Orange

[inner oak]

i need only fix one bug for gtk

its why it doesn't get events

[inner oak]

interesting

it looks like it some issue on libx11

cuz its keep getting messages with 120 len with recvmsg

and doesnt respond, just polls again

i need to try to use xev

cuz st works and xeyes works

also i want to do something like strace

(instead of serial doing debug printfs to parent process stdout)

cuz serial is slow asf

[inner oak]

meow

[inner oak]

things that ig are broken:

• poll
• sockets
• thread stuff

mm is fine

wtf

xev is fine

issue in gtk

[inner oak]

hmmm

it gets GenericEvent from xorg

but there's no mention about thiis in gtk

wait if i disable xi2 issue is fixed (with GDK_CORE_DEVICE_EVENTS=1)

maybe issue in that xorg is too new for gtk or idk

maybe ill just export this env

gtk3 port

[sly sandal]

look weird with two titlebars

[inner oak]

wdym

[sly sandal]

the main window has two titlebar bar

the one from twm and the one from gtk

[inner oak]

a

its some twm stuff

im going to port i3wm soon

[solid hearth]

kde when

[inner oak]

my poor initrd is not fan of this solution

[solid hearth]

nvme and ext2 when

[inner oak]

i already have them but it sucks

ill rewrite them

[jaunty oxide]

You can mask your shit sucking by adding a page cache

[solid hearth]

what if page cache sucks too

[cloud owl]

Add another page cache ontop

[inner oak]

i came to home only now

i was fucking 8 hours in school cuz some event

[inner oak]

i3wm and geany ports

i wanted to port gedit butt

it requires introspection

which is retarded and doesnt works in cross compiling

i want to port some terminal

[shell pivot]

[sly sandal]

Try to compile it with in your OS maybee

[inner oak]

i have one issue rn

echo something > somefile is broken on bash

its just dup2 instead of writing something to somefile and restoring stdout back and after just returns to shell

[inner oak]

wtf

only built in commands > is broken

ls or other command works perfectly

but i need to fix this cuz 1) i want to do self hosting later 2) maybe this bug will make dbus broken too

[inner oak]

found issue

i was cloning cloexec flag in dup/dup2

time to do rename()

its required by python, dbus uuidgen and mv

what the fuck

[solid hearth]

no close_range?

https://tenor.com/view/megamind-gif-26070434

[inner oak]

how i can say to mlibc to use close_range

[solid hearth]

idk

[inner oak]

issue was in getrlimit

i jsut hardcoded fd count to 1024

dbus requires getsockname so ill implement it tomorrow

[solid hearth]

so now it calls close 1024 times only?

[inner oak]

time to socketpair

[inner oak]

i forgot about freeing sockets

i need to implement this

[inner oak]

todo for tomorrow

also socketpair and freeing sockets are done

[inner oak]

grrrr

from hex to text its -1

and its uid

how does it gets -1 ????

[inner oak]

ig i can find issue if i look at dbus src

[solid hearth]

I need to figure out how sockets work

[inner oak]

wait im dumb

i didnt implemented SCM_CREDENTIALS for socket pairs, only for normal unix sockets

[final pilot]

how is that possible

isnt socket pair just 2 normal unix sockets connected to each other

[inner oak]

i made them different fd type cuz why not

[final pilot]

bruh 😭

they arent a different type of fd at all tho

[inner oak]

uuhhh

meow

[inner oak]

actually wtf

dbus never calls sendmsg on orange

but on linux dbus calls it

(im comparing linux strace and my syscall logs rn)

[inner oak]

when other dbus thread waits for SCM_CREDENTIALS for auth

maybe i found issue

maybe its some cross compile issue so ill hardcode it

[inner oak]

i have no idea why dbus doesnt works

it just says rejected external and after client fails to connect

also i still dont have vma for kenrel

time to stop using buddy for everything >4kb

[solid hearth]

ret = base
base += len
return ret

[inner oak]

sounds tempting

time to abuse clanker to know why dbus auth can fail

[inner oak]

all this time i had syscall log disabled for pthreads

i forgot about it

bruh

[inner oak]

debugging dbus feels like getting lobotomy

and bug 100% will be some random 1 loc bug

[inner oak]

uid changed gender

at least i got some hint

its 0 and after becomes -1

[inner oak]

bruh

i just

forgot

to add my os to these defines

[inner oak]

finally dbus is fixed

ill generate locales tomorrow

[inner oak]

so my todo is

• port cool stuff
• stop using buddy and use vma
• implement linux specific sysdeps like eventfd
• patch instrospection to be able to cross compile with my os

