#programming
1 messages · Page 7 of 1
LiDAR better than ultrasonics
but cost
true
first cheap sonar sensor i could find was 10 bucks and goes 254inches far, so about 6m in good units

that would actually be perfect, mind throwing me name of what you found?
High Performance Sonar Module LV-MaxSonar-EZ0 Maxbotix
its just the first one i found, you might find better ones
no problem
this weird function:
viewcorners = np.einsum("lj, jk -> lk" ,frustumcorners, lightview)
is apparently a lot faster than
viewcorners = frustumcorners @ lightview
lemme retime
https://ffmpeg.org/pipermail/ffmpeg-devel/2025-May/343646.html ffmpeg my beloved, 1600 lines of assembly 
how much? i wonder
i honestly dont know anymore how the performance botlleneck works
when zooming out it gets slower, makes sense, more is being rendered.
until suddenly you get a lot more performance
the performance to zoom graph is a curve
today I install gentoo on my laptop. wish me luck
goodluck
I'm not a big programmer person but that would make sense to me, when it comes into view it has to create the terrain, but once the terrain is created it's stored in RAM and way quicker to reference than to recreate every time
all of the terrain is already stored in vram
idk if that makes any sense, a thought would be if maybe you can keep it loaded for a period of time even if it's out of view, so if you're switching back and forth quickly of zoomed in and zoomed out
nvm then :)
I only know basics of programming, so not sure then, sorry
im using a heightmap
nah no prob
you were jsut trying to help
what I described makes sense in my head of the speed being a curve
maybe the vram isn't actually storing it properly?
aight, I wish u luck 
the ultrasonic sensors ended up being exactly what I was looking for, I just ordered 10 of a different brand :)

