Ethereal

[cyan phoenix]

Development of Ethereal mainly occurs in a different server, so updates here might be infrequent but I will post occasionally

More screenshots:

[cyan phoenix]

Hmmm trying to get my GCC port running is not going well

Porting Mesa was very easy once I fixed my headers

[solid stone]

interesting, good luck with the project 🙂

did you write a custom debugger for the kernel?

[solid cobalt]

Damn

Which libc are you using

Also a rare acpica user

[solid stone]

non-sleeping semaphores

[solid cobalt]

Fr?

[solid stone]

yeah I believe so, there is sleep_untilCondition() which looks like it can be used to implement blocking waits, but it doesnt appear to be used.

[muted patrol]

I think it's a custom libc

[cyan phoenix]

A while ago, but it's abandoned

[cyan phoenix]

Custom libc

[cyan phoenix]

My semaphores were actually made only for ACPICA, and were made before I even had a userspace 💀

Even then I think they're pretty abandoned, they are used nowhere in the kernel

[cyan phoenix]

Yup! Other sleep functions:
sleep_untilNever
sleep_untilCondition
sleep_untilTime

My sleep system probably needs a rewrite because I don't like it being a clock callback and would rather threads that call sleep_wakeup to actually iterate the queue and wake the thread up instead of relying on this async garbage

Apparently vbox doesn't like requesting a 1920x1080 framebuffer.. it renders extremely small

[solid stone]

lol nice, so how long have you been working on this?

[cyan phoenix]

about 8 months

it looked like this

Just a demo program in Hexahedron's fb

[torpid wolf]

using ACPICA is such an L for what is otherwise pretty much a completely NIH project... well i guess you also use mesa

whatever

your readme mentions you depend on grub but you also have a custom bootloader?

[cyan phoenix]

yea well I also have a mini ACPI driver I wrote, ACPICA is 100% optional

[cyan phoenix]

Yes, Polyaniline.
https://github.com/sasdallas/Polyaniline

[torpid wolf]

then why do you need grub?

[solid cobalt]

damn, gnu efi

[muted patrol]

looking at commits, polyaniline was only from about 3 weeks ago so probs just used grub first then moved over to custom (idk though just what it seems like)

[torpid wolf]

i see

[cyan phoenix]

It's also shit. UEFI only and was mainly for fun

[cyan phoenix]

Actually Poly was in dev for over two years, but I only worked on it for probably 20 days total, including pre-rewrite

Writing a bootloader is cool but I like the whole OSdev part

[torpid wolf]

use Limine then :3

[cyan phoenix]

Limine is ass

[torpid wolf]

why?

[desert seal]

i wouldn't say "it's ass"

well

yeah it's unstable sometimes

but imo it's a pretty solid bootloader

[gentle burrow]

me when i'm wrong

[torpid wolf]

what instability are you referring to?

like, specifically

[atomic marlin]

[desert seal]

haven't experienced instability myself but i could've sworn i've seen a little bit of people having issues

[torpid wolf]

like

sure, bugs happen, but like, do you have an example of some "instability" that has been persistent and wasn't fixed quite quickly after a reproducible example was provided?

[desert seal]

can't recall

[torpid wolf]

(or even an example of one that was fixed, for completeness sake)

alright, sure

[solid cobalt]

How are u not furious over people shitting on it for no reason

[torpid wolf]

idrc

i like if they provide evidence so i can fix whatever's wrong

if they don't... well, less work for me, and i am lazy

so that's a win

[solid cobalt]

Limine is unstable, btw I'm not gonna provide the evidence as to keep it this way

[torpid wolf]

that would be devilish

smh

[solid cobalt]

Lmfao

[atomic marlin]

also being mad isn't exactly going to solve those magic problems that exist so potently

[ripe kettle]

Only reason to not use the limine bootloader is custom boot GUI

Limine is 🐐

[atomic marlin]

custom bootloader running limine protocol

[muted patrol]

ah, makes sense!

[oak holly]

You didn't know, you responded that to limine's author, did you?

[torpid wolf]

@cyan phoenix hey, i am still waiting for an answer...

😔

[muted patrol]

😔

[cyan phoenix]

Was on a hiking trip

Anyways no I didn't know you were the author nor do I really care

I don't have the time to write Limine support if GRUB works fine and Multiboot was already a pain in the ass

Most limine build systems (at least from other OSes I've seen) seem shitty and hacky

But it's a good project and I will add support eventually, I'm just lazy

[gentle burrow]

brother

[cyan phoenix]

i am not putting it in gitmodules

[gentle burrow]

why would you

you just need the header for the kernel

[torpid wolf]

huh?

[cyan phoenix]

Every other OS that uses Limine, at lesst in my community, does

[cyan phoenix]

Limine Bare Bones at least

Again I also am just lazy

[torpid wolf]

you can use whatever build system, you can use meson or whatever you prefer

[gentle burrow]

do you have a distro builder or something

[cyan phoenix]

That's not what I was saying

[torpid wolf]

what is the shitty and hacky part in your opinion?

[cyan phoenix]

Mainly the gitmodules and calling a script from the Limine repo

[gentle burrow]

well you need something to build limine into your image

same way you need grub to make the iso

[torpid wolf]

what gitmodules? i am seriously confused

[cyan phoenix]

here's the OS I'm basing most of my Limine knowledge off of

https://github.com/UnmappedStack/TacOS

[gentle burrow]

that's not optimal

you should have a way of building an image

[cyan phoenix]

It's also similar to what Limine Bare Bones says I think - let me check

[gentle burrow]

limine bare bones includes kernel + bootloader

your kernel shouldn't need a limine build, just the protocol header

[cyan phoenix]

Nvm don't know where Unmapped got that from

[gentle burrow]

basically, you need a tool to build a distro

sooner or later

[cyan phoenix]

Then that's fine with me

[torpid wolf]

yeah

[cyan phoenix]

Although don't you have to setup a special linker section

[gentle burrow]

i do it in a generic way

[torpid wolf]

it's not mandatory

[gentle burrow]

i have the sections in my linker script regardless

i call them .boot{.init,.fini}

[cyan phoenix]

True, I already have a multiboot section.

If I can I'll give it a go

[cyan phoenix]

Start button art, will add it in tomorrow

[muted patrol]

oh god please don't use this as your view of limine buildsys

[turbid summit]

yeah use minse instead

[cyan phoenix]

I love stanix but that buildsys is very, uh, interesting 💀

Anyways I'm actually going to probably add a few gfx functions to go along with this:

• gfx_drawRectanglePattern
• gfx_drawRectangleGradient with GRADIENT_VERT AND GRADIENT_HORZ
• gfx_drawRectangleRounded

maybe add png/jpg support

[turbid summit]

why rectangle parttern?

[cyan phoenix]

To let cool patterns be drawn. Can also make gradients easier

[turbid summit]

oh cool

[cyan phoenix]

Hover icons

[oak holly]

What is Essence? Does it have something with nakst's Essence?

[cyan phoenix]

No, it is my own custom shell

All usermode programs for Ethereal, besides ports of demos like glxgears, are all written by me

[oak holly]

I see, I guessed that it's the cli name, the name reminded me about that cool project, that now sits dormant.