[inner oak]

i love when dbus doesnt want to launch when i disable debug mode

[inner oak]

i have 2 bugs rn

first i already found

[solid hearth]

2 bugs that you know of

[inner oak]

im returning ebadf on -1 fd

when i should just ignore it

and second i have no idea where it deadlocks without debug

[solid hearth]

store the owner in your mutex and print stack trace on double lock

that can find one type of bugs

It helped once or twice with my tty code

[inner oak]

it deadlocks in userspac

on some recv or poll idk

or in futex

[inner oak]

it requires pipe pollhup too

[inner oak]

gnome-terminal port

small fps cuz i ran it in debug mode

(syscall log)

i need to fix that deadlock without debug mode and pty stuff

[severe axle]

cool

Is this with linux bin compat or without?

[inner oak]

without

its mlibc

[severe axle]

Very cool

[inner oak]

i just need to fix some deadlock (my kernel bug) and ptys

[inner oak]

the hardest part is dbus

i had a lot issues with it

[inner oak]

it deadlocks in dbus again 😭

please dbus stop 🙏 😭

a

it just tries to connect and dbus-daemon doesnt accept it

and bruh

i need to finally do good disk loading

my initrd is fat asf

[inner oak]

its not futex issue

[inner oak]

and definitely not mm issue

it looks like without debug mode it gets pollhup from xorg ?????

[inner oak]

wait actually im blind

its at-spi connect

[inner oak]

find 5 differences

who is accepting it 😭

[velvet nebula]

?

[inner oak]

uhh

somehow

connect gets accepted by not at spi

when at spi polls for new connect

[inner oak]

im so bored of these undebugable race conditions with single core

maybe its cuz i didn't implemented some flags in recvmsg

like its impossible to have race condition with 1 cpu core

[inner oak]

no its not

i need to try to give to processes unlimited cpu time to see what will change (they will anyway yield in poll or other syscalls)

if it will be fine then idk issue in futexes

[inner oak]

day 3 fixing this bug

[inner oak]

im going to play hoi4

[inner oak]

i love abusing clanker

deepseek gave to me some hint

when i sent 5k syscall log to it

[inner oak]

i have no idea how to fix this

maybe if i write some sockets and futex test i can find bug ?

[inner oak]

hmmmm

maybe if i implement signals in poll it will work

[inner oak]

finally

i fixed this shit

now i dont have some weird deadlocks

issue was that i didnt implemented signal ret for poll()

[solid hearth]

didn't you say xorg needed something in sysfs for fbdev?

what is it?

[inner oak]

only fbdev directory

to detect it

[solid hearth]

okay

[inner oak]

/sys/class/graphics/fb0 (/sys/class/graphics/*)

[inner oak]

but glxgears have 500 fps

ig my scheduling is bit weird

but ill fix this in future ™

now i want to port more gnome apps and fix ptys

[inner oak]

and you actually can fill it if you want

[inner oak]

everything works with smp

[inner oak]

mate-calc port

i wanted to port gnome-calculator but its written in vala

idk how to cross compile this

[inner oak]

uncool news I really need to do instrospection and vala

at least cuz gnome stuff require it

my target is now gnome calculator

cuz its written in vala and require introspection

[inner oak]

holy shit my tmpfs is so fat

but ig its cuz im allocating everything with buddy

i need to stop this

i waste 5-6 mb on gnome-terminal after exiting it

its so terrible

[inner oak]

im going to rewrite my pty line discipline

[inner oak]

i had some break now im continuing orange development

[onyx matrix]

gtk apps but no job control 💀

[inner oak]

i didnt implemented SIGPAUSE

[onyx matrix]

ah

[shell pivot]

that depends

whats your minimum order?

[inner oak]

page size

[shell pivot]

im guessing youre allocating tons of small objects and you have a min order of like 8 or more, so the minimum block size it can allocate is 256

ah, then yeah

[inner oak]

issue is buddy aligns to nearest power of 2

like if i alloc 5009 bytes it will allocate 8kb

[shell pivot]

yes, thats a problem if the things you allocate are weird sizes like that

[inner oak]

tmpfs files are allocated with buddy

im going to change it to vma

[shell pivot]

i use buddy allocators for my BASIC interpreter instances, because often it is storing strings and theyre constantly realloc'ed

realloc that fits into the same buddy order is basically free

[shell pivot]

what about allocating a block device for your tmpfs?

then formatting it with e.g. fat32 or ext

this makes more sense for buddy allocator, and allows for a ton more flexibility than just a tmpfs made from an array and a tarball

basically, more like /dev/loop*

[inner oak]

gnome-terminal output is fixed

i forgot to add 1 to read_bytes when reading from tty packet mode

[solid hearth]

packet mode?

canonical/cooked?

