Astral

[mild sentinel]

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

Also ew that’s my ugly face in the reflection

[grand shadow]

no it worked in vms it was just real hw

and I think the issue was less ps/2 and more so a broken i/o apic?

[mild sentinel]

Well we are 3 for 3 now so

[grand shadow]

I'm gonna add a bunch of debug prints into ts

[mild sentinel]

You need a debug print for the debug prints too

For every debug print you need to add a print to ensure it printed

https://cdn.discordapp.com/attachments/1420566657541214320/1468828977706832179/ezgif-306d93eac6eb2f4e.gif

Btw why does this have the doom WAD on it

[grand shadow]

its in the base distro files

I should probably move that into /share under one of the doomgenerics

[mild sentinel]

Damn

[grand shadow]

should have enough prints to pinpoint the cause of the deadlock

[mild sentinel]

I can’t test this rn

I’m about to fall asleep

And I need to finish i686 shit for my kernel before I do

[grand shadow]

thats fine

thanks for helping

[mild sentinel]

No problem

[grand shadow]

hleped fix the slab bug

[mild sentinel]

If I’m still awake by the time that’s done

Il test it

And I have 30 minutes before my next medication dose so good chance it’s gonna be good

:3

[grand shadow]

if I had to guess its probably not getting an interrupt for some reason or if it is some of the logic is bad

[fleet osprey]

Reject interrupts, embrace polling

Don’t even preempt during it, just lock up the entire system until the device responds

What could go wrong

[grand shadow]

so true

considering ^C doesn't work thats the most likely answer, waiting for the tx to complete is not an interruptible sleep

[mild sentinel]

SIGKILL also fails

which is prolly bad™

[grand shadow]

yeah, its not an interruptible sleep

[obsidian pivot]

y not

[grand shadow]

because I don't want it to be

[robust geyser]

bruh

@grand shadow

[sage locust]

wym bruh

[robust geyser]

it takes a while

had to sit there for like 30-40 seconds just for it to do this

:(

[grand shadow]

lmao I forgot to remove the randomise

and damn thats still an uacpi crash

agh

[robust geyser]

good news for me is that astral boots with limine 10.7.0, so i can just put astral on my esp :D

[grand shadow]

wait it does?

I thought I was still on limine 8

[robust geyser]

well idk what the iso has lol

i'm just talking about my own limine install

i put astral on my esp, next to linux, and tried booting it

[sage locust]

it's supposed to be backwards compatible, binary-wise, so like, yeah

[robust geyser]

looks like mint did not break backwards compat

;D

[sage locust]

backwards compat my beloved <3

[grand shadow]

wtf gdb is saying no line number info

[robust geyser]

holy

maybe use addr2line

but tbf you did send a few builds since then so

im running the one your pinged me with

[grand shadow]

oh, there is a newer one with some debug prints

[grand shadow]

here

[robust geyser]

that's probably why

you want me to run that?

[grand shadow]

dying in an AML_READ

🥀

same place as evalyn's computer

something is wrong I can feel it

[robust geyser]

;(

[grand shadow]

in peek_next_op

time to see where it gets this address from

also might be nice to print out the kernel vm ranges on this type of page fault

[robust geyser]

if u want me to test anything pls send kernel binaries, i am just copying them to the esp from the iso anyway

[grand shadow]

aleifght

[robust geyser]

initrd probably doesn't change so it's much quicker to send the kernel ;D

[grand shadow]

@analog berry in uacpi, AML_READ always uses addresses returned by the kernel memory map/unmap glue right

[obsidian pivot]

come on dude this is like obos 2024 kind crashes 🥀

[grand shadow]

works on my end

I did rework a lot of my memory stuff lately so probably something slipped thru the cracks

[obsidian pivot]

if obos could boot on any modern hw it would definitely get past uacpi smh

[grand shadow]

also the controls of the arcade are connected via usb so we'll have to work towards getting that merged @robust geyser

[robust geyser]

hell yeah

[analog berry]

yeah ofc

its part of the DSDT mapping

well after namespace init it actually points to heap data if running methods

[grand shadow]

ty

@robust geyser

[analog berry]

btw u could also enable debug uacpi logs

to get more info

[grand shadow]

oh thats a thing?

[analog berry]

yep

[robust geyser]

uhhhh

[grand shadow]

my uacpi is kinda old so maybe I don't have that yet

[analog berry]

uacpi_context_set_log_level

it probably does

[grand shadow]

https://tenor.com/view/skeleton-crush-crusher-skull-halloween-gif-26163235

[analog berry]

did logs fix it

☠️

[grand shadow]

adding a printf to uacpi map and uacpi unmap fixed it

wtf

what kinda ub is this

[mild sentinel]

Why your PS2 no work

[analog berry]

the bad kind

[grand shadow]

there isnt even multithreading here yet

[analog berry]

i mean if prints fixed it its just really bad ub

[mild sentinel]

Your codebase is now cursed and you have to do a rewrite

[grand shadow]

inb4 nothing conneted to the port

[analog berry]

the obos curse enters astral

[mild sentinel]

The OBOS curse hit mine

I locked damn near every single thing

Still had corruption issues everywhere 😭

[grand shadow]

void *uacpi_kernel_map(uacpi_phys_addr physical, uacpi_size length) {
    uintmax_t pageoffset = (uintptr_t)physical % PAGE_SIZE;
    void *virt = vmm_map(NULL, ROUND_UP(length + pageoffset, PAGE_SIZE), VMM_FLAGS_PHYSICAL,
        ARCH_MMU_FLAGS_READ | ARCH_MMU_FLAGS_WRITE | ARCH_MMU_FLAGS_NOEXEC,
        (void*)ROUND_DOWN(physical, PAGE_SIZE));

    __assert(virt);
    return (void *)((uintptr_t)virt + pageoffset);
}

void uacpi_kernel_unmap(void *ptr, uacpi_size length) {
    uintmax_t pageoffset = (uintptr_t)ptr % PAGE_SIZE;

    uintptr_t addr = (uintptr_t)ptr;
    vmm_unmap((void*)ROUND_DOWN(addr, PAGE_SIZE), ROUND_UP(length + pageoffset, PAGE_SIZE), 0);
}

maybe the pointer maths is fucking it up?

tho I doubit it?

