#AuroraOS
1 messages · Page 3 of 1
Init subsystem coming in clutch.
I just stole Mathewnds 
That is shorter than mine.
each of my stages calls some functions
huh, this doesnt even need a loop
How’s that
void init() {
step1();
step2();
step3();
step4();
}
Yeah
Mine is an absolute bloat 
Well personally (no offense) I have a hard time understanding anything in your OS, that's not cuz it's bad just that its different and doesn't follow most OS's structure. It's also C++ which just adds to the confusion.
For me its just a blob of folders that mean nothing.
Indeed, but that's probably because of the weird coding style
And the lack of subfolders.
You don't categorise anything.
-# because obviously i'm stupid
i don't it's only one level deep
tho i promise my #1467786251510743061 would be WAY WAY cleaner then this
I program in C but I use clang to give me function overloading 
Yea, well exactly.
Personally I'm pretty happy with how the AuroraOS source has turned out.
I like how mine is but I still need to cleanup a lot of it
yea it's clean, well i didn't clean up my CSource tree for AxeialOS since when i first started its dev
Yours is pretty good yea
Once I get bash posted and then a few rewrites of a few systems and it should be perfect for a beta1 release
This code is pretty 
@quaint cape Whatcha working on rn
I'm in school.
No worries, social studies are boring asf anyways.
Fixing the remainder of bugs in my new scheduler.
Which have per-cpu runqueues.
And much more arch abstraction.
Fixed the main one yesterday.
it was my TSS being f'ed.
I'm also going to make a better paging impl.
I'm to afraid to touch my paging.
I mean x86-64 paging is pretty basic
Just some layered tables
Think of it like fixed sized radix trees
I have paging.
I'm to afraid to change it.
You can steal mine
It’s getting finished tomorrow
With VMA too
My paging is done except unmap and TLB shootdowns
I’m also gonna add support for 1GB and 2mb pages too
Cool.
I don't really have any VMA.
I'm to scared to change my paging for now.
Last time I did i f'ed up the whole OS for multiple months.
My VMA is also open source well will be once I finish it lmao
This is part of my VMA actually
Let me see your paging right now
It can’t be that bad
I also managed my pacing rework in a day
worked perfectly
Your gonna need one for MMAP later
Fixed all userspace-entry related bugs this morning.
Now my new syscall handler tweaks.
(had mixed up phys and virt in my paging permiters also used different constants in different files.
Added "extern asm" to snow
having fun commenting my own code
ima try to make snow semi readable unlike my other projects.
i'll commit snow in a week or two.
ima have to test everything before releasing it.
And fixing my messy backends.
and my semantic also needs to be reworked.
First syscall after rewrite done
and yes this is me trying to do something with mlibc
time to see what is messed up with archctl
Fixed most errors :D
though I get invalid opcode when running any mlibc related code.
Do you have SSE/FPU stuff enabled?
Wait after the systemcall?
What’s the instructions at all the RIPs
nope ima fix that
found out 1 min ago before this message
And make sure you context switch the fsbase which seems to be what this archctl is doing?
Fixed most issues now i have to make a proper mmap impl
Hmm might have to try and build and run bash after this.
mlibc does not need a full MMAP. It just needs annon pages
But it shouldn’t be too hard to support mapping tmpFS pages if you keep track of them
20k loc now'
ironing out the last bugs.
Rewriting print logic to use a ringbuffer.
Fixed some more snow bugs
Also made borrowing
And better semantics.
And some sort of half decent runtime support.
...so am I
ima have to fix issues in my sysfs now.
Stuff until bash:
- Fix sysfs.
- Fix exit syscall.
- DEbug a bit more.
It's bascially only sysfs left
and i need to actuallly run bash too.
sysfs? not needed for bash
Need some way to print stuff???
