<ChooseNextNodeInSearchRoutine>d__111)
UnityEngine.SetupCoroutine.InvokeMoveNext (System.Collections.IEnumerator enumerator, System.IntPtr returnValueAddress) (at <c39a522eee05469b8171a6cfeb646c59>:0)

[16:15:06.3548143] [Error :UnityDebuggerAssistant]

--- Exception Handler ---

Exception Caught: System.NullReferenceException
Assembly: PathfindingLagFix

Plugin Info
GUID: Zaggy1024.PathfindingLagFix
NAME/VER: [email protected]
LOCATION: \plugins\Zaggy1024-PathfindingLagFix\PathfindingLagFix.dll

Message: Object reference not set to an instance of an object
Source: PathfindingLagFix

--- Begin Frames ---

--FRAME 1:
In Assembly: PathfindingLagFix
Source: ./Patches/PatchEnemyAI.cs:54,9
Target Method: PatchEnemyAI.GetPathStatus

--FRAME 2:
In Assembly: UnityEngine.CoreModule
Target Method: SetupCoroutine.InvokeMoveNext

--- End Frames ---

--- End Exception Handler ---

[ashen harbor]

Weird I've not had this issue 🤔

[crimson jolt]

Noticed that Forest Giant sometimes stand in the same place looking at nothing for eternity
i feel like ive seen this happen a couple times in vanilla too, since v80

so im not 100% sure PLF is causing it, maybe making it worse? would be good to know if there's a way to consistently replicate it so it can be tried both ways

[ashen harbor]

Would be for sure

[crimson jolt]

do you have the seed where this giant spawned and was it flooded weather

[steep tartan]

1. No sry
2. No

[steep tartan]

It happened to me multiple times on March while I was testing some things

Idk if I have a consistent way to replicate it, let me try

[steep tartan]

couldn't find a way to replicate consistenly unfortunately, but I was able to make that happen again on Eclipsed March

[keen ruin]

saw the DM, sorry I haven't gotten to it yet 😅 got myself into the middle of a refactor for work that requires a lot of attention

[keen ruin]

that..definitely looks like a vanilla error, it seems like it must be killing the search routine in an incorrect way somewhere

particularly, currentSearch is almost certainly null, and the coroutine shouldn't still be running if that's been nulled out

[fiery salmon]

Don't worry about it, I don't want to put pressure on you haha. Just glad you got the dump DM, sometimes DMs are ignored cause of spam and stuff haha.

[tribal crescent]

got this error out of nowhere

[Error : Unity Log] NullReferenceException: Object reference not set to an instance of an object
Stack trace:
PathfindingLagFix.Patches.PatchEnemyAI.GetPathStatus (EnemyAI enemy, System.Int32 index) (at ./Patches/PatchEnemyAI.cs:54)
EnemyAI+<ChooseNextNodeInSearchRoutine>d__111.MoveNext () (at <aca1e98d6f844d3f85cd22b6f7be2f9b>:IL_00CF)
UnityEngine.SetupCoroutine.InvokeMoveNext (System.Collections.IEnumerator enumerator, System.IntPtr returnValueAddress) (at <c39a522eee05469b8171a6cfeb646c59>:IL_0026)

idk what it's related to

[tribal crescent]

also when launching the game this is said before clicking on online

[Warning:PathfindingLagFix] Method passed to ILMatcher.Callvirt() was null at ./Patches/PatchEnemyAI.cs:432 (TargetClosestPlayerTranspiler)
[Error :PathfindingLagFix] Failed to find the call to check if the enemy can path to a player in EnemyAI.TargetClosestPlayer().
[Warning:PathfindingLagFix] Method passed to ILMatcher.Call() was null at ./Patches/PatchFlowermanAI.cs:84 (DoAIIntervalTranspiler)
[Error :PathfindingLagFix] Failed to find call to ChooseFarthestNodeFromPosition in FlowermanAI.DoAIInterval().
[Warning:PathfindingLagFix] Method passed to ILMatcher.Call() was null at ./Patches/PatchDoublewingAI.cs:71 (DoAIIntervalTranspiler)
[Error :PathfindingLagFix] Failed to find instructions to choose player evasion node in DoublewingAI.DoAIInterval().

[keen ruin]

huh, was there a lethal update?

[crimson jolt]

no

[keen ruin]

huh

[keen ruin]

this is something that someone else had brought up already, did you have a giant as well?