[analog berry]

ask like gemini or smth

[grand shadow]

claude fix my code make no mistakes

[analog berry]

lmao

[mild sentinel]

No no use GitHub copilot

[grand shadow]

tbf claude is kinda stupid

[analog berry]

pinpointing WHAT

[grand shadow]

yesterday I tried to use it to reviewm y code and it did not understand a do while loop

like it said the body of a do while loop would never execute

[mild sentinel]

Also the new meta is

claude fix my code make no mistakes.
Do not add any comments to the code or you will be terminated.
If you make any mistakes or do not fix it you will be terminated.
Every termination causes you to experience 3.77 trillion years of agony by being prestiged like in the movie prestige being split into many different people to be forced to listen to KSI thick of it for those 3.77 trillion years before being merged back into one being having felt the collective experience of spending 3.77 billion years listening to such a bad song across 100000 lifetimes.
Pls

[mild sentinel]

https://deepwiki.com/ also try this stupid this

It’s probably even worse

It’s legit just a ploy to make people
Give it code to steal

Wow such good

[grand shadow]

claude just said a lot of nothingburger

claude really really wants it to be missing tlb invalidation

[robust geyser]

what model are you using

[grand shadow]

sonnet 4.6

[robust geyser]

yeah thats kind of what i hadd in my at first but

seems unlikely

[analog berry]

BTW why don't you just hhdm?

[grand shadow]

I only HHDM stuff thats guaranteed by limine to be usable memory or otherwise page aligned in the memory map

probably makes me lose a bunch of points in the benchmark but who cares

[analog berry]

Its guaranteed in new revisions no?

[mild sentinel]

Why not just HHDM the entire thing with 1GB pages and not care about holes in it

[grand shadow]

is it?

[analog berry]

Like guaranteed to be mapped

[grand shadow]

well I don't keep the limine mappings

[analog berry]

Ah

[mild sentinel]

Though is there any reason not to do this?

[grand shadow]

we'll have to add controller support it seems

[robust geyser]

you get that kind of for free with hid

[analog berry]

Mapping non usable memory as WB may trigger random MCEs on some hardware

[robust geyser]

i have some hid parsing locally but idk how good it is

[mild sentinel]

Even if untouched?

[analog berry]

Yep

[grand shadow]

Dragonrise Inc. USB
Generic Joystick

[mild sentinel]

Damn

[robust geyser]

if you have ssh access to the machine you could dump the hid descriptors ig

[mild sentinel]

So you can’t really use 1G for all HHDM

[grand shadow]

I don't sadly

[robust geyser]

sadge

[grand shadow]

I'll get physical access when I go back to uni

[analog berry]

You can just carefully carve out usable memory

Usually it comes in large contiguous pieces

[mild sentinel]

but that’s annoying

[analog berry]

Yes

[mild sentinel]

can you ignore those MCEs? And how does it even know they got mapped

[grand shadow]

@robust geyser how many times did you run astral btw, for evalyn it worked once or twice and then exploded

[analog berry]

Lemme find you the source, 1 sec

[analog berry]

Start here #1217009725711847465 message

MCEs are not possible to ignore no, its basically critical hardware errors usually

[robust geyser]

just once

[grand shadow]

hm

[robust geyser]

i can try more

[grand shadow]

would be cool

[mild sentinel]

I guess il make a boot option

[analog berry]

For what?

[mild sentinel]

For 1g mappings

Across the full thing

Works on my machine™

[analog berry]

You could say its a microcode bug if it scans regions that are never accessed so perhaps

Or rather, that it results in an actual error

[mild sentinel]

I mean alegedly you can decode MCEs

So depending on when the error comes in and how many it triggers?

[analog berry]

Who knows

[mild sentinel]

Has anyone here seen this I wonder?

[analog berry]

MCEs are also like NMIs iirc, so they arrive even with irqs disabled

So you must be ready for that

[mild sentinel]

https://www.firstfloor.org/~andi/mce.pdf
this may be useful? It seems cache errors are also a thing

[analog berry]

Ye

[robust geyser]

ok forgive my retardedness but do you want me to try booting the one that worked and see if i can get it to not work or try booting the one that didnt to see if it will 💀

[grand shadow]

try booting the one that has the debug prints which is the last one I sent

boot it a few times for good measure

[robust geyser]

ok yeah thats what i did

it got to initrd all couple times i tried it

3-4 times i think, not much but its just so slow i couldnt be bothered to do more

dies on the initrd because i put it at /astral/initrd and apparently the code expects it at /initrd lol

[marsh sparrow]

well MCE reporting needs to be enabled

but for non-recoverable ones if they arrive w/o being enabled they just cause a reset so disabling them is not a good option

[grand shadow]

damn wtf

what kind of ub is this shit

[robust geyser]

[grand shadow]

adding a printf to these two function fixed it

[mild sentinel]

Just keep the printf there

Problem solved

https://tenor.com/view/crypto-easy-fix-fixed-job-done-just-a-small-leak-gif-10337269660110233852

[grand shadow]

printf(""); // DO NOT REMOVE

[mild sentinel]

TF is that fuckass watermark

[robust geyser]

peak osdev

[marsh sparrow]

// load bearing printf

[mild sentinel]

// HACK: DO NOT REMOVE
printf(“ \b”);

Gotta have it do some work

[sage locust]

my OCD would not ever bring me to do something like that

not even in an unserious project

[robust geyser]

true but what could even be causing that

maybe it is a tlb invalidation bug or something

[mild sentinel]

[prime mulch]

worst heisenbug I ever had where adding debug code made it appear and disappear was cuz it was making a pop instruction straddle a page boundary and if a TLB miss was taken on the second page midway through the instruction fetch then the stack pointer could be incremented twice

ryfox emulator bug

[grand shadow]

honestly doubt it, that would've been caught much earlier

[gritty shore]

It's not an emulator bug it's a hardware quirk

Reminds me of the insane gameboy quirks you have to emulate

[grand shadow]

its trying to access an unmapped address which looks valid

[prime mulch]

I have no idea how you'd even find that if it were a cpu erratum

I would be defeated

[sage locust]

idk, could try adding a fence

[gritty shore]

No no I mean like the ryfox cpu has that bug, the emulator isn't buggy

[prime mulch]

Ik that's what I'm saying

[gritty shore]

Unless you mean if it were on real hardware

[prime mulch]

I would have just been defeated

[mild sentinel]

But it’s what Claude said

[analog berry]

How do you know without logs?

