Astral

[analog berry]

Sure but its not practical to enforce, like we cant force someone to run on their machine for every new os

[fierce forge]

I think I know what's going on with my timers

[gritty shore]

@analog berry https://github.com/nyx-os/gaia/blob/master/src/dev/acpi/acpi.cpp#L38 how much time would it take to convert lai to uacpi

[fierce forge]

I'm not pinning the measuring thread to a particular CPU

[gritty shore]

eh next kernel I will port it

[analog berry]

Oh god

[gritty shore]

what

[analog berry]

I thought you were asking me to convert lol

[gritty shore]

no

[obsidian pivot]

LAI 🤮

[gritty shore]

I wouldnt do that

[analog berry]

If you know what you're doing probably a few hours

[fierce forge]

No it's still off

Can someone look at my timer code?

It's off by x10 almost precisely

[obsidian pivot]

Reading the init code, it is impossible for that code to fault provided valid input

Before the log statement

[analog berry]

Maybe instead of reading it add a halt and a basic print?

Or just a halt for that matter

[obsidian pivot]

A basic print cannot happen until that first log statement

So I will halt it

[analog berry]

Yeah lets try that at least

Then we'll know whats up

[fierce forge]

I'm stupid

[obsidian pivot]

what

[obsidian pivot]

this iso hangs

first thing

Arch_KernelEntryBootstrap:
    jmp $
    ; blah

[hearty fulcrum]

dude writes a microkernel
"I'm stupid"
you are not

[analog berry]

Pls hang I dont want to debug hyper

[obsidian pivot]

btw if it does

the next thing that can triple fault

would also be a hyper bug

because it didn't give me the right ultra protocol magic

Arch_KernelEntryBootstrap:
    jmp $
    cmp rsi, 0x554c5442
    je .ok ; should triple fault on failure
    xor rax,rax
    jmp rax ; All hope is lost.
.ok:```

the next thing that could crash is a fucked stack

which'd be also hyper's fault

[fierce forge]

I've been setting pit to 10ms and doing calculations assuming it's 1ms

[obsidian pivot]

as the line right after .ok: is push rdi

after that, it's a page fault because the kernel was mapped incorrectly

[grand shadow]

todo: redo interrupt system before rewriting scheduler and redoing devices

for now while I dont do that though, its better

wtf that was such a big issue

yeah I really need to redo this

the other benchmark I have, glxgears, went up by like 50 fps

just by doing this

[analog berry]

uacpi benchmark paying off

[analog berry]

is this your final submission

[grand shadow]

nah my computer sucks balls and theres probably a thing or two that can be made better

[analog berry]

dont add for now then?

[grand shadow]

yeah I first need to see what more can I do and then get a test subject with a very good computer to run it

[analog berry]

lol

[gritty shore]

I'll get my friend with a 9800x3d

to run mine

[analog berry]

Oh you’re converting lai as well?

That's based

[gritty shore]

no

well yes

but in my new iteration ™️

which will not suck

as much

[analog berry]

Nice

[grand shadow]

here is it if anyone wants to try

I kinda do wanna see if it builds gcc faster now

probably not much

this time I will build it with --disable-docs

I think if I do -pipe it'll be faster

because of i/o being abysmal on astral

nvm

alright compiling

[carmine swallow]

getting close to managarm speeds

[grand shadow]

Wtf goated

[analog berry]

wtf

[analog berry]

maybe 2.4M is ok as a first submission:

[grand shadow]

@carmine swallow what is your cpu to you have a nasa computer

[carmine swallow]

i5 13600k lol

[prime juniper]

ha LLLLL I beat that on the 7800X3D (barely tho)

[supple hatch]

damnn

timer for sure is not the problem, im unsure why its inconsisent but hey

its not 150 ops per second

💀

[grand shadow]

@analog berry I think my submission will be qwinci's 2411598/sec on his i5 13600k

[analog berry]

nice

can u pls leave a comment under the PR

[grand shadow]

alright

[analog berry]

cool!

[grand shadow]

I cant outphd managarm

[supple hatch]

my submission will be mints

[analog berry]

one day

[hearty fulcrum]

ask nasa to lend you a pc for a bit :p

[supple hatch]

lol

lemme add a comment

to the pr

[analog berry]

yeah take your time i'll wait for everyone to get the best result they can

[supple hatch]

yea done

[grand shadow]

@analog berry it is done

[analog berry]

nice!

[craggy tide]

i am still not sure how this is possible

that my CPU is slower than a 13600k

[grand shadow]

Magic

[carmine swallow]

yeah idk, there are things like the host kernel config that could potentially affect it like I have all vuln mitigations disabled and whatever (and a slight undervolt which makes the perf of the cpu a little better but afaik that's mostly just under heavy multicore workloads where the temp would be an issue)

[fleet osprey]

I just use copiously overkill cooling to deal with that

I have a dual-fin-stack cooler with a chonky fan.... to cool an i3-12100f

it never goes above 50C under max load

[carmine swallow]

I have an arctic liquid freezer 2-fan 280mm aio water cooler even with it without the undervolt it gets to like 100deg with unlimited power limit

undervolted to like 1.2v or whatever its 80c under max load

[fleet osprey]

lol

all of my stuff has good enough cooling to not throttle under max load

my laptop gets right up against the limit (like 1-2C shy of the thermal limit), but doesn't quite throttle

Ryzen 7 PRO 4750U, best CPU

[craggy tide]

give me the same ISO and same qemu flags

[grand shadow]

@craggy tide
qemu-system-x86_64 -cdrom astral.iso -M q35 -enable-kvm -cpu host,migratable=off

[carmine swallow]

that's basically exactly the same as what I get

[craggy tide]

yeah that's about expected i guess

[hearty fulcrum]

Ilobilix #2 !!!!!!!!!!

