🪅-progaming

[strange matrix]

@fleet cedar i have a question about importing libs

is it something that usually devs do that they bring the function from the libs they use and just include them in their own code
to like cut down size?

like

if i'm using printf from stdio

should i just copy that and put it in a separate file because maybe i'm not using anything else from stdio

[fleet cedar]

The benefit with using code by others is not having to write it yourself

For complicated things like printf, that's a pretty big benefit

Also yes, the resulting executable becomes smaller because the implementation is in a shared dll

[median root]

Does statically linking then put the whole binary in or just used functions?

[strange matrix]

uhhh

i was more asking

can i just steal printf from stdio.c

and keep a copy

so my statically linked exec doesn't become massive

[fleet cedar]

Sounds like a terrible idea but go ahead

[strange matrix]

cause i've seen

like

there was someone

doing a hobby project

making the smallest linux or something

its on youtube

i think they did something like that

and i was wondering how practical it is

[fleet cedar]

Copypasting functions makes your exe bigger, not smaller

[spark tiger]

why does this fucking shit show my full name there

[fleet cedar]

To be creepier than necessary

[strange matrix]

wouldn't statically linking do that too ?

[strange matrix]

Hi, King Charles The Third, how can i help you?

[spark tiger]

i'm good

ok doesnt seem to show it aynmore thnakfully

[solemn ravine]

@woven mesa I may get back to working on clipboard app after I get my location simulator thing working

because after that I inherently have nothing to do lol

apart from experiment with new things

and I feel a lot more experienced now than ever

[woven mesa]

same!!

[woven mesa]

omg

[solemn ravine]

but also I'm thinking of switching to coredata

ikik scary

[strange matrix]

ahem
do not put your execvp() in a while(1) loop

lesson learned

[fleet cedar]

Eh, doesn't matter

[strange matrix]

not when my laptop hard crashes

[fleet cedar]

It'll still only be executed once

[strange matrix]

no its not

[fleet cedar]

Unless it fails

[strange matrix]

i had in one instance 1k instances of bash

[fleet cedar]

execve kills the current process and starts a new

[strange matrix]

wha

[solemn ravine]

swiftui was just slowing the app down ☹️ so I got demotivated on it

[strange matrix]

look heres what happen

i do pkill blood (my prog name)

and then it started making bash shells

so

jus don't put execvp in while(1)

i just cpu bombed myself i guess

poor laptop from 2016

[fleet cedar]

That's not how execve works

Show code

[strange matrix]

oki

lemme

[formal belfry]

also, is there a way to hide specific guild tags?

[strange matrix]

oh and wait

right

right

you're correct

i just found out my mistake

but lemme show you nontheless

[woven mesa]

lets use that one coredata wrapper i figured out how to use it

[strange matrix]

  while (1) {
    pid_t pid = fork();

    if (pid == -1) {
      perror("error");
      return;
    }

    execvp(comb_str, NULL);
  }

cancer code

[fleet cedar]

Well yeah forking is a different matter

[strange matrix]

don't be mad at me im stoopid

[solemn ravine]

which

[fleet cedar]

Why is whitespace a token

[solemn ravine]

you use coredata??????

[lavish frigate]

why not

[solemn ravine]

since when

[strange matrix]

prolly for a seperator

the seperator

[fleet cedar]

Because whitespace is not significant in most languages

The point of a lexer is to split into meaningful units

[lavish frigate]

hmm

[strange matrix]

right

[lavish frigate]

now that you mention it it is kinda smart to leave whitespace out

[strange matrix]

wait

WAIT

FORKBOMB

i forkbomb myself

that's why they call it a forkbomb

[solemn ravine]

I also may ditch the bottom panel entirely

[strange matrix]

holy crap

[solemn ravine]

in favor of a floating one like windows

[strange matrix]

+1 iq gained

[solemn ravine]

lc.i windows clipboard manager

[visual shell]

6 clipboard managers for Windows that will change how you work

Google Images • labsCore • Page 2/15

[solemn ravine]

maybe something similar actually that's kinda ugly

eh whatever I take back what I said about removing the panel

[lavish frigate]

@fleet cedar ty actually i removed the whitespace token now my parser code is at least 10x more readable (yes i did just ignore whitespace tokens inside the parser)

the

if matches!(stream.peek(), Token::Whitespace) {
  stream.next();
}

[woven mesa]

i dont but the wrapper is nice

i think its called grdb

[hoary sluice]

well, are you trying to do that too?

[strange matrix]

nope

making a shell

more discouraged after forkbombing myself

[lavish frigate]

😭

[strange matrix]

👍

[woven mesa]

i wanted that too but i wanted it to appear on the text caret

[hoary sluice]

then why does ur binary need to be statically linked and why does it need to be small

[strange matrix]

i want to ask that too

my brain

eh

dunno

[woven mesa]

i statically link my binaries only on linux

[hoary sluice]

why

[woven mesa]

why would i want end users to install swift toolchain

[hoary sluice]

unpatched nix support

[fleet cedar]

Why are you writing a shell in c of all things

[strange matrix]

i literally don't know

i wanted to pick something up

and i don't know

learn something

[hoary sluice]

write a compiler

[strange matrix]

turns out

not a good idea

[strange matrix]

for ?

[hoary sluice]

new language

[strange matrix]

so do what pyro is doing ?

[hoary sluice]

whos pyro

[strange matrix]

pyra**

mb

[hoary sluice]

whos pyr

a

[fleet cedar]

Or old language

[hoary sluice]

meh

[fleet cedar]

Resurrect cobol

[strange matrix]

isn't cobol already being used

?

by banks

or like

[hoary sluice]

oh chloe

[strange matrix]

financial systems?

[lavish frigate]

im not writing a compiler i am writing a new shell i told u before 😭 but its gonna have a language "inspired" by rust

[hoary sluice]

not already, still

[strange matrix]

you're making a "sh" clone

[lavish frigate]

no

[strange matrix]

well yeah

[strange matrix]

mmm

you just have basic shell funcs then

if i could figure out why execvp kills my original process

i feel like

i could make stuff way easier

for myself

[fleet cedar]

Because that's its whole thing

It replaces the current process with a different one

[strange matrix]

its yt time to see how people are running threads with C

[lavish frigate]

its gonna be a mostly expression driven language, im thinking of a syntax similar to rust but like:

let something = (echo "test");

if something == "test" {
  git commit -m "a";
}

i think thats pretty cool

[woven mesa]

@solemn ravine https://github.com/groue/GRDB.swift its actually quite nice

[strange matrix]

gerdub

what a name

[strange matrix]

made an alias

oh no another rust video
part 27: Ownership is cool in rust

im

should i already just switch to rust? @fleet cedar

like am i torturing myself?