as well as some extra wire 'cause I don't have the small enough size on hand
there was 1 that the sensor is about half the range, roughly 3m, but each costs about 3$ apiece, as well as the flight controller I have can already work with it
more than enough for what I need, only downside is I'd have to script it in Lua and I am not familiar with Lua 
a problem for next weekend when they come in :)
rip lua
i'm currently going through linux from scratch book 
People who main LFS should be feared
lfs?
Linux From Scratch
I'll make my own distro mindset
il make my own game-engine
ill make my own everything
i always forget just how many tests glibc has 
RIP centos 😭
too many os's
I fucking hate kernel anti cheat
fucking, GTA V and its stupid BattleEye anti cheat crashed and somehow took out my PC's ethernet
Like my ethernet controller is just dead I clean reinstalled Windows and it still shows "not connected" over ethernet wtf
that has nothing to do with kernel anticheat - your controller either works or doesn't
"What is written on Twitch as 'PC Specs?' refers to the Neuro runtime environment, or the spatial environment that Neuro uses to operate?"
not "not connected"
True dude
I did just that—in fact, there was once when I wanted a refund, but ultimately I found it was just the PCIe interface that was dirty
It crashed and my ethernet stopped working
the ethernet that has worked fine for literal years
you should probably check your router then
or whatever
And then GTA V crashes my PC and upon reboot it no longer works
Router is fine
Cable is fine
I plug the cable into my laptop it instantly connects to the ethernet
I plug it into my desktop, nothing
dead
And yet here I am, with Device Manager saying "This device is working" and not being able to connect to my network via ethernet
might have been a limit from your ISP?
My ISP doesnt impose any limits
I think simply, dont give dumb useless anti cheats kernal level access
it doesnt stop cheating, fun fact
the cable is fine
why should anticheats need kernel level access
that's crazy wtf
supposedly, to stop cheating because it can see everything running on the PC and in theory see any cheat programs running
In practice that lasted about a week until people made direct memory access hardware and can just insert modified code directly into system memory without Windows noticing
Hence why Im fairly positive thats what bricked my ethernet, since BattleEye was running inside the Windows kernal when it crashed its possible it broke something
Why my ethernet controller still wont work after a complete wipe and re-install of Windows I have no clue
oh
i liked this video https://youtu.be/lJYufx0bfpw
Head to https://brilliant.org/BreakingTaps/ to get a 30-day free trial. The first 200 people will get 20% off their annual subscription.
Watch this video ad free on Nebula: https://nebula.tv/videos/breakingtaps-the-algorithm-that-embraces-random-chance
Today we're looking at HyperLogLog,...
Yeah its properly messed up and I have no idea how to fix it
guys
I tried making chat gpt write a code
I can't send images for some reason so I'll just paste what it said
ETA: About 2–3 more hours for a fully tested, install-ready .mcpack. I’ll ping you right here as soon as it’s ready to download. Want me to notify you via email or another method too?
Is this normal?
ahh question
The urge to reply "it has been an hour how is it coming along"
I re-flashed my BIOS and my ethernet is still not working :l
ugh
last ditch effort to see if its actually truely dead Ill boot a linux live USB and see if that can connect over ethernet
Otherwise I guess Im buying a new NIC and never touching GTA V again
Google’s AlphaEvolve Breaks a 56-Year Barrier
AlphaEvolve, Google DeepMind’s new evolutionary coding agent, has made history by solving a matrix multiplication problem that has stumped researchers for over half a century. It discovered an algorithm that multiplies two 4x4 complex matrices using only 48 multiplications, beating the long-held record of 49.
This isn’t just academic — faster matrix multiplication can lead to faster computing hardware and major breakthroughs in AI performance itself. AlphaEvolve evolves its own code using LLMs, testing and improving it through a self-feedback loop. It’s already rediscovered 75% of known best solutions in geometry and combinatorics problems — and improved on 20% of them.
DeepMind’s CEO says that solving intelligence means we can solve everything — including every disease — within the next decade.
Ah the double edge sword of AI
On one hand, people are losing jobs to it, which is not a good thing right now
On the other hand, the ability of AI to just take massive amounts of data and find patterns and draw conclusions from insanely massive amounts of information could be a game changer in certain fields
Also our beloved AI streamers are also a perk
Old news
We had it two days ago
Personally don't know because it seems a infinite loop till get verifiable goals achieved
Which means it inherently can only be used on tasks that GRPO could be used to train on
Code, math and such, but not other tasks
Ok
G'morning
good morning stella 
Hi hyper!
gooood morning #programming
good evenign
Self-improving AI is here
https://www.youtube.com/watch?v=X37tgx0ngQE
If it's an LLM, who cares
I honestly couldn't care less about "assistant" style LLMs.
Its not even assistant style
its just if generates its own maths and coding problems then tests them, then solves them
Which is limited to maths and coding first of all
and is also very expensive because of how much retraining you'll need to do
And while it works, generating your own training data, even if it works for a while, is the definition of a positive feedback loop
Also self improving AIs have existed since the 1960s in genetic algorithms
If you want actually functional self improving AI, you gotta get with the times and give up on transformers
anyone have recommendations on how to explain to your parents that no, you cannot just ask ai to explain to you the assignment and then it'll just stick to your brain
because I'm starting to get sick of having to explain them that
Learning only works if your the one doing it, and AI can't even manage that as it has no soul.
unfortunately need something better than that
I get the approrach they are trying, and in certian cases it can work. That is my issue with your parent's opinion as it can generally work if used correctly.
The self improvement will literally 'wow look at this i printed a 2'
you have not even seen the video or understand it
All one can say is "That learning style doesn't work for me". Unless another chatter wants to go down the misinfomation route.
I have the same opinion that learning sticks if you do it yourself
The issue is that it depends on the person
luckily I got them to understand (for now) that if I'm gonna learn something the most I should use AI for is asking "hey, explain this concept to me" AFTER I give it a try myself
Some can learn from example, some can't. That is the line you'll have to defend
No problem
idk I just don't trust llms to tell the truth
why do you think so?
and that makes them very hard to learn from
I mean, I'm not a transformer fanboy but I'm curious
llms are great at gaslighting you into believing that you know something. But take it away again and ask a person to reproduce it on their own and odds are they will fail horribly
Seems like bred went with it, that was fast
Because transformers kinda just sucks for dynamic models, it's only really good for when the model doesn't have to adapt on the fly
I've also seen other much more successful non-transformer AIs that can actually improve over time on their own
Nothing public really exists outside of deep on Discord, the architecture is someone's private one
I mean on "transformers kinda just sucks for dynamic models"
that sounds kinda interesting if so
Honestly I've always thought it is because they eventually freeze
Maybe
could explain better
have to do so to release the model
It has to do with overfitting though I believe
if I'm to guess, based on my current knowledge on llms (which use transformers), the issue is that they get their performance by memorisation
like a kinda fuzzy knowledge base
hello

if clippy yells at me one more time for not using first() I'm gonna lose it 
You know it's right tho
if thats all I want, sure, if I also want the second element it looks both confusing and ugly
add second, third, fourth and maybe we can talk about it
There's two moments with clippy and kind of no in between:
Clippy: do this instead
Response 1: ya know, I hate you for it, but you're right. I should be ashamed I didn't see it myself
Response 2: the fuck are you talking about. Is this a nursery lint? 
but having first() and get(1) next to eachother makes zero sense
Why not do get(0..=1) then