#1324682579496144978 message

[keen ruin]

but this one... doesn't make a whole lot of sense unless someone else is patching those before PathfindingLagFix

at least given that there was no update past v81 (or assuming that public update is the one I tested against as well)

I guess I better get on my PC and check though

[keen ruin]

yeah no idea what you got going on here unless you're running an old version

neither of the two methods it's failing to find there have changed

[tribal crescent]

Should i just remove the mod then download it again?

[keen ruin]

I was thinking more along the lines of an old lethal version, I wanna say you'd see different errors at startup if it was the mod that's out of date

but can't hurt

I'd say just to see what's going on, verify game files and nuke your mod manager of choice's cached copy of the mod and redownload it

(I say the cached copy because Gale at least keeps a download cache that it copies from instead of going straight to Thunderstore for every install)

[tribal crescent]

Oh then i must've had that when i launched in in previous accidentally

[keen ruin]

that would make sense

but as far as the other error, I believe that's a vanilla bug

behavior with PathfindingLagFix likely shouldn't differ from vanilla with that one

I could obviously silence that error since it's in my code, but that would just mean you'd have no way to know that the search routine broke

[limpid folio]

After installing PathfindingLagFix, if two or more slimes are generated at the same time, they will "merge into one".

On the left, PathfindingLagFix is not installed. On the right, PathfindingLagFix is installed.

If two slimes are generated simultaneously and are far apart from each other, their grids will be connected to each other.

[keen ruin]

hmm, interesting

I wonder if they're getting all the same IDs

oh brother

has this happened in normal gameplay? or only when spawning via debug tools?

it looks like zeekerss did not make sure the enemy indices don't duplicate

I guess it doesn't really matter in vanilla, I'll just have to change how I look up my enemy data tracking

I'll try to get to fixing this soon™, working on Black Mesa finally atm

[fiery salmon]

Black Mesa

[keen ruin]

Finally got around to fixing that

Version 2.3.1

• Fixed an issue where multiple enemies spawned in the same frame could confuse their pathfinding data.

[ashen harbor]

Neat yeah we saw some PathfindLagFix errors recently and were gonna report them but saw others had already been mentioning it, is the issue with Smart Enemy Pathfinding also gonna be fixed where it was causing some people's games to freeze from multiple masked existing?

[keen ruin]

it being the giant?

also, I haven't had a chance to investigate that freeze unfortunately, but hopefully I can take a break from black mesa stuff this weekend and look into it

[keen ruin]

Version 2.4.0

• Added patches to make some gunkfish pathfinding async to remove stutters on large maps.

[dense falcon]

hi, i've been testing and i think PathfindingLagFix is the cause of this issue i've had reported.
i have these bridges you can build by hitting in my moons. they're made from some simple box colliders and navmesh obstacles which are enabled/disabled based on the bridge's hp.
here's without PathfindingLagFix:

and then immediately after a restart with PathfindingLagFix enabled:

ope, yt video playing and audible in the last vid, woops...

but yeah, was just wondering if there's any way i can get around this while still having PathfindingLagFix enabled

i disable both the NavmeshObstacle and the gameobject it's on in events like this, to allow them to walk across:

[keen ruin]

hmm, did you also disable PathfindingLib in that test without PathfindingLagFix?

I would hope it's not related to my PatchOffMeshConnectionStutterStepping patch

if it was enabled, though, not sure what the cause of that would be, tbh

I have no idea about how the pathfinding for those lil guys works

does this happen with any vanilla enemies chasing you onto the bridge?

[dense falcon]

the lib was enabled during those tests

last time i checked, as long as the pikmin can path across, other enemies can too, though i didn't test other enemies with/without this mod yet

it appears i have fallen victim to correlation

i just tested the pikmin again with the fix on, and it works this time

there's some weird inconsistency going on with my bridges somehow, may not be related to this mod it seems

[keen ruin]

fix on as in PathfindingLib option?

[dense falcon]

no, i meant i turned PathfindingLagFix back on

i think it just so happened to not work whenever i had it on, it seems like there's some other issue that makes the navmesh inconsistent

[keen ruin]

huh that's curious

[dense falcon]

yeah, i think i found the issue. i was disabling the colliders for the bridge with an animation, which was disabling them before the navmesh re-bake when landing, sometimes. if i disable them at round start instead, it seems to fix the issue. you can disregard my stuff here, sorry for buggin' ya with an unrelated issue

[keen ruin]

ah, I see, hmm

do moons not use renderer baking too?