[analog berry]

Congrats

[grand shadow]

Very nice

[grand shadow]

funni

I forgor I had to destdir

[grand shadow]

@analog berry in what situation would a mutex with a timeout be called

[analog berry]

AML

AcquireOp takes in a timeout value

[proper meteor]

[analog berry]

yeah its been a footgun and a source of confusion since the inception of acpi

they should've added TryAcquire and Acquire instead of this garbage

[grand shadow]

Ah well it will be a slow path

Ill loop on a try and sleep

And print a message "heres a nickel kid go buy yourself a better firmware"

[fleet osprey]

Simply install Coreboot, it's free

[analog berry]

its not used that much in the real world so thats fair

[carmine swallow]

there would actually be coreboot for the mobo that I have, I have been wanting to try it but haven't yet bothered

[grand shadow]

I am free from the chains that held me to the ground for now I shall be free until the wicked attempt against my free will once more

(I have a break from university

[obsidian pivot]

Same

[carmine swallow]

yeah me too for 2 weeks (though not from uni but from the civil service server management job that I am doing)

[sacred gyro]

Why do I want to torture myself with porting something like XFCE to Astral…

Why is this an urge I have

Send help

[robust geyser]

xfce is pretty easy to port iirc

probably just a matter of building it

and it isn't even that big

[sacred gyro]

Trying to port something as massive as GNOME or KDE sounds like pure suffering

[robust geyser]

gnome is pure pain

because of gobject-introspection

kde is manageable but it does have some linux specific apis it needs i think

[fleet osprey]

I'd love to see a way to run Astral on PowerPC (if basic PPC support were a thing I'd totally port it to the Wii )

[sacred gyro]

I seem to remember that being a pain for me while doing LFS

[robust geyser]

sounds like a mild skill issue tbf

[sacred gyro]

Yeah probably

[robust geyser]

managarm did have some work done towards kde6

and gnome too

managed to run mutter (or gnome shell??)

but it crashed

[sacred gyro]

Yeah, Mutter is the window manager for GNOME last I checked

[robust geyser]

yeah i think that crashed

it showed like a big message

that something crashed

on screen

[sacred gyro]

But I might look into XFCE porting when I get home

[robust geyser]

best of luck 🙏

[sacred gyro]

(I’d need to build an image probably, and then figure out what to do from there)

[robust geyser]

if you need recipes for xfce stuff: https://github.com/48cf/jinix/tree/trunk/recipes

a bit outdated but they should work

just keep in mind they are for linux

[sacred gyro]

Oh no

I see gobject in there 💀

[robust geyser]

nah that's for gnome

[sacred gyro]

Ah

[robust geyser]

thankfully it's not needed for xfce

i should really look into making gobject-introspection more cross-compilable for everyone

[pseudo shore]

As far as I can tell there is no proper audio yet, any plans to add it in the near future? I think that could be the last missing piece for this to be a proper desktop OS as far as I am concerned

[analog berry]

there is audio via HDA

[pseudo shore]

hda.h and hda.c on the rewrite branch look incomplete

Am I missing something?

[analog berry]

@carmine swallow was working on that so maybe he knows

[carmine swallow]

yes they are just the skeletons with basically no functionality, I had a more proper impl for it in a branch (that isn't merged) and I wanted to make it use uhda instead but haven't figured out a weird noise issue with it that happens specifically on astral

for reference if you want to look, https://github.com/Qwinci/Astral/blob/rewrite/kernel-src/io/audio/hda.c and https://github.com/Qwinci/Astral/blob/uhda/kernel-src/io/audio/hda.c

[supple hatch]

uhda on nyaux when

[carmine swallow]

after you have pci

[supple hatch]

yess soonn

[pseudo shore]

TIL that uHDA exists

[carmine swallow]

I should get back to it now that I have more time for a while again

[supple hatch]

btw i saw something called udrm

someone was working on

[carmine swallow]

yes Marvin is working on it

[supple hatch]

yes

https://tenor.com/view/marvin-gif-1644493607771721553

that is marvin leaked face reveal

[carmine swallow]

lol

[grand shadow]

if it would be usable thats another story

which I will be trying to improve

[crude agate]

https://tenor.com/view/marvin-gif-1644493607771721553

[supple hatch]

its spreading

[prime juniper]

It’s really not. You’re gonna want ConsoleKit (deprecated) or logins (systemd or derived). Or perhaps some BSD meme which surely isn’t much better

[robust geyser]

manageable in comparison to gnome*

[grand shadow]

which is why between gnome or kde I will probably do a secret, third thing in the future

(CDE port

[prime juniper]

Both need systemd (or logind split off) anyway. So meh. We almost have goi solved (FUCKING GRAPHENE SEGFAULT DURING LINKING)

[fleet osprey]

no new Astral progress in 2 days

[supple hatch]

flash urself some maidens

[hearty fulcrum]

talk is cheap, send patches

[grand shadow]

Ive slowly been studying something which I want to implement soon for speeeeeeeeeed

(it wont actually make that much of a difference)

[hearty fulcrum]

and what would that be?

[grand shadow]

the funni lock

tyr-glquake and tyr-quake are both perfectly playable on qemu now

big win

Broke: kmscon
Woke: xterm on fullscreen

[grand shadow]

lol I think the doc issue with the make install on gcc is due to me missing fucking perl

[carmine swallow]

btw one suggestion for astral, make astral/syscall.h a part of the astral package so other pkgs can use it without needing to copy the code from there and hardcoding syscall numbers to patches

[grand shadow]

I thoguth I installed it with mlibc in /usr/include/astral/syscall.h

huh apparently not

though yeah I will probably make it be part of astral so mlibc can use it

idk

[grand shadow]

where did this come from btw what are you cooking

[carmine swallow]

openjdk it builds fine at least (with basically the same patches that I already had for managarm with minor changes), idk if its going to work tho

[grand shadow]

ooooooo

if it needs some signal stuff like ucontext or siginfo it wont

[carmine swallow]

I am not completely sure about that but it might (at least for the jit, maybe the very slow non-jit version could work without)

[grand shadow]

wtf

probably somewhere is doing vmm_map with a size of 0

try a syscall log

you can enable it in the kernel makefile

(needs e9 logging too)

[carmine swallow]