[lavish frigate]

we are all using rust here

[fleet cedar]

Yes

[fleet cedar]

Up to you

[strange matrix]

😭

[strange matrix]

well hey

don't do that

absentee father what are you
just give me an answer

[woven mesa]

not me

[fleet cedar]

I'm an internet stranger

[lavish frigate]

you too

[woven mesa]

nooo

rust horror

swift is cute

[strange matrix]

swift has a massive list of dynamically linked libs

[lavish frigate]

swift is basically just apple flavoured rust

[strange matrix]

rust first-time compile takes 20 minutes to even wake up

[woven mesa]

swift is readable rust

[lavish frigate]

they have extremely similar syntax 😭

[hoary sluice]

sure but you should probably read up on how memory works first

at least a little

rust will throw you off otherwise

[woven mesa]

i dont mind rust but some of the syntax makes me explode

and also some of the naming conventions

[strange matrix]

i am following rustfully's explanation

so

[lavish frigate]

like?

[strange matrix]

i don't know i just feel like i'm weak not knowing how to make stuff work in C

[lavish frigate]

who gaf about c in 2025

[hoary sluice]

i probably cant either

u dont need to

but u should know what malloc does

[shrewd canopy]

Its not cute on windows

[hoary sluice]

nothing is cute on windows

[woven mesa]

horror

[woven mesa]

TRUE

windows sucks ass tho

[lavish frigate]

swift is a pretty good lang i think it should also be used for apps on windows & linux maybe then native apps wouldnt be so painful

[woven mesa]

that happened kinda

[hoary sluice]

replace swift with rust

[lavish frigate]

I hope Microsoft tries to make another api for native apps and fails that would be really funny I think

[shrewd canopy]

Suggest normal alternative to Windows where 80% of games work fine

Lol

[hoary sluice]

i hope microsoft files for bankruptcy

[hoary sluice]

linux

[lavish frigate]

Lmao!

[woven mesa]

linux

[hoary sluice]

and its way more than 80%

[woven mesa]

99.99%

of all bacteria

[lavish frigate]

Linux isn’t even close to 80% what are you on 😭

[woven mesa]

yea true

[hoary sluice]

yea its more like 99

[woven mesa]

i cant think of a single game ive ran on linux

[lavish frigate]

There has been a huge push cuz of the steamdeck but it’s maybe 50%

[fleet cedar]

Yeah 98% isn't very close to 80%

[hoary sluice]

everything that doesnt have kernel ac works

[woven mesa]

anything that has kernel ac isnt a game

its a rootkit

[hoary sluice]

yop

so 100%

[shrewd canopy]

Bro is glazing linux too much

[woven mesa]

i prefer macos

my server runs macos

[brisk bane]

                    'c.
                 ,xNMM.
               .OMMMMo
               OMMM0,
     .;loddo:' loolloddol;.
   cKMMMMMMMMMMNWMMMMMMMMMM0:
 .KMMMMMMMMMMMMMMMMMMMMMMMWd.
 XMMMMMMMMMMMMMMMMMMMMMMMX.
;MMMMMMMMMMMMMMMMMMMMMMMM:
:MMMMMMMMMMMMMMMMMMMMMMMM:
.MMMMMMMMMMMMMMMMMMMMMMMMX.
 kMMMMMMMMMMMMMMMMMMMMMMMMWd.
 .XMMMMMMMMMMMMMMMMMMMMMMMMMMk
  .XMMMMMMMMMMMMMMMMMMMMMMMMK.
    kMMMMMMMMMMMMMMMMMMMMMMd
     ;KMMMMMMMWXXWMMMMMMMk.
       .cooc,.    .,coo:.
lakhanlothiyi@Lakhans-Mac-mini.local 
------------------------------------ 
OS: macOS 14.6.1 23G93 x86_64 
Host: Macmini8,1 
Kernel: 23.6.0 
Uptime: 151 days, 20 hours, 47 mins 
Packages: 132 (brew) 
Shell: zsh 5.9 
Resolution: 1680x1050 
DE: Aqua 
WM: Quartz Compositor 
WM Theme: Blue (Dark) 
Terminal: WarpTerminal 
CPU: Intel i5-8500B (6) @ 3.00GHz 
GPU: Intel UHD Graphics 630 
Memory: 6262MiB / 8192MiB 

                        
                        




[woven mesa]

tf is using that much ram

[hoary sluice]

not hard to do when the alternatives are windows (garbage) and macos (good but its proprietary)

[nimble bone]

iutl if you have issues with linux gaming just play good games

[fleet cedar]

Loolloddol you say?

[woven mesa]

?

[strange matrix]

rexim does

[hoary sluice]

if you have issues running valorant on linux, ask yourself why youre trying to run valorant

[woven mesa]

idk what the hype is around valorant

why do everyone play it

all my frens play it

[hoary sluice]

the only good shooter game is cs and guess what it works on linux

[nimble bone]

ask @solid gazelle he spends hours a day on val

[woven mesa]

i hate bad mac software

[hoary sluice]

cant play fortnite on linux 😭 😭

[woven mesa]

me when i leave bad software running on my server mac

[hoary sluice]

mac server is crazy

[woven mesa]

fn is the only game i’d miss slightly but even then i havent played it in months bc i only ever play it with one friend of mine

[woven mesa]

nop

its cozy

[hoary sluice]

i have played fortnite once, for 10 minutes

in season 1 i think

[woven mesa]

i only started in s3 or s4

[fleet cedar]

Another win for linux

[hoary sluice]

yop

[woven mesa]

how long until epic releases mac releases of fortnite again

how did epic do anticheats on macos

[pseudo sierra]

according to protondb numbers there's somewhere between 15-25k games that do work on linux and only ~400 that are confirmed to not work so that 99% of games working is pretty accurate

[woven mesa]

@pseudo sierra

[pseudo sierra]

and before you ask yes over half of the games that don't work are cuz of ac

[woven mesa]

games with kernel anticheat arent games

they’re rootkits

so they dont count

[hoary sluice]

what are the other half

[woven mesa]

spyware

[hoary sluice]

its very hard to make a game not work on linux

unless you add a flag that crashes if ran on linux

in every file

[woven mesa]

just wait until i compile for ia64

[hoary sluice]

does windows work on ia64

[fleet cedar]