I don't have a devfs
I guess you could do so yes
ima try and fix this sysfs tmrw
Still no progress on sysfs :(
Fixed sysfs now it works
my exit implementation is broken smh
so yeee no bash for a while
get out of memory error smh
even though i run it at like 10gb RAM
ill have to investigate once i get back from todays skiing
skiing? fancy
it surely is
very cool , good luck with the os btw
thanks
you have any idea why puts outputs a lot of C++ code??
im genuinely have no idea how is panic is in C but theres std::span 
do you use c++?
mlibc does.
even then iy should not be possible
I dont.
might need to ping someone who knows mlibc internals
Also @quaint cape i realized that #1467786251510743061 has the same amount of stars as #1384888317597974600 lol
cuz i have no idea.
well this is a worse OS that is kind of dead tbh
could be, not probable though.
yeah, i mean if you didnt touch them it shouldnt break on its own
like why would only me get it when theres like 10 OS's which use mlibc
It's decent imo, i don't care about feature sets and code quality if the code is legit
and also you can run bash is already impressive
well feature wise auroraos might have the edge.
not in portability though.
i got it running before the rewrite for like 3 mins
then i decided ima rewrite everything
How many rewrites have you done
fighting back there now-
Nice, AxeialOS was rewritten only once and Modularus none
So still it's a decent and valid number of rewrites
time to fix exit impl and then get bash running once again.
struggling to build bash smh
time to rebuild my toolchain yeaaaaaaa
running into errors when building bash
finally was able to build bash
Tmrw I'll try and figure out why bash PF's upon entry.
Probably cuz I haven't implemented something.
Seems like you’re accessing garbage memory
Make sure your mmap() call is legit
Yea, well I'll take a look tmrw
I'm too tired for this shit today.
Does MMAP garuntee zeroed out memory?
No
No. There’s a flag for it
Ahh mlibc requesting anonymous pages requires zeroed pages though still
So if that’s not setup right
I’m not surprised at all page fault
I’m driving right now so im using voice to talk I’ll answer after
Yes anonymous pages should be guaranteed zeroed
I meant annon pages

I forgot to say that sorry
Yeah in the unix standard MAP_ANONYMOUS for mmap would give zeroed pages
Which if @quaint cape Isn’t doing i presume that’s fucking up bash badly 
I don’t even bother with porting anything
Even though my process api and syscall api is Unix compat
I don’t have a proper libc
You don’t even have one for your own apps?
No I have one for my own apps
But it’s in no way a full libc
Damn now my laptop panicked
Linux why 😭
I've actually never had a Linux panic .
What the f now it doesn't work
Welp f me I haven't commited my progress in so long.
just updated?
Negative. Hard crash
live?
Prob due to not being able to clear up disk.
did you touch anything recently
I'm running recovery mode to see if it fixes it.
Nope.
Okay, running older Linux version fixed it.
Zorin.
Ig I just gotta boot into grub each time I boot?
i think you can set default entry
you always?
just not the menu to configure?
didnt expect that
🙏
my first ever distro was zorin when i was 9 10 i think, but i have skill issue back then and cant install it
it was still tui back then
🙏
Its probably one of the most advanced ones nowadays.
zorin is clean ngl
prob better than mint
mint is overrated imo
ehh its a laptop i can just never shut it off
yep
ig thats a smart way
also, just asking
why codeberg
just because it doesnt have microsoft?
It's open-source.
It uses german servers.
It doesn't have microsoft.
over performance? and infinite infra money?
well its just ones taste really
personally a more practical person
i just use whatever works best
never really consider those
What does gh have that codeberg doesn't have?
performance
darker dark theme
huh?????
codeberg is slow asf
Slow????
the gui, git servers
its a nonsensical take, I wouldnt bother arguing the point
so in your opinion, codeberg is faster?
git remote, and gui interface
also GH stutters and now has bugs lmao
never seen one yet
ohh wow, I save a few milliseconds by using gh. woooowwwww
impressive.
those milliseconds really matter for you?
what are you talking about
Codeberg being "slower" than gh.
not milisecond
gui rendering is notably slower
Yes, thats what we're talking about.
on slower devices
it is your choice to provide your code and user stats to microsoft by using github, but you will be judged for it. There are good reasons people are moving away from it.
