AuroraOS

[mental otter]

Which i guess is vague

Does it include patch files that you need to modify a bunch?

Who knows

[quaint cape]

idk

ehh

yk which option im missing?

[mental otter]

You using mlibc master or release

[quaint cape]

master

[mental otter]

and did you use the script to download Linux headers?

[quaint cape]

maybe i should use release

[mental otter]

No I use master

[quaint cape]

which script?

[mental otter]

https://tenor.com/view/slam-smash-computer-slow-rage-gif-15806211

[mental otter]

It’s included with mlibc

[quaint cape]

ohh y mean that

yes i copied ur stuff

word for word

[mental otter]

Yeah yeah

Did the symlinks copy right?

[quaint cape]

ima have to check

else it might just be something w the config

wait do you need the linux option enabled?

[mental otter]

No

You do need to manually say no to them though

But I know mlibc likes having the headers

[mental otter]

What do they link to?

[quaint cape]

they aint links

symlinks*

[mental otter]

Copy the symlinks from the demo project

[quaint cape]

they have actual content inside of them

[mental otter]

That’s where I got my symlinks

Huh

That not right

[quaint cape]

Ima just steal astrals

[mental otter]

@quaint cape Did you also take my build script and patching system?

[quaint cape]

Yes.

[mental otter]

https://media.discordapp.net/attachments/766703671504797696/1198114808902451220/ezgif.com-video-to-gif_2.gif

Damn

[quaint cape]

Ik right.

[quaint cape]

Well i did rebuild and rename quite a bit

split it up into different files

[mental otter]

Remember never do this for major things or your knee caps say bye bye

[quaint cape]

🙏

[mental otter]

Try it without splitting it

[mental otter]

Plausible deniability /hi

[quaint cape]

well its still the same thing

[mental otter]

The folder structure though

[quaint cape]

its prob just my abi bits being broken

[mental otter]

It may get mad or some shit

[mental otter]

Perhaps

[quaint cape]

when i just started with mlibc i was stupid enough to delete some abi bits by mistake

and never fixed it

[mental otter]

It would be nice if you got my patches working so my kernel can be credited and others and be cool

[quaint cape]

so time to redo it right

[mental otter]

[quaint cape]

I will.

[mental otter]

My kernel is meant to also be very human readable to be reference

[quaint cape]

You will also get your very own place in the readme.

:D

[mental otter]

Because I may branch of my own kernels to others

:O

Oooo nice

[mental otter]

Like I may convert to a ų kernel at some point

And having a good base is good

[quaint cape]

its always good

[mental otter]

I also want to buy the physical copies of the Intel SDM

It’s like 200 bucks for 5500 pages

Like god damn that’s a steal

[quaint cape]

damn thats not bad at all

[mental otter]

And then when people ask for help I can take a picture of the physical book XD

[quaint cape]

loool

[mental otter]

It’s basically at cost

@quaint cape Btw feel free to ping me whenever you need a hand

I love helping out with my knowledge

[quaint cape]

Thanks.

[quaint cape]

<3

[ionic palm]

ngl build system can be annoying

[mental otter]

fr

[ionic palm]

how far wicks?

[mental otter]

Stealing my mlibc and gcc build scripts and patches

Which I got from astral for the psyches partly

[ionic palm]

i use newlib

[mental otter]

Mlibc is the most POSIX libc right

[ionic palm]

also just finished my kernel rewrite

[quaint cape]

Almost done.

[quaint cape]

i did ask you for permission though

[ionic palm]

almsot completed the OS??

[quaint cape]

nooo

mlibc

[ionic palm]

Oh nice

[quaint cape]

my rewrite is almost done too

[mental otter]

Yes. And Matt told me to look at theirs

[ionic palm]

i just completed and pushed the changes

[quaint cape]

well my rewrite is a big one

[mental otter]

I did try and do it manually but things were being fucky

[ionic palm]

git was having a stroke for some reason and refused to push it but fixed it smh

[quaint cape]

i got the total files i need to rewrite if i want to port it to another arch down to 8

its a maximum of 1k loc

[mental otter]

I’m gonna be in for a doozy when going to another arch

[quaint cape]

I've been abstracting hella lot lately.

[mental otter]

But il cross that bridge when I get there

And then abstract where needed

And same for SMP

[ionic palm]

i have no worries as my OS is uni-arch of x86_64

[quaint cape]

same, ive been preparing to do it for like 1 month now

[quaint cape]

im still not brave enough to even try

[mental otter]

I think il port to arm/risc-v whenever they have good desktop options that are open source

Mainly looking at the milkV titan with the VM extention

And or the PI5 once it’s a qemu target

[quaint cape]

I need to make a devfs before i can even think about porting to other arches tbh

[ionic palm]

also happy 1000+ messages

[quaint cape]

and maybe ext2 or ext4 drivers

[quaint cape]

:D yea

[ionic palm]

all bits set

[quaint cape]

At the same thing im maintaing this im a lead dev for a rblx game so funny

[quaint cape]

yee

Im also comitee for this club: were pretty active tbh

Only 44k messages

[ionic palm]

where?

[quaint cape]

And heres in the C++ server.

Together C C++

[quaint cape]

GG

[ionic palm]

oh i see

[ionic palm]

okay

[mental otter]

DevFS is really easy

[ionic palm]

only if you have proper VFS

[mental otter]

Just set the read and write function pointers to handlers for your drivers

[mental otter]

That’s also not too hard

[ionic palm]

ASTERISK is must here

[quaint cape]

It is.

but i reworked my whole fs

like everything

cuz i didnt have a vfs before

so now i have to rewrite a big portion of my code.

its fine though right

[ionic palm]

relatable

[mild shoal]

theres been a lot of work into making mlibc posix compliant, mostly for os-test memeing

I think its already better than glibc at this

[quaint cape]

🙏

[mental otter]

Honestly I don’t even think it’s a meme. It’s a good achievement

[ionic palm]

glibc gives me stroke when i try to read it

[quaint cape]

Mathewnd ehmm do yk what im missing

[mental otter]

I think it’s above Linux now

[ionic palm]

mlibc is way more readable

[quaint cape]

Yea

mlibc is the best libc out there.

[mild shoal]

u are missing that abi-bit

[mental otter]

For context they are basing thier mlibc/gcc/binutils port off mine which is based off yours

[mild shoal]

u need to get it from linux or write your own or whatever

[mental otter]

And I got no idea why it’s missing

[mild shoal]

probably was added recently and your patch is a bit old

[quaint cape]

ima just copy it from linux

[quaint cape]

yea

[ionic palm]

also big fact that mlibc is portable whereas glibc is not because of linux

[mental otter]

I think my patches are from let me check the commit

[mild shoal]

I'm working on getting a mlibc-linux ostest thing to compare against the other libcs with a common kernel

[mental otter]

https://git.evalyngoemer.com/evalynOS/mlibc-evalynos/commit/552440d5c9eccec94e44322de6b5d0c4865bbaa9
this commit

[quaint cape]

Linux doesnt either have that abi-bit.