And then you can deconstruct the given slice by assigning it to an array of the right size with named variables
because I want to match on the individual elements
I guess I'd need to see the code haha
But this is a classic case of slapping an expect on it and forget about it
get(0..=1) wouldn't tell me which element is not present in case of a None
might be second, first or both
You can match the slice
the slice is within the option
both [1].get(0..=2) and [1, 2].get(0..=2) just returns None. So you cannot tell whether one or two elements are missing
And what about matching the slice on [first, second, ..] or whatever that syntax was?
And if the things inside are options then you can add that to the match pattern too
Again I struggle to understand how the data you want to match looks like :D
anyway, point being that it sucks to use first when second doesnt exit
I mean true, but what I was trying to say was that if there're potentially more elements, there're other ways of retrieving them

Ah, building macros?
its a lisp interpreter
Ah 
the chain of events that got me here is insane
I finally got SSH working on my VM in VirtualBox so I can launch an instance of Visual Studio Code on it but it won't let me access my host machine's ports inside the VM

where do I actually download tkinter? idk why but nowhere on youtube says how to download it and just imports it into pycharm instead
Uhh, it should come with python already
when I try to run it it says "[Errno 2] No such file or directory"
wdym run, it's a library
well I import it into pycharm as a command, and when I run the code it gives me that error
either way, I don't have it
Hmm, I haven't used pycharm in years so I don't think I can really help with that then 
Damn.
You must have the module though, it's literally part of the python release
Maybe it fails to find the underlying implementation idk
Who knows
Is there more to the error than just that one line?
this is what I'm given
Oh it just tried to run a file in your project that doesn't exist
Interesting
Sounds like something pycharm specific then

I don't even know how tf pointers work in C/C++, now I'm going to recreate memory pointers/address in Blender's native geometry nodes

pointers are just references
ye I know
by how they work I mean their fundamental basics
in the assembly level
You just pass around the address and load it from memory as needed
Goodmorning #programming 

this both sounds cursed and very possible
but like, why?
yeah
its the "in blender geometry nodes" part that gets me
the main issue I have with the concept is you'd probably only have read access
actually I guess you have named attributes now
actually yeah those could work as cursed pointers
my beloved
(I didn't even realise this would be possible ngl)

Why does that pic lose more and more pixels
gon
it is very possible, I already have a theory on communication protocols between objects(node trees) based on a clock(simulation zone, which makes every frame a "step" or "tick"). Every tick, a centered global storage unit will check if other objects requested to store something(by checking if any named attributes have "Transfer" on "1"), then the center unit stores it, calls an async execution for the original object to delete those transferred attributes(happens on the next tick), use the above method to transfer the start address(vertex index) to another global address object which maps IDs(vertex index) to other addresses.
to make raytracer in raytracer but in 3D space

for my idea to work, I need objects to be communicating for their normals, hit points and hit angles, etc. So 
idk maybe
I feel like its a bit overcomplicated but I also have no idea what your specific goals are
I mean, making a communication protocol is the first thing and the only thing I come up with on this issue. If there's something simpler out there, I'll just accept the fate that I wasted 2 days on brainstorming
I don't have much geonodes experience
geometry nodes seem to be running in parallel, so tick-based executions are needed for async and stepped stuff. It's annoying with the limitation as well
at this point I'm just trying to burn my time 
and possibly to farm some videos on youtube
honestly I'm pretty confident whatever you're trying to do would work with semi-regular geonodes
sometimes you just want to do something because you can
there's "doing something because you can" and "doing something because you didn't realise there was a different way"
LMAO using this to represent program flows will be crazy
i love it
chat
just what we needed
The AI music REVOLUTION is HERE! Introducing the NEW FREE AI MUSIC KING ACE STEP that turns TEXT INTO FULL SONGS, lyrics, vocals, instruments, everything!
Say goodbye to paid services like Suno, because open-source AI is finally catching up... In this video, I'm showing you the ULTIMATE FREE AI MUSIC MODEL that can generate entire songs from sc...
local suno

mass slop generation
and mass song scraping now
why would I want to listen to ai music?
mass generated
why do artists still exist
ai is the future
chat
should I support linux for desktop application in flutter
this is a very important question
depends, what's the project and how much work is it to support it
that preview image gives me cancer
Linux users usually give you the best bug reports :D