[cyan phoenix]

Yeah I saw the project awhile ago and it looked interesting. Sad to see it died out

[oak holly]

That's fantastic actually. Great job.

[cyan phoenix]

Thanks!

[cyan phoenix]

rtl8169 MAC address getting works

xHCI still doesn't work on real hardware

My plan is to finish RTL8169 driver first so that I can have 3 different network card drivers

Maybe if I need it I'll do a PCnet driver but I don't have any machines that need that

After that it's profiling time to figure out why VMware is so slow single-core

[cyan phoenix]

RTL8169 working!

[dusk crown]

might also want to take a look how I did it, I use cmake which builds an image as part of its process but only builds what has changed, an image is always output if any depends of the image has changed

I don't know if what I did is considered "optimal" but it's a different approach so maybe you'll learn something different

(also it's using limine, so...)

[cyan phoenix]

yeah TacOS creator confirmed for me that is not optimal build system :D

limine support and fixing my broken mb2 is on my bucket list since RSDP finding doesn't work on some of my UEFI machines (as we all know since RSDP doesn't have to be in EBDP 😔)

i still have to fix RTL8169 sending packets/perhaps acknowledgement

[dusk crown]

personally i avoid acpi. havent needed to go further than the RSDP pointer limine provides for anything i do atm

its an eldtrich horror that thinks its an OS

if i did need to go deeper i'd probably use uACPI because it seems to be the most well supported way atm

[cyan phoenix]

I use ACPICA but due to some people not liking it (cough cough @torpid wolf ) I am writing my own AML parser as a side project

The initial idea was good in my opinion, before AML

[torpid wolf]

i mean, i don't like ACPICA because uACPI is just simpler and better lol

i would recommend against making your own (cc @solid cobalt )

[solid cobalt]

i mean give it a try since you're doing full NIH anyway, but it is a lot of work to put it lightly

it's a crappy recursive bytecode with vague unstandardized semantics

[cyan phoenix]

I don't honestly have much use for it right now, besides APIC PCI interrupt pins but having battery support would be nice :D

[cyan phoenix]

Yes, I do remember trying to read through the spec and wanting to kill myself

[solid cobalt]

yeah

it also either lies or just completely omits a lot of very important details

[dusk crown]

how many assumptions does uacpi makes about what your os has availabile? probably off topic. I'll go read up and if I have any Q's I'll prod someone on the relavent thread

[torpid wolf]

almost none

[oak holly]

NIH is "not invented here". So it means porting something, made by others and not vice versa, isn't it? If someone makes their own say acpi implementation, it'd be IH. Why is it called NIH? can somebody using this NIH termin explain?

[solid cobalt]

• Why not use X instead of writing your own?
• NIH.

[oak holly]

I see, it's a total opposite to how I perceived it. 😁

[cyan phoenix]

Inspiron 1420 1 day of runtime

[shell temple]

Would be cool to see it with a day of runtime under a heavy load

[cyan phoenix]

Yeah I should've launched more programs but the only reason it ran this long was because I forgot it was running

[cyan phoenix]

Ethereal can now run the Bochs emulator

SDL2 is coming along great

[solid stone]

cool

can you boot ethereal in bochs under ethereal yet?

[cyan phoenix]

Yesh I just had to fix some alpha blending. I'll get a screenshot later

[cyan phoenix]

GUI at v1.0.0

[topaz prism]

Ayy! Looks awesome! 👍

[solid cobalt]

p cool

[cyan phoenix]

Another photo of the GUI after adding a different wallpaper + font viewer

I probably should do a taskbar menu next

[solid cobalt]

wait wtf lol

[cyan phoenix]

^

Pulled together a small demo

[solid cobalt]

lol

[cyan phoenix]

Taskbar will load entries from /etc/taskbar/ or whatever path I want

Icons in /usr/share/icons of course

[topaz prism]

looks great! :3

[cyan phoenix]

I think I'll do a poweroff sequence nw

[cyan phoenix]

poweroff sequence is done

[solid stone]

Nice

I wonder how many kernels here do proper shutdown

[cyan phoenix]

Not a lot, I can gurantee

It's not like it really matters but I find it nice

[solid cobalt]

I don't think think any of them do

[torpid wolf]

Ironclad does

[solid cobalt]

Ah

[dusky dew]

Managarm can do it (well we’re don’t do sync yet but the kernel code has proper shutdown)

(And don’t try systemctl poweroff or friends, that shit broken for some reason )

[shell temple]

closest I have to proper poweroff is just sending SIGTERM SIGKILL and then sync and tell the kernel to poweroff

still missing stuff like telling drivers to denitialize etc but if you run poweroff (or press the power button) it should do all of that

[cyan phoenix]

xHCI is beating my ass

doesn't work on vbox/real hw but works on VMware + QEMU

on VBOX, the second device fails to init. real hw is mixed results but so far haven't gotten to actual CONTROL requests yet

[cyan phoenix]

also taskbar design

it loads files (sorted by 01-entry.ent) from /etc/desktop.d and parses their icons, names, exec, etc. from it.

[cyan phoenix]

Link to release 1.0: https://github.com/sasdallas/Ethereal/releases/tag/ethereal-1.1.0

[cyan phoenix]

kernel is more unstable now in thread sleeping/mem system but it looks nice

[icy junco]

holy shit amazing and I love that logo

[solid cobalt]

yeah good logo

[icy junco]

good luck dude

[cyan phoenix]

thanks! logo was designed by my girlfriend, she does most of the art assets

[icy junco]

ask ur gf to design some shit for my project 😭

[cyan phoenix]

just go date an artist

[icy junco]

bro thanks i can pull

how did dude pull vro

[cyan phoenix]

im 15 so got plenty of free time and social events

i got lucky lmao

[icy junco]

vro im 18 got no job

no bitches vro

[cyan phoenix]

the bitches don't matter the OS does 🥀🥀

[icy junco]

real 🥀

[hoary pine]

i apologize for calling you histrionic i didnt realize you were just 15

[cyan phoenix]

doesn't matter, call me whatever you want. I don't get special privileges in arguments

[hoary pine]

you have the special privileges that come with being 15

such as not being completely responsible for being annoying

I miss that. Now I get in trouble for being annoying :(

[tender verge]

just become the artist then

pretend like its the most essential part of your os and then try not to cringe at your previous artwork after 3 years after finishing it

[cyan phoenix]

nah though if i get into an argument i probably need to get called some shit lmao

dont feel bad :D

[oak holly]

V

[cyan phoenix]

xHCI success finally

[cyan phoenix]

redoing my interrupt transfers to be asyncronous, then HID driver, maybe Hub too when I do USBHubInformation_t structure documentation

[dusky dew]

if I may ask, can I get an invite to that server?