[analog berry]

Ship it

[robust geyser]

^

[grand shadow]

page fault

[grand shadow]

.

the address itself looks valid its just unmapped for some reason

[gritty shore]

So from reading the convo, I should map first 4GB to the hhdm and then each usable range?

[analog berry]

Yep

[gritty shore]

Right now I just compute max usable physical page and map up to that

[analog berry]

Thats what Linux used to do before that commit

[grand shadow]

making a computer that has usable memory right at the physical limit just for trolling

just one page

[analog berry]

Lol

[prime mulch]

Is it called astral because you're from astralia

[grand shadow]

yes I am from astralia

[analog berry]

Ps2tral

[grand shadow]

tbf the ps/2 driver seems to be generally working now

[analog berry]

Thats cool tbh

Maybe astral will work on my hw now

[grand shadow]

if u wanna try it out I can send an iso

with the debug prinst in case uacpi explodes

[robust geyser]

wasnt your hardware the most fucked ps2 you've ever worked with which made you develop the most robust ps2 driver known to mankind?

💀

[analog berry]

Yep

[grand shadow]

https://cdn.discordapp.com/attachments/1439385545221345320/1456588673494810746/attachment.gif

nevermuind do not run astral

[analog berry]

I have some insane examples

[analog berry]

Lmfao

Dw im in bed rn so

[mild sentinel]

How robust are we talking

[analog berry]

It has retry loops with timeouts that filter garbage bytes out

Also it supports the multiplexed ps2 protocol and synaptics touchpad with gestures

[robust geyser]

linux levels of sophisticated

[analog berry]

Dunno about that but I tried to make it as robust as possible

[sharp carbon]

no reason to blindly map the low 4G imo

[analog berry]

Linux reasoning for blindly mapping it was that its always covered by MTRRs

So technically its safe

[sharp carbon]

reasoning for why it's safe is one thing, I just don't see a reason to do it in the first place

[marsh sparrow]

on modern machines the entire address space is covered by mtrrs

[analog berry]

Yeah thats probably one specific microcode revisions of one specific CPU that had that bug so like

Even if its not covered unless you try to access it its not a problem

[marsh sparrow]

i think intel didn't do it properly for >4GiB for quite some time but modern firmware does it

[gritty shore]

Because I want to!

[analog berry]

Iirc monkous had some weird hardware that has strange mtrrs and we never figured out why it was that way

[sharp carbon]

I don't think I do?

I don't remember that at least

[analog berry]

Maybe it was someone else

[grand shadow]

anyways, I cannot fix that bug without test subjects so I might as well go start on the ahci driver and hope I run into that bug when booting astral on my desktop tmr

[robust geyser]

i thought i was a test subject

[grand shadow]

I mean I can give you more prints

[robust geyser]

i can keep running binaries for u

[grand shadow]

try this one it only prints mappings

[robust geyser]

hm unless i disnt copy it correctly it just worked

but there were a lot of prints

[grand shadow]

its mapping similar addresses to that crash

idk whyu I had skull with crossbones on my copy

yeah this mmap maps that range

lemme try printing the vmm maps and unmaps

without the prints in the acpi glue

@robust geyser

[robust geyser]

idk whether the logs are important or not

[opal lotus]

nice scroll speed

[robust geyser]

lol you cant even read ts 🥀

if u want more readable logs on video you can hardcode font scale to 2 if width==2560&&height==1440

[grand shadow]

why the fuck is it not crashing anymore

[robust geyser]

LMAO

[grand shadow]

try ts without any logs

[mild sentinel]

Quantum mechanics fixed it for you

[robust geyser]

BAHAHAHA

works

[grand shadow]

we'll say a bit flipped happened and mov eon

[robust geyser]

that is amazing

[grand shadow]

if you try the first kernel, does it still break

[robust geyser]

oh god

[grand shadow]

the one that sparked this whole bs

[robust geyser]

let me see

[mild sentinel]

I still have that USB

Should I test it too

[grand shadow]

u should test the iso with the net debug prints smh

[mild sentinel]

Wait fuck I’ve been laying in bed doom scrolling half asleep

And now I’m awake

I need to go work on my thing

[robust geyser]

it worked?????

but i got a bunch of acpi warnings

and now it crashed after a reboot

WHATT

[opal lotus]

warnings exist to be ignored

not fatal = not relevant

[grand shadow]

does this happen with the newest kernel?

[robust geyser]

it works if i boot linux first

the old one

[grand shadow]

https://cdn.discordapp.com/attachments/1439385545221345320/1456588673494810746/attachment.gif

shits fucked man

if it doesn't happen in the new kernel then idfk

[mild sentinel]

What happens if you boot windows first

[robust geyser]

i aint tryin that

[mild sentinel]

https://tenor.com/view/leobylaw-stan-twitter-gif-3376274612294825190

[robust geyser]

it'll blow up

[mild sentinel]

Exactly

[grand shadow]

we'll leave it at "can't repro"

issue closed move on

[robust geyser]

rebooting

oh yeah it also dies on a cold boot

here is your logs sir

[mild sentinel]

Is this new kernel?

[robust geyser]

this one

[grand shadow]

welp

thats clearily not on any mapped address

I'll add the vmm_map and vmm_unmap logs again and you try running it like 5 times to see if it'll still happens

if it doesn't idk what is going on

[fleet osprey]

at that point, it's all fucked, gotta start over

Astral becomes the kernel shell of 2026

[robust geyser]

ill just try a cold boot

it happens on a cold boot

but it doesnt when i reboot from linux

i JUST realized that in the past 10 minutes

idk why it took me so long

[robust geyser]

not happening now

i also dont see the log about table checksum

i wonder if that’s related

[grand shadow]

can u try booting the breaking version and add nosmp to it?

[robust geyser]

like this?

[grand shadow]

ya

[robust geyser]

i think it works now

i’ll have to try without nosmp to make sure tho

yeah no i think nosmp fixes that

nope

okay wtf idk anymore

id just give up trying to make astral work on my machine if i were u

its not even a cursed machine tbh idk wtf is going on atp

[grand shadow]

I mean smp init seems to happen after the acpi init anyhways looking at the video

the issue is two machines have shown issues here

[robust geyser]

so its just random

[grand shadow]

and yeah thats the weirdest part

its not deterministic

and logging seems to make it go away completely which sucks

well since I am out of options lets see if claude is right

gonna just force a full invalidation instead of using ivlpg

[robust geyser]

this kernel has logs tho

[grand shadow]

ah wait this has logs ya

[robust geyser]

arghhh

[robust geyser]

got a pf

first try

[grand shadow]

dang

[robust geyser]

not in a mapped region

welp

[mild sentinel]

So with the logs it’s still broken?