*The first element of the table is the size of the array
I read something about this a while ago. Company makes a thing available on linux and their bug reports skyrocketed. but turns out about 95% of the reports weren't related to linux issues itself
so if you can deal with the remaining 5%, it's probably worth it

what
for memory
do explain :o
so i need to convert each example into it's KV cache representation
and that is 3gb
for each example
and i need ~640K of these
that's probably going to take more than my entire life's internet data received times 10
uuuuh... opposite of ram disk and then SSD goes brr?
oh right mmap
oh
its 1.92 petabytes
nevermind
so what make it infeasible
rn I'm thinking reroll for chance of extra mutation
usually if you mutate them all you have a chance of collapse. but if you keep around k top phenotypes you should be fine
I'm following the NEAT paper
oh you meant several mutations on a single genome
kinda
yeah
yeah have you checked the date it came out? we've learned a bit since then
hahaha
I only read the paper and then implemented my own genetic algorithm from what I understand
and it works quite well
nice
can send you a link if you want to check it out
sure, why not
haven't touched it in a while
what are the modern advancements anyways
I know of HyperNeat
and Neat with self attention
https://gitlab.com/bit-refined/meiosis
I mostly used it for neuroevolution, but it's written in a way to support generic genomes and phenotypes
how you evaluate fitness is still up to you
rust, huh
:D
i would need a gazillion terabytes

void EvaluateGeneration(Task& task) {
for (auto& s : species_population) {
for (auto& g : s.genomes) {
g.fitness = task.Evaluate(g) / s.genomes.size();
}
}
}
skill issue just get more memory 
download more ram
I recommend checking out the examples at first
gives a good insight into how I designed this
will do
cause I usually write the stuff I wanna do first, and then build my API around that
struct Algae {
age: usize,
// Our little algae has no other properties apart from its "brain" that require genetic encoding,
// so we can use the recurrent phenotype directly.
brain: Network<4, 5, f64, Recurrent<f64>>,
energy: f64,
genotype: NetworkGenotype<4, 5, f64, RecurrentType>,
}
Ah yes, algae 
but you can see, even the neural nets I implemented are generic over the weight type
smart
mine are hardcoded rn
I'll first test it before I make it generic
yeah way smarter
the issues I ran into were fun tho. "oh this is generic, guess I need yet another trait..."
Sometimes i just read clickbaity news for the funny comments
what are traits, rust thing?
eeh.. interfaces?
ah
what are the comments on it?
that's a lot of ::

I think I'll put the project down for today
I'm starting to go delulu
tomorrow!
it shall be continued
made good progress though

konii is slowly gathering fingerprinting data of us 
sounds like the ideal programming assistant
rubber ducky assistant. just never have it answer
true
idk programmers are pretty good rubber ducks, even if they don't understand what you've written they can look at it an go "what does this do" and suddenly you realise you're stupid
Literally me yesterday
I usually explain the issue to a coworker and that resolves it without them even having a chance to talk back
yeah that's one option
occasionally I manage to get some words in whilst getting used as a rubber duck
Also does anybody know how to prevent the kernel from always killing my processes?
It always complains about my 20gb ram usage and kills it
use less ram 
🤯
I mean it just means you ran out of ram. not a lot you can do but use less
buy more ram 
download more ram
hm
pretty sure the kernel doesn't consider swap when triggering OOM
cause it likely couldn't switch stuff to swap before RAM filled up
That's pretty stupid
no it's rather smart because swap is very slow compared to ram
No yes ofc. But if it's going to basically die then it shouldn't care
causing a crash is a lot worse than slow archive data
idk is there anything in the system logs about why it was killed
kernel SIGKILLED it
idr how much info there actually is
likely just lists the process that was killed and thats it
I think it's memory leaking 😂
what makes it condescending
it's not closing the files
yeah, process uses too much, kernel kills it in its entirety. managing assigned memory efficiently is on the program. kernel doesn't give a fuck what you do with the memory unless it runs out of physical space.
give it googly eyes
Yes ik
you should see my solution from before I had a real fan haha
oh im kinda dumb it's loading a lot more than 8gb of text... it's loading like 40gb
I assume that's ooming because it can't get 40GB of contiguous memory?
I got 32gb swap and 16gb of physical
I put the air intake into a cool room and pumped it into hot rooms 
though I thought that would be a allocation issue not an oom killer one
my system only uses like 2gb max for other processes
cruel
😔
Fs
you could display all sorts of things on its body because its made up of pixels :D
get 41gb of ram, simple
redeems to "hack" the display
:D
I mean could be set things like the redeem idea bred has
not "arbitrary user input"
that's pretty cool looking
is that RGB or BGR 🤔
looking at the gaps rgb
Are you sure that's enough? Loading 40gb can and will take more than 40gb
It's worked before just randomly
just mmap the whole file
sometimes it won't, sometimes it will
Unironically yes
probably due to external memory usage from the os
Add more swap just in case
white - cyan
blue blue
blueish cyan
It's partitioned on a full drive
atp all I can do is pray that my system doesn't decide to start caching everthing into ram or something idk
remind yourself that you're not a fish
it works 100% of the time 50% of them time