[ripe kettle]

Same

[torpid wolf]

same

[icy junco]

baller

[icy junco]

same

i want an invite too

[upper bronze]

I'm also interested

[cyan phoenix]

@icy junco @upper bronze @torpid wolf @ripe kettle @dusky dew

official dev server: https://discord.gg/cGeEGTuE

the community where Ethereal's dev thread is: https://discord.gg/r4dSnpkC

[icy junco]

baller

[ripe kettle]

thx

[dusky dew]

thanks

[icy junco]

np

why am i saying no problem

crazy

[solid cobalt]

Lol

[cyan phoenix]

god my GUI runs like shit without kvm

still have to work hubs into my USB stack

probably will just add a HubInformation struct in the device and then have a hc->init_hub_port method fo xHCI

UHCI/EHCI I think are fine

[icy junco]

to be expected

does ur GUI use X

[cyan phoenix]

not at all

custom window server

[solid cobalt]

are u gonna add hw accel support?

[icy junco]

^

[cyan phoenix]

everything in ethereal is custom, minus a few things I will fix later like ACPI + libm + ttf

[cyan phoenix]

like a DRM api?

or do you mean SSE?

[solid cobalt]

well that and kernel drivers for gpus

[cyan phoenix]

SSE is already there

[solid cobalt]

and i guess userspace drivers since you're fully NIH

[cyan phoenix]

DRM is a bit away

[cyan phoenix]

wha

[solid cobalt]

yeah

[cyan phoenix]

my drivers are kspace

[solid cobalt]

a huge part of gpu acceleration is userspace drivers

for generating shaders and command buffers

[cyan phoenix]

oh sure

oh yeah 👍

[cyan phoenix]

im not exactly the best at graphics programming, libgraphics was probably the best thing I've done (?) graphics related so itll def take me a while

right now the priority is fixing some kernel scheduler bugs (and maybe improving the very very shitty algorithm) and USB

[solid cobalt]

u could make your kernel drm compatible and port mesa

[cyan phoenix]

i already have OSMesa but i'd also like to do my own DRM api

NIH :D

[solid cobalt]

OSMesa?

[cyan phoenix]

Mesa with framebuffer

[cyan phoenix]

it's a rendering backend

[solid cobalt]

oh u have a mesa port?

[ember hazel]

OSMesa is not that

[cyan phoenix]

what?

[ember hazel]

OSmesa is mesa without OS support

[cyan phoenix]

off-screen rendering

which means rendering into a user allocated framebuffer

https://mesa-docs.readthedocs.io/en/latest/osmesa.html

i need to switch to llvmpipe probably instead of using the default

[solid cobalt]

yeah llvmpipe is way better

[ember hazel]

llvmpipe does need llvm though, which if you are doing NIH may break a line

[cyan phoenix]

it would be a port

im not actually going to use llvmpipe/Mesa to render my GUI

i already have some SSE-accelerated graphics functions

[ember hazel]

So are you gonna write your own stack for desktop use?

[cyan phoenix]

i did

[ember hazel]

For hardware accel in the desktop I mean

[cyan phoenix]

if you're talking about things like graphics API, yes
window manager API, widget API, etc

[cyan phoenix]

probably will

[solid cobalt]

do u have posix compatible syscalls?

[cyan phoenix]

yes

[solid cobalt]

ah

[cyan phoenix]

feel free to take a look: https://github.com/sasdallas/Ethereal

[ember hazel]

Why would you port mesa then

[solid cobalt]

i mean even in a nih os u want ports

although ig not mesa

[cyan phoenix]

because its fun?

and you get glxgears

[ember hazel]

like, an OS only needs 1 GL stack

you could have glxgears with your own stack

[solid cobalt]

libgl.so from nvidia

[cyan phoenix]

back when I ported Mesa it was just to port things like SDL + rendering acceleration without having to write my own damn renderer. libgraphics is definitely not a replacement OpenGL

however I don't need OpenGL in the kernel

[ember hazel]

you dont need any of this in the kernel, the kernel shouldnt have to concern itself with graphics

[cyan phoenix]

*in the OS

[ember hazel]

you do need GL on the OS unless you want to make your own GL competitor for hardware accel

[cyan phoenix]

now what would I need GL for? i have "a GL" but it's not a replacement for OpenGL

feel free to look yourself, the API contains graphics utility functions

[ember hazel]

complex effects and apps, your apps will need to talk to the kernel's graphic drivers somehow and GL is the standard for that

[cyan phoenix]

uh sure.

ill implement the GL api at some point

[cyan phoenix]

severe kernel instability is occuring 💔

i need to rewrite my ASM memcpy to use 32/64-bit writes first

non-KVM qemu really likes 32-bit writes

[cyan phoenix]

yay the kernel instability is in a system that commonly fails

this is actually good

because i think i know where it fails

[tender verge]

yay

[cyan phoenix]

WOKEN_UP_ALREADY race condition🥀

i know damn well that this is poll's fault

its always fucking poll

ill just add a mutex around my waiter structure

[torpid wolf]

is your kernel built with -mcmodel=large or is it PIE?

[cyan phoenix]

-mcmodel=large

[cyan phoenix]

i didn't really feel like doing PIE

[torpid wolf]

why not -mcmodel=kernel?

[cyan phoenix]

wait, let me make sure its mcmodel large

yeah it is

[torpid wolf]

it must be given your load location unless it's PIE

[cyan phoenix]

iirc i switched to using mcmodel=large when i moved to higher half, and mcmodel=kernel resulted in a black screen

[torpid wolf]

hm that's weird

[cyan phoenix]

yeah, it was pretty weird.

i guess I kinda forgot about it

[torpid wolf]

i am saying because, other than the fact that it's convention to load kernels at the top 2GB of the address space, -mcmodel=kernel generates better code

better in the sense of more efficient

[cyan phoenix]

let's try again actually, ill recompile with mcmodel=kernel

[torpid wolf]

you'll need to move the kernel load address

[cyan phoenix]

oh

in that case maybe later

[cyan phoenix]

the entire memory system is pretty fucked to be honest

definitely not my best work but it works well enough

and now ethereal doesnt build at all

apparently my GCC's cpuid is no longer liked by Ethereal

mutex does nothing to prevent the double wakeup

i love how the only way to trigger this bug is to grab a window and jerk it to the side really hard

only on non-KVM

this bug is not poll related its socket related!

or mutex, haven't figured it out

its circbuf related, which i believe is UNIX sockets

[cyan phoenix]

yup

i fuck up

still not sure where

too tired will fix tmr

anyways so:

• yay xHCI, but disabling slots is being buggy and leaking memory (and after detaching a device and freeing the DMA for device input ctx I get a parameter error when BSR = 1 on reinsertion)
• HID parser coming along great
• need to implement USB Hub logic
• need to fix that damn non-KVM bug and put out v1.1.1 bugfix release
• need to finish up NVMe, should take me like 15 minutes
• fix PS/2 stalling on my Lenovo Legion

[cyan phoenix]

the yes is off center im gonna kill myself

[tender verge]

i dont even see it

the yes button looks fine

[formal phoenix]

notice how it says that its inside of the abort function, did you forget a closing }

[cyan phoenix]

Oh shit you're probably right. I'll have to take a closer look

[cyan phoenix]

another day another bug

this is catching my eye, why is it losing its thread structure? this node was never supposed to wakeup

cpu_fpuInitialize is wrong too lmao

[cyan phoenix]

its pipes

specifically mouse events

[cyan phoenix]

this bug has been fixed! while the GUI runs like shit on non-KVM, it (shouldn't) crash

[cyan phoenix]

after a lot of fucking around, finally HID reading

[cyan phoenix]

another demo photo

doomgeneric port releasing later

[tender verge]

doom generic

[icy junco]

baller!!!

also can u ask ur gf how she did the shading on that sphere

i needa know

[cyan phoenix]

yeah sure ill lyk what she says

ok flanterm is actually pretty cool

[icy junco]

epicc

[icy junco]

very much is

[cyan phoenix]

i wrote my own terminal driver before it

and im kinda glad i did since the project is barely on the edge of NIH

but this will be a cool backup to test out Vim + ncurses

[icy junco]

NIH?