syscall: pid 37 tid 38: execve: path /usr/lib/jvm/java-17/bin/java uargv 000000000057EF60 uenvp 00000000009FFE00 (444 cached pages, 1007374 free pages)
syscall return: pid 37 tid 38: 18446744073709551615 ENOMEM (446 cached pages, 1007374 free pages)
``` does this mean that the execve itself failed with ENOMEM?

[hearty fulcrum]

https://downloadmoreram.com/

[grand shadow]

Yeah

Wtf

Cna you do a readelf of the PHs of the elf

[wild marsh]

inb4 1TB p_memsz

[grand shadow]

GNU STACK

ARGH

Yeah thats it I think

Size is 0

[carmine swallow]

but why would you do anything with that as its not PT_LOAD?

[grand shadow]

Wait true

Wtf

[carmine swallow]

ig Ill try to add prints to figure out where exactly it comes from

[grand shadow]

Why is it trying to load at 0

[carmine swallow]

its a pie

[grand shadow]

I dont do relocations in the kernel

[carmine swallow]

you don't need to either, they are done by the ld

[grand shadow]

Oh

Blegh

When qemu did that it always complained

Ig you can add a check if its a pie and set the base in elf load

Just to see if itd work

Cuz it loading to 0 breaks the load logic

[wild marsh]

yeah for pie you just pick a base and add it to all virt addrs

like you do for ld.so

[fierce forge]

I urgently need to port mlibc to pmOS

[grand shadow]

I can see if I add support for them if you want @carmine swallow

I wont be home for a few hours

[carmine swallow]

what I did for now is checking if the supplied base is zero and the file type is ET_DYN and if they are then just unconditionally applying a static base

I ran into an another problem tho and its that LD_LIBRARY_PATH's $ORIGIN is broken because of no AT_EXECFN so I am trying to add that rn

[grand shadow]

Ahhh nice

[carmine swallow]

at least it now starts

[obsidian pivot]

Based

[analog berry]

the kernec gets me every time

[grand shadow]

Wonder if its the ucontext meme

I dont have it implemented

[carmine swallow]

it might be, Ill try with the slow non-jit one if it makes any difference

[grand shadow]

Its a feature now lmao

[carmine swallow]

apparently not

so yeah ig it needs ucontext regardless

[grand shadow]

probably not that hard to implement, just need to define the struct and push it into the signal stack info

and then convert the astral context_t into an ucontext

like a function that does it

[carmine swallow]

yeah

and to restore just convert it back

wine also wants that, maybe it would work after that too

[grand shadow]

it'd still be missing siginfo I think

[carmine swallow]

might be

also one other thing that it needs is defining MLIBC_MMAP_ALLOCATE_DSO in mlibc meson.build or else you are going to get funny bugs because of mlibc allocating dsos at fixed addresses overlapping the ones that wine uses using MAP_FIXED

[grand shadow]

my mmap works so that probably would work just fine

I already use MLIBC_MAP_DSO_SEGMENTS and MLIBC_MAP_FILE_WINDOWS

do you want to implement the ucontext stuff or can I do it

[carmine swallow]

you can do it if you want

[grand shadow]

alright

my signal code is a mess anyways I dont want you to suffer

[carmine swallow]

lol

[grand shadow]

in the meantime you can probably add a CTX_PRINT in mmu.c to see where it segfaults

or was it PRINT_CTX

bleh

[carmine swallow]

I think this is a vsyscall for getcpu and its using that because astral doesn't implement the mlibc getcpu sysdep

ig I can just add that as a stub in mlibc for now to make it not use that vsyscall

[grand shadow]

yeah

what does getcpu do

[carmine swallow]

gets the number of the cpu that the thread is currently executing on

[grand shadow]

oh

yeah for now just stub it

eventually I can add a syscall or something

[carmine swallow]

now I think its running to the ucontext stuff, the ip is sane and rdx is 0

it also wanted CLOCK_MONOTONIC, I just made it the same as CLOCK_REALTIME (and MAP_NORESERVE mmap flag but that can just be ignored, just had to add it to the KNOWN_FLAGS define)

[analog berry]

a little trolling

[grand shadow]

for CLOCK_MONOTONIC you can just return like

timekeeper_ns_from_boot

or whatever the macro was

or function

oh wait

timespec_ns(timekeeper_timefromboot())

[carmine swallow]

wouldn't plain timekeeper_timefromboot() be fine too as it returns a timespec?

[analog berry]

i added that

[grand shadow]

oh wait yeah

[grand shadow]

alright

I went ahead and also fixed another signal bug lol

when a syscall got interrupted the signal return rax and rdx got trashed

or rather

the signal rax and rdx

so the ucontext pointer was wrong because of it

but now its fixed

[carmine swallow]

meanwhile I took an another look at the hda stuff, I think the buffer size was too small as it seems to pretty much work when I make it bigger (though cancelling the write and then restarting doesn't really work very well)

[grand shadow]

oooo nice

I can take a look at the code later to see wrt the cancelling stuff

if its something astral related

alright, ucontext stuff pushed

[carmine swallow]

nice, Ill try it

is it restoring the context from the uctx?

[grand shadow]

No

I probably should do that

Ill eat dinner and then implement it, should be quick

[carmine swallow]

also the fpu state inside mcontext should be a ptr

I also took a look at the java code to see what exactly it wants from the signal stuff and it looks like it wants siginfo too tbh I don't think its too bad as it basically only uses si_addr which is cr2 for sigsegv

or well idk if it even needs that, maybe just having the struct present as all zeros is enough

[grand shadow]

wait what

[grand shadow]

maybe it only uses it for like printing an error or something?

idk if it would go as far as do map stuff on a page fault

[carmine swallow]

it looks like it uses it for things like stack overflow detection, null exception and some other stuff like that

but yeah I don't think its necessary as managarm doesn't populate it either afaik

[grand shadow]

hmm yeah I can probably just memset it to 0

I already allocate it in the frame anyway

[grand shadow]

@carmine swallow done

now it takes the context and fpu stuff from the ucontext and also has a dummy siginfo memset to 0

[carmine swallow]

nice, though it looks to still be faulting. Ill take a look at it more tomorrow

[grand shadow]

damn, should try to find out what the rip of the fault is

[carmine swallow]

apparently sometimes it faults within sys_sigaction when memcpying the act pointer to a stack copy of it (with cr2 pointing to the stack) but its not consistent, sometimes it also says "vmm: no range"

[grand shadow]

Waaiit

Blegh what

Why is the stack not mapped

[analog berry]

memory not kernec

[obsidian pivot]

obos curse has finally caught up to you, I see

[grand shadow]

@carmine swallow what address does it use as the stack

By default I map it at the end of the user space for the main thread

[carmine swallow]

its def not that, Ill see what it is

[grand shadow]

Could be that its in another thread and that handling is borked

Maybe too small of a stac

?

[grand shadow]

Why is it aborting so much

[carmine swallow]

idk, maybe I will try the jit version again if it would have less spam

[grand shadow]

I think it might be aborting on loop or something? Idk

The stack looks fine