then suffer >:c
this should help
There are many situations in the kernel that won't allow it to sleep in order to swap memory for itself. If you don't have any direct memory available at that point the OOM will kill something no matter how free your swap is
is very cute emoji I don't see the issue :D
Ah so basically if the kernel can't get enough memory for itself then you get killed?
yea
yes
this programming business sounds very violent ;-;
and thanks to virtual memory your program can allocate terabytes of memory cough electron cough, but only when physical memory becomes too little to continue operating the kernel will it start killing the biggest users of memory to sacrifice the minimal amount of programs

haha in my mind I heard the "correct answer" sound neuro can make when those emojis popped up hahaha
you can change oom_score_adj to try to discourage the kernel from killing your process but then all best are off whats going to die instead
luckily it also considers process trees, or else systemd being killed would be very bad :D
imo we should consider killing the kernel before my processes
what good is the kernel anyways
surely that'll go great 
On the x86 computer architecture, a triple fault is a special kind of exception generated by the CPU when an exception occurs while the CPU is trying to invoke the double fault exception handler, which itself handles exceptions occurring while trying to invoke a regular exception handler.
so its exceptions all the way down
where is my quadrouple fault handler
thats when your cpu just catches on fire
I am the quadrouple fault handler
how did you actually make this I'm mildly curious now
ic
I couldn't tell if it was a render or simulation or something, but I was just impressed by its representation of the blooming in the blues 
but if its based off a real image then it makes sense that the blues look like they have real bloom
at some point I kinda want to check if discord supports non-srgb colour profiles for profile pictures
iirc slack allowed hdr emojis at one point
my beloved
what the fuck
I need this, now
if you don't have HDR enabled, do the two emojis look the same?
probably
it really depends on how your system and browser stack handles colour
I think that filter stack is deepfrying it a little
lemme check
god I love wget when you're downloading tiny files
yes surely I am downloading at 200MB/s

(thank you chat for giving me a distraction from revision once again)
first image might be less deepfried
well
second now
thanks discord
for some reason djv refuses to open those, only my browser will
doesn't even display
download the image and drag it into your browser
looks normal in my browser
I was gonna say it looked weird in mine but its probably getting messed up somewhere in the remote desktop pipeline
which browser / are you using hdr?
Orig looks washed, alt looks normal haha
firefox and yes HDR is enabled
ah
I even have the brightness up 
ok as much as I love firefox anything to do with colour is not a good idea
😭
oh god hdr on linux too
haha :D
istg never touching windows related stuff
not that OS, but just like app window
yeah best wishes
I mean, works for me
trust me windows is actually really sane when it comes to colour in my experience
nah not the OS mb
oic
it's just like app windows
just never use anything other than standard srgb for images and you'll be fine
there are so many exceptional cases that I am not able to handle
but anyway it kinda works
Look at the quality of the background image. Looks so good
oh camera is shit apparently
oh wait for clarity the _orig one is the one from the article
oh well :D
not the unmodified image
also I noticed that HDR over HDMI looks different than over display port
but oh well :D
ima just rest for a bit and turn on unity
what is this, tiktok? 
yeah they showed us at least 3 videos like that in the lecture
anyone knowlegable about linux here currently?
how knowlegable are we talking
enough to answer wtf is going on here
Bingo not working an hour before stream 
nothing is working an hour before stream on my end 

welp, theres your answer 

now to figure out how i can ssh into it without it being able to ping google.com
that's even better, it's dns resolving but not sending network traffic 
which is very weird

because dns itself is network traffic
restart the machine and pray it comes back up
(I have made this mistake many times when remote desktopped)
it's almost like a reboot broke this to begin with 
looks like it's a Docker container?
might be Docker messing with the firewall then
docker could be messing with the firewall, true
what's your host os?
fedora i think?
ic
oh god
no idea how fedora works ngl
barely

fedora uses firewalld and docker just does random nftables shit on top of it. It can lead to..... interesting situations
wait i should check selinux to see if it enabled itself
if routing goes down while you were connected, usually you can stay logged in
logging out and back in is gonna be an issue tho
podman works better on redhat systems since its a redhat product and basically a dropin replacement for docker
yeah, Docker is notorious for fucking up firewall configurations in various ways
idk how to fix it though, I haven't encountered it in quite a while and I use podman now
DNS working is also weird
solution is to reinstall and start from ground up
surely i can do that in the 1 hour before stream and have time to configure the bingo

why is this machine slow
i'm gonna go take a look at it and it's gonna be literally on fire isn't it
well, either way it looks kinda cooked 
surely you can fix it in like 5 minutes 
i blame the people who installed a chunkgen mod on it and then decided to chunkgen like 1 million chunks 
oh that's hilarious
so you are missing internet connectivity within docker but its all fine on the host system?
i got a response from cloudflare dns