[cyan phoenix]

not invented here, it means the project is self standing

i have a custom libc + custom kernel + custom wm + etc. but ACPI, font renderer, and libm are stolen

[icy junco]

😭

[cyan phoenix]

i wonder if flanterm has better scrolling

probably than whatever the shit i was doing before has

[icy junco]

only one way to test

cat /dev/urandom

if u have such device

if not, on x86 u can use funny rdseed

mess around with that and very epic things happen

[cyan phoenix]

waIT

it didnt crash

no way

unresponsive though

[icy junco]

ummm lol

[cyan phoenix]

I did a light to dark gradient, bounce light on the bottom, additive layer for the shine at the top, rainbow and extra shine as an overlay layer

that's her response

[icy junco]

epic

tell her shes an epic gamer

[cyan phoenix]

done

[icy junco]

[cyan phoenix]

here's the ASCII version btw

[icy junco]

very epic gamer

https://tenor.com/view/what-are-you-what-an-idiot-sandwich-gif-16965482

[torpid wolf]

that's speedy :3

btw, idk if you know this but it is likely that it has hanged for something completely unrelated to Flanterm itself

[cyan phoenix]

yeah it did

[torpid wolf]

i (and @amber gyro ) fuzzed Flanterm with the LLVM fuzzer for days

[cyan phoenix]

it was unrelated

interesting

[torpid wolf]

like to make sure it has no bugs, especially nothing security critical

[cyan phoenix]

yeah this same thing happens on my custom emulator, and Flanterm hanged for unrelated reasons

[torpid wolf]

yeah

[cyan phoenix]

but very cool project

[torpid wolf]

ty

i appreciate

[cyan phoenix]

do tell though if there's a way to use ARGB

like i mentioned in Unmapped Nest

because I had to modify Flanterm's code to mask with 0xFF000000

[torpid wolf]

ARGB in the sense of transparency? no, because Flanterm doesn't do alpha blending

or any other form of transparency

[cyan phoenix]

no i meant

[torpid wolf]

mostly for performance reasons

[cyan phoenix]

my window server does alpha blending

the pixels stored in the framebuffer have alpha calculated using ARGB

flanterm by default uses 0x00 alpha so its totally transparent

[torpid wolf]

you want a mode where it prints pixels out with 0xff000000 ORed to them, or what? i am trying to understand, sorry

[cyan phoenix]

yes, exactly.

my framebuffers are all ARGB and by default flanterm uses a zero alpha meaning any pixels drawn are transparent :D

I'm fine with patching it just curiously if there was a builtin method

[torpid wolf]

i can add it

it sounds useful

but isn't the way alpha blending works the other way around?

as in, 00 = fully opaque/0xff = fully transparent

[cyan phoenix]

Nope

0 = fully transparent, 255 = solid

[torpid wolf]

i see, then if that's the case it sounds reasonable to implement this

[cyan phoenix]

Perhaps something similar to SDL_MapRGB

[torpid wolf]

does passing a palette colour set with 0xff ORed to them not work?

[cyan phoenix]

Haven't tested it yet

Only really played around for around 15 mins

Can try tomorrow though

[torpid wolf]

sure no worries

[torpid wolf]

i kinda got confused because Limine's config takes a "TTRRGGBB" value for the text background, which is explicitly not AARRGGBB and works the other way around

i am not sure why it was done like this when this stuff was implemented, but it's been like 6 years so i don't really remember

maybe it's handier for a config writer

though it makes me wonder whether it should be changed to actual AARRGGBB with ff = fully opaque

(probably not worth the hassle)

[icy junco]

is mint ice cream adding alpha blending to flanterm real

[cyan phoenix]

yeah this is practically just a me thing and I doubt anyone else using ft will have issues

[torpid wolf]

no, that's not happening, lol

[cyan phoenix]

What you could do is just make it 0xFF if some config value is set or 0 otherwise

[icy junco]

ok mint ice cream

[cyan phoenix]

Nah my window manager alpha blends automatically so flanterm's 0 alpha makes it transparent

[torpid wolf]

yeah, i was basically looking into doing just that

[torpid wolf]

ok Krispy Creme

[icy junco]

😭 💀

thats diabolical

[cyan phoenix]

is this some weird pet name thing

[torpid wolf]

it's a reference to AI Deltarune

[icy junco]

lol

[cyan phoenix]

lmao

[torpid wolf]

also because their pfp is Kris

[icy junco]

yes

[torpid wolf]

yes

[icy junco]

i agree

[cyan phoenix]

i understand

anyways it's like midnight for me and I have a decent sleep schedule for the time being lmao

[icy junco]

i should not even be typing this i should be trying to sleep 😭 🙏

[cyan phoenix]

gn yall

[icy junco]

goodnight!!

[torpid wolf]

gn!

[icy junco]

decent sleep schedule imagine

[torpid wolf]

same

it's almost 7 am here

lol

[icy junco]

6am here lol

barely can sleep

[turbid summit]

use blender you can do this

[cyan phoenix]

Actually she did this on Ibis Paint

[turbid summit]

oh i didn't know this thing

i do everything in blender

[cyan phoenix]

its actually just a paint app im scared by her skill

ethereal would be incredibly ugly without her

[turbid summit]

next step : do the same as the guy who make photo realist things in paint

[turbid summit]

you would have to use "programer art"

[cyan phoenix]

working further on HID parsing

finally works on real hardware + can receive interrupt transfers

just need to implement the extraction, driver logic, and write a quick keyboard driver

for drivers i won't worry about Ethereal's USB stack weak_bind priority setting and just trust HID drivers

[cyan phoenix]

HID keyboards are done

mostly

I still have to do auto-repeat and more but I'll get to that later

works on real hw too

time for HID Mice

[solid cobalt]

Cool cool

For modern hw u probably want hub support

[cyan phoenix]

yeah agreed

[solid cobalt]

Since most stuff is behind internal hubs

[cyan phoenix]

i have to rework a decent bit of my USB stack though

[cyan phoenix]

tested on my 2024 Lenovo Legion and it worked somehow

but yeah my USB stack was designed with UHCI/EHCI in mind and has some xHCI hacks

would need to modify it to call a specific xHCI method to set the initial slot context up correctly

[solid cobalt]

I was talking more about PCs, you know how they have multiple ports e.g. on top of the case? That's connected via one wire to the motherboard so iirc it appears as a hub

[cyan phoenix]

and add an init_on_hub_port method

yes it does, you are correct

hubs are next on the agenda

VMware also emulates hubs so for full compat gotta do that too

[solid cobalt]

Nice

Hubs are notoriously hard, managarm still doesn't handle them that well

[cyan phoenix]

yes, ive had working hub drivers for UHCI/EHCI so hopefully wont be too much of a challenge (?)

afaik its just configure the initial slot context and configure the hub port endpoints differently

[solid cobalt]

Dunno, I just know its a lot of pain

[cyan phoenix]

eh maybe xHCI is easier?