Negative.
beside its oss, what is the actual reason
reason(s)
ok, reasons
Big difference right?
its yours machine, not mine, and specifically, im talking about code search performance and file indexing
codeberg is much slower here
.
you're supporting microsoft and allowing your code to be scraped for AI usage (you'll note codeberg protects against this), codeberg is also not used as a marketing platform. This may or may not bother you, but github being a US company is a concern for many. I'm also fairly certain codeberg is a not for profit, which again is a personal choice but its a nice thing to support.
hmm
for the sake of transparency here I use both (for now), but I wont be hosting any new projects on github and generally try to avoid the github site whenever possible.
i can try to mirror
like currently have a gitlab mirror
yeah I currently have codeberg as my main mirror and it automatically pushes to the github repo for me 🙂
it has built in support for that 😉
really?
wrt to the performance, not sure what thats about - I've always found codeberg faster for everything. The exception being when its offline, which is something I've never personally encountered but people here have complained about it a few times. If you're looking at more than a file for a project imo its nicer to clone it and have the full feature set of your favourite editor to jump around.
you're looking for "Mirror Settings" in the project settings. You could alsod do it via an action script I guess (codeberg has its own CI called woodpecker and also supports the forjego one - it has free runners for both).
quick question, do you know if bash requires dynamic linking or if its possible to only have static linking
@quaint cape Did you see the thing about annon pages? Btw mlibc does give you really easy dynamic linking if you need it. And you can use the “ldd” tool to check if it’s a static binary
does it
how
also the binary doesnt even run as it requires
Do anon pages give out zeroed pages
I don’t think that’s compiling right?
If it’s asking for ld Linux
i think so yea
smh
this is my config
malloc*
Huh
Do you have dynamic linking?
I do not and I’m burnt out heavily and stuff
I’m taking a break from OSdev and I’m gonna hopefully return to the prism thing I got organized and stuff
isnt prism just a dist around ur kernel
Ive always done a custom userspace so I don't know much about porting bash :p (this is changing with my current iteration) . What's the specific issue?
Also you'll get more eyes on your issue if you ask in an on-topic channel rather than a progress thread. I would suggest going there.
.
This isn't directed at you in particular but its something I've noticed lately, that people are asking for help in their project threads and not the channels. The on-topic channels are there to ask for help if you need it!
Nope 
What is it
I fixed the corruption issues 
@quaint cape How’s bash coming along
Making dynamic alloc for mlibc rn
Sometimes it happens to me while I’m debugging it and I’m sort of rubber ducking to my channel/void and it helps and then people walk in so times with advice
Nice nice
thats understandable! I was talking about the case where people start by asking a question 🙂
sorry lol
nah you're fine lol, it was just an observation
more shared library stuff yeaaaaaaaaaaaaaaaaaaaaaaaa
I’ve been asking in my proj channel a lot yeah
continuing with bash port today
Can I ask what syscalls it was complaining about?
mprotect
i wonder if that can be stubbed for bash?
maybe?
i just implement them all
that are easy
wait cant i just return 0 for sys_getuid since theres only one user?
@mental otter do you know if i can just return 0 for getuid and getgid?
whats the difference between getuid and geteuid
you should be able to
nothing afaik 
man 7 credentials
should give you a quick rundown of all that stuff
sure lol
btw does bash need sys_clock_get?
or can i just stub that with return 0 since its not needed
I got into a bash prompt with these:
extern syscall_print
extern syscall_mmap
extern syscall_openat
extern syscall_read
extern syscall_seek
extern syscall_close
extern syscall_archctl
extern syscall_write
but to get a proper working shell u'll need a bit more
i have a hard time believing that sys_clock_get could possibly be needed
for bash its not
No fork 
i just noticed that I didnt implement my time stubs yet, but a simple ticks++ is enough for getting started and getting bash and basically a bunch of stuff to run
do i even need signals?
cuz when i try and stub them mlibc just panics cuz it wants a real signal
No
.