[mild shoal]

I mean it can be done its just a huge ass codebase

mintsuki had to use an llm to sift through it and the patch file is like 17k lines

[ionic palm]

if the OS is linux like

[mental otter]

God damn

[quaint cape]

yk what ima steal it from u thanks :D

cuz evalynos has it for some reason

lool

[mental otter]

Has what?

[quaint cape]

ur os is better than linux just so yk

[quaint cape]

That abi bit

[mental otter]

LMAO XD

[ionic palm]

that would've taken more than 6 months if was done manually

[mental otter]

Honestly good use for LLMs as long as a human verifies it

[ionic palm]

LLM can be used to automate lots of stuff, which saves time

[mental otter]

Just don’t ask it to make an IDT

[quaint cape]

high quality idt i see

[quaint cape]

17 k lines waaa

[mental otter]

@quaint cape How’s it going

[quaint cape]

Terribly.

[mental otter]

What error

[quaint cape]

Outdated mlibc stuff or sum

[quaint cape]

.

even though ive added the abi bit to my abi bits

now ive decided to clone ur mlibc and insert the auroraos sysdep to see if it helps

it might be something with the mlibc version ig

[mental otter]

Hmmmm

What sysdeps do you have?

Did you delete any of mine or set them to to stubs

That can break things

Maybe do a fzf for that file @quaint cape ?

[quaint cape]

shouldnt be a rpoblem:

it is right there

sigset_t

[mental otter]

h u h

And how does my kernels regular build script go

[quaint cape]

but when u check the sysroot they aint there

[mental otter]

What happens if you do a clean build

Delete the headers build folder

And start from ground zero

[quaint cape]

Ima do a full rebuild ig

ill update yall when its finished

[mental otter]

https://youtu.be/wbO3p7_Mf30
@quaint cape Put this song on while compiling. It may or may not have a slightly probability to increase the change of it not certainly failing

[quaint cape]

Ima go make myself some food

When ill be back its time to check on it again

[mental otter]

Sounds good

[quaint cape]

Still same issue.

[quaint cape]