@dusky dew pinging because you'd know

[solid cobalt]

xhci is the thing managarm struggled with

Idk the state of other hcis tho

[cyan phoenix]

then maybe im fucked

but so far xhci is working pretty god

*good

[cyan phoenix]

UHCI/EHCI operate totally differently than xHCI

they use QH/TDs

[solid cobalt]

Yeah I know they're a lot more manual

[cyan phoenix]

surprisingly i found them easier

its just QH/TD

[solid cobalt]

U have to allocate addresses, flip bits etc

Xhci kinda does that for you

[cyan phoenix]

yeah but you also do that on xHCI

for the TRBs

[dusky dew]

xHCI is decent if it has sane hardware to my knowledge

now the quirks, that's the fun part

[solid cobalt]

Yeah

[cyan phoenix]

as with everything lmao

i wonder what quirks linux does

[solid cobalt]

It has a quirk mask for xhci

With 99999 quirk types

[cyan phoenix]

https://github.com/torvalds/linux/blob/master/drivers/usb/host/pci-quirks.c#L1158

[dusky dew]

all of em

[cyan phoenix]

oh that's only PCI

[solid cobalt]

Thats pci

[cyan phoenix]

yea

didnt realize this was porn ok

so far what i can see is
https://codebrowser.dev/linux/linux/drivers/usb/host/xhci.h.html#1569

[solid cobalt]

I think it has a separate file for xhci quirks

[cyan phoenix]

very possible idk

[solid cobalt]

Not at my PC atm so cant check

[cyan phoenix]

dw about it

anyways back to mouse driver

[cyan phoenix]

i just finished writing the shittest HID driver i think mankind has ever seen

complete with bad hacks to reverse linked lists

and shitty UsageId stacks

[icy junco]

i thought the thread got deleted

😭

was scared for a sec

[cyan phoenix]

nah we cant kill ethereal

i just forgot to update lol

latest things that are happening:

• unix sockets are corrupting the heap ‼️
• the entire thing will crash very quickly ‼️
• but USB mouse and keyboard work on my hardware

[cyan phoenix]

ethereal's sleep queue has now officially fucked itself

completely randomly

[cyan phoenix]

this was not random actually lmao

I fucked up a memcpy

[cyan phoenix]

My window manager feels a bit unstable on real hardware

The size of my Legion's screen is insane so could be related

its actually a 2560x1440 panel but Ethereal looks very tiny on it

Here's a picture of Ethereal running (the moment I took this photo it crashed)

that weird background is because i dont scale (didn't want to implement that into my GL yet) and the default lines bg is too small

new debug information in Celestial as well

RPS, while accurate, there are no clips, so it's just "loops per second"

my wm doesn't use poll() for its sockets (it REALLY should, but my poll() is broken and I keep forgetting to fix it)

Now the wm crashes more because FreeType sucks, klange was right

[cyan phoenix]

IRC clients + ping

once i get a web browser maybe ill log onto this Discord

[dull crater]

ok bro

only managarm has flexed that

#voice-0 message

[cyan phoenix]

i have an in progress Netsurf port :D

[dusky dew]

yeahhh

about that

[gentle burrow]

do you like rust

[dusky dew]

highly unlikely that netscape can run that

[cyan phoenix]

yeah i've been more focused on other things so it mainly consists of initial build support

but to be honest there's also the discord API so it's not that bad

[cyan phoenix]

wait did i say netscape?

i meant netsurf oh my god

[dusky dew]

Sounds like custom clients. That gets you banned over here (ToS violations), and risks your account in general. Not the smartest move probably

[dusky dew]

Whatever is in Debian sid does not run discord in the browser (this is latest release)

[cyan phoenix]

Not a selfbot, instead a normal Discord bot (just as a PoC)

I'm not an idiot.

[cyan phoenix]

oh well :D

[dusky dew]

That’s gonna require cooperation of every server you wanna join tho. Nasty. Would work for a PoC, creative thinking I must admit

[cyan phoenix]

yeah, just in a test server ofc. nothing like that

Discord API requests can be sent over curl iirc if you set the right headers, and considering i have a curl port then

[dusk crown]

it's also against tos to use a bot token as a user

just throwing that one out there while putting on my "ackshaully" discord dev nerd face

best thing to do is either get electron apps to work, or discord in browser

[topaz prism]

port chromium

[dusk crown]

lol I guess? sounds like a lot of work

[topaz prism]

I was mostly joking lol

does indeed, sound like a pain

[cyan phoenix]

really? i mean its just sending API calls though

how is that against TOS

[dusk crown]

it's against API to use the bot API for interactive user apps like a custom client, as it's against the rules to automate user tokens

just is, discord be weird

I guess they don't want competing clients

[dusky dew]

that

and if we as a server find out, and don't act on it, we are at risk

[dusky dew]

have tried, have actually ran shit, is a lot of pain, took me a solid month (also it's fucking massive)

[topaz prism]

[topaz prism]

wow! very impressive, and does indeed, sound very painful 😅

[solid cobalt]

We need this back

[dusky dew]

Yes yes

But Managarm is currently on a bug hunting and correctness spree outside of the secret projects ongoing

[solid cobalt]

Understandable

[icy junco]

FUCKING BALLING

wait no

this is old

[cyan phoenix]

v1.2.0 is preparing for release

the remaining things to fix:

• some minor HID bugs that certain mice dont like
• auto-repeat in HID keyboards
• calculator GUI app :D
• GPT partitioning support
• maybe some more work on R/W ext2?

also window manager optimizations as always :D

school starts next Wednesday and then this thread will slow down

[cyan phoenix]

a man can dream 😔

impressive work

does Chromium by default use those Windows 7 styled buttons? i cant recall it doing so

[tender verge]

this looks like windows vista if you look very far

[cyan phoenix]

design inspiration

[cyan phoenix]

been considering doing more ports but so far this is the list

goodnight everyone :D

[dusky dew]

Yeah old

[dusky dew]

It does with my patchwork on that version

[dusk crown]

"so that's what life would be like if id invented the finglongerported chrome"

[gentle burrow]

gotta port rust

[cyan phoenix]

How is the hell

[dusky dew]

bad

but not currently working on it

[cyan phoenix]

...okay then

Anyways good morning

[cyan phoenix]

Checklist for today

[cyan phoenix]

I got completely distracted because I felt like writing an AML parser, but that's hopefully going to be completed + integrated into the kernel :D

(OSI test from uACPI)

other than that, I worked a bit on wm protocol and some bugfixing (such as ports of Bash and Vim)

Looks like there's still some annoying process bugs in Ethereal but I'll have to debug them further later

I also did a bit of alpha blending in the compositor to test it out (sadly the compositor does not like it )

Fixed signal bugs + TTY bugs as well :D

Gonna keep working on AML parser using uACPI tests

[solid cobalt]

cool

[cyan phoenix]

oh wow hi infy

[solid cobalt]

hello

[cyan phoenix]

as i was saying i dont know what you're talking about with copyright headers and folder structures?