Bad action

Wait

Write to a read only address possibly

I forgot what the flags mean

[analog berry]

well astral does a lot more soo

try running java on obos

[obsidian pivot]

give me a couple months

[analog berry]

sure

[obsidian pivot]

technically, I have more LoC so that OBVIOUSLY means I do a lot more

[analog berry]

indeed

[carmine swallow]

on the jit version at least java somewhat works, javac dies tho

I wonder if it works enough to be able to run minecraft server

apparently not, it complains about corrupted jar file

[grand shadow]

Funky

Wonder why

[carmine swallow]

idk Ill have to add printfs

[grand shadow]

Nice work

[carmine swallow]

I think its caused by SEEK_END not taking offset into account

[grand shadow]

Oh shit

Youre right

[carmine swallow]

at least I got it to run until the eula part (though sometimes it hangs somewhere, idk why)

also it ran into some kind of memory meme again before getting this far, I just worked around it by making mmap always return rwx memory

[analog berry]

thats pretty cool

[carmine swallow]

yeah

[obsidian pivot]

Based

[grand shadow]

Epic

[carmine swallow]

it looks like it worked after I added eula=true to the file manually before starting it for the first time idk if I can connect to it tho, time to try ig

bruh it crashed the moment I said that lol

[analog berry]

whats going on with these logs

[carmine swallow]

I launched java with -verbose to diagnose the hangs

the spam about it using online processor count is always there though

[analog berry]

what is agh$... agh

[carmine swallow]

watchdog 💀 I think this is related to the auth stuff, maybe disabling online mode would make this not happen

[grand shadow]

Souece agh

Its in pain

[carmine swallow]

obfuscated class/method names

[grand shadow]

Did you run netd vionet0 before

[carmine swallow]

no

[grand shadow]

It might be stuck on name resolution or something

You need to run the dhcp stuff

For some reason the routing table doesnt return a failure in udp when theres no route

And so the resolution just hangs

[carmine swallow]

Ill try running that to see if it fixes the thing so I can try connecting to it

apparently it didn't ig I can still try disabling the online mode

[grand shadow]

Weird

[carmine swallow]

or maybe its just that one tick takes so much time that it decides to kill it, ig I can try disabling the max tick time

though I don't think that's actually about the tick time and rather that auth stuff

[grand shadow]

Probably yeah the backrrace is o auth stuff

[carmine swallow]

at least I managed to join it for like 5s before the server died disabling the online mode didn't help either

[robust geyser]

holy shit is that a minecraft server running on astral

[carmine swallow]

yes

[robust geyser]

that is so cool

any idea why it's dying?

[carmine swallow]

other than its because of some kind of watchdog no

and I have no idea about the random hangs that sometimes occur when starting the server either

[grand shadow]

Could probably try to run with syscall logging on

Would be slow but bleh

Worth a try

Crazy that it even connects

[carmine swallow]

yeah, on managarm I didn't get it this far (though it might have been because of some bug that was fixed after that or alternatively my tcp server impl)

[analog berry]

Daaamn should I host my gtnh server on astral

[grand shadow]

Probably would explode your computer

[analog berry]

its fine i run it in a vm

[obsidian pivot]

I should daily drive astral

what could possibly go wrong?

[grand shadow]

The vm would explode

[obsidian pivot]

I explode

[analog berry]

we should host the entire world infrastructure in obos

[obsidian pivot]

imagine being a pilot flying a plane, then you suddenly page fault

midair

[carmine swallow]

just make it automatically restart the os

[analog berry]

just call uacpi_reboot in the pf handler

[obsidian pivot]

fr

[fierce forge]

(broken firmware)

lmao

[fierce forge]

(I'm not laughing about obos)

[grand shadow]

Wonder what subsystem even causes the hangs

[carmine swallow]

it looks like its futex waits + clockget + nanosleep that its stuck doing

or well it just does those in a loop

[grand shadow]

Huh weird

Does any of these rrturn any weird codes

[carmine swallow]

I don't think

[grand shadow]

Blegh

Some thread might be getting stuck?

Can you give me a bit of the looping logs

[grand shadow]

wonder if tghats in mlibc or in java

[carmine swallow]

well that's easy to check ig, I can just add a print in the java futex fn that does the astral futex syscall

[grand shadow]

true

try to save the logs into a file and see what other threads do and if any of them get stuck in a syscall

[carmine swallow]

the syscall log?

[grand shadow]

yeah

[carmine swallow]

that's going to be big lol

[grand shadow]

I've done it before its not that bad

[obsidian pivot]

does astral have anything special in it's pf handlers

other than stuff for CoW and file mapping

and crashing

[grand shadow]

Wdym

[obsidian pivot]

like

what features does astral vmm have

[grand shadow]

Probably the bare minimum

Mapping, cow, lazy allocation, signaling on bad access, etc

File mapping

Its actually a pretty small function you can just read it yourself

https://github.com/Mathewnd/Astral/blob/rewrite/kernel-src%2Fmm%2Fvmm.c#L524

[obsidian pivot]

ye I did

we just love the 1k line file called vmm.c

[carmine swallow]

64mb of syscall logs 💀 how do I identity anything important from that

"only" 651280 lines

[grand shadow]

You can send me it and I can take a look in a bit :p

I am used to the trouvle

Trouble

[carmine swallow]

60k futex calls in there btw

[grand shadow]

Ty Ill take a look

[carmine swallow]

and I think its from mlibc as I didn't see any printf spam on the screen

[grand shadow]

syscall: pid 45 tid 60: ioctl: fd 38 request 35090 argument 000000001A9A5E50 (46019 cached pages, 978676 free pages)
syscall return: pid 45 tid 60: 18446744073709551615 ENOTTY (46019 cached pages, 978676 free pages)

last thing it seems to try to do

then it closes the socket and just

does nothing more

(tid 60)

and tid 61 keeps spinning

so I wonder if it was waiting for a response from tid 60

and I suppose the bit of activity at the end is the watchdog?

seems to be SIOCGIFCONF

@carmine swallow can you take a look if anything uses SIOCGIFCONF in openjdk

[carmine swallow]

idk about that, this log was it getting stuck before even getting to the world loading

[grand shadow]

weird

[carmine swallow]

yes https://github.com/openjdk/jdk17/blob/4afbcaf55383ec2f5da53282a1547bac3d099e9d/src/java.base/unix/native/libnet/NetworkInterface.c#L1187

[grand shadow]

huh its meant to throw an exception but it doesnt

so something is getting locked up before that I assume

I also see these pthread_getattr_np(): Implementation is incomplete! but I dont think it should be an issue

wonder if its because of a missing sysinfo?

probably not too

I really need to make a tool to parse this stuff

yeah I need to make at least something like "find last syscall or return of every pid+thread pair"

[grand shadow]

1_2: syscall: pid 1 tid 2: openat: dirfd -100 path /tmp/.initcmd flags 0 mode 0 (373 cached pages, 1009020 free pages)