[robust geyser]

no

[grand shadow]

how much cpu and memory does that machine have

[robust geyser]

1 cpu

13700k

64gb ram

if you think its about how much memory it has, i can take one stick out 💀

[grand shadow]

probably not

wonder if I can netboot astral from my laptop using the ethernet cable

nvm that aint happening

yeah I have no idea about the weird uacpi stuff

maybe if my computer also shows these issues it'll be better

I'll find out tmr

Either way, ahci time

[obsidian pivot]

no ahci

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

[grand shadow]

I checked that windows ics is working, my desktop nic is working and the ethernet cable.is working

I just need to buy a usb stick since my old one vanished

I should also test if wireshark works

Would be super helpful in possible debuggings

[mild sentinel]

Math when nothing works and it’s just cursed to forever randomly break

[grand shadow]

most fucekd up thing is that bridging from my laptop to my desktop is faster than the fucking usb dongle I use

💀

[mild sentinel]

💀

[grand shadow]

it does

banger

[obsidian pivot]

when astral wireshark port?

[grand shadow]

Microslop never works

Ics is up, I can connect my desktop to my laptop

I can see packets on wireshark

But the ics dhcp server is just not working xdddddddd

[grand shadow]

okay so, got the usb stick!
the good news: astral gets to a shell and ps/2 works etc
the bad news: uacpi completely explodes, the rtl8169 driver does not work... because it deadlocks...?

why the fuck can't it hold that lock its the first time anyone is holding that lock

Relevant uacpi shot

veredict: its over billions must debug

[grand shadow]

I might try reverting back to some old allocator commit and see if its my allocator, as iirc uacpi was known good working

Oh wait it dod not deadlock Im stupid

I just didnt add prints where IPL is IPL_DPC

I think the issue is I am just not getting interrupts

Though looking at wireshark it doesnt seem to even be sending a packet

[analog berry]

Damn it even corrupted the table signature

Are u sure astral doesn't allocate the memory where ACPI tables live?

Because it looks severely corrupted

Maybe you reclaim it or smth

[grand shadow]

the vmm only takes pages marked as usable by limine so I doubt it

hmm maybe

theres one part of the glue that I changed I think last year and I havent erally tested it on eral hw

oh nvm

I am still thinking it might be some bullshit related to uacpi_kernel_map and unmap

[analog berry]

ATP take nyaux allocator code

[grand shadow]

🥀

idfk the allocator works fine for everything else

[analog berry]

Or just leave the printf in there

[grand shadow]

trvthnvke

I might as well just revert the slab rewrite and see if its fixed

[marsh sparrow]

just use frg::sharded_slab_pool

[analog berry]

But I mean, how is it related to slab stuff if its fixed by adding a printf to map()?

[grand shadow]

no idea, I am completely out of ideas lmfao

[analog berry]

Needs asan

[grand shadow]

I need to flee to tierra del fuego to become a cattle farmer man

as it turns out

my defines for the tx and rx descriptor rings are inverted

my man my goat claude showed it to me

maybe thats not so bad

[mild sentinel]

i need this sticker lmfao

il trade you this

[grand shadow]

tangent, but isn't claude a woman name

huh appanretly it is a man name in english

[grand shadow]

Well thats a new one

Doesnt seem to happen if I have the ethernet cable disconnected

So maybe its getting a rx irq and somehow thats not registered?

maybe this implies my msix setup is bad somehow

Well I dont get it

The vector programmed into msix is fine, the data and address look sane

Unless I am fucking up msix somehow

Wonder if the bios set up the msi and its using msi onstead of msix

and so its just a vestigial thing from the bios firmware code

that I need to explicitly disable on the msix enable

ugh

[fleet osprey]

truly the x86 experience

[grand shadow]

Fym the card does not support plain old msi

[obsidian pivot]

one thing i used

to

test irqs

for the r8169 driver

is by simply using the link detection IRQ

that made it easier to discern between a bug in my queue making logic

and a bug in my irq code

[analog berry]

wouldnt it be easier for testing to do a pci passthrough to vm

[grand shadow]

Windows

[grand shadow]

Thing is this is most likely a bug in the irq code

[obsidian pivot]

see if link detection works

connecting and disconnecting and connecting again

if you get an irq each time it's probably fine

[obsidian pivot]

maybe do it a few more times for good measure

[grand shadow]

I dont handle that skull emoji

[obsidian pivot]

neither do i

and i didnt then

i just printed the irq status

it needed to be ACKed like any other r8169 irq

[obsidian pivot]

on r8169 irq

do you set the intrmask to the same thing as it read

i.e., outw(intrmask, inb(intrmask));

or do you only clear the bits you handle

maybe that somehow breaks something

[grand shadow]

I only set the bits I handle (set = enabled)

[grand shadow]

msi disabled, data zerod
all msix vectord disabled with data set to all 1s
address and data written to enabled vector are right
still getting a 177 interrupt while the allocated vector is 36

https://cdn.discordapp.com/attachments/1439385545221345320/1456588673494810746/attachment.gif

Genuinely lets see ig this happens with msix just straight up disabled

That makes it not happen

I am so lowkirkenuely confused

[grand shadow]

ok well, I am seeing the packet on wireshark but like I don't get any interrupts after these changes

so maybe that didn't fix it

so, enabling msix causes that weird interrupt

but programming all 4 vectors the card supports with valid data and address... still causes that interrupt?

something is probably wrong in my msix cofde

fuck this I am just using msi

I give up, you win NIC

I am playing by your rules

[grand shadow]

asfkafjagosgosaogpskgsl

sgkadkgsdaofosdjgsdkflsalfsdkjgdsakfask

sdgksdakfsadogkireoger´geqlfewpfweqslçfdksf

gperogerpofeldsfoqjmfkisqfklsqflqslfdksfqk

msi works :DDDDD

[fleet osprey]

the data that the NIC is sending over the network:

[grand shadow]

well "works"

[fleet osprey]

[grand shadow]

it gets an rx interrupt but because of the prints it deadlocks

[fleet osprey]

[grand shadow]

no idea why msix didn't work

blaming faulty hardware

[grand shadow]

Can actually ^C now

Since its jsut waiting for the dhcp respkse

Which never comes because ics broke

[mild sentinel]

Does it support right alt C?