When I ported it, it really didnt need signals
But that was also like 3 years ago
Worst case you success stub it
Or just patch out that ensure in your local patches or w/e
or spend a week writing an ipc implementation
well, im starting to consider maybe writing my own window compositor and porting x11
both as alternatives
since i can use sdl for the window compisitor
and then bash as the ground up shell
huh well need bash ported anyways
imo the worst part of porting bash (initially) is pselect
posix select? i remember a time few if any os's fully implemented it because it was supposed to wait for so many different things. that was several standards versions ago, though
Bash versions?
posix versions. it was 20 years ago i think
Ahh, well yeah. It's honestly not too bad. I don't have signals yet so I just implement pselect with select which itself is implemented by poll
got carried away
lmao
just done syscalls for everything i have
and i made some sort of experimental ipc implementation
What syscall does bash currently give up on?
(only signals, pipes, message queus)
dunno ima rebuild it with my new ipc implementation
ima also have to spend half an hour writing mlibc implementations for all of them
How does the "experimental ipc implementation" work lol?
well it always takes 1 rewrite for me to be happy with any of my implementations
so before my next rewrite ill class this ipc implementation as shit
lmao
that's honest! 😄 i wonder how my first implementations are going to look 😕
For my non POSIX IPC that il use for my own tools I’m thinking of shared pages + a few syscalls to wait and signal
welp this might take a few days for me to patch all syscalls
i also forgot about .bashsrc and everything else i need setup
or well make all syscalls needed
i already got 50 total syscalls btw
55
huh why does it immideatly exit?
@mild shoal do i need to define those constants for it to run?
progress
No
trying to fix bugs just made it worse
input is broken though
Did you end up implementing polling?
Why polling? I’m pretty sure they have a keyboard driver. They just need the charater device to be opened right on bash so it can read characters from the terminal like how it’s outputting things
I worded that badly. I meant a polling system i.e. i/o multiplexing
(poll, select and their signal safe variants)
nope
Then how did you satisfy the pselect mlibc sysdep if you don't mind me asking?
I dont.
Interesting, then how do you get to the prompt? Are you doing bash with or without readline?
without readline
Ahh then that makes sense
well a bit buggy but works
next i gotta debug why it throws out of memory
found it
my paging_copy impl was broken
before i commit ima have to commit quite a bit
what is paging_copy copying here? the page tables?
well before for some reason i blew my whole pmm implementation by erasing kernel bindings
now it just copies the user PML4s-
dont ask about my code its crazy bad
I know my page table copying code would probably have hundreds of bugs and race conditions so I just don't copy anything. except the mmap() areas.
and lazily populate/CoW and faults.. I tried copying on an older OS, I got random memory corruption and kernel panics all the time..
well im struggling
the thing is that as i said i mark everything as shit before being rewritten atleast once
when i get whole bash working ill rewrite the syscall logic and the proc logic
and some pagin g
just rewrite it until it works.. I'm at OS attempt 5 right now 
lol well for me a rewrite is just rewriting files
i never change os
well, none of them were something serious, basically just testing grounds for allocators, paging, fork, ...
sounds good
you learn smth each time
yes, I learned not to do ACPI the next time..
why?
doing acpi with uacpi rn
Hmm anyways guys should I use meson or stick with Make?
Which is better?
I wanna settle with one before my build scripts get too outta hand.
why does uacpi require fflush?
going good with uacpi 🙏 (not)
#GP in uACPI?
4.0 already? didnt know that
personally uses cmake
meson didnt play nice with me
and makefiles just left my source tree garbled up with objects files
gnu make can build in a separate dir. see info make page 4.5
Directory Search (GNU make)
time to fix acpi driver yee
still #GP?
yep
https://www.tomshardware.com/software/operating-systems/california-introduces-age-verification-law what will this mean for hobby OS?
If they wana fine people just don’t pay. This shit is ridiculous and I seriously doubt it would hold up in court especially against small hobbyist groups
Also wonder what will happen with Linux, can't Linux just prohibit anyone in California downloading it.