[inner oak]

other

TIOCPKT

its just normal tty but with additional byte in beggining (usually 0)

i need to install some dark theme

my eyes hurt when i see white theme

[inner oak]

my eyes are not hurt now

installed orchis-theme and changed gnome-terminal to dark mode

[turbid bridge]

This is goonable

[inner oak]

real

[onyx matrix]

hell yea

[inner oak]

i want to implement proper EFAULT returning

[onyx matrix]

what do you do rn

[inner oak]

bug fixing

[onyx matrix]

no like

when userspace gives you a junk pointer

[inner oak]

nothing

i just check is it in kernel space or not

but im not checking is this valid pointer

[onyx matrix]

bruh lol

[cloud owl]

I mean I also do that, but I use special functions that the kernel on a fault returns an error instead if it can’t handle it any other way

It is one of the more efficient methods if done right

[onyx matrix]

well yeah i also check if it's kernel space or not but i also handle pfs when copying from/to userspace

[cloud owl]

I’ve seen people lock the page tables and manually look them up and verify

Which just feels partly redundant when the CPU can do that for you anyways lock free

(Well mostly in the case of no fault)

[onyx matrix]

ye

why put in all the effort when the cpu does it for me

[feral vortex]

https://tenor.com/view/dio-brando-gif-25280711

[inner oak]

i have some issue

gtk page faults randomly

[inner oak]

ig its some gtk bug

cuz it pagefaults only on exit

anyway ill just send sigsegv to it

[inner oak]

bruh

i cant continue work on orange cuz gitlab.gnome.org is down

im downloading tarballs from it

[inner oak]

python3 port

its required by introspection so why not

[narrow mesa]

If python 3 is so good port python 4

[inner oak]

i hate gobject instrospection

[severe axle]

if you're using mlibc with the linux abi you can do what Managarm does

[inner oak]

im doing it but it segfaults somewhere

im fixing it rn

[inner oak]

im dumb

fix was just replacing some arguments in linux thread entry

[inner oak]

gnome-calculator port

[onyx matrix]

now port the entire gnome stack

[inner oak]

its my goal

[inner oak]

porting gnome rn

i want to port gnome 43 at least cuz mutter 43 doesnt require colord or glycin

[inner oak]

itttt requires gtk4

[inner oak]

im patching it to not use egl rn

[inner oak]

actually its possible to port gtk4 without egl

it looks like it can fallback to glx

gtk4 port

[solid hearth]

gnome when

[inner oak]

but

some gtk4 stuff is slow

i need to optimize it later

[inner oak]

ig its because it uses gl stuff for rendering

[inner oak]

😭 😭 😭

gnome requries firefox (spidermonkey)

😭 😭 😭

it means time to setup rust cross compiler

[oak cobalt]

Told ya so

Oh and do yourself a favor

Get your rust setup working with something smaller than fucking spidermonkey

Like ripgrep or eza

[inner oak]

dddkfjjdejej

i completely dont know how rust works

so im going to steal some patches from assertos

[severe axle]

lol

when you get firefox or gnome to work, i'll be curious what kind of patches / features we're missing on the Managarm side

[inner oak]

i still dont have networking

ill do it after userspace adventures

[inner oak]

did some optimizations fps in glxgears increased from 900 to 1000

gtk4 became bit usable

time to do shm stuff

iirc it should make mesa stuff faster

[inner oak]

i got race condition 😭

with 12 cpus

when thread page faults and gets sigsegv

[inner oak]

how

it jumps to already killed thread

[velvet nebula]

why not move a thread to a dedicated killing queue

so cpus won't jump to it

[inner oak]

im so dumb

i forgot to switch cr3 before paging free and it was freeing current cr3

i want to do setitimer

[solid hearth]

it's not hard

[inner oak]

ik

mesa wants it so much so ill do it

[feral vortex]

I free the cr3 by freeing the physical frame

[sly sandal]

But didn't you run MESA without it ?

[inner oak]

it can run without it

but it spams with it and gets enosys everytime

[sly sandal]

Oh okay

[inner oak]

getitimer/setitimer is done time to do shm

[inner oak]

shm is done

[vast lake]

cool!