the copyright headers are just what i wrote that clarifies the BSD 3-clause and name of the component

[amber gyro]

libpolyhedron looks quite similar on first glance to toaruos' libc, with some folders missing

[cyan phoenix]

let me see

at a glance i see it yeah

[amber gyro]

like, its not the exact same, and im not claiming it is lol

[cyan phoenix]

you might find the contents of the folders are different

[amber gyro]

yeah

[cyan phoenix]

i mean i just picked that folder structure because i liked organizing them in header files

[gentle burrow]

i wouldn't say copying a folder structure is stealing code

[cyan phoenix]

well to be fair some of my libc is ToaruOS code

(attributed)

[amber gyro]

yeah, but seeing toaruos copyright headers made me compare them lol

[cyan phoenix]

its a fair assumption

[amber gyro]

basically what im saying is that going around claiming everything you wrote is NIH while having very much taken code from other projects in not small amounts does come across a bit weirdly

[cyan phoenix]

yeah actually i remember the @file @brief structure from there

[cyan phoenix]

id say small enough? i mean its just libc and libgfx afaik

maybe kernel has some refs but if so that's probably really old by this point

i really dont remember

wait i can just check

hm that's a bit more than id like

[amber gyro]

yeah lol

libgfx is the window manager or something, right?

[cyan phoenix]

thanks for bringing it to my attention, ill get most of these removed

libgfx is my graphics system, celestial is wm

libcelestial is library for wm

[amber gyro]

do try to remove them by not just like, removing attribution lol

[cyan phoenix]

yes i meant to rewrite the functions :D

[amber gyro]

you dont have to remove it ofc, but if you really want to NIH everything, do it properly lmfao

well, "properly"

nothing wrong with how youre doing it rn

just like, dont call something you didnt NIH "NIH"

[cyan phoenix]

afaik its just little things like some flow control in E1000, ATA configuration code, TCP signatures, and the most grievous being printf

[cyan phoenix]

i thought almost NIH covered that but yeah this is a bit much, i thught it was just like one or two files

thanks!

[amber gyro]

np :3

[hoary pine]

do it in your own language and evade any accusations of plagiarism forever

[cyan phoenix]

another task:

• remove ToaruOS

so i think ill start with kernel on that one since it barely has any refs to ToaruOS (just long replaced concepts, then libc, then gfx (since it has some stolen MMX alpha blending code that I think will be a pain in the ass to rewrite)

[cyan phoenix]

write it all in Zig, then no one will steal it

[solid cobalt]

jackal

[dull crater]

lmao @ when NIH master turns out to not be NIH

[cyan phoenix]

?

man i did not realize this much of Ethereal was taken lmao

tomorrow im removing this shit

no need to rub salt in the wound 😔

[solid cobalt]

i mean acpica alone is probably bigger than the rest of your os

[cyan phoenix]

actually ACPICA is totally optional in Ethereal

I wrote a MinACPI subsystem that just does basic MADT parsing and it does still work

ACPICA is literally just there for poweroff

i was gonna do _PRT rerouting with it but i guess ill just finish my AML parser and do it custom :D

[solid cobalt]

yeah thats until u depend on more stuff from acpi

[cyan phoenix]

^

otherwise besides checking PS/2 FADT bit (and maybe scanning for a PS/2 device) I have no use for it

so technicaly I still get a NIH point

[solid cobalt]

_PRT is gonna be fun

[gentle burrow]

when dtb

[cyan phoenix]

musl is only needed for ports

[solid cobalt]

because it also requires _CRS which requires parsing all acpi resources

[cyan phoenix]

aarch64 is long abandoned

[gentle burrow]

riscv64

[cyan phoenix]

libfreetype is the actual only required external component sadly

i dont have plans to write a TTF rasterizer yet

ive been planning on removing ACPICA entirely and this AML parser is a good excuse to shave off a good like fucking 30s of compile time

[topaz prism]

I would love to also go mostly NIH, but I refuse to write an AML parser

[cyan phoenix]

That's valid I thought I was being special using ACPICA

it was never actually ethereal's plan to attempt to be be fully NIH

[hoary pine]

define "fully NIH"

[cyan phoenix]

i was planning on using X.org but then when I realized the support needed by libc I thought it was just "go nih or use external everything"

[cyan phoenix]

edited

mb

[topaz prism]

ye that makes sense

[oak holly]

Instead of removing stuff in panick, maybe pay less attention to aggressive admirers?

[cyan phoenix]

eh i respect my critics and augustsky and electronode are right, i cant call myself NIH until fully all external components are NIH

[cyan phoenix]

dunno what @dull crater 's problem is but id rather take a break and go NIH first

*almost NIH, I still don't have plans for a TTF rasterizer

[topaz prism]

wait they changed their name again 😭

[cyan phoenix]

but ACPICA + musl + others should be able to be kicked out

including the ToaruOS code 💀

[dull crater]

the part of misrepresenting stuff is what rubs me wrong

[cyan phoenix]

for the record though i never called myself fully NIH

i called myself almost NIH

[cyan phoenix]

im not misrepresenting, i said almost NIH

[solid cobalt]

wdym almost

[cyan phoenix]

the thing i said in #when-your-os-goes-crazy is true.

[cyan phoenix]

ACPICA, freetype, and now this

acpica is easy to kill but freetype will be a pain

[solid cobalt]

libc, gfx, some kernel code as well right

[cyan phoenix]

yeah but that's easy shit to remove, i was just lazy and then forgot about it

[hoary pine]

for the record theyre like 14

[cyan phoenix]

after looking over the kernel code there isnt really anything there anymore? Its just old concepts no longer in use really much

[cyan phoenix]

again my age doesnt matter

i deserve to get called out for shit

[topaz prism]

almost: an adverb;
not quite; very nearly.

[cyan phoenix]

i wont hold it against anyone who criticizes me justly

and they are valid to criticize, this is a concerning amount of external code in an OS that's supposed to be pretty NIH

i thank electronode for bringing it to my attention :D

[solid cobalt]

15 iirc

[cyan phoenix]

yeah 15

but it doesnt fucking matter guys

if im old enough to comprehend C then im old enough to comprehend NIH

[solid cobalt]

are you gonna do an aml interpreter now?

[grizzled coyote]

lmao

[cyan phoenix]

yeah, i will finish ToaruOS removal tmr

[amber gyro]

AML is very complex

[solid cobalt]

u wanna do it in a day?

[amber gyro]

i would consider very carefully if you really wanna go there, given that making a AML interpreter (that is good and works on actual hardware) is not a small task

it is a task that, from what i heard / looked at, seems similar to making a entire OS

[cyan phoenix]

yes, i know. i understand AML

[cyan phoenix]

nah multiple

[solid cobalt]

aml speedrun

[cyan phoenix]

but i understand AML and ACPI enough to the point where I do have a basic interpreter

[topaz prism]

dang

[cyan phoenix]

right now im working on making it through the uACPI tests starting with OSI to get a grip on how i shuld structure my parser API

[cyan phoenix]

its really not that bad

i did HID bytecode before, and while this is structured differently its just bytecode at the bottom.