Or just the main function

[pseudo sierra]

random bs or ancient games that use really obscure calls that behave weird in wine/proton

[woven mesa]

its intel itanium

[hoary sluice]

well you can always patch it out

[woven mesa]

how do people manage to use “obscure” calls

guhh

[brisk bane]

Ayame Stats

Bot Started

<t:1748736387:R>

CPU Usage

0%

Memory Usage

1.5 MB

Swift Version

Swift 5.10

Jun 1, 2025, 1:29 PM

[woven mesa]

@hoary sluice

[hoary sluice]

if half the game relies on access to the windows kernel then its gonna be hard to just patch it out

[woven mesa]

fr

[hoary sluice]

1.5 MB is crazy

a barebones openbsd install uses 10

[fleet cedar]

At that time things were often done with raw asm

[hoary sluice]

if ur game is written in ia-32 asm its not gonna work sorry

[strange matrix]

who

has coded in ia-32 asm

in their life

well actually nvm

beside some people

who has written a game in ia-32 asm

  // code below is from an llm + me. don't trust it

  if (pid < 0) {
    perror("fork failed");
    exit(EXIT_FAILURE);
  } else if (pid == 0) {
    char *args = {};
    execvp(bin_name, NULL);
    perror("execvp failed");
    exit(EXIT_FAILURE);
  } else {
    int status;
    waitpid(pid, &status, 0);
    printf("Child process finished with status %d\n", status);
  }
  return;

i've stooped low
i'm sorry

[pseudo sierra]

scrolling through the top 1000 games on protondb it's basically all ac problems and the only one I've seen so far that wasn't was some funny problems with 32bit prefixes or smth but the last report was 3 years ago so it's probably been fixed

[lavish frigate]

crazy thing to say for someone who cant use pdn without vm/dualbooting

paint.net my beloved

[strange matrix]

y not use krita tho

feel like its good enough

@fleet cedar
how do i allocate memory for a substring if i don't know the length of it?
do i malloc and then realloc or ?
i wanna know what you'd do

[fleet cedar]

Find out length, then allocate

Or just allocate a bit too much

[strange matrix]

mm

okay

but i don't know length

[lavish frigate]

because krita and paint.net are 2 entirely different pieces of software with different goals and different target audiences

[strange matrix]

they can both edit images

that's good enough for me

😭

[lavish frigate]

and krita is made to be good for drawing and paint.net is made to be simple

nothing compares to the speed at which i can edit stuff in paint.net

[strange matrix]

y not

uh

use wine for pdn

then

[lavish frigate]

it opens in less than a second, i have my image edit with simple tools and a simple ui and then i close the app

[lavish frigate]

doesnt work thats why i said "without vm/dualbooting"

there are some people trying to make something comparable for linux but of course it is still early

[strange matrix]

wait y

do you know y?

[lavish frigate]

...because wine doesnt support everything..?

it doesnt start up

[strange matrix]

i mean

more specific i was asking

[lavish frigate]

you can find that out yourself i do not use linux desktop

[strange matrix]

ohhh

i did nawt know dat

[lavish frigate]

i am on glorious windows

[woven mesa]

i didnt see the ancient games part of that message until just now

but yeah nvm

[strange matrix]

tsk

uses pro

[woven mesa]

yeah what about it

[strange matrix]

ltsc iot enterprise

[woven mesa]

no thanks

[strange matrix]

but y

it doesn't have the usual bloat

[lavish frigate]

why would i use ltsc iot enterprise

[strange matrix]

gives me access to some more stuff too

[lavish frigate]

i do not have the usual bloat

[hoary sluice]

why would i want to use pdn

[strange matrix]

well yeah you can reduce weight quite easily

[woven mesa]

who cares

[lavish frigate]

exactly what a non pdn user would say

you cant value it until you try it

and as a linux user you cant try it so you are missing out

[woven mesa]

@lavish frigate

[hoary sluice]

i literally never draw anything

[lavish frigate]

pdn isnt for drawing

[hoary sluice]

you think i havent used it?

[hoary sluice]

then what

[lavish frigate]

image editing

[lavish frigate]

you have no clue about it so yes

[hoary sluice]

i never edit images either

[hoary sluice]

ive used it for drawing

[lavish frigate]

& opening it once doesnt count

[strange matrix]

y not use photopea*?

[woven mesa]

@hoary sluice fix

[strange matrix]

LMAO

[lavish frigate]

because photopea is photoshop lite and not paintdotnet

again

entirely different goals and target audiences

[strange matrix]

maybe i don't know what pdn is offering

[lavish frigate]

its so great because its native so its fast and its simple

[hoary sluice]

its like saying boo you cant play valorant without dualbooting

why would i want to

[woven mesa]

it offers paint on the net

[lavish frigate]

absolute perfection

[strange matrix]

yoooo

on da net

[woven mesa]

true

dotnet is funny

[hoary sluice]

either way gimp is better

[lavish frigate]

LMAO

https://cdn.discordapp.com/attachments/1029315212521771020/1370052911593423000/attachment.gif

[strange matrix]

where's your shape tool dumbass

[hoary sluice]

my what tool

like i said i dont draw

[woven mesa]

@lavish frigate

[strange matrix]

oh my god GIMP is so infuriating sometimes

[fleet cedar]

Imagine rectangles

[strange matrix]

Imagine Rectangles >>>> Imagine Dragons

[hoary sluice]

average windows user response when cant think of response

[lavish frigate]

you just cannot ragebait me lil bro

[hoary sluice]

try not to immediately prove my point challenge

[strange matrix]

im a lil bro : 3

[lavish frigate]

https://tenor.com/view/linux-linux-users-gaming-computer-computers-gif-23121479

[hoary sluice]

this is a 2018 nvidia issue

[lavish frigate]

right

[strange matrix]

tbh nvidia on windows is fucky wucky too

i have 2 monitors

i play 2 videos

one monitor instantly crashes

great drivers

[lavish frigate]

update yo damn drivers

[strange matrix]

i have the latest

my dear friend

what am i doing

i NEED TO SUBSTRING

[lavish frigate]

then something else is going on

[strange matrix]

ARGHH

[lavish frigate]

&str

[hoary sluice]

windows issue!!!