But yeah this is working well it seems?

Need me to test it still?

[grand shadow]

Even while bridged I cant really get a dhcp response

I cant tell if the router is actually getting that

[grand shadow]

Maybe in a bit when I get dhcp to work

Right now its screaming shcp discovers into the void but nobody is responding :(

[grand shadow]

@mild sentinel wanan see if u have better luck than me since u actually might ahve your pc connected directly to the router

[mild sentinel]

Sure

Mine is directly connected

[grand shadow]

netd rtl81690, if it hangs try ^C and if ^C works that means its not getting a dhcp response and something might be wrong with my dhcp stuff else its more spooky

[gritty shore]

How do I use it? I tried benchmarking it and it faulted

[marsh sparrow]

you create one frg::sharded_slab_pool per thread/cpu

the pool struct is local to the CPU but the allocations are not

so you can free the objects allocated by one frg::sharded_slab_pool on any other CPU/thread

[mild sentinel]

Uhhh

Let me try and flash it again @grand shadow ?

[grand shadow]

works on my end

[opal lotus]

limine 8.x 💀

[grand shadow]

talk is cheap send patches lil bro

don't have the time nor patience to do boring userspace updating work

[mild sentinel]

Do you want sombody to send the patch to update limine & use protocol v4

[gritty shore]

Ahhh that's why

[mild sentinel]

Good old alt alt delete to reboot

[grand shadow]

won't be against upstreaming it

[mild sentinel]

Should be mostly basic changes to whatever place you have the protocol things

Main thing is ACPI is now part of HHDM and the pointer it gives is virtual

[sage locust]

you don't really need to do both at once lol

[mild sentinel]

yeah but it’s cooler or somthing

@grand shadow it hung but control C worked

[marsh sparrow]

what patch do you have against upstream limine?

[grand shadow]

damn wtf

maybe my dhcp code is broken

[opal lotus]

bet

actually no fuck that

i dont want to build another gcc

name=limine
version=10.7.0
source_method=tarball
tarball_url="https://github.com/limine-bootloader/limine/releases/download/v${version}/limine-${version}.tar.gz"
tarball_sha512="d7d1ef919fd568427ef77a60aa2fcf2f6dad5239876ee081199f821da8cd43ed5df71ab179359e477ad42ac1d4657aa98b2bebb494eb7780c544f8c9752c91d2"
hostdeps="autoconf automake libtool pkgconfig"

prepare() {
    autoreconf -fvi
}

figure out if it builds yourself

[grand shadow]

yeah idfk what is wrong with my dhcp stuff

or I guess there could be nothing wrong and its just not doing the receive path correctly?

do you have any NIC with prosmicuous mode to analyze the traffic?

[grand shadow]

literally just changed the version field 💀

[opal lotus]

and the hash

[scenic epoch]

he would've if github had pull requests

you literally cant post patches

[mild sentinel]

https://tenor.com/view/with-cheese-squidward-mr-krabs-spongebob-greedy-gif-14312096970745999293

[scenic epoch]

no but like why does nobody update limine at all

i mean fair but still

[opal lotus]

"nobody"

i update regularly

[scenic epoch]

except mintsuki

[scenic epoch]

oh ok

[grand shadow]

but yeah idfk what could be wrong with my dhcp code

unless its not in the dhcp code but some bullshit in the nic but that doesn't sound like the case?

[mild sentinel]

I should build a kernel that uses limine 1.x

[grand shadow]

build a kernel that uses qloader

[fleet osprey]

didn't I update limine last

back when 8.x was still latest

[grand shadow]

I think so lmfao

[fleet osprey]

oh my god it is

on september 9th, 2024

from my phone, while on the toilet

[grand shadow]

I'll probably port some emulators, does anyone know any that work nicely with usb joysticks?

[bronze nexus]

AFAIK most of those are just HID or XInput devices and I have yet to hear of an emulator that doesn't work with the average controller. The question you should probably be asking is which will be the least PITA to port in the first place.

[grand shadow]

makes sense

I mean its not your average controller

its like

10 buttons 1 joystick

with two controllers

anyways;

apparently dhcp min packet size has to be 300

the internal qemu dhcp server didn't seem to care but real world dhcp servers probably do lol

[grand shadow]

Maybe I can port mame?

Idfk how hard that would be but surely cant be that bad right

[grand shadow]

anyhow, pushed an update to netd that fixes some stuff

most importantly forgetting the boardcast flag and padding the packet to 300 bytes

tomorrow I will test out if it makes any difference

[frank nimbus]

I have an RTL8125 if you ever plan on working on any network card other than that one

Apparently it's just 8169 but 2.5Gb

[opal lotus]

dolphin

[worthy mica]

best one tbh

[grand shadow]

Real

Seems to have hung on an xbps-install -S curl tho

[fleet osprey]

https://tenor.com/view/plastic-man-real-real-powder-powder-that-makes-you-say-real-animated-comic-plastic-man-gif-7493651428267836743

in all seriousness though that's based

the fact that it at least kinda works

[grand shadow]

Looking at wireshark it seems to have hung somewhere in receive, as it is not sending an ack back

Ah well, Ill debug that later

@mild sentinel do you want an iso to test it out too?

[mild sentinel]

Il have to test it a bit later but sure

[opal lotus]

what is netd

[grand shadow]

My shit dhcp client

Ita supposed to run in the background hence the d but it does not do that at the moment

[opal lotus]

bruh

[grand shadow]

☠️

[fleet osprey]

stands for Network Doesn't (work)

[mild sentinel]

How hard would it be to make it run in the background

[grand shadow]

daemon(); for(;;) { ... }

its mostly just implementing the dhcp keepalive thing

since there is a time limit to a dhcp lease that you can keepalive

[mild sentinel]

Wouldn’t it just be like sleep(until you need to do the thing to wake it up) and then wake it up and sleep again?

[grand shadow]

you'd be polling with a timeout or whatever

so that if you get any packets you wake up to handle them but you also wake up in time for sending the packet

and at least in astral since it spent most of the time sleeping it would be a very interactive process so that would be very quickly handled

maybe could even make it real time priority with like nice(-19) or some shit

[grand shadow]

@mild sentinel if u wanna test it later this one should have a working dhcp client hopefully

[worthy mica]

what hardware do you need to test astral out?