check_exception old: 0xd new 0xd
     1: v=08 e=0000 i=0 cpl=0 IP=0028:ffffffff800008e5 pc=ffffffff800008e5 SP=0030:ffff80003ff95f80 env->regs[R_EAX]=0000000000000028
RAX=0000000000000028 RBX=0000000000000000 RCX=ffff800000000000 RDX=00000000000003f8
RSI=0000000000000031 RDI=00000000000003f8 RBP=ffff80003ff95fa0 RSP=ffff80003ff95f80
R8 =ffff80003ff63000 R9 =ffff80003ff69088 R10=0000000000000000 R11=0000000000000000
R12=ffffffff80015160 R13=0000000000000000 R14=0000000000000001 R15=0000000000000000
RIP=ffffffff800008e5 RFL=00000046 [---Z-P-] CPL=0 II=0 A20=1 SMM=0 HLT=0
ES =0030 0000000000000000 00000000 00009300 DPL=0 DS   [-WA]
CS =0028 0000000000000000 00000000 00209b00 DPL=0 CS64 [-RA]
SS =0030 0000000000000000 00000000 00009300 DPL=0 DS   [-WA]
DS =0030 0000000000000000 00000000 00009300 DPL=0 DS   [-WA]
FS =0030 0000000000000000 00000000 00009300 DPL=0 DS   [-WA]
GS =0030 0000000000000000 00000000 00009300 DPL=0 DS   [-WA]
LDT=0000 0000000000000000 00000000 00008200 DPL=0 LDT
TR =0000 0000000000000000 0000ffff 00008b00 DPL=0 TSS64-busy
GDT=     ffff800000014f18 00000037
IDT=     ffffffff8001cc40 00000fff
CR0=80010011 CR2=0000000000000000 CR3=000000003ff60000 CR4=00000020
DR0=0000000000000000 DR1=0000000000000000 DR2=0000000000000000 DR3=0000000000000000 
DR6=00000000ffff0ff0 DR7=0000000000000400
CCS=0000000000000008 CCD=0000000000000000 CCO=LOGICB
EFER=0000000000000d00
check_exception old: 0x8 new 0xd
``` yeayyy when i try to run the rewrite.

funny thing is that it fails in CPU which has never failed on the old system.

[ionic palm]

extremely relatable

[quaint cape]

damn my fs tweaked out

[quaint cape]

@mental otter do u yk of any freestanding printf implementation that i can borrow

im starting to get anoyed of my current system

[mental otter]

Nanoprintf

Check how I use it in src/kernel/libc/stdio.c

[quaint cape]

shouldnt be that hard tbh

[mental otter]

Honestly just copy paste it

It’s really simple

https://media.discordapp.net/attachments/766703671504797696/1198114808902451220/ezgif.com-video-to-gif_2.gif

[quaint cape]

ik right

[mental otter]

https://tenor.com/view/shrimp-simple-its-shrimple-gif-25735431

Go Wild have fun idc put a comment if you want

[quaint cape]

its lit included here

[mental otter]

Yup XD

Although I implement a few other functions to make them wrap the same

There is a way to do it without a buffer like that too

[quaint cape]

Happy new year from the AuroraOS devs

-# or ig im the only dev so ye happy new year from me :D

Ill try to focus more on AuroraOS next year, maybe get xorg ported

so i can finally do something better than this colored bs screen

[mental otter]

Ooooo nice

Is it a working mlibc?

[quaint cape]

That screen no.

I made the OS run itself after the rewrite though

I run into the same mlibc error about that missing abi bit.

[mental otter]

Hm? In what way

[mental otter]

damn

[quaint cape]

Ik right its crazy

[quaint cape]

Lemme push it to codeberg and u can download itself and try running it to see if its just something about my machine.

I have a hard time beliving that though

[mental otter]

why not try my OS build in your machine?

[quaint cape]

Did it, it works then its something wrong with my setup somehow ig.

[quaint cape]

now im getting even more confused:

its lit right there

[quaint cape]

hmm attempt 2

think i found what was wrong

[quaint cape]

@mental otter do yk whats wrong?

[mental otter]

Did you ever compile mlibc as sudo before?

[quaint cape]

no

do i need to do that?

[mental otter]

No you shouldn’t

That can break your hosts headers

[quaint cape]

hmm think i fixed it

[quaint cape]

progerss

now im on the mlibc building part

anyways i cant manage to be up more

its 2 in the morning

gn

[mild shoal]

Youre missing some abi bits definitions, make sure u have it symlinked and in your sysdep's meson.build

[quaint cape]

Ehmm Mathwend how would i modify the bits to include the ABI bit sigset_t without changing it every time manually

[quaint cape]

@mental otter do yk how this is even possible

missing it even though its right there under abi-bits/limits.h

[mental otter]

no 😭

wait

is that the right abibits folder

[quaint cape]

what is the right one:

its also in the auroraos abi-bits folder

[mental otter]

that should be right

wtf

show me the meson file

[quaint cape]

if not no_headers
    install_headers(
        'include/abi-bits/auxv.h',
        'include/abi-bits/auxv.h',
        'include/abi-bits/blkcnt_t.h',
        'include/abi-bits/blksize_t.h',
        'include/abi-bits/clockid_t.h',
        'include/abi-bits/dev_t.h',
        'include/abi-bits/errno.h',
        'include/abi-bits/fcntl.h',
        'include/abi-bits/gid_t.h',
        'include/abi-bits/ino_t.h',
        'include/abi-bits/limits.h',
        'include/abi-bits/mode_t.h',
        'include/abi-bits/nlink_t.h',
        'include/abi-bits/pid_t.h',
        'include/abi-bits/seek-whence.h',
        'include/abi-bits/signal.h',
        'include/abi-bits/stat.h',
        'include/abi-bits/uid_t.h',
        'include/abi-bits/vm-flags.h',
        'include/abi-bits/wait.h',
        'include/abi-bits/riscv-hwprobe.h',
        'include/abi-bits/sigevent.h',
        'include/abi-bits/sigval.h',
        'include/abi-bits/sa_family_t.h',
        'include/abi-bits/sockaddr_storage.h',
        'include/abi-bits/sig-limits.h',
        'include/abi-bits/suseconds_t.h',
        'include/abi-bits/access.h',
        'include/abi-bits/socklen_t.h',
        'include/abi-bits/socket.h',
        'include/abi-bits/poll.h',
        'include/abi-bits/resource.h',
        'include/abi-bits/in.h',
        'include/abi-bits/rlim_t.h',
        'include/abi-bits/utsname.h',
        'include/abi-bits/fd_set.h',
        'include/abi-bits/sem.h',
        'include/abi-bits/time.h',
        'include/abi-bits/ipc.h',
        'include/abi-bits/statvfs.h',
        'include/abi-bits/fsblkcnt_t.h',
        'include/abi-bits/fsfilcnt_t.h',
        'include/abi-bits/shm.h',
        'include/abi-bits/termios.h',
        'include/abi-bits/msg.h',
        'include/abi-bits/mqueue.h',
        'include/abi-bits/utmp-defines.h',
        'include/abi-bits/utmpx.h',
        'include/abi-bits/sigset_t.h',
        subdir: 'abi-bits',
        follow_symlinks: true
    )

[quaint cape]

it has it too

[mental otter]

is the symlink valid

whats in that file

[quaint cape]

seems valid to me

[quaint cape]

First sign of life after rewrite:

Only problem is that the fs is cooked

Also added Limine wallpaper.

So ye I mainly have two problems to solve.

1. The mlibc ahh weird issues.
2. The FS malfunctioning ten times. (and yes with the fs i mean both tmpfs, vfs and tarfs)

[ionic palm]

no VFS???

[mild shoal]

U need to add it manualy

[quaint cape]

The vfs needs no init lol

fs.c and fs.h is vfs

just that i dont call it vfs

[quaint cape]

really.

[quaint cape]

Another stupid question, but how can this even happen:

../options/ansi/include/limits.h:161:10: fatal error: abi-bits/limits.h: No such file or directory
  161 | #include <abi-bits/limits.h>
      |          ^~~~~~~~~~~~~~~~~~~
compilation terminated.
[14/311] Compiling C object options/ansi/libmlibc-musl-math.a.p/musl-gen

[quaint cape]

. If you consider this conversation

cuz i aint for sure missing a header

[mild shoal]

Maie sure u have both it on meson.build and the symlink and tbat it points to a valid file

U might need to reconfigure meson too

As in, recreate the buiold dir

[quaint cape]

its crazy as i have it in my meson.build, its symlinked points to a valid file, its in the sysroot/include/abi-bits dir.
I've already tried rebuilding it, (full rebuild)

[mild shoal]

Hm

What is your cmdline to build it

[quaint cape]

 meson setup --cross-file=${SRC_DIR}/auroraos.cross --prefix=/usr -Ddefault_library=static -Dno_headers=true build

[mild shoal]