[fleet cedar]

Use a language that has strings

[strange matrix]

my strings are on the heap

[fleet cedar]

Or build them yourself

[strange matrix]

i am attempting

[lavish frigate]

not a fan of char[]?

[strange matrix]

void bs_trim(char *str) {
  bs_trim_start(str);
  bs_trim_end(str);
}

void trim_start(char *str) {
  size_t start = 0;
  while (str[start] && isspace((unsigned char)str[start])) {
    start++;
  }

  if (start > 0) {
    size_t len = strlen(str);
    for (size_t i = 0; i <= len - start; i++) {
      str[i] = str[i + start];
    }
  }
}

void trim_end(char *str) {
  size_t len = strlen(str);
  while (len > 0 && isspace((unsigned char)str[len - 1])) {
    str[len - 1] = '\0';
    len--;
  }
}

void char_tokenize(char *str) {
  printf("{");
  for (size_t i = 0; i < strlen(str); i++) {
    printf("[%c]", str[i]);
  }
  printf("}\n");
}

[strange matrix]

look i have nothing against the language

i have shit against myself

i am the stoopid

okay?

language is perfect

(im coping)

[hoary sluice]

no its not

[strange matrix]

HOLY YES

ARGH

@fleet cedar i have made
substrings... kind of

look at my beauty

char *bs_splitargs(char *str) {
  char *substring = (char *)malloc(1024);
  size_t iter = 0;
  size_t len = strlen(str);
  while (iter < len) {
      if isspace(str[iter]) {
          break;
      }
      else {
          iter++;
      }
  }
  for (int i = 0; iter < len; i++) {
      substring[i] = str[iter];
      iter++;
  }
  return substring;
}

no llm involved here HAHA

[woven mesa]

@lavish frigate barely 24fps at lowest settings

[unkempt ember]

Genuine question : Why breaking the while loop instead of doing something like

while ( iter < len || !isspace(str[iter]) )
{
  iter++;
}

?

[strange matrix]

uhhhh

dunno

me stoopid :3

ty for telling me

i've had bad experiences with multi-rule loops in C#

so

yeah bad habits

[unkempt ember]

No problem ! I didn't know multi-rule loops were problematic in C#, what problem do they have ?

[strange matrix]

uhh

apparently sometimes

if one of the rules is true**

it doesn't check the next rule

or backwards

didn't matter if you used AND/OR

just was like that

me and a classmate were stuck on it for a while

eventually we just did nested loops

[unkempt ember]

What the - ?

[strange matrix]

oh my god it happened again

literally

okay so

i have this right?

  while (iter < len) {
      if isspace(str[iter]) {
          break;
      }
      else {
          iter++;
      }
  }

[strange matrix]

i put this and now it doesn't work

[unkempt ember]

That's my fault !

[strange matrix]

yeah i was slowly realizing 😭

man

y coding be awkward

[unkempt ember]

with || it continue even if you are out of bound because when iter >= len then !isspace(str[iter]) will be evaluated

[strange matrix]

yeah OR be like that

question

aside from

C

what was the hotkey to

uh

make text like this

[unkempt ember]

It's inline code : `Your text`

[strange matrix]

i know i know

what's the shortcut

like

Ctrl+B bold

Ctrl+I italic

[unkempt ember]

I don't know if there is one actually

[hoary sluice]

why is vencord gay again

[fleet cedar]

🌍 🧑‍🚀 🔫

[strange matrix]

it has always been

[unkempt ember]

To get talkative people talking - and it works

[strange matrix]

man i wish i wasn't one

[fleet cedar]

I don't think you're allowed to use discord if you're under 14

[strange matrix]

i am 16 sir

June 19 2009

how so

was it june or july

i forgor

[fleet cedar]

You said you are one, which os below 14

[strange matrix]

you know sometimes my iq drops below one
because i had to think like 30 seconds straight to realize you were saying

[hoary sluice]

how do you forget your birthday

[strange matrix]

i don't

i have to convert calenders

sure i can say

[hoary sluice]

from what calendar

[strange matrix]

28th Tir, 1388 HS

[fleet cedar]

I always confuse which is which of june and july too

[strange matrix]

Hijri Shamsi

[strange matrix]

its gay month now

so

june now

and

uhh

like

17th Khordad

so

20 more days

probably june yeah

19th of June 2009

i was born

[hoary sluice]

how am i just now finding out theres a solar hijri

[strange matrix]

we have

Hijri Ghamari (Moon)

Hijri Shamsi (Sun)

Saudi uses Moon

we use Sun

[hoary sluice]

whats the point of using a solar calendar that isnt gregorian

youre just making conversion confusing for no reason

whats the advantage of using solar hijri over lunar hijri

[strange matrix]

i have no idea

just tradition ig

we've been using dis shi since islam came

so

[hoary sluice]

well no that would be lunar

[strange matrix]

i think shamsi is

days since

[hoary sluice]

unless persia started using solar immediately?

[strange matrix]

and ghamari would be

nights since

[strange matrix]

naw

uhh

it was

Abu Bakr

that ordered invasion of persia

and then

he ded

Omar came

and then he took over persia

[hoary sluice]

in the 11th century ce

so like 500 years after

[strange matrix]

yeaj

@fleet cedar https://i.imgur.com/wWrS0Y4.png

i love strings

you know that

[hoary sluice]

oh they started using solar cause of agriculture

ok makes sense

[strange matrix]

look atp you know more than me about our calanders

[hoary sluice]

lol

[strange matrix]

i don't care i just need to conv this shi like daily

[hoary sluice]

why not just keep track of both dates

[strange matrix]

uh

my phone only does gregoryian

and i don't have patience to make waybar count for me

[hoary sluice]

so do u use gregorian in daily life

[strange matrix]

no i generally don't have use for calender

i forget days daily

only to know that i have physics exam tmrw

[hoary sluice]

lol

[strange matrix]

https://i.imgur.com/Zt3EqbI.png

YES

ohhhh

satisfaction

(i don't want to think about how i did this)

are you proud of me :3 ?
@fleet cedar

[hoary sluice]

now put ur code in an llm and ask it to scream at you

[fleet cedar]

Did you check for memleaks with valgrind?

[strange matrix]

with what?