here's the fun part
the whole machine is failing at sending requests EXCEPT for DNS requests
but the dns is cloudflare which is external.... thats odd
yep!
what happens if you like curl cloudflares DoH server
I would say maybe your tcp is fucked since dns is udp but your ping is also not working and thats even lower in the stack with icmp
curl --http2 --header "accept: application/dns-json" "https://one.one.one.one/dns-query?name=cloudflare.com" --next --http2 --header "accept: application/dns-json" "https://one.one.one.one/dns-query?name=example.com"
oop
no response
what does ip route show say? Maybe something killed your default route
wait that's two requests
default via 192.168.0.1 dev enp2s0 proto dhcp src 192.168.0.41 metric 100
10.0.0.0/24 dev wg0 proto kernel scope link src 10.0.0.2
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
172.18.0.0/16 dev br-23723b24b7ad proto kernel scope link src 172.18.0.1
172.19.0.0/16 dev br-e627f4b2c93b proto kernel scope link src 172.19.0.1 linkdown
172.20.0.0/16 dev br-98bd4f004a46 proto kernel scope link src 172.20.0.1
172.21.0.0/16 dev br-c69d517469d2 proto kernel scope link src 172.21.0.1 linkdown
172.22.0.0/16 dev br-57fc1c14c9a7 proto kernel scope link src 172.22.0.1 linkdown
172.23.0.0/16 dev br-0ccfb4d6cdcf proto kernel scope link src 172.23.0.1
172.26.0.0/16 dev br-472d28bda5c8 proto kernel scope link src 172.26.0.1
192.168.0.0/24 dev enp2s0 proto kernel scope link src 192.168.0.41 metric 100
@stiff mica
oh god docker routing I hate it so much
you could listen with tcpdump on the main interface to see if the packets uses the right way
I mean on the first look the routing looks okay... weird
very dumb question but have you tried just restarting it? :P
yes
oh
and a restart caused this to begin with, so whatever is going on, it's been broken for a WHILE
damn
and just didn't apply until restart
and have you isolated it to this host specifically?
cause I had a similar issue once and turns out it was my OpnSense firewall :P
i am using discord on a machine that's connected to the same network
hmmm
maybe look thought the dnf history to see if someone installed something weird
can you freely talk to local devices?
also audit log for selinux stuff
thanks selinux
here's the fun part
smh my head
since dns works can you ping the ip of the dns server?
heck, local devices which can directly interact with home assistant installed onto the same machine are behaving properly
right, but outgoing from the device? like say ping your firewall or something else on your network?
just trying to isolate the issue, sorry if I'm asking annoying questions :D
pinged this machine
it seems to be specifically outbound connections
which makes me think..
oh why is my router refusing connections
interesting
yeah that sounds like a firewall/router issue
try to stop the docker service and see if that fixes the outgoing connectivity. At least that might narrow it down
do you somehow have a misconfigured gateway address or something?
idr much about networking
to have routing tables and any other tables return to "normal" I assume?
heh
it's just weird
yea that would shutdown all the weird bridges
they shouldn't really matter tho
but they do sadly, nftables does a lot of internal routing
oof
docker is stopped but docker shall continue running, just to fuck with you
there's only one thing I'm slightly confused about
is the ip route show output deterministic?
the socket should be ok if all container are stopped
all the docker routes are still there
cause I would've expected host routes to come first, followed by docker stuff, and not have it be in between
this machine might as well be cooked
ok, stop docker, disable the service so it wont autostart, and reboot. that has to kill all the bridges
in theory....
ps aux | grep docker and then kill -9 <docker pid>? sounds scuffed btw
anyone know about this? @noble zodiac? Cause the only important thing for ordering is the metric right? doesn't matter what I check, the route that is listed last in cloudburts output is always below the default route for all of my servers. so just trying to verify this isn't anything.
i'm
about that wg0 network
you could potentially take down wireguard
well, what are you doing with wireguard? :D
I always forget how wireguard routing works
I think so but the internal firewall might still block stuff inbetween
the machine is still refusing connections
how does it show in ip link, is it down or up in there?
it's been a bit too long for it to still be doing that
oh boy
i'll tell you when the machine cooperates
😔
for reference, here's my NAS
default via 10.0.0.1 dev br0 proto dhcp src 10.1.0.0 metric 1005
10.0.0.0/8 dev br0 proto dhcp scope link src 10.1.0.0 metric 1005
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1
and my second server
default via 10.0.0.1 dev eno1 proto dhcp src 10.1.0.1 metric 100
10.0.0.0/8 dev eno1 proto kernel scope link src 10.1.0.1 metric 100
172.17.0.0/16 dev docker0 proto kernel scope link src 172.17.0.1 linkdown
172.19.0.0/16 dev br-5f688005dbc1 proto kernel scope link src 172.19.0.1
172.20.0.0/16 dev br-09003b30cdbd proto kernel scope link src 172.20.0.1
172.21.0.0/16 dev br-80528a4db2da proto kernel scope link src 172.21.0.1
172.22.0.0/16 dev br-2bafb91f31c2 proto kernel scope link src 172.22.0.1
172.23.0.0/16 dev br-682b3a166f96 proto kernel scope link src 172.23.0.1
172.24.0.0/16 dev br-d8d0b3c48115 proto kernel scope link src 172.24.0.1
172.25.0.0/16 dev br-4dc065c070ee proto kernel scope link src 172.25.0.1
172.26.0.0/16 dev br-8841971fdefc proto kernel scope link src 172.26.0.1
so as you can see, main route is always directly after the default route.
But if this doesn't matter then idk. Have you tried attaching wireshark?
eeeeh
I wonder if you're routing your traffic to that peer or something weird
wasn't there a command to see what route is being used for a certain type of request?
got it
shouldn't matter
I think it's just ordered by the network prefix, with the default route always being first (makes sense since it's the same as 0.0.0.0/0)
ip route show to match try this, followed by a public IP
this will give you the route being tried
that's neat
I really want to learn how the ip command works beyond just ip addr and ip link lol
yeah working with critical infrastructure I've seen lots of fucked up networks customers have set up :D
(ip is a public server, no worries about it being leaked)
luckily Cloudburst doesn't use HSR with Lane A and Lane B traffic 
it's going the right way, but something's stopping it
chuck a traceroute at it I guess?
atp a reinstall would prob work better 
but might as well figure out what's going on so it don't happen again
nope
this also works apparently
ip route get 1.1.1.1
and it tells you where it pulled the info
oh replace the IP I guess