31_32: syscall: pid 31 tid 32: read: fd 3 buffer 00007FFFFFFFFE5F size 1 (370 cached pages, 1007549 free pages)
33_34: syscall: pid 33 tid 34: waitpid: pid -1 statusp 00007FFFFFFFF9EC options 10 (578 cached pages, 1007375 free pages)
41_42: syscall: pid 41 tid 42: waitpid: pid -1 statusp 00007FFFFFFFF9CC options 0 (590 cached pages, 1006052 free pages)
45_46: syscall: pid 45 tid 46: futex: futex 000000000047F9FC op 0 value 0 timespec 0000000000000000 (3091 cached pages, 1001598 free pages)
45_47: syscall: pid 45 tid 47: futex: futex 000000000067E818 op 0 value 2 timespec 0000000000000000 (46026 cached pages, 970627 free pages)
45_48: syscall: pid 45 tid 48: nanosleep: time 00000000186AB740 remaining 00000000186AB740 (46026 cached pages, 970666 free pages)
45_49: syscall: pid 45 tid 49: futex: futex 000000000067E288 op 0 value -2147483588 timespec 0000000000000000 (46026 cached pages, 970627 free pages)
45_50: syscall: pid 45 tid 50: futex: futex 00000000009FB538 op 0 value 2 timespec 0000000000000000 (3237 cached pages, 1001288 free pages)
45_51: syscall: pid 45 tid 51: futex: futex 000000000087FFB0 op 0 value -18446744071562067968 timespec 0000000000000000 (46025 cached pages, 978656 free pages)
45_52: syscall: pid 45 tid 52: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970666 free pages)
45_53: syscall: pid 45 tid 53: futex: futex 000000000067F198 op 0 value 1 timespec 00000000193FFB40 (46026 cached pages, 970666 free pages)
45_54: syscall: pid 45 tid 54: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970666 free pages)
45_55: syscall: pid 45 tid 55: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970666 free pages)
45_56: syscall: pid 45 tid 56: futex: futex 000000000067F518 op 0 value 1 timespec 00000000196FF8D0 (46026 cached pages, 970666 free pages)
45_57: syscall: pid 45 tid 57: futex: futex 000000000067F098 op 0 value 1 timespec 0000000000000000 (46026 cached pages, 970666 free pages)
45_58: syscall: pid 45 tid 58: futex: futex 000000000067D998 op 0 value 1 timespec 0000000019B857F0 (46026 cached pages, 970666 free pages)
45_59: syscall: pid 45 tid 59: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970666 free pages)
45_60: syscall: pid 45 tid 60: futex: futex 000000000067E698 op 0 value 207 timespec 0000000000000000 (46026 cached pages, 970627 free pages)
45_61: syscall return: pid 45 tid 61: 0 OK (46026 cached pages, 970705 free pages)
45_253: syscall: pid 45 tid 253: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970626 free pages)
45_254: syscall: pid 45 tid 254: futex: futex 000000000087FFCC op 0 value 111 timespec 0000000000000000 (46026 cached pages, 970627 free pages)

quick and dirty python script never fails

seems like tid 48 is sleeping idfk for how long

idk what 48 is

61 is doing something?

its not in a syscall

the rest are all on futex_wait

I'll take a look in a bit if that tid 48 sleep is recent

ok yeah that nanosleep is recent

makes me wonder if theyre waiting on a thread which exited?

there are so many to look through tho ugh

there are some threads that seem to do literally nothing which wtf

bleh idk

wonder if its something related to pthread_getattr_np not having a proper implementation?

[carmine swallow]

I kinda doubt that as I didn't run into this kind of stuff on managarm

[grand shadow]

hmmm

yeah I have nooo idea

there are too many threads and moving variables for me to have any idea wtf is going on

this spawns over 200 threads

@carmine swallow what are all the cases it hangs/crashes?

like this one is before the world loading

[carmine swallow]

the other one was also sometime close to this one, idr at which point exactly

and the crash was just sometime after the server is fully started

and the mmap stuff is a separate issue on its own, I just worked around it by making it always return rwx memory

[grand shadow]

what exactly was the mmap stuff doing?

[carmine swallow]

I am not completely sure, it creates errors like this

[grand shadow]

its trying to write to a page which doesnt exist?

wtf

returning it as wrx should have no effect on that

so weird

[grand shadow]

can you recreate this and send me a syscall trace

[carmine swallow]

and its the same cr2 + error

I think this is also the same thing that javac runs into as it works just fine when mmuflags is changed to always be rwx

[grand shadow]

wtf maybe obos curse has hit me

syscall: pid 41 tid 43: mmap: hint 0000000100000000 length 65536 prot 3 flags 50 fd -1 offset 0 (4078 cached pages, 999361 free pages)
syscall return: pid 41 tid 43: 4294967296 OK (4078 cached pages, 999361 free pages)

it allocates 64k there

[carmine swallow]

I wonder how does changing the mmu flags fix it tho

[grand shadow]

maybe its like some defrag stuff I do

if two ranges are right next to each other and compatible they get merged into one to save memory and processing speed

https://github.com/Mathewnd/Astral/blob/75a24d30269e810b6767982616d1fde89689eaca/kernel-src/mm/vmm.c#L184

here

I've had bugs here before and I dont doubt there might still be 💀

[carmine swallow]

shouldn't range->next->prev also be updated inside the if that's inside the while loop?

else its still going to point to range which isn't correct when newrange is between it and range

I think that was it, it doesn't crash there anymore

[grand shadow]

shit youre right

where does it crash now?

[carmine swallow]

well it doesn't really crash anywhere, there are the hangs and possibly the watchdog thing

[grand shadow]

well at least thats one more bug fixed

javac works now I assume?

[carmine swallow]

yes

you can compile and run a java hello world

[grand shadow]

EPIC

gj qwinci

[carmine swallow]

I also found one futex-related bug, vmm_getphysical always returns page aligned addresses (it does try to add the offset within the page but it aligns the address before that so it ends up doing nothing)

didn't fix the hang tho

[grand shadow]

damn nice catch

I think I had just put that align there cuz I forgot to align it back but then forgot I undid it at the start 💀

I dont think this sentence was understandable

I need more hours of sleep

[analog berry]

no it makes sense actually

[obsidian pivot]

does that mean I can run my minecraft server on astral now??

[carmine swallow]

if you get lucky you can run it for like 5s (also breaking blocks don't drop anything in that time)

after that its going to get killed by the watchdog

[grand shadow]

I wonder if those openjdk fixes would make qemu work properly

right now it hangs while loading from limine

[obsidian pivot]

damn

astral is 🆒

[carmine swallow]

ig you can try, I can send them in a moment

[grand shadow]

I can wait until we can figure out the futex stuff

[carmine swallow]

oh also I forgot to mention, it wanted F_SETLK64/F_SETLKW64, I didn't implement those properly yet

[obsidian pivot]

is this a Mathewnd-approved statement?

[grand shadow]

astral is cool indeed

[grand shadow]

what do these even do

[carmine swallow]

apparently they lock segments within open files

ig its like flock but for specific range of bytes

[grand shadow]

I dont think this would have a big impact

[carmine swallow]

yeah likely not

[obsidian pivot]

@grand shadow I found a possible way you could optimize your sys_tcb_set function