[strange matrix]

oh my god it will

[woven mesa]

gm chat

[fleet cedar]

Valgrind

[fleet cedar]

Sleep is not real

[valid jetty]

@hoary sluice i just had an insanity idea

generic functions can be hoisted right

[strange matrix]

let me see

[valid jetty]

because their body is defined at the point when you call them

because they need to be specialized

what if i

do that for every function

not just generic ones

qbe allows hoisting

and it would not only allow hoisting in elle but it would also act as dead code elimination directly

functions that arent called are never generated

[fleet cedar]

Which definition of hoisting? Js style?

[valid jetty]

no, just

calling a function before its definition

fn main() {
    foo();
}

fn foo() {
}

this currently doesnt work

requires forward declaring

[fleet cedar]

Oh, you mean thing that hadn't been a problem in any language developed the last 40 years

[valid jetty]

well

its not a problem in elle either if i make explicit return types

but i dont like that for prototyping

and this kinda solves it indirectly so

[hoary sluice]

can you not call non generic functions before theyre defined?

[valid jetty]

yes iirc

because theyre not immediately computed theyre kept somewhere to be specialized when you call them

actually idk but at the very least generic functions can call other generic functions before theyre defined

[hoary sluice]

i present to you: a language developed in the last 40 years where this is a problem

[valid jetty]

ok but thats an interpreter

the problem makes sense in that world

[hoary sluice]

well no i could just push all expression statements to the end of the ast

[valid jetty]

benefits

• dead code elimination
• hoisting is good
• compilation times will be faster

cons:

• mutually recursive functions will probably stack overflow (though i can catch that i think)
• this doesnt play nice with the lsp because its essentially lazily evaluating the bodies of functions, a function's body will only have lsp support when that function is called in reachable code, which obviously isnt ideal

idk how to approach the lsp problem, this is already a visible problem with generic functions

what if its not a problem

i can just do the rust thing of no lsp support unless that function is exposed in a crate

[hoary sluice]

evil

[fleet cedar]

Where does rust do that

[valid jetty]

but rather than it being exposed in a crate its whether its called in visible code

[valid jetty]

you get nothing, not even autocomplete or hover support, nevertheless errors

its as if youre typing into a text file with syntax highlighting

[fleet cedar]

Ah yes, no diagnostics in files not part of the project

[hoary sluice]

@valid jetty i was trying to make readln and came to the conclusion that making io without monads or any types at all is stupid so now im watching typed lambda calculus videos

[fleet cedar]

Not sure what that has to do with declaration order

[hoary sluice]

function definitons will be forced at the call site, and if there is no call site then the definition will never be forced

[valid jetty]