i think it's works without sudo
why's it using wg0 lol
hahaha
but I guess that would explain the issue
how the fuck 
:D
lmao
I guess but isn't it literally just saying "you should've used root so we're gonna ask you for it instead"
it was wg0 this whole time 
my guess is you had wireguard configured to vpn through a peer and now that peer is down
before you go can I add you as a friend
:D
i mean
it's uses sudo anyways
i just kinda blew it up with nmcli
no idea why dns was leaking that though
it's too late for me to configure bingo i think tho

arguably it was doing the right thing
I do wonder how it can have a peer tho if it was set to "unknown" as a state lol
hell yea, go get it
last minute setup my beloved
holy hell wait, if it always went over wireguard, does that mean all outgoing internet traffic went over the tunnel? hahaha
poor little endpoint
should probably hang it the other way around 
99% of prompts now shall have this as the explanation
No explanation provided, most likely the prompt is self-explanatory or Cloudburst is a lazy bastard
Ye its upsidedown, the bottom has buttons
I know I was making a joke
god I'm so bad at humour 
the giraff tho
I just have autism

what kinda 3d does it have
You need those glasses
ah so its likely the polarisation ones yeah
dual play mode sounds like a neat concept tbh
One eye is polorized horizontaly, the other vertically.
Or player 1 horizontal player 2 vertical
It is imo, just never executed correctly
Linus tech dude made a video about these
Im suprised i managed to find one
These are rare
Its not the expensive one in the video exactly, but it has the same features
This one is just not oled nor 240hz
That's actually crazy
Unfortunately I still haven't managed to grab a 3090 for training NeuroSynth on
wait I remember uncle snoek (in the VC) also had some in his room. is this a netherland thing? lol
I had a talk, and apparently it's not
just a weird coincidence :D
Idk
Im not in the netherlands
Ah
oh then I'm just stupid
my memory isn't the best :D
Dementia 
define.. "imploded" :D
it deadrd
yeah but I'd rather just explicitly use sudo myself
I mean I figured haha
this time it's not my fault tho
I really don't like programs that interactively handhold when you do stuff like that
taking its sweet time
it doesn't
$ ping nwero.net
PING nwero.net (104.21.64.1)
64 bytes from 104.21.64.1: icmp_seq=1 ttl=55 time=3.56 ms
64 bytes from 104.21.64.1: icmp_seq=2 ttl=55 time=3.64 ms
I can ping it
plenty of IP's behind the DNS entry tho
cloudflare
well, with that many IP's I dunno how it selects the right one
lemme check them all
they're all cloudflare, it'll get where it needs to go
from what i can see internally, everything should work
i feel like this might be cloudflare s fault now
it could still be your fault
nvm my router is on fire
yeah they all work
I've had this previously when I had my backend ip set wrong
no idea why it times out instead of failing fast but oh well
Rip router
oh okay it's swapped places one sec
quick 8 min adventure of setting up opnsense? :D
router restarted itself
that moment when init runs out of memory 
just checked my firewall and it's using 22GB of memory for ARC. what the fuck
zfs on a firewall? 
arc?
default for opnsense
zfs cache basically
I mean fair enough if its not being used elsewhere
just struggling to understand what the fuck it could be caching
surely its reclaimable
I way overspeced my firewall
I wonder why they do that, seems incredibly unnecessary given that it shouldn't really serve anything from the disk
should only need to write logs and statistics and all that can't be cached anyway
worth it tho 
ok whatever's happening, it's probably cloudflare s fault
you can do more on opnsense than just firewall stuff
check where your dns records are pointing in the cloudflare dashboard
ik where they're pointing
as long as you've checked they're right 
I know that you can, it has tons of plugins and stuff, I just don't see why you actually would use OPNsense for that instead of Proxmox or TrueNAS Scale or something
maybe I'm just underestimating it 