Do u install headers properly and if so did you rerun that step?

[quaint cape]

Thats the thing thats weird.

The headers are properly there.

[quaint cape]

. Look at this SS, this is from the abi-bits dir in the sysroot. You can clearly see "limits.h"

[quaint cape]

and when you click on it you get.

[mild shoal]

Hold on let me get to my computer

Give me 15 min

[quaint cape]

Want me to push the current state to codeberg?

[mild shoal]

Sure

[quaint cape]

Damn it wont let me have my git repo

pushed

Also keep in mind that you need to modify the sysroot file /bits/sigset_t.h and add #include <abi-bits/sigset_t.h> each time you rebuild lol

[mild shoal]

Why would u need to add that include

[quaint cape]

would be good if you fix it.

its to prevent the missing sigset_t types

[mild shoal]

That doesnt happen on my end nor in ci

[quaint cape]

thats not mlibc thats gcc build

[quaint cape]

wait have you built it?

[mild shoal]

No like not on astral nor in mlibc ci this happens

When u upload the code Ill take a look

[quaint cape]

its pushed already

clone dev btw

and the build script is ./ports/mlibc/mlibc.sh (from root)

[mild shoal]

Can u send the link I dont jabe it

[quaint cape]

https://codeberg.org/aurora-org/AuroraOS

git clone https://codeberg.org/aurora-org/AuroraOS -b dev

[mild shoal]

Ok so found an issue already

U dont set the include dir anywhere

rtld_include_dirs += include_directories('sysdeps/menix/include')
    libc_include_dirs += include_directories('sysdeps/menix/include')

For example the menix ones

I dont see that in your code anywhere

[quaint cape]

I could reinclude those im pretty sure I removed them.a feel free to copy astrals and then replace astral with AuroraOS, that might an issue.

[mild shoal]

Im not even in my laptop Im just skimming over the source

But yeah that is an issue

[quaint cape]

Ima try rebuilding with those defined.

[quaint cape]

Didn't make any change.

[quaint cape]

finally.....

@mental otter :D

[mild shoal]

What was the issue?

[quaint cape]

Absolutely no idea.

I just cloned the main branch and dropped in my sysdeps.

And it worked.

[quaint cape]

Damn i really succeded at making a great fs: 😭

[mental otter]

@quaint cape The elf’s a loading now?

[quaint cape]

No the FS is broken.

[quaint cape]

but elf loading and scheduling and shi was done before

so its already implemented.

[quaint cape]

Development paused for 5 days.

Ima be away from my computer

Still can respond to DMs from my phone though.

[quaint cape]

First day of camp guys, it's been kind of crazy tbh. Great lads though.

[quaint cape]

Back today guys, can anyone remind me of what I've been doin.

[mental otter]

your filesystem

[quaint cape]

Ahh ye, fcking fs.

Well that needs to be fixed yes.

[mental otter]

you were also gonna sign all legal rights of your kernel to mr snrub

https://tenor.com/view/snrub-gif-20270577

[quaint cape]

ofc

[quaint cape]

Mr snub the greatest.

[quaint cape]

wtf is wrong with this fs

hmm does directory creating work or not

okay path handling scuffed.

/init.elf works

but not /init/init.elf

[quaint cape]

hmm this darn fs

[ionic palm]

we gotta make a new file system named 'DarnFs'

[mental otter]

@quaint cape I wonder if you can make the FS as a userspace app on your system and run test suites on it to better get it working?

Then port it back

[quaint cape]

maybe.

i know whats failing, just not why its failing lol

[quaint cape]

Guys what does uacpi actually need?

[quaint cape]

Could you take a look at the fs if you have time, right now I'm really stuck cuz i have no idea why it's failing. Would be good to have a second pair of eyes.

[mental otter]

leme finish my merge conflict rq and il take a looksie

[mental otter]

@quaint cape link to code?

[quaint cape]

VFS, tarfs, tmpfs. To get the full errors i think you need to clone the repo and run make run

[mental otter]

this looks really fuckin sus

[vital summit]

it's weird. you don't need strlen

iterate the string once.

normalise paths before basename so the end / isn't a problem and just keep a ptr to the last / and then the basename is all after that ptr when you see \0

[vital summit]

assuming the path is stored normalised so you only normalise once not in each function call:

static inline const char *basename(const char *path){
    const char *base = path;
    const char *p = path;

    while(*p){
        if(*p == '/'){
            base = p + 1;
        }
        p++;
    }

    return base;
}

[quaint cape]

Will test once i get home.

Our school had a fire alarm so freezing my butt off rn standing outside.

[vital summit]

https://tenor.com/view/bruh-cold-gif-13280369

[quaint cape]

Unfortunately didnt fix anything

if you look on the bright side as long as its not in a long path it works.

[quaint cape]

like aslong as its not really a path it works fine right :D

/text.txt works

but not /txt/text.txt

[mental otter]

Wait

What about /a/a.txt

[quaint cape]

negative.

its something with the dir setup

it can create dirs

but cant create anything in them

[mental otter]

It can’t do nesting

[quaint cape]

It should be able to do it.

it complains about create method not being there

[mental otter]

Hmmm

[quaint cape]

like look at this 🙏 I do populate it with create func

[mental otter]

Are inode flags set to dir though

[quaint cape]

yes and no, i didnt look into that before as i just assumed that process did work but whenever a dir has a file inside of it all of a sudden its marked with file 🙏

[mental otter]

Well there it is

[quaint cape]

then there has to be something wrong with my tar fs or sum

how tf did /bin/ become a file

ohh wait no, its a dir

my debug outputs are so messy they are unreadable.

wait im suspecting everything is flipped

all files are created as dirs and then all dirs are files

@mental otter is it only me but this looks completely flipped, all files are dirs and all dirs are files

[quaint cape]

Hmm idk... This just seems so weird.

[quaint cape]

Reworked a bit more in the arch directory.

Now I realised i prob need to rework my logging implementation as it's messy asf.

[quaint cape]

I have to port bash before this summer at least.

Also thanks to everyone who starred AuroraOS, I'm hoping to hit 10 stars as I had on my old gh repo.

[quaint cape]

Reworked my whole printing system

Now it uses nanoprintf

How it looks now:

[quaint cape]

damn guys look im able to write negative bytes to a file loool

anyways its cuz everything is flipped

ima fix it today hopefully.

[quaint cape]

got sidetracked impl sysfs now

@mental otter do yk how capable your kernel needs to be to run uacpi

[mental otter]

i think its mostly PMM VMM and other stuff related?

[quaint cape]

ohh my

astrals glue

[quaint cape]

@mental otter Rate my contributor introduction: https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/docs/contributor_introduction.md I spent the last hour writing it lol

[mental otter]

Based

Do you think I am smart enough to contribute XD

https://cdn.discordapp.com/attachments/1323419288668995657/1439390497687867562/me_on_ur_mom.gif

[quaint cape]

🙏 ur one of the better people that has even been interested at looking at AuroraOS lol

[mental otter]

If I’m not too busy with my own I can take a looksie

Maybe I can port my new timer drivers

[quaint cape]

lol would be great i only have PIT rn

[mental otter]

I know what I must do

https://media.discordapp.net/attachments/766703671504797696/1198114808902451220/ezgif.com-video-to-gif_2.gif

But from myself

[quaint cape]

Implemented sysfs

Also fixed all fs issues yeeeee

Im hopeful i fixed all issues but im unsure.

@mental otter you have any ideas on how i would organize my syscalls

[mental otter]

No and mine are shit

I’ve been cooking some ideas tho

[quaint cape]