well if you do this

fn main() {

}

fn foo() {

}
``` the main fn will be compiled because its the entry point but as the foo fn is never called from any reachable code, its never compiled

[valid jetty]

^^^

[fleet cedar]

Sounds horrible

[valid jetty]

there will never be any hover identifiers to check or errors

i could uhhhh

idea

[nimble bone]

Elle will bundle Grok to complete your code

[valid jetty]

if a function wasnt forced through reachable code

its just compiled after

with the same method of reachable calls

fuck this might need multiple passes

1. compile main - the entrypoint, force functions as needed
2. functions which werent forced are compiled seperately after, forcing any calls inside as needed
3. repeat until no functions are left (???)

i dont quite know that much of the implementation details i would have to try it

[pulsar elk]

oop

wrong chat

[valid jetty]

it is june !!

[hoary sluice]

is the lsp separate from the compiler

[valid jetty]

well technically yes but its very intertwined

[hoary sluice]

#[cfg(lsp)] or something for compiling unused functions?

[valid jetty]

even if it was completely seperate, unused functions being compiled out-of-order would cause the exact same error lol

i can do that for lsp only tho

[hoary sluice]

why would they be compiled at all

[valid jetty]

for lsp support

[hoary sluice]

since when does the lsp compile

[valid jetty]

the lsp goes through the whole codegen to find errors and hover stuff, just doesnt go as far as to generate any files

[hoary sluice]

oh

[hoary sluice]

are steps 2 and 3 only done when its the lsp

[valid jetty]

with more information about the impl details, yes

[hoary sluice]

maybe if i watch enough videos on lambda calc i will eventually know formal math notation

[valid jetty]

do you not

[hoary sluice]

i never learned it

i dont know all the symbols

[valid jetty]

husk

[hoary sluice]

i just know the ones that i had to use or had to know the definition of somewhere

"if the sky is blue, then the sky is blue" https://en.wikipedia.org/wiki/Modus_ponens

[valid jetty]

[valid jetty]

why do you need that stuff just make it a builtin to rust

[hoary sluice]

horror to maintain

i wanna read from files

how do i read a file line by line

readLn filename line = -- builtin``` ???

this probably means i open the file a lot

or make it ugly and hacky

if its a monad i can do

file = fopen "filename"
readLn file
readLn file
readLn file
readLn file
readLn file
readLn file
readLn file
readLn file

[valid jetty]

fn main() -> Result<()> {
    let path = Path::new("foo.txt");
    let file = File::open(&path)?;
    let reader = BufReader::new(file);

    for line in reader.lines() {
        println!("{}", line?);
    }

    Ok(())
}

[hoary sluice]

ok now youve read a file in rust

wow

[valid jetty]

can values in icypeas not hold arbitrary values from rust?

[hoary sluice]

they can

well could

if i want them to

i can make Value::BufReader

[valid jetty]

then why cant a variable hold a BufReader and then you just call readLn on it

[hoary sluice]

thats literally just a monad but i stole it from rust instead of implementing it properly

[valid jetty]

well yeah i guess

[hoary sluice]

what if the user wants to make their own monad

[valid jetty]

tough

[hoary sluice]

they cant just edit the compiler

[valid jetty]

ok fair enough

[hoary sluice]

and this whole thing breaks down if i wanna compile

[valid jetty]

are you planinng to go further into it and add applicatives and functors and stuff

[hoary sluice]

im planning to make a type system and have everything else come for free

[valid jetty]

uhhh i dont think thats exactly how it works but sure

[hoary sluice]

a functor is a morphism from a category to another right

ok it is

dont functors come for free

its literally just a function T -> U

[fleet cedar]

A functor is just a covariant generic type

[hoary sluice]

for free as in doesnt require changes to the interpreter

[fleet cedar]

(Technically doesn't have to be covariant but the abstraction is more useful with that constraint)

[hoary sluice]

i was about to say

[hoary sluice]

well it has to be a function

[fleet cedar]

No?

Vec is a functor

[hoary sluice]

is a functor not a function of a category that returns another category

[valid jetty]

this statement has been living in my head rent free ever since you said it

idk why

it just has

[hoary sluice]

"Functors are structure-preserving maps between categories. They can be thought of as morphisms in the category of all (small) categories."

how is Vec a functor

[fleet cedar]

You can turn any Fn(A) -> B into a Fn(Vec<A>) -> Vec<B>

[hoary sluice]

then T -> Vec<T> is a functor, not Vec

Vec itself is an object

[fleet cedar]

And guess how that (type-level) function is spelled in for example haskell

[hoary sluice]

oh theres a difference between Vec and Vec<T>

[valid jetty]

is this idiomatic??

gen.tree.iter()
  .filter_map(|stmt| match stmt {
      Primitive::Function(FunctionSource { name, .. }) => Some((name, stmt)),
      _ => None
  })
  .collect::<HashMap<_, _>>();

[fleet cedar]

Vec<T> is just a convenient way to refer to Vec applied to an unspecified type

[hoary sluice]

yes ok i get it

[hoary sluice]

its spelled []

[fleet cedar]

Yep

Not [T]

[strange matrix]

at least i don't know how both work

[hoary sluice]

Vec is of type T -> U and Vec<T> is U if i understand it correctly

[hoary sluice]

but in that case, its the vec constructor that is a functor, which is a function

[fleet cedar]

Haskell spells those kinds (that's what type-of-type is called) * -> * and *

Vec::new is completely unrelated though

Functors are fmap, not pure

[valid jetty]

have you guys ever used a multimap

[hoary sluice]

any function T -> Vec<T> is a Vec constructor

[hoary sluice]

yes in aoc

[fleet cedar]

Yep and any such function is irrelevant for the topic of functors

[hoary sluice]

which includes Vec::new if T is specified

[hoary sluice]

why, Vec::new is a vec constructor from an empty set of T to a Vec<T>

[fleet cedar]

Functor means any function A -> B can be mapped to a function f A -> f B

A -> f A is irrelevant

[hoary sluice]

i think i confused oop constructors and type constructors

i get it now

a functor needs a T -> f T and a f T -> f U

[fleet cedar]

No

Functor is only fmap

[hoary sluice]

well in a program you need a way to create f T

[fleet cedar]

Having A -> f A is called a pointed functor, but haskell just folds it into Applicative

[fleet cedar]

Sure but not as part of Functor

[hoary sluice]

ok

where did you learn all of this

[fleet cedar]

Haskell

[valid jetty]

are hashset difference operations fast??

my intuition tells me its O(n) where n is the length of hashset A

but im not sure

[fleet cedar]

In theory they can skip all the hashing stuff, but since rust hashmaps have different hashers per instance it's just essentially {x for x in a if x in b}

Still O(n) though, yes

[valid jetty]

i see

[hoary sluice]

@valid jetty ```hs
println "outside comment"
{- inside comment -}

-- inside comment

{-
inside comment
-}

{-
{- inside nested comment -}
-}

{-
{-
inside nested comment
-}
-}
println "outside comment"

try not to rip off haskell challenge

i can actually just use the haskell lsp

[balmy lintel]

kill yourself what

who writes comments like that

[fleet cedar]

Being able to comment out a block is very useful

[valid jetty]

the neferious /* */

[fleet cedar]

Ever wanted to comment out part of your css?

[lucid trail]

Insane language written purely in C without a separate backend??https://nature-lang.org/

I really like its match

[valid jetty]

what’s special about it

it’s just like any other match lol

@hoary sluice

(not my screenshot)

[fleet cedar]

The no-arg match seems like it could have uses

Very few though, else if is a thing

[valid jetty]

is that not exactly the same as rust ???? 😭

am i missing something

[hoary sluice]

Is this a Mandelbrot in sqlite

[hoary sluice]

you can do it in rust but its ugly

it looks bettes than an else if chain in languages that support it

like kotlin

in rust its

match () {
_ if ...
}

[hoary sluice]

toxic 😦