[inner oak]

time to getaffinity stuff

for nproc and etc.

[inner oak]

wtf

why gtk4-demo breaks when i setup cairo as gsk renderer instead of gl

when gtk3 works fine when its cairo handled

maybe its just some unimplemented/i dont have drm or something issue

[inner oak]

so now when my kernel not suck in opengl so much i can continue porting pain

[inner oak]

maybe its because of my patch that removed egl from gtk4 or some version mismatch

cuz i dont think that its kernel bug (or its missing EINTR return)

[idle forge]

or sysconf

[inner oak]

i tried to lazy allocate 4 pages instead of 1 page for 1 page fault and it increased glxgears fps from 1300-1400 to 1700-1800

[onyx matrix]

zoom

[inner oak]

im building llvm again

[solid hearth]

https://cdn.discordapp.com/attachments/1366384847538487397/1376123845009735710/caption.gif

[inner oak]

i forgot

that i had aarch64 and riscv64 ports

and they are really poor

omg i fixed nvme

i was writing uint32_t instead of uint64_t in nvme write64

[inner oak]

100 syscalls

[inner oak]

fixed

my writev was returning EAGAIN when there already was written some data

[idle forge]

only?

[solid hearth]

I'm at 195

[onyx matrix]

i only have 75 (only 70 implemented 😔)

[inner oak]

finally i setup clang cross compiler

[idle forge]

how does your syscall table look

[jaunty oxide]

astral only has 105 system calls

[idle forge]

i have 137

[inner oak]

how to post it

[idle forge]

upload the .c

[solid hearth]

send a github link

[inner oak]

im trying to build rust cross compiler rn and port it

[onyx matrix]

sys_enosys_stub

[inner oak]

meoww

[inner oak]

i have some important asf biology exam tomorrow

[onyx matrix]

i saw my biology teacher going for a walk in town today in school hours

i went home cause i didn't have class

most of the teachers are manifesting and apparently they do this by hanging off bridges??

(random pic from some random article)

[inner oak]

i hate rust

why it uses my headers with host c compiler

[inner oak]

finally

time to try port something in rust

trying to port xshmfence and bruh

time to patch it

[idle forge]

W

how did you do std?

linux_like?

[inner oak]

(i stole some managarm patches)

[inner oak]

am i memory safe now

[inner oak]

vulkan

[onyx matrix]

does it run better than opengl when cpu rendering

[inner oak]

no

it runs really buggy

and slow

[onyx matrix]

rip

[inner oak]

its buggy on linux too bruh

on kvm

maybe its cuz of swrast

or lavapipe idk how it named

im porting something cool rn

on rust

[jaunty oxide]

on rust fr

[idle forge]

on rust fr

i ported fish

[jaunty oxide]

use ksh93

[idle forge]

trash93

[jaunty oxide]

"I enjoy kicking pregnant dogs" what the fuck marvin why would you say something like that

[idle forge]

https://cdn.discordapp.com/attachments/403372275270942722/1512829004657201372/netanyahou_tech_tips.gif

[inner oak]

compiling rust rn again

i forgot to enable linux target

[idle forge]

[hexed matrix]

i ported new versions of rust-libc and rust-rustix

now im porting rust-rustix0.38

[hexed matrix]

i looked at managarm patch for rust-nix and im not going update this

[hexed matrix]

jirpt]hjgr]phoptrehjekb]bkekbkgfbfglbkdl;bkdflvbmfdvflkd;gldkg

i forgot that i am linux_like in rust

so i can just ignore patches and do it only for errno

[inner oak]

finally i can type

[inner oak]

my honest reaction when x11 crate requires another version of nix

[inner oak]

idk why but i love rust now

i want to start doing drivers in rust for orange

[solid hearth]

rustlover0

[inner oak]

i will keep kernel in c++ but drivers will be in rust

[idle forge]

https://tenor.com/view/rust-my-beloved-my-beloved-heart-lock-gif-26944656

[inner oak]

i still porting something cool

its pain to port every crate that require some platform settings

[inner oak]

today i had some important asf russian exam

[inner oak]

im doing epoll rn

[inner oak]

epoll is done

i want to move xorg to epoll now

[inner oak]

time to do locks in fcntl

sqlite requires it

[inner oak]

something cool is zed

editor

[idle forge]