Il just use Linux from scratch if I have to
it means nothing. because:
- an open source project cant force you to upgrade to the version with the age check
- online centrally managed user accounts arent a thing
- there is no mandatory "app store"
- the age check can be patched out by the end user
- open source projects arent going to pay for the commercial age verification apis theyre peddling
- theres no 'linux inc' to send a fine, especially not ine in silicon valley
- the law cant take effect internationally
so, just ignore it and carry on as normal
whole thing is half baked by someone who thinks: osVendor == (apple || google)
its not even enforceable on windows, unless they enforce secure boot only and microsoft store only
true
bash stuff?
fork being broken
Feels familiar, I had some fork bugs too...
well my fork is always broke
i hate fork
its my num 1 enemy
and context switching
thats also hell
hatred of fork is why my os exists at all
in 2009 when I was learning from tutorials I was using the jamesM tutorial
it's fork and paging model was just massively scuffed and left broken
not as an "exercise for the reader", was just ... bad.
I hated it so much I went to read the spec on it
the I decided who in their right mind builds a process model like that in the 21st century, and walked my own path instead.
Holy this fork shit is making me go crazy
reject fork return to modernity
lol
i used to hear fork criticised regularly. industry professionals really hated it lol
checking for BEOS mounted file system support functions... no
configure: error: could not determine how to read list of mounted file systems
``` huh
me trying to build coreutils
why the
when you get a message like that from configure, it's usually the last thing it tried from a list
if you search the configure script, you can find the list and what tests it makes
df needs a list of filesystems. du i think has a "stay on this filesystem" option which might need the list too
Mathewnd was nice enough to clarify to me
ima just enable the linux option in my mlibc
bash commands finally work
or well kind of work
and yea i have some logs enabled
😭
i spammed logs soo much
better version without logs:
looks like it
theres a lot of shit thats broken rn
yeah, bash and coreutils are extremly dependent on POSIX-correct syscalls. You miss 1 edge case and everything breaks...
Well I'll have to varify all my syacalls to be posix correct.
lol ima have to update my history file today
proper acpi
ofc
i go mainstream
i like how auroraos is turning out to just be another astral clone 
hmm
is this in tcg (asking as 6063 ops/s seems awfully slow)
why is your .bashrc EPERM instead of ENOENT though?
unsure why.
good question
the whole OS is just getting slower and slower with the days passing as i dont care about optimization anymore
why dont you care tho
What’s your memcpy/memset functions look like?
There is an easy optimization you can make on x86-64
(rep movsb/stosb)
what is it?
oh nvm
If I don't have anything that runs on it why care about optimization.
Once i get coreutils ported (will send SS of progress later) I'll actually have something to run on it
maybe thats a point
anyone know how to get mlibc to not freak out once i use malloc or free etc..
Iirc malloc/free are not available in ld.so. you have to disable that code using a macro or or allocate in some other way
@quaint cape Why are you making one giant commit lmfao
lol
dont bother commiting
so i just commit a months progress each monthg
and before that i do five days of rewrite
aka cleaning up logs and shitty stub impls
Commits are good incase you fuckup lol
well i did somehow almost delete my whole fs
which made linux panic
i waas able to recover it though
ZorinOS recovery mode
clutch
💀
yk why coreutils fails like this?
.
I can’t even get bash to compile dawg
🥀
Everything wants to have strange ass errors
huh????
I’m lowk gonna just report mlibc
bash is easy to compile
send me the errors
also idgaf if we talk about ur OS in here so feel free to speak
huh?
That’s what I said XD
that's why I'm doing my own libc. Everything just works. (well, at least everything I have implemented)
well i use bash 5.1
My mlibc port is based on a 4 month version so 
I should update it anyways and refactor a lot of it
firstly, use new mlibc
secondly make sure you have
posix
Yes I have posix BSD and glibc
glibc
and bsd
same shit
just decided to do ironclad approach and make it myself
And I ended up having fucking uhhh one of my mlibc patches be just adding more headers
Then everything page faulted at addr0
💀
you patch mlibc???
which headers?