i feel like its super messy rn as they are all in one single 1k lines file

yk what ima just put them all in seperate files in a folder.

call it a day

[mental otter]

What’s all in one files like that

[quaint cape]

I got a lot of questionable things in my os lol

welcome to unreadable bs

[ionic palm]

absolute pathetic

[quaint cape]

ehh this os sucks tbh

anyways ima try to make it suck less

so ill switch them up.

i think this rewrite was really needed.

[ionic palm]

https://github.com/VOXIDEVOSTRO/AxeialOS/blob/pre-main/CLibrary/userland/syscalls.c

[quaint cape]

ohh damn thats even more unreadable than my bs.

[ionic palm]

570 lines of nonsense syscall stubs for newlib

[quaint cape]

you barely need 5 for mlibc lol

[ionic palm]

newlib

[quaint cape]

fixed it now they are all in seperate files.

[mental otter]

You need like 3

[midnight rampart]

Take a look at the kernel api header in uacpi, thats what you need to provide.

Its the core stuff a kernel needs really, some basic virtual memory stuff, spinlocks and mutexes, heap allocator, and raw pci access.

https://github.com/uACPI/uACPI/blob/master/include/uacpi/kernel_api.h

[ashen plume]

What is that formatting …

[vital summit]

I copied evelyns

[quaint cape]

ik right

[vital summit]

fact of the day: linux kernel formatting just says you should indent with 8 characters. it does not say what that character is. brb, submitting a change to the linux kernel with 8-tab indents.

[celest shoal]

lol

[mental otter]

That’s not my formatting 🥀

[mental otter]

I mean if it’s an actual good important patch if it comes up just say they don’t specify and have them fix it

[vital summit]

i copied your screenshot

[mental otter]

Like it’s not the strangest thing to happen on the mailing lists

[quaint cape]

Even more work: https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/docs/technical_details.md

It's kind of mid actually but it works atleast

now my documentation is a tiny tad better.

[quaint cape]

these classes are so boring i just write docs the whole time since that is the only thing you can do on your school chromebook

[quaint cape]

Damn my proc logic is broken smh

now proc_exec wont work 😭

it just goes to idle and complains about no new proc being found.

Fixed.

It was me somehow commenting out thread create

😭

[quaint cape]

Now i dont even know what the issue is anymore.

[quaint cape]

I'm going to try to fix the scheduler tmrw ig. Still unsure what the issue is.

[ionic palm]

relatable

[quaint cape]

Damn might have to rewrite some more proc logic.

[quaint cape]

after that i can hopefully run bash

[quaint cape]

Even more docs work: https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/docs

If you're ever bored feel free to read the AuroraOS docs.

[ashen plume]

I need to document my kernel

I only have in code comments as documentation

And even those are minimal in a lot of places

Like for example my vmm isn’t commented at all

[quaint cape]

Damn still unsure about what this sched issue is

[quaint cape]

Ehmm well i dont document a lot of the code

thats more general kernel information.

[quaint cape]

i like how my IDT doesnt even catch this exception

something has to be really messed up

[quaint cape]

Best way to fix the issue is to rewrite the whole scheduler right!!!

So that's what ima be doin.

[mental otter]

What was your old scheduler btw

And was it the logic on switching tasks

[mental otter]

Like the code to build the iretq frame and stuff

Making it crash

Or the picking of tasks

BTW it may break if the next task is the same as the old task so should check for that case and return normally rather than attempt a switch

That was an issue I faced

[quaint cape]

I'm unsure anymore tbh, it just crashes when entering anything so im thinking that its the iretq, the weird thing is that it double faults.

[quaint cape]

no, the first one didnt even return.

[quaint cape]

Looks like some sort of PF, i haven't been debugging it more than a few hours.

[mental otter]

Double fault would mean your fault handler is fucking up

[quaint cape]

it doesnt handle double faults ig idk

havent looked at it for a year

[mental otter]

If you don’t handle a double fault it goes to a triple fault

[quaint cape]

Yes, i know.

it goes all the way to a triple fault.

[mental otter]

You could steal my scheduler code

For task switching

It’s only like 7 pounds of glue and duct tape

It’s really fucking picky

I think I found out why a while ago

I think I was fucking ip alignment

But I CBA to fiddle with it when it works

[quaint cape]

🙏 ima make a double fault handler.

but first ima watch jetlag the game

[mental otter]

Wait you don’t have a double fault handler 😭

[quaint cape]

no

[mental otter]

Literaly just make it run panic()

That’s it

[quaint cape]

ohh damn i do have double fault why tf is this failing

[mental otter]

Your double fault is faulting

[quaint cape]

ohh my god why 😭

[mental otter]

What does it do

[quaint cape]

check_exception old: 0xffffffff new 0xe
check_exception old: 0xe new 0xe
check_exception old: 0x8 new 0xe

The thing interesting is that it's all just pointing to userspace RIP so it's impossible to debug lol

[mental otter]

Is your IDT pointing to user code?

[quaint cape]

How would it be doin that lol

[mental otter]

I mean you’re saying it’s pointing to userspace RIP?

[quaint cape]

Right now I'm just looking at the qemu debug traces.

The exceptions occur as soon as i jump into userspace.

Maybe I dont map the IDT into the pml4? idk

[mental otter]

Wait wait wait

How are you cloning page tables

[quaint cape]

Ima just give u the relevant code: https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/kernel/task/exec.c https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/kernel/arch/x86_64/paging.c you are good at finding mistakes

[mental otter]

Il just tell you how it’s meant to be done

Okay so the PML4 is one page

Right

You get a new page copy the entire thing

And then zero out the user half

So that the kernel mappings are the same

Aswell you need to ensure the entire upper half is filled with PML3s

And that they will never be deallocated by your VMM

That way your kernel always stays mapped across all mappings

[quaint cape]

Welp, let me check my stupid paging impl.

anyways theres my page code.

[mental otter]

This nukes kernel mappings

You memset the entire thing

[quaint cape]

ohh shit thats true

i didnt realise that

[mental otter]

https://tenor.com/view/dancing-family-guy-peter-gif-20560406

Told ya

[quaint cape]

i just thought i were supposed to memset the whole thing to wipe it

[mental otter]

The kernel needs to stay mapped to everyone through right

Its also why you do higher half kernels

It’s an optimization to do it like this

And also makes checking if it’s a user pointer easier

[quaint cape]

Yes, it was me being stupid. Or well my whole process related code is stupid asf. It was written a while ago.

[quaint cape]

ohh wait that didnt work

[quaint cape]

Welp time to get to work writing more docs

[quaint cape]

my main function is so short nowadays

[mental otter]

imo id inline that or replace it with arch spesific things for asbtraction

[quaint cape]

huh wdym

its a full fletched init routine stolen from Mathewnd, so i can add or remove whatever I want.

[mental otter]

like have the main fucntions call the init stuff

the starting rotuines

[quaint cape]

what

[mental otter]

put this function inside the main function

[quaint cape]

it doenst matter what i put in there

[mental otter]

instead of calling it

[quaint cape]

just a mess

[mental otter]

really?

[quaint cape]

the good thing with the init routine is that it doesnt matter what arch i have

[mental otter]

really?

[quaint cape]

yea

[mental otter]

how does it work internaly

[mental otter]