[amber gyro]

id trust whatever @solid cobalt says over me though given that they wrote uacpi

[solid cobalt]

yeah basically mine took a few years on and off and like 20k loc iirc, also a lot of windows reverse engineering

[cyan phoenix]

yeah i know

[grizzled coyote]

find . -name '*.c' -or -name '*.h' -exec sed -i '/[Tt]oaru(OS)?/d' '{}' ';'

[amber gyro]

i have written bytecode interpreters too, specifically for amd atombios

[cyan phoenix]

Yours is also designed to be a commercial framework

[amber gyro]

thing with AML is that it is totally shitfucked basically lol

[solid cobalt]

[cyan phoenix]

yes this is true

[cyan phoenix]

🤔

[cyan phoenix]

thing with ACPI is that it is totally shitfucked

[solid cobalt]

yeah, basically the spec omits half of the things and lies about the other half

[cyan phoenix]

yes it does

[solid cobalt]

thats true

[cyan phoenix]

mine is just designed to be a simple ACPI interpreter

not designed to run on Windows at all

anyways if anyone else wants to roast the code then feel free to let me know what else i have to remove to be fully NIH

[solid cobalt]

well simple or not you still want it to not hang/overheat your pc

[cyan phoenix]

of course

[solid cobalt]

anyway if you can pass most of the tests and not die on the --large test suite then it probably works

[cyan phoenix]

yeah and then i just have to cleanup the shitfucked API

shouldnt be too hard though

right now im just slowly working my way through it

[solid cobalt]

--large basically just feeds like a thousand real hardware blobs

[cyan phoenix]

yes I read the docs

er actually maybe that was banan telling me

[solid cobalt]

👍

[cyan phoenix]

🤷

anyways if my code has been roasted enough i should get back to work

[amber gyro]

means you cant call yourself running any ported software fully NIH either i guess

[solid cobalt]

it should be easier since tests basically verify nt semantics i've RE'd etc so you dont have to do it yourself

[cyan phoenix]

running ported software isnt ever NIH

[amber gyro]

although not sure if that is a important metric

[amber gyro]

yeah

[cyan phoenix]

my kernel uses nothing from libm though

i think it was just to build some random ass port? idk

so i can just drop musl with no consequences

[solid cobalt]

well math helpers are pretty hard to get right

[dusky dew]

Well true, but I’d draw the line at the libc level if you’re actively doing ports. Hell even that can be addressed, if properly marked as such. I wouldn’t expect a fully NIH kernel to have a NIH libc, unless the readme says otherwise (for example)

[cyan phoenix]

README does not mention anything about NIH iirc and it should not. NIH was never a goal of Ethereal it was just something i realized that i was doing

[dusky dew]

Oh I’m not calling you out there

That was a generic example

[cyan phoenix]

yeah im just saying that NIH wasnt ever a goal but i was wrong to keep calling it almost NIH if it was this far

[dusky dew]

Yeah

[cyan phoenix]

i mean sure but im capable enough to figure it out?

[solid cobalt]

yes

[torpid wolf]

mlibc also uses libm from musl

which i believe in turn comes from the sun microsystems libm

so basically everyone copied each other

lol

[solid cobalt]

dammn mlibc is not truly NIH

[cyan phoenix]

is anything truly NIH?

[solid cobalt]

@hoary pine 's work

[cyan phoenix]

if you didnt make the resources you used to create it

is it arguably NIH?

[cyan phoenix]

MINTIA is probably one of the coolest things ive seen in a while

[solid cobalt]

what do you mean by resources?

[cyan phoenix]

any sort of documentation/tutorial/whatever

im mainly joking

[solid cobalt]

well partially yes

but it's hard to make anything that's not unix basically so it will have to use the same underlying algorithms and general structure etc

[cyan phoenix]

slowly getting through it

god damn the encoding on CondRefOp looks annoying as shit

[solid cobalt]

most ops are this way

they return a value and also have a target

e.g. Add etc

[cyan phoenix]

yeah i know

im just saying reading the AML spec page

CondRefOp requires SuperName which needs SimpleName which needs Arg op, Local op, etc.

[solid cobalt]

btw you should probably start with something simpler, like the return_* tests

u can implement that and actually pass that test in a few hours

[cyan phoenix]

sure

i just figured _OSI would be easier

[solid cobalt]

well it requires named objects, condrefof, conditional logic etc

so i wouldnt say so

[cyan phoenix]

im up for a challenge :D

[solid cobalt]

sure lol

[grizzled coyote]

uacpi wasn't built in an afternoon btw

[cyan phoenix]

why does everyone keep joining the thread to tell me

im not remaking uACPI

[solid cobalt]

// Name: Return Integer (Byte 0xCA)
// Expect: int => 0xCA

DefinitionBlock ("", "DSDT", 2, "uTEST", "TESTTABL", 0xF0F0F0F0)
{
    Method (MAIN, 0, NotSerialized)
    {
        Return (0xCA)
    }
}

These ones are nice to start with

[cyan phoenix]

if this explodes in my face at least you guys get a good show so id just let the show go on

[cyan phoenix]

yeah i just need to fix my MethodOp incorrectly moving the pointer for some reason

fiexd

alright im gonna take a break this was a very pleasant way to find out i have a ton of shit to fix

[solid cobalt]

wdym?

[cyan phoenix]

realizing how much ToaruOS junk is in my OS

by #when-your-os-goes-crazy callout

[solid cobalt]

lol

[cyan phoenix]

hmm FreeType seems to rely on Musl

(Musl's math library)

That means removal will be a bit more challenging than I thought but I'm dedicated

[solid cobalt]

maybe you could just accept that it's not completely nih and be okay with that

[cyan phoenix]

no way i feel bad for saying that it was almost NIH before

i have a guilty consciousness

[amber gyro]

It's okay, dw

[cyan phoenix]

nah i also would like to be NIH

so many OSes here are very impressive like #1230349543623757845 and #1350438975386423296 but the only up I have on them is being more NIH

so i gotta maintain it

i work best under competition

[amber gyro]

Does it matter, though?

What matters the most is having fun with your project

[cyan phoenix]

i work better under competition. ethereal has grown the most when i was competing

i have plenty of fun :D

just also hate to see the ToaruOS code considering the main thing I knew about was the inspired clock system and thats the only real thing I wanted to be taken

[amber gyro]

If you have fun NIHing everything, do so, but if it gets too much to NIH everything, no one is gonna dislike you for having a dependency or two or a billion

[cyan phoenix]

i mean i might just port Newlib for fun to test my kernel's system calls

i can hold off removing musl and freetype

but ACPICA + ToaruOS code has to go

yeah actually first im gonna port newlib

[solid cobalt]

why newlib specifically?