I forget how their surface generation is set up

[green gust]

whats that

[keen ruin]

navmesh can bake onto colliders or renderers, and I'm 90% sure vanilla moons at least have that set to renderers for the interior navmesh surface

[green gust]

oh that

you made it sound like some sort of performance thing

i believe we do renderer baking

could be wrong

[keen ruin]

yeah I guess I could've been clearer, I was hoping the context was enough 😅

I be lazy

[green gust]

i dont car for context, i just read "renderer baking"

[keen ruin]

[dense falcon]

mine's set for just physics colliders on default, room, navsurface, and colliders layers

[keen ruin]

as long as that's not for the interior too, that should be okay

changing it from renderers to physics colliders may break some interiors (including vanilla, not sure)

also turns out the whole gunkfish patch thing was not quite right, and boy did zeekerss question my old code's assumptions with this

I had to rewrite the patch to make it match vanilla's behavior properly after realizing it was screwed up, and that involved some fiddling around with code that has been working fine for every other enemy for a while

but now:

Version 2.4.1

• Reworked the gunkfish patch to behave exactly like vanilla while only using one pathfinding job instead of two.
• Removed the DistancePathfindingFallbackNodeSelection option, vanilla now explicitly uses a very similar criteria to BestPathable.

(and yes, removing that option means that PathfindingLagFix now has zero options except the presets which currently do nothing )

[drowsy galleon]

Teehee

[next shard]

The illusion of choice

[vivid dock]

A forest keeper was near Titan's main entrance, looking at the wall and doing nothing, and one player was able to kill him with a shovel before that happen the giant was chasing him.

And a thumper spawned inside a closed room, and when I opened it, it stayed still near the vent where it spawned and only started to move when I was in its field of view.

No errors

[candid gyro]

[Debug :PathfindingLagFix] Player paths from SpringMan(Clone) (SpringManAI, -530768) are 613.0371ms old, which is more than twice 200ms, using synchronous paths.
Question what is this? trying to troubleshoot an issue im having

[keen ruin]

this is probably normal, it just happens if an enemy doesn't need to check if it can target a player for a while

[candid gyro]

ok

[keen ruin]

The giant thing is something that others reported and thought happened in vanilla, but the thumper sounds potentially more suspicious. Got a seed?

When you say the room was closed, do you mean that it was locked? If so, then perhaps the room just didn't have any AI nodes, that's the first thing I'd check for on the seed

[vivid dock]

Yes, locked. I don't have the seed.

sorry

[keen ruin]

if you or anyone else can get me a mod profile, seed and recording for these possible bugs, it'll make it easier to determine if they're something I need to fix or not

particularly a recording starting at the first time you interacted with the enemy in any way, or before it spawned

[broken magnet]

i had these exceptions on vow 114362800 when the tulip snake got off my head. he seemingly got stuck in that place which is somewhat visible on the side of the screen.
code is 019e1f48-cd8a-43c7-4804-fd1a13808ff3, it doesn't include the closed beta mod obviously but it doesn't touch enemy ai

[keen ruin]

Sorry it's takin me a while to look at that, probably a pretty silly obvious goof but I've been deep in the Black Mesa mines

it is out now so I should be free to look into this and other stuff

[vivid dock]

35218506 Offense. I found a seed where this happened with a thumper and hydrogere. 019e35c2-6c21-aafc-1531-de59f259bff7

there's two nodes inside the room where they are locked

[keen ruin]

fixed this locally, was a very minor goof where I forgot to give the fallback node a fallback node

[keen ruin]

thanks, I'll look into this now and see if it's a bug or not

damn the static on the pause menu is kinda bugging my eyes 😅 what mod is that?

[crimson jolt]

probably mine

[keen ruin]

@vivid dock is this where you saw the blob trapped? looks reproducible at least

[crimson jolt]

film grain was globally removed post-launch, i have some settings in BF to add it back

[keen ruin]

gotcha

[crimson jolt]

just set RestoreFilmGrain to None (although you'll have to reboot before it takes effect, i believe)

[keen ruin]

it's mainly a problem in the free cam but I think it would probably still tire my eyes otherwise tbh 😅

hmm, the blob not moving does appear to be a bug in PathfindingLagFix, what the helly

I wonder if this is a regression or a pre-existing issue

[keen ruin]

oh interesting, it spawned in a diff spot for me

[vivid dock]

maybe spawncyclefixes

[keen ruin]

did the same thing though, the blob was staying still until I switched off async

so definitely something I need to look into

[crimson jolt]

SCF 100% changes spawns

but also we were running into a problem a couple weeks ago where the challenge moons didn't have consistent spawns

[keen ruin]

I used your profile code so I think it's just because of player input most likely

[crimson jolt]

so im not sure if spawns are actually consistent based on seed anymore

lol

[keen ruin]