[hoary sluice]

haskell does

i want to eventually allow arbitrary operators, and // is way more likely to be used as an operator (probably will be built in floor division) than -- or {- -}, same goes for #, i want to use it in syntax and not waste it on being a comment

[valid jetty]

uea

[hoary sluice]

and in comes with the added benefit of Haskell compatibility

so i can procrastinate writing a tree sitter parser for longer because the Haskell one will work

[hoary sluice]

i dont have {} blocks

but probably will in the future

so thats another benefit

[hoary sluice]

the only way -- is reasonably conflicting with an operator is decrement, but i refuse to add that either way

mainly because there are no mutable variables but also because -= 1 is better

[lucid trail]

css not having // is a annoying af tho

[hoary sluice]

if foo-- is equivalent to foo - 1 then just write out foo - 1

[hoary sluice]

using /* */ for multiline and not // for signle line would be confusing

gn

[valid jetty]

it is not equivalent

lol

[hoary sluice]

ik it mutates foo

but if you want -- in an immutable language then itd be equivalent to - 1

[valid jetty]

and then another distinction

this is postop, it mutates afterwards

foo-- will yield foo lol

not foo - 1

[hoary sluice]

yes rosie i know how -- works

if you cant mutate foo then foo-- is either foo - 1 or noop

noop wouldnt make sense

[valid jetty]

foo - 1 also makes no sense lol

[hoary sluice]

neither would - 1 but we need some kind of argument for not using -- for comments

[valid jetty]

[hoary sluice]

unless foo is a monad and -- calls map(-1)

[woven mesa]

gm

[ivory heath]

in what code are yall using match statments so much the way its done matters

matches alone are ugly and add a lot of indentation

[lavish frigate]

real question is why are you not using more match statements, they are more powerful than if statements while looking cleaner, can simply code and most of the time only add a single level of identation

[ivory heath]

why do you have lots of branching logic

branching logic is bad for performance and the human mind is bad at holding a lot of conditional logic

the only real use case i can think of for a match statment is when dealing with enums

enums themselves however can usually be a value that means something

or, well why do you have that many enums

[valid jetty]

you can do things like matching against 2 enums

match (x, y) {
    (X::Foo(a), X::Bar(y)) => x * 2 > y
}

among others

for single-case matching, it’s useless

but for stuff like this it’s really clean

if you need single case matching use if let X::Foo(a) = x {}

[ivory heath]

please do not use rust for examples

only 3 people can read it

[lavish frigate]

well for one the performance cost is negatable due to compiler optimisations (unless you use a shitty language? but well why use that then)

match statements are powerful exactly because they help you deal with conditional logic. grouping together possible branches (since the compiler should tell you when a branch isnt covered) with most of the time the only viable alternative being switch statements (which are ugly and often handled in awful ways like implicit fall-through) or if-else ladders which come with their own problems. lots of conditional logic is also sometimes unavoidable depending on what you are trying to accomplish. like the people here are making their own languages ffs
i.e.
this:

let status = match status_code {
  200 => "OK",
  404 => "Not Found",
  ...
  _ => "Unknown",
}

is much better than:

if status_code == 200 {
    println!("OK");
} else if status_code == 404 {
    println!("Not Found");
} else {
    println!("Unknown");
}

good languages that force you to cover all branches also make exhaustiveness checking easier (something which if/else doesnt do so you can easily run into an uncovered branch). if you use enums and in the future add to the enum the compiler will instantly spring up telling you to cover it which makes refactoring easier. really useful if you need to convert between values & especially guards

if you use rust as an example match statements can also be used on structs & tuples which can be incredibly helpful:

let point = Point { x: 0, y: 100 };

match point {
  Point { x: 0, y } => println!("X is zero! Y is {y}"),
  ...
}

[ivory heath]

well for one the performance cost is negatable due to compiler optimisations (unless you use a shitty language? but well why use that then)
compilers cannot optimize branching conditions without using PGO or bolt passes. I have no idea what world you live in. even then if the hot path isnt hit, the cpu has to walk back the entire pipeline

forcing you to implement all conditions of an enum is nice. ill give rust that. however how many times are you using or forced to use an enum or match statment that how its written matters.

[valid jetty]

and pattern matching isn’t even hard to do, you’re just traversing 2 asts simultaneously and once you meet a catch-all pattern you just allow it to be equatable to the whole subtree of the expression you’re matching

“for this interview you will be tasked with traversing a tree” * 2

[ivory heath]

im still flabbergasted at "well for one the performance cost is negatable due to compiler optimisations"

[valid jetty]

until recent years, compilers would not optimize if statements into direct branches like they did for switch/match cases

i remember looking at a video on this in c++ a while ago

https://youtu.be/LJ_irGs5ly8?si=4EIWVWgrYUroZBj-

[ivory heath]

compilers have done it for almost two decades.

also compilers dont usually use jump tables, instead a binary search on a dictionary

because if we have large gaps in our data... lets say 3 to 500, then we have a lot of zeros filing the jump table

and an unexpected value will segfault or cause a stack overflow

[lavish frigate]

im comparing the performance cost of match to if-statements specifically of course branching still costs performance
match statements allow the code to be laid out for better optimisation from the compiler

[ivory heath]

they get converted to basically the same thing in LLVM IR or get mathed out in POLLY

no compiler thinks in C/Rust

[ivory heath]

oh because hes using gcc as an example lol. altho im suprised gcc IR didnt already do it....

[ivory heath]

the example where the comiler optimizes the whole loop out is this BTW
https://llvm.org/docs/Passes.html#indvars-canonicalize-induction-variables

[lucid trail]

oh my god

interesting, if there's no semicolon it sees it as an assignment

good feature?

man i really have no idea where to start with codegen

[runic sundial]

LLVM Ir giving me flashbacks

Anyway, bring benchmarks and your IR/bitcode output to the backup claims that "X IS FASTER" or "Y IS SO MUCH BETTER"

At the very least go throw it at godbolt

Amen

[valid jetty]

yesss good

i have this too

[fleet cedar]

What about let x = false && (let y = 2) == 2;?

What is the value of y afterwards

[hoary sluice]

2

the value of x is false

[fleet cedar]

Even though the y = 2 is not executed?

[hoary sluice]

you can make it not short circuit declarations

[valid jetty]

fuck that's a good point

[celest lodge]

x = ((false) && ((2) == 2))

x is a boolean then with value false since the boolean and isnt satisfied

[valid jetty]

yea but then what's y

[celest lodge]

do you want to do scoping?

[valid jetty]

y is defined inline but won't be set if the rhs isn't reached

so y will be potentially uninitialized

i think rust will throw an error about that

[celest lodge]

personally id rather not let inline declarations be a thing

[fleet cedar]

Yeah inline declarations seems just weird

[valid jetty]

it's useful when you don't wanna make sugar for declaring multiple variables at once

x:= y := z := 0;

[celest lodge]

walrus,,

[ivory heath]

I think everyone know branchless is faster

It’s also kinda hard to show it’s faster in just IR

[runic sundial]

Branchless is faster up until you have a CPU which expects branches and you're thrashing the branch prediction with your funky logic

Faster is PGO

done by hand.

Write code YOU can read

Measure the impact