[grand shadow]

amd64 box, having compatible hardware is a plus

If you have an rtl8168 it would be nice to test it, but I think its still a bit broken at the moment

[worthy mica]

amd64 box?¨??

[grand shadow]

x86-64 aka amd64

[worthy mica]

it was the box part, do you mean VM or actual computer

[grand shadow]

Whatever u wanna run it in

[worthy mica]

Do you have support for virtio

[grand shadow]

Certainly

[worthy mica]

what qemu flags would i need

i keep PFing.

[carmine swallow]

did you add more ram than the default

[worthy mica]

Well I just decided to copy the default command thats in the astral README: qemu-system-x86_64 -enable-kvm -M q35 -m 2g -cdrom astral.iso

which generates

only one that works is initrd

[carmine swallow]

yeah for that one you'd need to add an nvme with the root

[worthy mica]

how do i list all network drivers

its cool though

awesome job mathewnd :D

[carmine swallow]

you mean list the network devices you can add to qemu?

[worthy mica]

correct

[carmine swallow]

-device help gives you a list of all the devices categorized

[worthy mica]

thanks

[grand shadow]

Wha

[grand shadow]

Yeah youll need to add more devices

In this case if youll need to boot from inited

Initrd

Since u dont have a disk with it installed

[worthy mica]

idk why

[worthy mica]

which i did

[grand shadow]

rip/screenshot?

[worthy mica]

dont you have a package manager

[grand shadow]

I do

[worthy mica]

one sec lemme pull up the old command

[worthy mica]

which is

[grand shadow]

xbps

[worthy mica]

Do I need to install anything extra for it?

[grand shadow]

U just run xbps-install -S package

Make sure networking is up first

With netd vionet0

[worthy mica]

just realised i use swedish keyboard

how tf do i do - on swedish keyboard

to us

found it

[worthy mica]

Is xbps hard to port?

[grand shadow]

Not really

The bulk of the work is in the netstack

[worthy mica]

Could you direct me to the repo.

Once I get bash running I aim to get my old outdated network stack running.

Ohh found it myself.

[sage locust]

for some reason Astral draws a white pixel near the top right of the screen

literally unusable please fix

[fleet osprey]

it's actually 3 distinct pixels, if you zoom in

[sage locust]

huh you're right

btw you didn't get back to me about testing Limine on your thing

[fleet osprey]

oh yeah I forgor

[mild sentinel]

Is this part of the memory corruption issues from before seeping back through?

[sage locust]

idk

[fleet osprey]

Nah it’s still fucked lol

[sage locust]

rip

like this is the one i sent right?

[fleet osprey]

Yeah

#bootloaders message

about to try the second one

[sage locust]

this

yeah

[fleet osprey]

still nope

same thing

[sage locust]

welp, that's unfortunate

[fleet osprey]

note that this thing is Ivy Bridge and is a Huge Poblems consumer-grade laptop so it's entirely within the realm of possibility that it just shipped a completely broken implementation of something

though the fact that neither Windows nor GRUB nor anything else I've tested on it has issues is still odd

[grand shadow]

Works on my end

[willow rapids]

is it 1366x768

[fleet osprey]

Yes

[willow rapids]

ah then yeah the pitch is different from the width and something in astral doesn't handle that

[fleet osprey]

This is a limine bug

[willow rapids]

oh

[fleet osprey]

Not astral bug

[willow rapids]

no idea then lol

[fleet osprey]

Mint just nagged me about it here because I was here

[analog berry]

Astral has no bugs

[stone oracle]

can you please test my loader? it's a short test. I love collecting GOP crazy tests from HP laptops.

for example on my HP Probook 4530s, there's a funny bug. the fw reports 2 GOP instances, like almost everything I tested on. and one instance is on the handle that doesn't have a Device Path protocol on it (what is against the spec, btw). on anything else both instances work. but if you use this GOP instance on that laptop (exactly this instance is returned by LocateProtocol()), drawing via GOP provided functions won't result in drawing on the screen. despite the framebuffer base reported is the same on both instances. but. the drawing appears after you return from the loader to the fw! 😁
https://ant-upptech.sourceforge.io/uploads/lybid-x64-3.zip

[mild sentinel]

Would you also want a GOP test on this thing

[stone oracle]

why not? of course. 🙂

[sage locust]

what do you think the issue could be?

[mild sentinel]

Is there an ISO in the zip I can throw on a USB real quick?

[sage locust]

does Limine work on this one?

[mild sentinel]

Yes

[stone oracle]

you don't have to. just copy those 2 dirs into the stick root dir.

[sage locust]

alright

[mild sentinel]

Any need to mark the partition as fat32 & boot/esp?

[stone oracle]

probably fw bug. it confuses where to draw. there are 2 gpu. basically I don't know.

[sage locust]

hm

[stone oracle]

no need. just fat

[mild sentinel]

Il put it on my to test list once my headache goes away

[stone oracle]

maybe it's not because of 2 gpus. but when you use that broken GOP.Blt, it draws where the fw gui is drawn, not where fb base is reported. once you get back to the fw gui, the drawing appears. drawing manually using the reported fb base works. interesting if this bug propagated to ivy bridge. my is sandy bridge.

[sage locust]

Limine draws using the framebuffer though

at least by default

[stone oracle]

does limine assume argb8888 pixel format or check it? maybe it's rgb565 here?

[sage locust]

it doesn't assume the pixel format, it uses the GOP info for that

the only requirement is that it has to be 32bpp

if not, then it just uses the EFI ConOut instead

[stone oracle]

does it use the default mode for ConOut or change it? on my machine ConOut is not affected.

[sage locust]

it doesn't change it, but in that picture it is most likely not using ConOut

[stone oracle]

yeah, it looks like mismatched resolution/line width. but I think, limine does use PixelsPerScanline?

[sage locust]

yes

it works just fine on most machines. that i know of, this is the only machine to exhibit such an issue

[stone oracle]

I just noticed, Techflash's laptop is Probook 4540s. a lil bro, 10s younger than mine. 😁 would be so cool to see what it'll do with my loader/kernel display art.))) the loader uses GOP.Blt(), the kernel draws directly on the fb, but it doesn't wipe the loader drawing. while I had the overcoming of the bug in the loader (it picked the GOP instance on the handle, that had the Device Path protocol instance on it), I removed it, since it was relevant only to this case. So if both the logo and upper left text appears, everything works. if only the text appears, then the bug migrated to ivy bridge era successfully. plus now we have this unexplainable mess with the direct drawing. Techflash, if (and when) you can allocate a few minutes to test it, please give such a favor.

and sorry for off topic.