Instead of always syscalling, use the wrfsbase instruction, if it exists, which does the same as your archctl syscall, but without a syscall.

Note that it needs to be enabled explictly

https://www.felixcloutier.com/x86/wrfsbase:wrgsbase

[grand shadow]

That sounds like a non issue

[obsidian pivot]

yes

it's a microoptimization

[grand shadow]

If you set the tcb that much you deserve whats coming for you.

[obsidian pivot]

fair

[prime juniper]

Fuck that’s based. Can Managarm do that?

[carmine swallow]

yes, though javac hangs at the end (it still writes the file though so you can just close the terminal and run the produced class in another one)

[prime juniper]

Interesting

[grand shadow]

My new mutex has been laid out on paper, now I just have to make it realoty

[analog berry]

lol

can you post a photo of said paper

[grand shadow]

No you would not understand my handwriting and it was my company name on it

[supple hatch]

Mathew evil incorparted

https://tenor.com/view/фуфелшмерцпакостин-фуфелшмерц-gif-25654193

MUHAHAH PERRY THE PLATYPUS PREPARE TO SEE MY LATEST INVENTION

THE ASTRAL INATOR

[analog berry]

keep your secrets

[supple hatch]

IT WILL WIPE EVERY DISK DRIVE ON THE ENTIRE TRI STATE AREA AND INSTALL ASTRAL ON IT!

[analog berry]

[supple hatch]

[grand shadow]

Soon

[grand shadow]

Oh my GOD I left it at my desk at work

[prime juniper]

Would’ve been nice to have that picture now huh

[prime juniper]

Nice

[grand shadow]

true lmao

[prime juniper]

That’s what you get for being a selfish prick and not sharing

[fierce forge]

Are you sharing company secrets?

[grand shadow]

yes I hold the key to a femtosecond lock, the fastest lock ever invented by humankind. the discovery of this lock will allow us to travel the stars

[supple hatch]

can u let me *~~steal ~~*oh i mean use it in nyaux

[fierce forge]

MIT - Midway Incorporating To pmos

[grand shadow]

well since I had already finished how I was gonna do it I'm just doing it from memory lol

[grand shadow]

@carmine swallow can you test this to see if its faster now

its using my new locks and it made absolutely n ochange to anything

[gritty shore]

It may not be about locks but rather how they're used

What does the contention look like?

[grand shadow]

there wasnt even contention it was just one thread running

[prime mulch]

have you looked into turnstiles

oh youre doing pushlocks

[fervent fjord]

yooo welcome back will

[prime mulch]

hello streaks

[fervent fjord]

long time no see, how is life

[prime mulch]

i got FAT! i got OBESE!

not really

im chilling

[fervent fjord]

I did get fat

lol

nice to hear you are chilling

[prime mulch]

ur just becoming more like sealkind

[fervent fjord]

true that

more eugh eugh

[prime mulch]

https://tenor.com/view/seal-clapping-animals-cute-applause-gif-17525701

[grand shadow]

Omg its funny hyena man hi

[prime mulch]

hello MATHEW

[grand shadow]

Hru

[prime mulch]

im gooood

u should do priority inheritance in ur locks

[grand shadow]

Yeah it sounds like something thatd be good, especially when I do ule

[prime mulch]

ule has a fun extra wrinkle which is interactivity inheritance

[robust geyser]

how does that work, do you have to specify a priority of a lock when locking it?

or like even when declaring it

do you just assert(new_priority >= last_priority)

[prime mulch]

its where if you have two threads A and B where the priority of thread A > priority of thread B

[robust geyser]

ohhhh

[prime mulch]

and thread B owns a lock, and thread A blocks on the lock

it raises B to its priority temporarily

[robust geyser]

damn that's cool

i assume that's to allow thread B to finish quicker and release the lock?

[prime mulch]

its to prevent priority inversion

[robust geyser]

or, well, hopefully let it finish whatever it's doing quicker

ok yeah that stuff goes way over my head

[prime mulch]

i mean its basically just where an interfering thread(s) with a priority inbetween A and B never blocks to allow B to run and so A waits for lower priority activity

which is not ideal

this prevents that by allowing B to immediately run and preempt all those other guys

[robust geyser]

ah okay that makes sense

yeah i understand the concept of thread priorities, i just didn't understand the reason behind bumping B's priority

but that is smart i like it

[prime mulch]

for priority inheritance to be a "complete" solution it has to be transitive tho

like

you can still get priority inversion if you have a chain of threads waiting for locks

and you only boost the first thread in the chain (the immediate owner of the lock youre blocking on)

and not further threads down the chain

ie, threads that own locks that are being blocked on by threads that own locks that are being blocked on by etc etc

u have to boost them all

before u block

and there are other kinds of inter-thread dependencies other than just locks and they all have the potential for priority inversion

so its a complicated problm

[robust geyser]

yay more stuff i have no damn clue about, fun

unfortunately i do not have a phd in schedulers

[fervent fjord]

priority inversion is when you have 2 threads, one is just a low priority daemon, and the other is a high priority watchdog that, if it doesnt run in X time, will panic the system, and the low priority daemon takes the longest lock ever when the watchdog guy needs it

[robust geyser]

yeah i get that, but i don't really understand the dependencies of dependencies things

it just sounds like a really complex problem lol

[prime mulch]

yeah thats the two thread scenario

priority inheritance doesnt work there the only good solution is to somehow break up or remove that dependency

[fervent fjord]

priority inheritance can band-aid it

ideally you do several things

[prime mulch]

well if theres just two threads and the higher priority one blocked then affecting the priority of the other one is ineffectual

itll run anyway

the real problem is the length of the critical section

in ur scenario

[fervent fjord]

changing the priority of the low priority task or changing the length of the section is the same, both things lead to lower execution time, its just that in one you make it so the blocking task gets to run more, and thus take less in the grand scheme of things, while in the other you make the actual code take less

you can also do things like rando boost that windows loves as another solution

at least used to love, idk if they do it anymore

[prime mulch]

windows has full priority inheritance now since the late '00s apparently

they might still do random boosting to catch some of the more complex nearly unsolvable "other" inter-thread dependencies i mentioned

especially with stuff like memory manager page-in and whatever where the first faulting thread "owns" the page IO and subsequent threads have to wait for it to complete that and this can chain as well and whatever

[fervent fjord]

it sounds dumb and cool I hope they do

[prime mulch]

VMS did priority ceiling where when a thread holds a kernel mutex it is raised to the highest priority

[analog berry]

Welcome back

[prime mulch]

this makes priority inversion impossible but is more costly bc you have to do a priority raise/lower every time you acquire your first kernel mutex or leave your last one