site works
oh in general zfs? yeah idk. does it prevent bitrot? may be a reason

@rough bloom idk why the think emote, but bitrot is a concern for non-ECC systems haha
well, and any normal hard drive
I know btrfs prevents it, but unsure about zfs
ZFS will catch and repair bit rot transparently
yeah, I'd guess so, just not sure if/how zfs would help there
yeah, the likely reason then
thats actually pretty neat then
I found it, itdoes some wonky interlacing
Pretty cool
now you can watch [Filtered] the way it was meant to be watched
What?
wait nvm misunderstood how it worked
You need the glasses yo only see 1 of the 2 parts
Without you see both at the same time
oh right, OPNsense uses FreeBSD which doesn't support btrfs but has really good support for zfs I think?
makes sense why they're using it then 
So i wouldnt watch [redacted] on this. Very sus bred
You need ones that can switch between the 2 sides
Relative screen space SMOL
Cuz 3d is mixed per glasses, dual play is one direction per glasses
Oh ye
what scaling is windows running on that
who even pinged me
@stark needle
#announcements? :D
300%
NEVERMIND IT WAS THE STUPID ANNOUNCEMENT PING
I HATE THIS


I forget that people get those pings

tbf I've messed up my notif settings here pretty bad
the channel is muted + hidden + disabled all mentions
Is the Ai learn over time or I am just dumb by asking that
I don't get the notif sound if people ping me, just the icon
generally no
I always run 100% (anything else breaks stuff)
300 is insane for a tv
tbh on 11 most stuff runs fine unless you're using obscure software
I'm talking stuff like Pxls or Neurolings
Those things hate scaling != 100%
Also I'm on Windows 10 and never touching 11 on my own hardware
pxls was fine for me but I can see why neurolings would struggle
11 is fine. Just not great
idk why you're complaining about scaling on 11 if you don't use it
I consider it unusably bad
Lot of little things stacking to make me not even consider it as an option
I'm probably gonna go to Linux Mint or something next
It's been almost 4 years Microsoft, where is my windows 12 
I truly can't wait
Last I heard 25H2 is supposed to be ai focused
So that's fun
Don't quote me on that though
The ruin everything update part 2
24H2 was already bad enough, 25H2 is gonna make it even worse
I forgot what 24h2 was even about
I don't even notice 90% of the major updates
as long as my desktop looks identical I don't care really
I still look at the changelog when I catch there being an update
Never know if there's anything interesting
ngl I've never looked at the changelogs
I guess making the OS as garbage as possible to force people even more to have a Microsoft account
Hello fine people and goeie dag
I already had a ms account for other things so this never really bothered me
Thanks for the input last night @olive sable ,it helped me a lot with my direction 🍻
Ah cool because I didn't change anything 
same here tbh
I think the only place I use my Microsoft account is Xbox services though
Lol jokes, nah it helped me widen the audience a bit which is a nice add to my prototype
okay looking at the actual changelog for 24H2, most of those changes don't even apply to anything that isn't a copilot+ pc and the rest is just eh whatever
there's a new ai components menu that popped up in the add or remove programs part of settings
idr when that appeared though
Except maybe stuff like breaking all the things
Like audio and other drivers
yeah nothing broke for me either
interesting
I've heard a lot of people got audio issues and other drivers issues with 24H2
that sucks
my decently scuffed audio setup works fine 
windows in general has the tendency to reset its audio subsystem after major updates
Never had problems with audio
yeah for me it resets itself sometimes after updates (though hasn't happened in a while)
Everytime i join a discord vc, or even fortnite parties, i need to turn it off and back on
it hooks itself pretty deep into windows
huh
i've never had that
are you using it directly or via peace




ye






surely it works