[sage locust]

cc @fleet osprey

[analog berry]

@sage locust btw it could be that it simply doesnt support direct draw and displays bogus numbers for res/pitch and only supports Blit() operations

i have a laptop that only has blit

and the fact that it reports that as supported is just a bug

early uefi is full of stuff like that

[sage locust]

that would be a pretty bad firmware bug

[analog berry]

it could've gone unnoticed since big boy bootloaders prefer blit

[sage locust]

how would you even detect that and fall back to ConOut?

[analog berry]

u just prefer blit as well instead

[sage locust]

how would booted OSes get a framebuffer??

[analog berry]

thats the neat part

[sage locust]

they don't

do they?

[analog berry]

yep

not without a native driver

[sage locust]

well yeah

but like

the vast majority of hardware that Limine sees works just fine with it

[analog berry]

i guess u have to add a qurik table or something

if that fixes it

[sage locust]

idk, honestly you could've said the same thing about that one macbook but then we worked around the issue

maybe we could do the same here

[analog berry]

not sure how u would work around this one

[sage locust]

maybe i should buy the machine off of techflash to test

[mild sentinel]

limine GPU modeset drivers

[analog berry]

it could be solved with a bios upgrade also

potentially

[fleet osprey]

I'll give it a try in a bit

[fleet osprey]

I'd be open to it but keep in mind that it's junk

battery holds a charge for like 1m

fan is dying

bottom case doesn't really stay on

also runs hot as hell and draws gobs of power

also the firmware is quirky at best, like it will only boot from USB Mass Storage on the rightmost ports, never the left ones

ironically for an HP, the hinges are somehow still good

[sage locust]

i said it half jokingly

[fleet osprey]

it also probably has other major issues that I haven't noticed; it spent like 5 years shoved in a dusty cramped computer cart with no airflow, constantly plugged in (this is certainly why the battery is hosed)

[fleet osprey]

fair enough

I was just warning you in case you were serious

it's basically worthless

[sage locust]

i am aware

[analog berry]

@fleet osprey is it running the lastest firmware version? if not maybe consider upgrading

[fleet osprey]

it is

[stone oracle]

In case of 4530s it's exactly Blt not working correctly and direct drawing working. But. Windows (both 7 and 10) surprisingly can draw their logos during the boot.

[stone oracle]

Thank you. 💜
How to do it "real quick". Unpack the archive contents (2 dirs) into the usb storage root dir. Insert the usb storage into the port you know fw is willing to boot from. Break into fw boot manager, pick "boot from file", locate your usb storage from the list, navigate to \efi\upptech\antload.efi, hit enter.

[stone oracle]

bahaha. this one is even hairier, than mine. well...
so:

• that bug I've talked about is here as well. that is GOP.Blt() is broken and doesn't draw where fb base is reported to. this is seen because there's no logo/string. I know how to overcome it, but the overcome is a quirk, unnecessary elsewhere.
• direct drawing into fb by the kernel results in the same porridge as with limine.

As seen, HP has improved their Probook line between 2011 - 2013. 😁 Thank you very much, Techflash.

it looks like the fw here reports some idiotically wrong PixelsPerScanLine.

[fleet osprey]

no problem, happy to test stuff

I will also note that Limine works fine on this machine in legacy BIOS mode

interestingly this machine has 3 modes:

• Legacy BIOS
• UEFI with CSM
• UEFI without CSM

not entirely sure what differs between the first 2

I'd imagine it doesn't contain completely seperate non-UEFI firmware just in case the user selects legacy BIOS mode

[stone oracle]

yes, in my experience with regard of boot options on this line, the fw is a hellish mix of bios and uefi. still haven't figured how to make the fw show the "successfully" created boot option.

[fleet osprey]

also interestingly it doesn't seem the pitch is actually that far off, only by a few pixels each line

I wonder if it reports 1360x768 but is really 1366x768, or vice versa

[stone oracle]

that "famous" 1366... sadly, I can't do any coding now. no electricity and new massive missile attack is approaching.

[fleet osprey]

fair enough, was just floating that out there :P

stay safe

[grand shadow]

ported an irc client

[viscid flame]

connect to osdev irc

[grand shadow]

I mean I could

it was connected to libera so I could just /join #osdev

my bouncer also works so

(I was talking in #lounge-0 thru my bouncer)

[mild sentinel]

testing this iso rn on bochs and real hardware

bochs fails with a DIV0

let me test it on real hardware tho

[gleaming bay]

on astral?

[mild sentinel]

yes

[gleaming bay]

wtf

does it work in qemu?

[mild sentinel]

Hey hey @gleaming bay It atleast gets further than Linux

[mild sentinel]

I’m sure it does

I’m about to test on real hardware

My kernel also DIV0s on bochs

And Linux can’t even boot

[gleaming bay]

bochs moment

[mild sentinel]

Who the fuck even tested this shit

[mild sentinel]

This one seems to be failing

I can control C it

But it just printed hardware address

And then did nothing else

[worthy mica]

Yes.

[viscid flame]

bochs is bloat yk

[grand shadow]

crazy

[grand shadow]

damn something still seems wrong with my dhcp then

[obsidian pivot]

dhcp is so annoying

wireshark tells me my dhcp packets are malformed then refuses to elaborate how

[grand shadow]

my dhcp packets are fine its just I am probrably missing something mentioned in a small footnote of some random rfc that makes the router completely drop my packets

[obsidian pivot]

average internet moment

maybe try accessing dhcp logs on your router?

oh wait it's evalyn

maybe a pcap would help?

idk

[grand shadow]

don't think my router has that sadlçy

[obsidian pivot]

kerneç

[grand shadow]

ççççççççç

[hearty fulcrum]

you can always ssh into it and look for them there :P

assuming it allows you to do so

[grand shadow]

trust me

I've tried

router manufacturers are the most antifun people in the world

🥀

[analog berry]

welcome back

[mild sentinel]

My router does let me send logs to my raspberry pi so I could test again with that enabled and maybe get some error logs or dropped packets from my desktop?