not to mention its not very fine grained

the windows kernel was originally spec'd to do this as well and i confirmed this behavior exists in some of the earliest builds but by release in 1993 theyd switched to random boosting and discarded this approach

prob bc it sucks for a gui-oriented OS

[fervent fjord]

it sounds a bit bad for responsiveness yeah

[prime mulch]

u use VMS over a command line where nobody notices anyway so they still do that afaik

wow it took me like 1 hour to bring up VMS or NT

new record

[fervent fjord]

shouldve made your first word when you joined VMS

for the memes

anyways nice to have you back, interesting talk

[prime mulch]

thx glad to be back

[obsidian pivot]

hyenasky is back

[grand shadow]

How did they solve it with pushlocks

[supple hatch]

oh hey ur back!

check out my thread

#1230349543623757845

daily nyaux shilling complete

[grand shadow]

alright so now before I redo the irq system I decided I will go around and make the souce code a bit better to read as to not drive other people away from it lmao

@robust geyser other than the fucked up naming convention what other parts did you think was bad

[prime mulch]

they used "autoboost"

which is some complicated shit i dont understand

[robust geyser]

lack of formatting rules

i usually like clang format but

if not then i tend to try and follow the style

but yeah a clang format is usually nice because i dont have to think :^)

[grand shadow]

hmmm I will look into it

[grand shadow]

I see

[prime mulch]

i switched to turnstiles bc autoboost seemed wacky

[orchid fog]

What's the deal with it?

Effectively is priority inheritance according to https://learn.microsoft.com/en-us/windows/win32/procthread/priority-inversion

[analog berry]

@grand shadow so is the warning still there?

even with this patch

[grand shadow]

ok so

I found a way that removes the warning

[analog berry]

what is it

[grand shadow]

adding #include <uacpi/platform/libc.h> to kernel_api.h

[analog berry]

huh

[grand shadow]

idfk

[analog berry]

im confused

[grand shadow]

but the zeroed define comes from libc.h

and the kernel_api uses it

[analog berry]

wait why would it come from libc

[grand shadow]

where am I supposed to define it

[analog berry]

in your makefile...

[grand shadow]

BRUHJHHHHHHHHHHHHHHHHHHHHHHH

kms

[analog berry]

-DUACPI_NATIVE_ALLOC_ZEROED

bruh

of course its not propagated everywhere

not every c files includes libc.h

[grand shadow]

https://tenor.com/view/scrim-gif-25728233

I may be stupid

just might be

just a maybe

perhaps even

[analog berry]

lmaooo

at least u got free contrib points

[obsidian pivot]

@grand shadow ur thing might be broke

on latest astral

[grand shadow]

you have to do make clean because I updated all of jinx

[obsidian pivot]

fuck

[grand shadow]

do you really just not want an iso

[obsidian pivot]

can you send me that minimal astral

iso

I can't be bothered to build gcc again

(x2)

for astral

[analog berry]

Inb4 astral ps2 driver bricks your laptop

[obsidian pivot]

still borked

with teh astral iso I have

from a long time ago

yeah the self tests failed

for both ports

[analog berry]

Bruh 💀

[grand shadow]

probably the same will happen, I didnt really touch the ps2 code since then

cant really debug when the only computers I have for debugging it works

[obsidian pivot]

yeye

[dim arrow]

I can test it as well if you want

[grand shadow]

I mean if you want

dont expect it to work

[obsidian pivot]

good news and bad news

good news: self test did not fail

bad news: still no input

[analog berry]

Its probably a race

[grand shadow]

it works on my laptop which has a real controller, it works on my desktop both on real controller and on emulation

its just a curse from the heavens

[obsidian pivot]

btw the ops/s is 400k/s

for uacpi

[grand shadow]

heres a nickel kid, go buy youself better aml

[dim arrow]

my usb is so slow, so it is taking a while to copy

[obsidian pivot]

3849 ops

[dim arrow]

I haven't turned my laptop on in 5 weeks, so I hope it has some charge

and it is dead

it can't mount the root with the default boot selection

which makes sense

I'll try the initrd option instead

it works

typing works fine

[grand shadow]

WHAT

https://tenor.com/view/vegeta-majin-dragonball-gif-7829830

[dim arrow]

so maybe oberrow's laptop is cursed...

222k ops/s btw

my laptop is very slow though

only 1.10GHz

[obsidian pivot]

a blessing from the heavens!

glory be to god

[dim arrow]

4925 ops in 22ms from 12 AML blobs

[dim arrow]

caps lock light doesn't come on

caps lock doesn't work at all actually

shift does

[grand shadow]

yeah caps lock is unimplemented on the terminal

tho on xorg it works fine

(except the lights)

[dim arrow]

pressing the power button did shutdown my laptop after waiting a couple of seconds for init to kill the tasks

[grand shadow]

it only takes that long because bash is an annoying lil bitch who doesnt die on a SIGTERM

it sets up a custom action to do idk what

[dim arrow]

I'd already exited out of bash

[grand shadow]

interesting

[grand shadow]

today I'll probably just unshit the style of the x86-64 code

I really dont feel like going 200 files but my action have consequences or something

[fervent fjord]

just update the style as you modify the files

no need to be insane

[grand shadow]

true

[supple hatch]

mathew

u awake

[grand shadow]

I think so

[supple hatch]

do u have 5 mintues and a laptop

or a pc idrc

[grand shadow]

Not rn

[supple hatch]

fair

[supple hatch]

no astral progress in

2 days

😡

https://tenor.com/view/get-back-to-work-gif-10302248683148027648

get BACK to work /j

[grand shadow]

Soon

Also what the fuck is that gif

[supple hatch]

no idea

💀

i searched

get back to work in tenor

[fierce forge]

Where do you find these gifs?

[grand shadow]

That looks like youtube kids slop

[supple hatch]

this came up

[supple hatch]

it is

[supple hatch]

[craggy tide]

no Astral progress in 25 minutes

literally dead project

[hearty fulcrum]

dead os xd

[fierce forge]

[prime juniper]

https://tenor.com/view/whipped-manager-gif-12488413

[fierce forge]

dead os xd
I petition to remove Astral from the uACPI leaderboard

[grand shadow]

why does lto not like this assert __assert(page->refcount == 0);

bruh

[fierce forge]

SMP?

[grand shadow]

this happens in very early init

[fierce forge]

My kernel doesn't even enter main? lol

[grand shadow]

wha

wrong thread?

[analog berry]

There should be a policy, something like if there's isnt a commit for a year it gets removed or smth

[prime juniper]

Brb setting fallback commits to bootstrap-Managarm for something simple each year

[analog berry]

Lmao

I'm sure Managarm will remain active lol

[prime juniper]

Yeah lol

I would be sad if it died out. Especially if with it all its subprojects died too. Losing mlibc would be a big loss

[obsidian pivot]

if anything if we lost mlibc someone would fork it and continue maintaining it

and maybe add a backdoor or two

[prime juniper]

XZ moment

[prime juniper]

Likely

[fierce forge]

With lto

[obsidian pivot]

skill issue

you're welcome

[grand shadow]