💀 💀 💀 💀 💀

[inner oak]

what's wrong

[idle forge]

it's that even remotely portable

[inner oak]

yes

i already built this

[idle forge]

it pulls in chromium

what

[inner oak]

freebsd have zed port

where it disables webrtc stuff

[idle forge]

ah

yeah webrtc is stupid

[inner oak]

im implementing some stuff that required by sqlite like ftruncate rn

zed uses it

[idle forge]

btw how did you port rust libc

[inner oak]

linux_like

[idle forge]

ah okay

that's easier

[inner oak]

yes

i can just copy values from real linux cuz i follow linux abi

[inner oak]

yesterday i had last exam

im free now

[turbid bridge]

free(exam);

[onyx matrix]

segmentation fault

[inner oak]

meow

[inner oak]

i had some break after rust adventures

my lazy ass will do orange today

[inner oak]

my lazyass

finally started doing something

im fixing flocks rn

[inner oak]

i

did something

but

https://tenor.com/view/ralsei-deltarune-meme-gif-17328629134228745439

[inner oak]

i completed deltarune chapter 5

[cloud owl]

Chop chop back to porting gnome

How's that been going btw

[jaunty oxide]

https://tenor.com/view/linux-arch-gnu-gnome-nerd-gif-12255644138566808999

everybody loves systemd

[inner oak]

im fixing some kernel bugs

for sqlite

for zed

after ill port firefox

and after firefox i will be able to compile some random gnome library

[inner oak]

i got some ub bruh

[inner oak]

maybe if i setup watchpoint i can know why memory gets corrupted

[inner oak]

im so dumb

i setup stack to start of allocated memory instead of end

(i wrote this code year ago)

[onyx matrix]

lol

[inner oak]

i want to update mlibc too

tf is aio

[turbid bridge]

Asynchronous IO

[inner oak]

sqlite requires normal file mmaping bruh

[cloud owl]

Time to implement a page cache and 70 vulnerabilies

[inner oak]

😭

[turbid bridge]

Shii im here for it

[inner oak]

it looks like it requires file mmaping with map_shared

[inner oak]

ill make some kernel thread that will sync mmap allocations with vfs

[inner oak]

im thinking how to implement ts

also im thinking about implementing swap

it shouldn't be hard cuz iirc its just getting unused pages and saving/loading this to disk

in some random thread

[inner oak]

im sick

[solid hearth]

that's sick

[west bloom]

ah sweet summer child

[inner oak]

i was playing deltarune today again

but with mod

[inner oak]

me when another memory corruption

https://cdn.discordapp.com/attachments/1045309080069214244/1484593761571832008/togif.gif

[inner oak]

something something

[onyx matrix]

meow

[analog spruce]

Mandarine when

[inner oak]

in future i want to add good aarch64 support and without uefi for rpi

[solid hearth]

limine on rpi when

[onyx matrix]

i thought you could get uefi working on rpi

[inner oak]

after day of playing gta 5 i finally figured out how to sync ts

[inner oak]

i want to do rbtree

[cloud owl]

You can use my library for it if you want

It's a bit like kgz since you already use it

[inner oak]

im at the sea now

[vast lake]

Irl?

[inner oak]

im doing some shit rn

[onyx matrix]

!@!@$#@

[solid hearth]

https://klipy.com/gifs/qbert-game

[inner oak]

its only part of my asserts

masterpiece of asserts

[onyx matrix]

💀

[analog spruce]

Managarm in a nutshell

[idle forge]

😭 why no stdint

[inner oak]

std::uint32_t looks more beatiful than just uint32_t

[idle forge]

no like

why long long

😭

[inner oak]

a

idk

idr why i did this

[idle forge]

smh

i wonder if there is a clang-tidy lint for raw int vs stdint

[inner oak]

i implemented map_shared with disk sync and sqlite still doesnt works

bruh

time to debug ts and read sqlite src

[inner oak]

need to try it on linux with mlibc first

to know is this my kernel bug or other

it works on linux with mlibc

its my kernel bug

[inner oak]

it looks like i have issue in F_SETLK, F_GETLK

hmmmm

maybe its because i dont automatically unlock file ??

time to read man

[inner oak]

holy shit

sqlite works with shm mode

(with some test made by clanker)

fym mlibc panic

maybe its because i didnt implemented MAP_ANONYMOUS with MAP_SHARED

[inner oak]

https://klipy.com/gifs/flowery-my-beloved

[inner oak]

im continuing zed porting

time to find out why ts doesnt creates sqlite tables even on linux

[inner oak]

mlibc update finally fixed sqlite shitt

[inner oak]

wtf rust