Well I have a seperate source tree
I just copied astrals setup for them
well exactly which headers
Because I did not wana find like all 40 of slender man’s pages XD
im so confused rn
Also another thing
ehh
Let me get the patch file lol

Why bash 5.1 tbf
Idk my shi is cursed lmfao
@quaint cape Ping me when you get that commit in lol
So I can port in doom
if i work my ass off i can get it in later today
dont over work yourself pls
ehh dw
btw if i commit it today it will be with a broken waitpid impl
if you feel like youre up for a challenge feel free to try and fix that
perhapss
What’s the build system like?
ehmm manual .sh files
if you feel like youre up for the challenge change that tooi
huh?
git submodules
i dont use them btw
ehmm #lack of build system
dev

ty
oops
pls make a build system for this
i really need a better build system
doom is gonna make that thing huge
steal this lol
why did i comment these
💀
like why did i remove that out
well i do have that
its outdated asf though
look at meta/ramdisk.sh
there are none right
isnt that the initrd?
also im gonna shill for jinx https://codeberg.org/Mintsuki/jinx
my build system aint that bad
./ports/mlibc/mlibc.sh
./ports/bash/bash.sh
then copy the sysroot/usr into initrd/usr
fym new branch
and done
git clone https://codeberg.org/aurora-org/ports
fwiw this can also handle host packages too like gcc/binutils
firstly ima switch make for meson
and then ima consider using jinx
AuroraOS got a new build-system.
@mental otter hows it going w porting work=
im working on fixing my PS.2 driver
ahh lmk if you need help
I changed my ps2 code
It now deadlocks on real hardware
💀
I did do a fix for IRQLs though
Sounds like my average change.
Nah it’s my IRQLs
And shedular spinlocks

Well I think
If I undo the “fix” it works again
Reworking my ports impl to use jinx
fun
cant build gcc now that i use jinx
yeeeaaaaaaaaaa
@mental otter btw hold off the doom port work until i get jinx working in maybe a week
and meson running etc
should be easier with jinx anyways
after its setup*
and initramfs generation
because 
distrobuting the wad may be a nightmare
i have my script yoink it from IA
ill just do like astral and place it in my initrd permanently.
huh something is broken with my jinx impl
time to spend an hour debugging some random shit
time to pick up AuroraOS dev
Fixed meson
RM'ed RFd the makefile now
meson is better tbh
Backspaces now work
still have some fun page faults to fix though
@quaint cape How’s your jinx rewrite going?
If everything is setup hopefully I can figure out your build system and get doom going
Because I just finished mine and I have ports file I can port
Great tbh
Still ironing our some bugs in my impl
Might push today or tmrw.
Rewrote small part of task stuff
hmm auroraos has so many bugs ngl
claude please fix all the bugs in my OS make no mistakes
maybe i should try that
might create 1k other bugs though
such is life
you have a cute Limine wallpaper, you may be interested in the new Limine Flanterm FB params request
it allows you to keep it into the kernel if you want
Really :D
Always wished for a background.
Naaa
yes! it's quite easy to use, just see the spec, it's at the bottom
Will do once I get home
I can just load a file
my kernel didn't like updating to the newest flanterm version:
mind sending me a link to the specification?
cuz i am having trouble finding it
you will need to translate \n to \r\n manually in newer Flanterm revisions, there is a reason behind it but the tldr is "don't worry about it kitten"
🙏
wait since when did limine have built in Flanterm???
Limine has always used Flanterm internally, but this request is new, it allows the bootloader to optionally pass down to the kernel the Flanterm initialisation parameters so things look continuous and cute
you should not rely on the respose existing of course, but if you wanna use it if it's there, then yeah
Doesn't this response also give the benefit of also deriving the background?
Or wait thats what we've been discussing 💀
it's honestly really easy to do, i doubt you'd find Claude helpful here beyond wasting tokens lmao
-# i dont even have claude code btw
.
This might seem like a stupid question but should i even bother having my old manual Flanterm implementation as a backup. I feel like if Limine can't produce Flanterm surely my kernel can't (or am i stupid)?
i am not sure what you mean, all that request does is provide the parameters you can optionally pass to Flanterm initialisation instead of whatever else you'd default to
the request doesn't need to be implemented by the bootloader, so you should handle it not being returned
Currently I'm using Flanterm without any Limine abstraction.
Should I keep this implementation in case if the bootloader doesn't provide the flanterm request?