bruh ubsan is silent about whatever is happening in the pmm

useless piece of

[analog berry]

Yeah

Ubsan only catches most blatant ub

[grand shadow]

well at least I have ubsan now after like almost 2 years

(I stole most of it from @elfin turtle ) ((I gave proper attribution))

and it caught some stuff so

[fierce forge]

I added it, and it made the bug not manifest itself anymore

[elfin turtle]

lmao

[hearty fulcrum]

Astral disappeared from the active threads list

[grand shadow]

dead os xd

[crude agate]

need to save astral in active threads

[prime juniper]

sets timeout to 1 hour…

[lean kettle]

sets timeout to 1 week...

[grand shadow]

I think Im just gonna do devices next. I really wanna do device stuff

Like a proper driver subsystem

[prime juniper]

Demotes you….

[lean kettle]

🥺

[obsidian pivot]

I was looking into why I had so little storage on my PC left

and it turns out

astral takes 16G

(this is a minimal build)

[prime juniper]

cries in Managarm source tree

[tulip zinc]

is this dead

[obsidian pivot]

no

kinda /j

main dev got addicted to factorio

cc @grand shadow

[tulip zinc]

valid

[grand shadow]

wrong game

[wild marsh]

cogmind moment

[grand shadow]

truly

[grand shadow]

Worth procrastinating, this was a super good run. Usually I do stealth but I tried a treaded combat run and it was very fun and got like 47k points

Died due to alert spiral right next to the fkin exit 💔

[prime juniper]

Start again

[crude agate]

.

[grand shadow]

dead project

(I've just slowly been thinking of how I'm gonna approach the whole device driver stuff)

with hotplug and modules and etc

[weary scaffold]

Come back bro this shit Is so fire, giving me inspiration for my project

[unreal merlin]

hi @grand shadow! I made a small PR for Astral to use the IA32_APIC_BASE MSR to get the LAPIC address instead of grabbing it from MADT, if you can have a look

[proper meteor]

why is it better to do it like this?

[obsidian pivot]

the MADT's LAPIC address can be bogus

sometimes

although idk of any cases where it has been

but always remember the golden rule of dealing with anything osdev: firmware devs are braindead

[proper meteor]

lmao

[unreal merlin]

neither do I, but in my opinion it's better to trust a MSR over an ACPI table

[analog berry]

TLDR: one is static data that is (hopefully) up to date, the other is actual dynamic source used to update it as well

[tribal dust]

yea the MSR can be updated to move the LAPIC address

[grand shadow]

Hi I will look into it soon tyy

[grand shadow]

@unreal merlin merged ty

[analog berry]

A day with a new astral commit is a good day

[grand shadow]

@carmine swallow is the openjdk work in any fork?

[carmine swallow]

I don't think but I can push it rq

[grand shadow]

Would be cool if you could make a pr

[unreal merlin]

ty too

[carmine swallow]

I made a pr, it turned out to be kinda big because of all the other ports and stuff tho

[grand shadow]

Merged tyyyy

I asked for it both cuz I wanna see if I can get openjdk to fully work and if some of these bugfixes fix qemu

[carmine swallow]

here is also an archive of the files that I put to distro-files for the mc server (mostly just eula.txt, server.properties with online mode disabled though I don't think that really helped with anything and a script that I used to start it) , you also need -netdev user,id=net0,hostfwd=tcp::25565-:25565 in Makefile QEMUFLAGS (the hostfwd part is the only change)

[unreal merlin]

are you trying to run minecraft on Astral?

[carmine swallow]

the server already kinda works (though iirc sometimes it hangs while loading waiting on some futex) but if it loads then you can get to the world on a client on your host system for a couple seconds before the server dies because of some auth related watchdog timeout

[grand shadow]

Not minecraft itself but rather the minecraft server

Though with openjdk working I wouldnt doubt mc works

[grand shadow]

Yeah I had to do something with hostfwd when I ported openssh

[unreal merlin]

nice

how long did it take to port openjdk?

[carmine swallow]

to astral specifically not that much, like a few days as astral already had basically everything that it needed

originally I ported it to managarm and at that point it took somewhat longer because there were some mlibc bugs iirc

[warped plinth]

https://cdn.discordapp.com/attachments/1061407633745125397/1284644688787279914/astral.gif

[supple hatch]

its been

2 days

😡

[lean kettle]

Astral is dead

[crude agate]

nooo

If the astral is dead, who will stop nyaux ?

[prime juniper]

Managarm of course! We’ll send the to him and then he is doomed to get a PhD first

[analog berry]

nyaux himself

[crude agate]

nyaux's shitcode is a big threat to humanity

[crude agate]

quackos is a new nyaux killer 🙄🙄🙄

[obsidian pivot]

I will of course

[crude agate]

legendary battle

[fierce forge]

who pagefaults the last

[supple hatch]

😡

[supple hatch]

die.

[supple hatch]

[fleet osprey]

Astral thread is archived, it's so over

[grand shadow]

Ive had cogs on my mind lately

My plans for when I go back into working on this is finally unfuck my device driver handling which is something I have wanted to do for a while

[fleet osprey]

maybe Astral will finally have working PS2 support from that

[grand shadow]

Nah thatd go against the laws of nature

[obsidian pivot]

Obos will take over

[supple hatch]

3 days no progress 😡

[grand shadow]

FACT: 90% of osdevs quit right before replacing windows

[robust geyser]

mintia on x86 more progress than astral

tragic

(pls come back i miss astral)

[analog berry]

ikr i miss astral progress

and i wanted to use it for the future project

but its dead now

[bronze nexus]

Which project would you use it for?

[analog berry]

PoC 3d acceleration on an igpu

[bronze nexus]

Damn, Astral must be quite far along then

[analog berry]

it has a mesa port so yeah

[grand shadow]

Its asleep it wont die dw dw

I still have plans

[analog berry]

[fleet osprey]

https://tenor.com/view/mistamandrew-plans-i-have-plans-i-cannot-share-haters-gif-15485001350530694977

[prime juniper]

Managarm is ready for you

[analog berry]

userspace drivers + c++ + already existing drivers etc etc

this is more of a fun poc to see what i can achieve

if i make a proper library similar to udrm then ofc

[prime juniper]

Fair, can be C, which drivers

[analog berry]

lol

[analog berry]

@grand shadow i have downstream stuff to make astral work with uacpi 1.0

i can update if u want

[grand shadow]

Sure go ahead

[grand shadow]

I was planning on doing some more work this weekend but I ended up doing something completely unimaginable

I touched grass

Crazy I know

[fierce forge]

What's bad about userspace drivers?

[analog berry]

Nothing, I just dont want to touch them personally

[proper meteor]

No way

[hearty fulcrum]

teach me this power

[obsidian pivot]

what's this grass thing your mouth speaks of!

kill the witch!

[obsidian pivot]

when astral

@grand shadow

[grand shadow]

soon + a few business days

[obsidian pivot]

[sacred gyro]

At least it’s not soon + a few business years

[analog berry]

a few business centuries

[hearty fulcrum]

too soon

[obsidian pivot]

when will mathewnd port debian to astral

debian mlibc/astral