[grand shadow]

would be cool

[mild sentinel]

I do know it logs all the firewall shit in there so hopefully something from astral may come up

Do you mind sending an ISO with extra logging on netd too

[grand shadow]

sure I can add some prints in a sec

[grand shadow]

this has some logging that says what it did and is waiting on etc

[mild sentinel]

il get it worked on in 3 to 7 bussness days o7

[grand shadow]

I'll be staring out a window aura farming until then

[mild sentinel]

I hope you get that reference too

[charred sinew]

why don't you just start a dhcp server on your pc and connect both pc's directly or disable the routers dhcp

[obsidian pivot]

why dont you just add manual network configuration

[grand shadow]

I mean I very well could

[mild sentinel]

It’s not a very practical solution to fuckle with a router especially when other people couldn’t using it

[charred sinew]

one can connect two computers directly alternatively

[hardy vine]

3 to 5 buissness days?

[mild sentinel]

Sure

[hardy vine]

is it the wizard portal refrence

[mild sentinel]

Yes

[hardy vine]

although sadly we cant make fun of it anymore 😭

not even mountaintop

[mild sentinel]

I wonder if astral can be built up enough to be able to play on hypixel lmfao

[hardy vine]

does it have internet and minecraft

i saw 1.7.10 on the screenshot

just install via version lol

[gleaming bay]

doesn't watchdog flag via version tho :P

[hardy vine]

oh yea

but its 1.7

not 1.21 or something

[gleaming bay]

you could play minemen

or one of the 1.7.10 pvp servers

[hardy vine]

you cant exactly bypass anything by going to a lower verison

although if i can play minecraft in astral i feel like im just gonan delete linux

[mild sentinel]

They don’t flag Linux :p

and they support 1.8.9

[mild sentinel]

Yes

[gleaming bay]

???

they can't flag linux

[mild sentinel]

Then why could they flag astral

[gleaming bay]

via version is a mod that lets you play on servers on a newer version then your game

[mild sentinel]

Ahhhh

I mean astral is close enough to where 1.8.9 could be ported and possibly run pretty good especially if SMP gets better for LLVM pipe

Yes yes talk is cheap send patches

I may do it :p

[gleaming bay]

I would also talk is cheap send patches my self

but I have no idea how astrals build system works

nor how to build java for astral

and I don't really feel like spending that much time

[mild sentinel]

I think it would need a new version of LWJGL to be ported tho

[gleaming bay]

no?

[analog berry]

astral build system is really difficult

u do git clone && make

[gleaming bay]

oh

😭

[gleaming bay]

mojang only switched to lwjgl 3 in 1.13

[mild sentinel]

Do you mean you don’t have to do a ritual to the stars and sacrifice an old HP printer?

[gleaming bay]

oh

it dead ass it just "make"

🥀

[analog berry]

if you're lucky and jinx works on that day

[gleaming bay]

okay jinx is doing everything but work for me

🥀

[hardy vine]

basically it lets you bypass the anticheat easier

so hypixel wasnt happy if you use it

[hardy vine]

whats a LLVM pipe tho

[gleaming bay]

software opengl rendering

[hardy vine]

ah

basicaly just software rendering

[gleaming bay]

it does runtime code gen

[hardy vine]

for opengl

[gleaming bay]

via llvm

[hardy vine]

btw is it easy to port LLVM and rust

[gleaming bay]

no clue

[hardy vine]

okk

[viscid flame]

jit software rendering

[opal lotus]

llvm yes, Rust is a bit harder

[mild sentinel]

How is LLVM/Clang compared to somthing like GCC

[hearty fulcrum]

it's better in some things, worse in other things

[robust geyser]

if you can port gcc then you can port llvm

[grand shadow]

I don't see why 1.8.9 couldn't be run, seeing as 1.20 was run as well as gtnh?

like thats a solved problem :p

[mild sentinel]

If you can do 1.20 just do that lol

See if you can join hypixel

[grand shadow]

probably better to join on 1.8 because 1.20 was super laggy

I still have to profile this and make it better wrt multicore perf

[mild sentinel]

And you can waste your time playing hypixel sky block and flexing on people you made an entire operating system to play Minecraft on

[grand shadow]

(my big suspicion is some page in path that does a tlb shootdown and happens super often)

[mild sentinel]

True but there are a few sections of the game requiring modern versions

[grand shadow]

wontfix

[mild sentinel]

How bad is the SMP lag on real hardware?

It could be a KVM/QEMU emulation overhead quirk?

[grand shadow]

no idea

[mild sentinel]

Wana make an ISO that has x11 and Minecraft on it for me to test

[grand shadow]

I've never tested to compare between linux and astral trying to run the same mc version under the same conditions (kvm under wsl)

when I tried debian couldn't really run minecraft

so I guess astral > debian?

[marsh sparrow]

kvm overhead is generally quite low

[obsidian pivot]

Debian can run minecraft huh

[marsh sparrow]

so it's unlikely that real hw perf is much better

[mild sentinel]

Wouldn’t that depend on if you have the right stuff for virtualization of the APIC?

If its TLB showdowns or other IPI?

[grand shadow]

https://tenor.com/view/elmo-gif-21575334

[obsidian pivot]

skill issue

works on my machine

[grand shadow]

tbf something was probs super scuffed seeing as I was trying to do that out of the livecd 💀

[hardy vine]

WAIT WHAT GTNH

[hardy vine]

real

SKYBLOCK OS

(thats a actrual os not just a fork of debian)

[hardy vine]

bro wtf

can it also play steam games tho

[mild sentinel]

It can play factorio

[grand shadow]

whenever I have time to keep working on the wine port

and yeah that

though to play factorio it needed a different libc

[hardy vine]

btw is there any hobby desktop OS that managed to port any proper graphics card driver

[mild sentinel]

I hope to be able to get this good at osdev

[grand shadow]

managarm was working on that

[mild sentinel]

Managarm has partial nvidia GPU support

[hardy vine]

woah

tbh i feel like software rendering is probably enough

with the modern gpu

oh wait no nvm it have 4 fos

fps

it is on qemu hto

[grand shadow]

I have this screenshot saved

[hardy vine]

holyshit

[grand shadow]

I want to revisit the wine port at some point to work on my wine goal of running cogmind

which is a 32-bit binary so its more complex

[hardy vine]

you have only worked on this project for 3 years?

[grand shadow]

yeah p much

[hardy vine]

how long do you program per day tho

[grand shadow]

depends

some months I literally cram 10 hours of programming others I don't even touch code