i know how id arch abstract this though

[quaint cape]

macro magic, basically you define dependencies.

i have one dependencie "arch_ok" which signals the arch stuff is done.

And that one is defined in each arch.

[mental otter]

zamm

math is smart

[quaint cape]

whatever is before arch_ok doesnt matter, its arch dependent.

[quaint cape]

does this look okay?

[mental otter]

use the HHDM request

not a fixed thing

wait what is this for?

[quaint cape]

proc

exec

[mental otter]

loading a user proces

yes

it looks fine

[quaint cape]

https://codeberg.org/aurora-org/AuroraOS/src/branch/dev/kernel/task/exec.c# could you take a look at this file, its only like 70 lines so it should be fast.

see if anything is terribly wrong.

[mental otter]

where is it crashin

[quaint cape]

as soon as the proc runs.

userspace RIP

[mental otter]

what fault

what error

[quaint cape]

looks like PF

my handler doesnt catch it somehow

also why tf is RIP 00000 all of a sudden

[mental otter]

bad stack?

https://osdev.cc/

put the error code in here for PF

[quaint cape]

hmm

[mental otter]

Reserved right?

[quaint cape]

yea

something is up with my paging.

[mental otter]

Yes

You wana steal my fancy new paging code XD

Just dosnt have page table freeing yet

[quaint cape]

idk, i dont think its the paging.

i think its the setup

[mental otter]

Maybe

What even is the reserved bit

[quaint cape]

Thanks grok :D

[mental otter]

Your paging code is probably fucked then

[quaint cape]

might be lemme check

for sure hope its not.

[mental otter]

Bit 3 (RSVD) indicates whether a reserved bit was set in some page-structure entry

[quaint cape]

Damn this things actually good

[mental otter]

Yeah it is!

[ionic palm]

ngl, to fix my problems i rely on my kernel exception or just manual tracing

looking at code obviously

[quaint cape]

Same tbh

[quaint cape]

Welp my parents just told me to use Windows instead of Linux (dw i will still keep linux on my laptop) so now I have to translate all build instructions to windows.

or use jinx or sum

[hollow girder]

wsl

🙏

[quaint cape]

Possibility

[mental otter]

WSL should just work

You just may not have KVM

[quaint cape]

yall should I do a full sched rewrite?

I feel like it might be nessecary

I might have to i just realised how much bs it is

[quaint cape]

ehh so i did rewrite quite a bit of the scheduler

ima have to rewrite my fd system lol

its so scuffed

[quaint cape]

okay so i fixed a lot damn

also arch abstracted quite a bit

[quaint cape]

it didnt fix the page fault though

current todo:

• fix this PF thing
• rework it to be per process specific fds

reworked stack setup so its in assembly

[quaint cape]

Might need to focus on fds

[ionic palm]

what policy?

[quaint cape]

huh?

[ionic palm]

what scheduler policy you use?

[quaint cape]

round robin

might have to remake that too

implemented per cpu run-queue

[quaint cape]

Which policy would you recommend

GREATTT

always love when the PID is -2321312312312

[ionic palm]

CFS

completely fair scheduler 🤓

well personally i use the MLFQ

just that CFS is way better

[quaint cape]

hmm

I like how in the latest rewrite the arch.c file has become the trash file where i throw everything which doesnt have a place in.

[ionic palm]

lol

[quaint cape]

Just enabled logging each page that maps

its taking so long to even map the initial memory

@mental otter How long do you think this will take

[mental otter]

Oh it’s gonna take fucking ages

I did this too for a log dump

I had to lower the memory to 96MB

To make it bearable at all

[quaint cape]

ohh shit

i just realised i have 1 gb memory

[mental otter]

💀

[quaint cape]

I ended up lowering it to 25 mb

[mental otter]

also you could maybe speed it up by disabling flanterm and just spitting out to E9

[quaint cape]

well its pretty fast as it is

[mental otter]

[quaint cape]

like its only 25 mb

How little memory can you kernel run in?

[quaint cape]

I'm pretty sure I tested AuroraOS a while back and it couldnt run on anything less than 20

mb

[mental otter]

IIRC the lower I went it had a random chance of crashing with limine OOMing

Maybe my newer version can handle some things a bit better

[quaint cape]

tbh ima take a break today

i'm sleep deprived so ye

gn everyone

[quaint cape]

Fixed the first page fault it was me passing phys addr instead of virt addr into my map_page func

now i get another one yeee

[quaint cape]

What should I do guys?

• Keep the name AuroraOS.
• Change to snowix

There's so many people confusing AuroraOS with arch linux and the linux dist AuroraOS

I we're also able to highjack the codeberg org name snowix: https://codeberg.org/snowix

@ionic palm what do you think

[ionic palm]

If it's confusing people, ig change?

[quaint cape]

1 vote for change.

Anyways ima try to find a good name for Axeial.

[ionic palm]

Thanks

[quaint cape]

What do you want for theme

For example I live in sweden so i like winter.

[ionic palm]

I want something legacy-like

my OS is unix-like

[mental otter]

I have never heard of that Linux distribution

Keep it the same

[quaint cape]

tbh the reason i want to change it is cuz i think aurora-org looks ugly

https://codeberg.org/snowix

This is much nicerrr.

[mental otter]

https://tenor.com/view/kris-noelle-dance-deltarune-ice-gif-3559165176338577380

True true

Though I’d maybe pick a diffrent image

[quaint cape]

maybe yea

it's just a snowflake in ascii art though

[mental otter]

It looks really strange on mobile

Though I kinda see it if I zoom in

[quaint cape]

Blue theme doesnt work well.

[chrome cosmos]

good new name. i knew i'd heard auroraos somewhere else, but figured it was one of those names which was inevitably going to be reused a lot.

[vital summit]

i should have just called my os "OS"

i mean nobody else has

[quaint cape]

yea

https://codeberg.org/snowix/snowix Still some stuff broken but this is the rough overlay

[chrome cosmos]

i'm sure someone has 😂

sweet ascii snowflake

wait, yahoo mail is still working? i migrated and lost my login details years ago because i thought it was shutting down

good building instructions. i normally don't bother trying oss unless they have a disk image, but this had me thinking about it. but i don't really want to try building on cygwin when i'm half asleep ☺️

[ionic palm]

why your repo has 0 stars?, lemme give one

[quaint cape]

Thanks :D

It's brand new so it's appreciated.

[ionic palm]

Nice

[quaint cape]

Ehmm so do I open a new thread for it?

yk what ima maintain AuroraOS until I get bash running on snowix

Then ima quit this thread and close it

[chrome cosmos]

oh? snowix is a whole restart, not a rename?

[quaint cape]

Rename + Rewrite.

Not a whole restart

I'll prob restructure a bit but not remake the whole thing.

[chrome cosmos]

ah yeah, rewrites can be good

[quaint cape]

I'm thinking of creating my own programming language called snow, and then rewriting the os in that

just a crazy idea on my mind

[chrome cosmos]

that'll up the workload. you won't be bored!

[quaint cape]

it will.

Well, theoretically it wont be thaat hard if it just is a transpiler and not compiler.

[chrome cosmos]

yeah, you won't need to optimize

[quaint cape]

Then the question is if it should transpile to C or LLVM?

[quaint cape]

exactly.

[chrome cosmos]

though with modern cpus essentially being optimizing machine-code transpilers themselves... idk 🙂

[quaint cape]

lol

[mental otter]

honestly C might be cooler imo?

[quaint cape]

yea

it would also be better

[mental otter]

probaly can get optmized better too

and intercompatability is easier

[quaint cape]

would be good for osdev too

[quaint cape]

yea i can still use uacpi, flanterm etc

[chrome cosmos]

ya, c optimization is nuts

[quaint cape]

well thats the game plan then