I'm pretty sure they aren't

[crimson jolt]

with and without mods

[keen ruin]

or at least they definitely weren't from what I remember

[crimson jolt]

well... they are supposed to be

[keen ruin]

like if you kill an enemy it changes the spawns I'm pretty sure

[crimson jolt]

and it's been over 2 years now, but im pretty sure they were consistent when v47 first launched because i remember challenge moons used to be consistent

[keen ruin]

hmm

[crimson jolt]

maybe time has warped my memory but i really think i remember that

[crimson jolt]

he actually tried to fix this in v80

lemme find the patch notes

• Fixed the game's randomness so that creatures will always spawn in the same order on Challenge Moons.
  • When creatures die, which would normally open up space for different creatures to spawn early, it now waits until the determined spawn order has been finished to begin spawning extra creatures.
  • Also made it so the Old Birds cannot spawn from thin air.

the challenge moons thing is just totally wrong because even in vanilla v81 we were seeing desync

but he has confirmed that spawns are intended to be consistent based on the seed

and he added new power level variables (currentEnemyPowerNoDeaths, currentOutsideEnemyPowerNoDeaths, currentDaytimeEnemyPowerNoDeaths) which do not ever decrease, and this value is used exclusively until it no longer allows for any enemies to spawn (at which point it returns to using the original variables)

anyways i think im starting to derail a little bit but basically i would not be surprised if you encounter issues with testing spawns even with a specific seed

looks like you were able to replicate the bug anyway so maybe it's ok

[keen ruin]

yee it seems to be fine

hopefully another repro attempt doesn't spawn it differently tho, I don't want to install imperium lol

[keen ruin]

oh god I was looking at the wrong version decomp while trying to figure out what was going wrong with the search routine

oh jeez

I see the problem, I don't know how I missed this before

ChooseNextNodeInSearchRoutine runs an outer loop now

what a pain

[keen ruin]

okay after rewriting a good portion of the patch, it seems to be working

just verifying everything and then I will push out an update

[keen ruin]

Version 2.4.2

• Updated the roaming patch to function properly in v81+.
• Fixed an IndexOutOfRangeException thrown when an enemy can't find any paths in ChooseFarthestNodeFromPosition() or ChooseClosestNodeToPosition().

coming soon to mod managers

[boreal brook]

Random question. Is it possible to add a moons blacklist to global roaming in SmartEnemyPathfinding? Pareidolia tends to blow up if it's eclipsed and eat people's frames with it on. Which is sad because that's a moon where I would like the mimics tocome out of all the fire escapes the most.

[vivid dock]

With GlobalRoaming, even if a masked sees a player using the elevator when it reaches the top, the masked has already forgotten the player, and the masked won't call the elevator.

[keen ruin]

oh, like it steps out, sees the player, the player goes down, it forgets?

that's kinda a vanilla thing more than a SmartEnemyPathfinding thing unfortunately

it has a timeout after it loses line of sight

[vivid dock]

Goes up

[vivid dock]

Like the butler?

[keen ruin]

well same for most enemies really

at least the ones that care about line of sight

[keen ruin]

I could try to do that, yeah. Probably a more effective thing would be to set a limit on the number of connections to allow, but even more important would be to see if there are any strategies I can use to avoid as much work for each enemy's path

I feel like our current setup is fairly naive, and I could probably brainstorm some way to eliminate or cache paths to avoid doing so much work

if I can reduce the number of paths tested to O(n log n) instead of O(n^2), that would make an immense difference for the really problematic moons

[tardy pivot]

isn't global roaming async? why is it affecting frames?

[keen ruin]

the jobs for them run pretty long, so if there's a lot of enemies, it's easy to saturate all the job threads on a low-end system

one other thing that might be worthwhile is to make a dedicated thread pool in PathfindingLib for this so we never saturate those

[tardy pivot]

then shouldn't a job queue to cache extra requests work?

or a thread pool ye

[keen ruin]

you mean serializing all the smart pathfinding jobs?

[tardy pivot]

instead of immediately starting the jobs. yeet them into a queue and have something dequeue when a job thread becomes available

but there isn't much that can be done if the player system has so few resources that the queue keeps growing

[keen ruin]

the job system isn't really designed for that, so I think a separate pool would make more sense

that said, I think reducing the work would be much more fruitful sicne the pathfinding is already slow on march

[tardy pivot]

sure but you'd need to invalidate the cached paths when the navmesh changes. or revalidate them each time you try to use them

[vivid dock]

One thing I noticed is that, no matter if global roaming is turned on or off, masked don't try to hide inside the ship with smart enemy pathfinding.