If Limine fails making an Flanterm window surely I won't be able to make one manually.
you should keep that, this request provides an extra way of getting initialisation parameters for Flanterm, it's not an abstraction
you should handle a response not being returned because bootloaders may not implement this and it's not something essential for a kernel to run
It's not replacing your Flanterm implementation. It allows your Flanterm implementation to inherit the color, wallpaper, etc that Limine uses.
ahh.
Now I'm starting to understand.
Maybe should read those docs?
np
so its basically this?
yeah
the new names are better, though note that the ABI is still the same so it's just an API change, but older kernel binaries will keep working as-is
why tf doesn't inputs work with the new system lol
the background doesn't transfer either
fixed input
background still stupid
@zealous girder sorry to disturb, but what did i forget?
shouldnt this produce the expected behaviour or are there some kind of extra-step needed to get the background working?
you need to give Flanterm a proper allocator else the canvas won't work
I do have a question. Does having a proper allocator also speed up flanterm in any way? Or does it just give things for stuff like the canvas etc
Okay
Noted
Ima fix that tmrw
Could you also write some sort of docs for flanterm lol.
pretty awesome!
speed up? no
the allocator is only using during initialisation, not at runtime
and the internal bump allocator is crazy fast by virtue of, well, being a bump allocator
so if anything your allocators will make initialisation slower
but that's not the point, the point is that the pool of the bump allocator is limited so you cannot use bigger screen sizes, canvases, etc, and also it uses a bunch of preallocated space in .bss obviously, which some may wish to disable
ahhh makes sense
it is really awesome :D
my biggest and most anoying bug is this:
mlibc panic for no reason
@quaint cape does running help twice work for you at all?
because the second help
decided to repeat like 20 times
🥀
Ehmmm
Considering I can't run one command.
I have a hard time believing I can run two.
Is bash --version built-in?
Cuz that's the only thing I've tested
Hm yeah
I did end up getting mine working now tho
Just by stubbing every thing possible 💀
I had to add some functions into the stubs too
But bash even things I have job control

why
I just like emulate just enough
bash can run without
Because I was getting tired of all the errors
And just wanted to get it to a clean state
🥀
which errors?
So I stubbed things out properly
Scroll up in my channel
Thing was acting cursed AF
having no job control doesnt trigger errors
what
how
what did u do?
do you just loop run bash each time
wait thats genius
who needs waitpid
just run bash in a loop for each cmd
then wtf
💀
🙏
🥀 how did u even get to that point
Desperation
yk what is making me go crazy
my fcking jinx shit is bugged
you using jinx jet
yet*
would love to see
yea
soon ima start printing all fcking paths
i cant either see what jinx is doing cuz im using meson
so yea
cooked
i havent looked at jinx for so long though
might just let it run and see if all problems just magically go away
time for fundemental build system rewrite since Evalyn pointed out my shitty approach thanks btw <3
this is like the third time??
Do a mono repo this time too
🗣️ 🔥
Oh btw @quaint cape
I ended up using makefiles still
Instead of meson
Still way cleaner
Because now the makefile only makes the elf
since its just better
me ❤️ meson
i just think all programmers as boys
well
yall are quite rare ❤️
lol
1/100
not rly
i dont use arch btw
Btw I bet the mlibc panics are from your syscalls mucking up some state?
prob
ima just see wherever i return -ENOSYS in my kernel
there's quite a lot of places actually
some syscalls are stubbed with me just returning enosys
which might freak it out.
well my sys_get_ms is stubbed
wonder if that could be the reason?
huh
UNIX epoch
well tbf my error codes and stuff are shit
0 time
ima have to rework that completely
Why return ENOSYS when you can stub it easily
when i get bash working its the next thing
what time string do you return.
It’s not a string?
<t:1774529618:D>
true
also what is that sys thing
i just wrap it in sys_
new mlibc thing??
@mild shoal is it true mlibc wants +ENOSYS not -?
do you manually clear the screen to black at boot?
cuz if you don't it'll look nicer
and boot faster
No.
I'm investigating why that happens.
Or well, rephrased I shouldnt clear my screen at boot. Might be old bits of code left that does it though
👍
now i just need to update mlibc sysdeps right
what is new?
are there any like update details?