Ima leave this thread as AuroraOS

but I'll prob be busy making the language.

[quaint cape]

Starting with transpiler

Ima see how far i get within a week

[chrome cosmos]

have fun!

[mental otter]

@quaint cape you got this

[quaint cape]

thanks it will take a while

but its gonna be worth it

[quaint cape]

i think i got a basic version working

[quaint cape]

Is this readable?

I think it's pretty okay.

[ionic palm]

looks like *lexer for some programming lang?

[quaint cape]

*lexer

[ionic palm]

*lexer

[quaint cape]

its for snow

[quaint cape]

This is the syntax im aiming for:

import builtin.stdio


func main(i32: argc, string[]: argv) -> i32 { // 1 arg = name
  stdio.printf("Hello, %s", argv[1]);
  return 0;
}

[quaint cape]

Therefore the main function in snowix would be:

import arch.arch as arch
import lib.print as io
import core.init as init
import task.exec as exec
import task.proc as proc
import task.sched as sched

func main() -> ! {
  init.run_all_routines();
  
  proc.proc_t * init_p = proc.create();
  exec.exec(init_p, "/init/init.elf");
  
  sched.start();

  io.kprint(ERROR, "Ohh no something terrible happened :skull:, looping forever.");
  
  for (;;) {
    arch.hcf!() ;
  }
  
}

This is just me purely dreaming on how a perfect lang would be

@mental otter Does this look good?

[ionic palm]

looks very C

with the semi-colons and braces

[quaint cape]

its a mix between C, rust and Python lol

[mental otter]

I personally hate rust and python syntax and the -> stuff but it’s not too bad

[ionic palm]

Python syntax is fine, indented but rust is pure BS

[mental otter]

you should make a void retirn implicit tbh

[quaint cape]

I find some stuff in rust pleasing.

I love the -> but python also has that

Some stuff in rust is garbage though

[ionic palm]

what would it be for "in your OS?"

Scripting?

[mental otter]

Iirc it would transpile to C

[oak charm]

So apparently I can specify the domain of my values now in terms of the types.
It's not bad per se but your usage of : is kinda weird

[oak charm]

in rust -> exists because it borrows it of ML family languages, function types are denoted as a -> b
e.g. in OCaml:

let rec factorial : int -> int = function
  | 0 -> 1
  | n -> n * factorial (n - 1)

Although usually you'd write it as

let rec factorial (n : int) : int = match n with
  | 0 -> 1
  | n -> n * factorial (n - 1)

Not sure what motivated Graydon to use ->, even in ML languages you'd not see many people write functions that way (some times in special cases), however rust does allow named arguments (and it doesn't curry as well) with this syntax so it's a little weird. IMO it would have been fine if it did the Zig thing.

[oak charm]

please don't follow C conventions, it's better to have a primitive infinite loop {} construct

also use void instead of !, try to minimize symbol soup

[quaint cape]

Its a transpiler.

Not a compiler.

[mental otter]

What’s wrong with while(1) {}

[oak charm]

that works too

idk this for (;;) {} just bugs me >:M

[mental otter]

Honestly anything that’s not C like pisses me off

C++ gets a pass because it’s still mostly just C

[oak charm]

fair enough

I guess you won't like my language at all then

lel

[chrome cosmos]

this. when i saw ! { i wondered why and how we were trying to not a code block. 😉 (i must have forgotten lisp, where a code block always has a return value.) otherwise, i think i could learn to read it as well as i ever do

for(;;) bugged me when i was new to it, but that was long ago. i'm fine with it now, though i prefer breaking out of always-infinite loops. what i don't like is when languages have all possible loop options because it's a lot of mental clutter and they're never enough, lol.

[oak charm]

another option is unit if you're a pedantic type theory person like me

wouldn't want to be doing ex falso quodlibet in our type systems

[chrome cosmos]

what kind of type is unit?

[oak charm]

uhhh, you know category theory? the basic is that unit is a type with only one inhabitant, the empty tuple

[chrome cosmos]

oh i see

i'd call that 'empty', of which 'void' is pretty-much a synonym

except the type itself isn't really empty, is it? ponders

[oak charm]

the more theory bs is that:

• in category theory it's called a terminal object, every object has a unique outgoing morphism to 1 (e.g f : X -> 1)
• in type theory it's a type with a single inhabitant, a () or any arbitrary literal that the implementor decides
• in formal logic it corresponds directly to truth (⊤)

[oak charm]

no it isn't

[oak charm]

in fact there is such a type as well!

we call it bottom, void, 0 or false

[oak charm]

same thing as this just reverse it

[chrome cosmos]

interesting

[oak charm]

the more theory bs is that:

• in category theory it's called a initial object, there's a unique outgoing object from 0 to every other object (e.g f : 0 -> X)
• in type theory it's a type with NO inhabitant (Bottom in haskell IIRC)
• in formal logic it corresponds directly to falsehood (⊥)

some people like to call it Void or Empty because it's like a empty set

even tho it isn't

[chrome cosmos]

oof. my brain's imploding, but i'll remember that there are types which aren't empty although containing only 1 empty object, and i can understand there are reasons for that even though i can't (at the moment) understand what the reasons are.

[oak charm]

empty tuple is just a pragmatic choice for familiarity AFAIK

you can have a nonsense literal as well

[chrome cosmos]

fun

[mental otter]

I’m too used to C and void just refers to nothing and void* is a pointer to an unknown type

[chrome cosmos]

i'm grounded in bits and bytes. (c fits right into that, though i started with basic and asm.) theory is like another planet which looks intriguing but getting there seems like a big journey. 😉

[oak charm]

you would find this really odd as an advice, but what helped me was turning off my brain

[oak charm]

in rocq:

Inductive unit : Type := tt.

hehe
You can do the same in haskell:

data Unit = Tt

and the same thing in any other FP

and uhh in typed racket as well:

(define-type Unit 'tt)

i think

[oak charm]

I'm not too deep into analytical philosophy but as far as I am aware ⊥/0/Void symbolizes nothingness or chaos

so uhh C does that perfectly fine IME

[chrome cosmos]

when my brain turns off, no new understanding is going to happen. 😄 what works is accepting that my previous understanding is only one way to look at it.

[oak charm]

Fair, hehe, I say this is because this sort of theory is more symbolic than other things, like flip these symbols, move this here and there, that's why I found turning off the brain useful, you feel like a child playing with those shapes and trying to fit circles in place of triangles

[chrome cosmos]

yeah 😄

[oak charm]

anyways we've gone pretty far, I guess I shouldn't scare Wick X)

[chrome cosmos]

hehe yeah

[quaint cape]

what has been happening

I've just been hanging out with friends lol

[chrome cosmos]

just theory discussion 🙂

[quaint cape]

And it resulted in that u should change?

[chrome cosmos]

i'm a little confused, (seriously fatigued today,) but i think it resulted in me describing how i change

[quaint cape]

Okay, damn I'm getting a lot of projects to maintain.

AuroraOS, AuroraDB, Snowix, Snow, Snowerr, RoofNut...

I'ma drop everything exept AuroraOS, Snowix, Snow and Snowerr.

And I'll give some more updates to AuroraOS

[oak charm]

too weak

be me and have a thousand tiny projects

well uhhhh currently focused on maelstrom: diagram

greyed out means done, yes I have only completed one
I don't compile it to anything yet tho

[quaint cape]

Well tbh AuroraOS is 100 hundred small project in itself, also I have more that I didnt include cuz they are tooo small.

[quaint cape]

Anyways this is how far I've come:

import random as random 
import lib.print as io