[cyan phoenix]

from my experience its the easiest to port

although i do recall it not having some sys headers

uh oh here comes mlibc guy

[dusky dew]

It’s also incomplete as hell

Last I checked

[cyan phoenix]

fine ill do mlibc

[dusky dew]

Do whatever you want, I’ll let the record reflect I did not advise anything here, just made an observation

[cyan phoenix]

whatever you say

but yeah i guess i wouldnt mind giving mlibc a go, it worked well for Nyaux

[amber gyro]

@cyan phoenix is it okay if I DM you for a second? I have a couple questions I'd rather ask in DMs about your project

[cyan phoenix]

yes

[dusky dew]

This thread is (temporarily) archived as the OP has left the server. Normally we wouldn’t put out a notice like this for threads, but this one had gained some traction, so a note seemed like a good move. This thread will be locked in the meantime to preserve its contents

[dull crater]

welcome back

[cyan phoenix]

oh ok good

we up

many changes since:

• mlibc ported and old libc dropped
• vm, vfs rewritten
• page cache yippee
• xhci fixed
• rtl8169 fixed
• new window server rewrite
• cool window manager shit
• custom widget framework
• half life port

[solid cobalt]

welcome back

[cyan phoenix]

file explorer based on widget framework

[cyan phoenix]

still working on my page cache, it seems to give "decent" performance

much better than not having it ofc

[cyan phoenix]

butterscotch on ethereal

[cyan phoenix]

my tcp driver is trying its best for ClassiCube

[topaz prism]

oh hey you're back!

[cyan phoenix]

yes we are back

[cyan phoenix]

debugging GTK3, glib seems to have some corruption issues on its poll queue

shouldnt be too hard as i made a small reproducable

famous last words

damn, i was casting a short to a long and writing it

[dull crater]

btw running glib tests was pretty nice to do

I haven't done that in a while on managarm, but last time we did, it uncovered a bunch of issues that we fixed

[cyan phoenix]

hey thanks that's a great idea

i was just writing my own tests i forgot they had a testsuite

did fix above issue and now GTK runs and can be interacted, but its slow as shit

i kinda just want to see if webkitgtk even launches

then i can start adding these ports to ethereal's ports repo

[cyan phoenix]

That glib test suite helped me catch so many bugs that was great

[cyan phoenix]

how the hell do you build libicu

this buildsys is ass

[gentle burrow]

#! /bin/sh

name=icu
version=77.1
revision=2
tarball_url="https://github.com/unicode-org/icu/releases/download/release-77-1/icu4c-77_1-src.tgz"
tarball_blake2b="7ee4457dbb9abfee543aa91314e052282fb1c73e5fd9d9b21777d2ea32ae29c3eb3d83a78aafc7d00530435c014aa0f6196faf82ed67593c844e28395ab4720e"
source_imagedeps="build-essential pkg-config"
hostdeps="gcc pkg-config"
deps="core-libs"

prepare() {
    autotools_recursive_regen
    cp source/config/mh-linux source/config/mh-unknown

    mkdir -p cross-build
    cd cross-build

    ${source_dir}/source/configure \
        CFLAGS="$HOST_CFLAGS" \
        CXXFLAGS="$HOST_CXXFLAGS" \
        LDFLAGS="$HOST_LDFLAGS" \
        --prefix=/usr/local

    make -j${parallelism}
    cd ..
}

configure() {
    config_data_packaging=library
    if [ "$JINX_ARCH" = riscv64 ]; then
        config_data_packaging=archive
    fi

        configure_script_path=${source_dir}/source/configure \
    autotools_configure \
        --with-cross-build=${source_dir}/cross-build \
        --with-data-packaging=${config_data_packaging}
}

build() {
    make -j${parallelism}
}

package() {
    DESTDIR="${dest_dir}" make install

    post_package_strip
}

👍

[cyan phoenix]

same result 😔

[gentle burrow]

works for me

[cyan phoenix]

good for you

maybe its this ver of icu

nope its the same one

dammit

[gentle burrow]

oh you need to patch it btw

[cyan phoenix]

ohhh

wheres the patch

[gentle burrow]

diff --git icu-clean/source/data/Makefile.in icu-workdir/source/data/Makefile.in
index 266ad7a..71f7da2 100644
--- icu-clean/source/data/Makefile.in
+++ icu-workdir/source/data/Makefile.in
@@ -240,6 +240,7 @@ ifneq ($(ICUDATA_SOURCE_IS_NATIVE_TARGET),YES)
     @-$(RMV) $@
     $(INVOKE) $(TOOLBINDIR)/icupkg -d $(BUILDDIR) --list -x \* $(ICUDATA_SOURCE_ARCHIVE) -o $@
 else
+    @mkdir -p "$$(dirname $@)"
     @echo "$@" > $@
 endif
 endif

you're welcome

[cyan phoenix]

let me test it

[cyan phoenix]

no change 😭

but thanks anyways

[gentle burrow]

F

you did regen right?

[cyan phoenix]

i ran autoreconf -f -i unless im stupid

[cyan phoenix]

where did you pull this patch from

i dont see it on the bootstrap repo

[gentle burrow]

gloire

wdym it's in the bootstrap repoo

patches/icu/

[cyan phoenix]

the managarm one?

[gentle burrow]

no

mine

zinnia

[cyan phoenix]

i mean if my regen command was correct

i just ran it and it exited with success so

yeah it did regen

wait

ah thats why

[cyan phoenix]

my build env had $TARGET set

which was interfering with the buildsys

thats pretty stupid

[gentle burrow]

dam

all build systems have some flaw

[cyan phoenix]

damn i wonder why my og message got deleted in this thread

makes it look particularly ugly

[cyan phoenix]

all desps for webkitgtk compiled

its time

[cyan phoenix]

keeps trying to include epoxy/egl.h even though i have egl off weird

and keeps trying to execute $SYSROOT/usr/bin/unifdef 😔

Okay back to building

Anddd nevermind

webkitwebdriver compiled

😭

What do I even do 😭😭

Maybe j1

C++ file of doom and despair

Yeah alright we're just gonna j1 ourselves through javascriptcore

[cyan phoenix]

this is like the 4th time its tried to use EGL

from epoxy

what do i even do

i guess i need to build with libdrm and a DRI GLX backend in mesa

unfortunate

[cyan phoenix]

@shell temple thank you for mesa patches :)

now i can finally keep building

[cyan phoenix]

finally got it to compile, fixing it rn

[cyan phoenix]

k webkitgtk boots up now

it doesnt complete its connection probably because of those earlier gtk bugs

JSC is bugged too

[dull crater]

btw did I ever mention os-test to you

https://sortix.org/os-test/

[cyan phoenix]

yeah ive seen it before

[cyan phoenix]

i thought it moreso tested libc and not syscalls

[dull crater]

more, but not exclusively

[cyan phoenix]

adding credential + fd passing to my unix socket implementation

i really should do another rewrite of unix socket but i dont really care allat much

[cyan phoenix]

not dead

working on webkitgtk still

finally i can pass some more tests in glib

including the ones with SCM

[cyan phoenix]

i will taking a break from trying to port WebKitGTK in order to improve my existing GTK applications and port some other fun stuff

since the sysroot is incredibly big and i dont have root support i am tired of it

[cyan phoenix]

pushing all of my ports soon

[cyan phoenix]

got window resizing working in my homebrewed window manager

[cyan phoenix]

larp overdrive

[shell temple]

Bro is using old mlibc 🥀

[cyan phoenix]

can old mlibc not run fvwm3 🥀 🥀

[shell temple]

Its bugged