Focus on making shit maintainable so you can fix it when it's INEVITABLY slow

[runic sundial]

And with the IR/bitcode etc you can see if it's actually turning your switches into luts

[lucid trail]

Hmm i can’t get away with this by evaluating it right to left can i

[fleet cedar]

What about let x = (let y = 2) == 2 && false;

[valid jetty]

that breaks any invariants a developer may have about logical operators

it’s supposed to be lazy

if the lhs isn’t true the rhs isn’t even computed

[lucid trail]

Yeah short circuiting

[valid jetty]

so doing it rtl essentially breaks that principle

or well

flips it

gives unexpected behavior

[lucid trail]

x would be false from true && false. For the other case if I go by short circuiting then y won’t be assigned

[fleet cedar]

Yes and then what would be the value of y

[lucid trail]

Why wouldn’t it be 2

it’ll definitely be LTR

[fleet cedar]

Why would it be 2? It hasn't been assigned

[lucid trail]

((let y = 2) == 2) && false would still evaluate the lhs first

I haven’t actually thought this through

[hoary sluice]

but false && (let y = 2)

would see the false and immediately exit

[hoary sluice]

@valid jetty does #cast return option

[valid jetty]

no

if it cant cast its a compile time error

actually

interesting

this makes sense with how the codegen is set out

inline variable declarations are compiled before the variable itself, so y := 2 is essentially inserted at the top before any condition occurs

not sure if thats intended behavior in this case

but at least the value isnt uninitialized

actually no

we had this the wrong way round lol

or

hmm

😭

[lucid trail]

Yeah, can’t guarantee y is declared

[valid jetty]

im so fucking confused

y is set to 2

but

the rhs branch is never hit

because nothing is printed

so how tf is y being set

is it garbage data/? idk

[lucid trail]

which case is this

[valid jetty]

codegen for

use std/libc/io;

fn main() {
    x := false && (y := 2) == 2;
    io::cprintf("%d, %d\n", x, y);
}

[lucid trail]

What do you mean by this

[valid jetty]

that statement is only true for particular cases

not in this case

[lucid trail]

What if you swap lhs and rhs of &&

should be the same right?

[valid jetty]

yeah why

itll just be in reverse order

outputto

[valid jetty]

im just confused because logically in this case, y should be uninitialized garbage but it isnt

let me examine the assembly

holy shit qbe optimized it away

.text
.balign 4
.globl _main
_main:
    hint    #34
    stp    x29, x30, [sp, -16]!
    mov    x29, sp
    mov    x0, #16
    sub    sp, sp, x0
    mov    x0, #8
    add    x1, sp, x0
    mov    w0, #2
    str    w0, [x1]
    mov    x0, #0
    add    x1, sp, x0
    mov    w0, #0
    str    w0, [x1]
    adrp    x0, _.21@page
    add    x0, x0, _.21@pageoff
    bl    _printf
    mov    x0, #16
    add    sp, sp, x0
    mov    w0, #0
    ldp    x29, x30, [sp], 16
    ret
/* end function main */

qbe detected that its doing 0 == 0 and just removed the branching

wtf

[lucid trail]

the compiler did not preserve short circuiting?

[valid jetty]

elle itself did, qbe didnt (apparently?)

im still looking into this assembly

this is weird

[lucid trail]

Yeah that’s what I meant. There’s 2 str instructions

[valid jetty]

that assembly is the exact output of this code without the print on L14

strange

[lucid trail]

after like an hour of reading docs i got a function that adds 2!

[valid jetty]

yay

[lucid trail]

definitely largest sense of accomplishment so far

[meager bobcat]

I found this in the discord code
this always returns true, right?

[fleet cedar]

Might be some sort of browser compatibility check

[meager bobcat]

thats a =, not a ==

[fleet cedar]

Yes

[meager bobcat]

isn't length a read only property

[fleet cedar]

It assigns to an unwritable field and thus should throw a typeerror

[meager bobcat]

oh alr

[lucid trail]

Should I even have unsigned integers? It's pretty annoying since cranelift does not have unsigned integers

Yeah I'm removing unsigned types, they add a lot of complexity for me

[shrewd canopy]

Dont remove unsigned types...

[fleet cedar]

If you want to be a language feasible for practical use, you will have unsigned integers

[celest lodge]

if you want to be a good language

[dense sand]

is there any k8s devops mage?

trynna use the @kubernetes/client-node, but im getting unable to verify the first certificate

the config code is literally

import { KubeConfig, CoreV1Api } from "@kubernetes/client-node";

const kc = new KubeConfig();
kc.loadFromDefault();
export const k8sApi = kc.makeApiClient(CoreV1Api);

export default k8sApi;

[valid jetty]

no IR has unsigned types because they typically have the same memory representation

they usually instead have unsigned operations like comparison and equality which you must do differently if it’s unsigned

[fleet cedar]

At the IR/ASM level, types are not signed

Operations are signed

[valid jetty]

^^^

unsigned types are just front-end sugar, along with using an unsigned version of operations

[fleet cedar]

Addition, subtraction, and multiplication turn out to be the same for signed and unsigned, so there aren't different operations for them

[valid jetty]

the only thing that types themselves tell you is the size, alignment, etc at an IR level

[valid jetty]

yep

[fleet cedar]

Division and comparison have separate signed versions though

[valid jetty]

as do floats

because obviously the memory representation is very different

[supple whale]

yeah i always use @kubernetes/client-node in prod

its goated

works rly well

[hoary sluice]

Nope I always do (float) ((int) f1 / (int) f2)

[valid jetty]

well i mean yeah

typically you promote integers to floats when there is a float in the operation

[dense sand]

seems too ass

[valid jetty]

not the other way round

[dense sand]

like bruh

[hoary sluice]

it was a joke

[valid jetty]

i mean not necessarily lol

sometimes you need that behavior

[supple whale]

you're confusing k8s being ass, with the client being ass

its good

[hoary sluice]

these letter number letter acronyms are annoying

just say kubernetes

[valid jetty]

like in rendering for a vector2 but you have a position on a grid which is an integer

[hoary sluice]

does (float) 17 return something close to 17

[lavish frigate]

do you not love a11y, i18n, &c?

[hoary sluice]

i meant keep the bits, reinterpret as an int

probably need pointers for that right

[lavish frigate]

a11y secretly stands for ally because of pride month but dont tell coorps that

[fleet cedar]

/run ```rs
println!("{}", f32::from_bits(17))

[rugged berry]

Here is your rs(1.68.2) output @fleet cedar

0.000000000000000000000000000000000000000000024

[hoary sluice]

im an o1p

[valid jetty]

that’s a pun cast 😭😭 ofc that’s not gonna work

[hoary sluice]

o6t

[dense sand]

yea i mean, the issue is in minikube probably, cause i cant even access the api from browser

[lavish frigate]

o1p

[valid jetty]

m2w

n1a

c6d

[hoary sluice]

e2e

[lavish frigate]

t2s i0s s0o a1s

[valid jetty]

that’s not how you 😭

husk

m0eow

[lavish frigate]

thats how i do it

[valid jetty]

i was searching for e yesterday and it literally just wasnt there

i thought i went insane but i have a fucking screenshot

[supple whale]

ah yeah, node might be binding to the wrong network interface

and it just doesnt see minikube

[lavish frigate]

what does kubernetes even do

[supple whale]

average minikube shit

i had a full dev cluster in my job so i didnt have this prblem

[dense sand]

should i just wrap it to docker container or something

[supple whale]

no clue tbf, i never deployed a k8s box in my life

i was just a junior YAML engineer

[valid jetty]

isn’t it more work to count the number of letters in between kubernetes than to just type it out

[supple whale]

if you work in the industry you know k8s is k8s

every1 names it that

[valid jetty]

😭 ok

[supple whale]

its not a choice

[lavish frigate]

please use the appropriate abbreviation "e6e" for everyone instead of every1

[valid jetty]

p4e u1e t1e a9e a10n “3” f1r e6e i5d o0f e41

[dense sand]

who says internationalization instead of i18n

[lavish frigate]

me