fn main() -> i32 {
   
  i32 random_num = random.random();

  if (random_num != "21") {
    io.printf("Yeaaaa");     
  } else {
    return -1; 
  }

  return 0;
}

This works

Ima clean up the codebase and commit later today or tmrw.

I have to add a better semantic analyzer and error system though.

And make it cleaner and less libc dependent.

[oak charm]

nice

[quaint cape]

Ima rewrite the messy backend tmrw and also work a bit on maybe adding classes and doing some general bug fixes and improvement, maybe adding some runtime support too

[quaint cape]

implementing stuff

classes etc

[oak charm]

what is your object system like?

[oak charm]

this kind of looks like rust prehistory, in that rust, you also declared values this way:

let i32 x = 3;

https://github.com/rust-lang/rust/blob/d6b7c96c3eb29b9244ece0c046d3f372ff432d04/src/test/run-pass/i32-sub.rs

[quaint cape]

lol

[oak charm]

umm not exactly prehistory, that's a different repo

[quaint cape]

i am inspired by rust yes.

[oak charm]

don't understand the “lol” factor here. lol X)

[quaint cape]

its funny that it looks like rust yes.

[oak charm]

quite very much

[quaint cape]

its a mix between C, C++, Rust and Python.

That just transpiles into C.

[oak charm]

so it's a glorified preprocessor? X)

[quaint cape]

ofc what did you think

[oak charm]

Still it's nice, what are your goals outside of improved-syntax

in your opinion

[quaint cape]

Better OOB.

[oak charm]

for Aurora as well

[quaint cape]

Aurora will stay the same, few updates.

[oak charm]

that's.... every C-like hobby lang ever?

still fine though

gl

[oak charm]

you aren't working on it or are you focusing on snowix?

[quaint cape]

I will rewrite Aurora source code in snow and call it snowix

:D

[quaint cape]

I will be focusing on snowix

Once I get a bit further with snowix I'll open a new thread.

[quaint cape]

ofc.

What should I do to make it different.

Try to get better runtime support?

[oak charm]

no, really I was just saying, don't mind me
was looking forward to something different maybe

[quaint cape]

what could i do to make it different.

[oak charm]

not sure how that'd help, but if you ever want to, check out https://github.com/jart/cosmopolitan

[oak charm]

that depends on what you want to do to make it different, what interestests you etc etc

I'm not trying to sound discouraging btw, sorry if I sounded like that

[quaint cape]

im asking what you would like lol

[quaint cape]

it didnt

[oak charm]

all of what I'd like to see will make your language finish in 2074
but in general what I'd like to see is a different memory management approach, although that's a pretty hard task, something I'm kind of struggling with my self

[quaint cape]

well, that's yea....

memory management would be quite hard to change.

since it's only a transpiler.

[oak charm]

another thing are some niche new academic topics, like effect handlers, or stuff like that. but that'd get too out of topic nvm

[oak charm]

you can do it but you'd have your own library for memory allocations

[quaint cape]

Yea.

I need to create my builtin library.

[oak charm]

and whatever static analysis method (RAII, regions, lifetimes, substructural, borrows in the compiler)

this way you perform analysis on the source, and compile it to C according to the analysis (allocate value here, move here, free here) etc

[quaint cape]

Even more bugfixing.

Anyways snow is pretty capable nwo.

[quaint cape]

https://codeberg.org/snowix/snow/src/branch/main first commit lol

This is how far it has came.

import builtin.math as math

extern C {
#include <stdio.h>
}

fn main() {
  i32 result = math.add(1, 20);
  printf("%d", result);
}

sometimes it hangs just randomly lolll

[oak charm]

that is concerning lol

the compiler or the generated program

[quaint cape]

parser.

thats just on my hand though.

some more bugfixing needed.

[quaint cape]

Fixed extern statements being buggy.

Ima also have to fix the parser infinite loop soon.

[quaint cape]

Anyways, might take up AuroraOS development for now.

I'll meanwhile work on snow and fix bugs and develop the language. After a month or two I'll rewrite AuroraOS in snow.

I need some time on this language, dont want to rush it and then be needing to change everything each 10 minutes.

I might try to make some sort of borrowing system.

[quaint cape]

So far this even I:

• fixed elf loading issues. (auroraos)
• fixed some vfs issues. (auroraos)
• fixed types being broken (snow)

[quaint cape]

fixed timer issues

[quaint cape]

still fighting with PF's.

Also binge watching AWP all of a sudden.

[quaint cape]

Adding more stuff to snow now ig.

[quaint cape]

Thinking about adding some sort of config file in my initrd module to allow me to have better config.

@mental otter Do yk what format I should use? I thought about TOML, it might be good.

[mental otter]

Why not a kernel cmdline?

Like how Linux and other things do it to config the kernel

[quaint cape]

Could do, my initial thought process was to have the init config state if you're supposed to boot into the init proc or kernel cmdline.

And also regulate debug prints etc..

[mental otter]

The cmdline can be parsed right away before mounting an initrd

[quaint cape]

True.

[mental otter]

Before paging and crap too

You could have both? And have your init program read the config file and then tell the kernel certain things

[quaint cape]

I could use kernel cmdline and then store it in the config file, and then having some presets in the config file.

[mental otter]

I mean that would duplicate things no?

A cmdline parser is really simple too

Depending on what your confinging

If it can wait to initramfs mount

It can wait for the init server to tell the kernel and parse the full file

A TOML or other parser would be heavy in your kernel

Look at how my kernel handles it

[quaint cape]

Kind of yea.

[quaint cape]

There's a prebuilt library for it :D

toml-c

[mental otter]

Still like stack size limits or something

[quaint cape]

And now comes the hardest part, which file should the config parser be in 😭

[quaint cape]

FInally fixed all mlibc build error.

So now you dont have to do 3 hours of manual setup when trying to build mlibc

[prime river]

Why do you restore non callee-saved registers in your scheduler?

If I understand correctly, RAX, RCX, RDX, R8, R9, R10, and R11 shouldn't be saved by the scheduler

[quaint cape]

looks like old code.

Check out newest commit.

https://codeberg.org/aurora-org/AuroraOS/src/branch/dev

@prime river

[prime river]

Ah I see

Actually I might have been mistaken

I'm implementing my own scheduler, that's why I was looking at yours. I think you do it right

[quaint cape]

Well yea.

And I would have to say that after the rewrite my scheduler is actually semi decent.

[midnight rampart]

it depends. If you're calling your context switch code like a function you can rely on the ABI to save some registers for you, in this case they're doing it inline so everything must be saved.

I saw someone use ⁨⁨[[gnu::no_callee_saved_registers]]⁩⁩ for their context switch function the other day too, thats a neat trick if you dont need access to them (the attribute is target specific, maybe dont do this).

[prime river]

Ah I see. Yes I'm ⁨⁨call⁩⁩ing to it from C. That means my initial presumption of leaving out those registers from the stack frame in my case is correct?

[midnight rampart]

yes

[prime river]

Thanks

[midnight rampart]

np 🙂

[quaint cape]

yea

I'ma have to work a bit more on my scheduler tbh

It still crashes.

Once jumping into a userspace program.

[quaint cape]

imm so stupid

im a fcking idiot

i marked something as static instead of extern in a header

thats why the PF happens 😭

Im just that idiotic

ohh and now I get another PF-

fun :D

[ashen plume]

mine is too