#Astral
1 messages · Page 36 of 1
womanagarm
kilogram
managramme
electroencephalogram
alakazam
managarararam
midazolam
managerarm
I keep reading it as mammogram 
tbh this is one of the most impressive os projects just behind managarm
mintia has fallen so behind
it's over
it's so joever?
joewarida
I'm gonna be working on a generic posix psutils
since that is one of my goals for astral + zinnia is going to use it because sysctl and no /proc linux brainrot
W
my ragebait shitpost might reach my wine post I am cryin
Can it beat ShrekOS though
nothing beats shrekos
If I can get enough calc studied today I can probably get pgrep or pkill imppemented
Why do they make computer science students study calculus when our whole job is to ask ai to make reactslop smh
Actually no thats wrong
Our job is not to make reactslop its to be an uber driver
fr
Are U doing sysctl KERN_PROC for zinnia process stuff
lock in bro get me the triple integral of your mother
Should be pretty simple to add it to the psutils code

just run c2rust on astral to make zinnia
or mrustc to make astral from zinnia
Claude, rewrite astral in rust make no mistakes
Czapek did this once and claude just said I aint doing allat
How much was Claude pushed into doing it btw? I'm sure there is some way to make it try

It did do it for other large projects before
I think he didnt push it
This sounds like a multi-week task — I apologize for the trouble
Its just claude being bad
Codex wouldve been like "yes sir I will rewrite this into rust AND zig"
Mint isnt even here for me to be throwing anticlaude rage bait like this
🥀
Calculus is the backbone of AI
Nuh uh AI is just water + electricity + gpus
What kinda calculus are you doing
We had multivariable calc this semester and to finish it off we are doing series
Yeah it was cool
Yeah the website Ive been using to help me study has it as calc 2 while multivariable is calc 3
Yeah that's closer to what we did
Calc 1 was derivatives, calc 2 integrals and series, calc 3 multivariable calculus
Damn you have an entire calc class for derivatives?
We had just one that had derivatives, integrals and some analytical geometry stuff
Well, it was limits and derivatives I guess
And optimization problems using those
So kinda like an intro to calc
i had a single course that covered derivatives, integrals, infinite series, ordinary diff equations
during the first semester
Ah that makes more sense
Damn
I am thinking of taking the diffeq class but its an elective for cs but I know I dont commit to studying until like a few days before the exam so I dont want to suffer more
Ngl I am thinking of trying to learn ML during winter break next month just so I dont forget calc 3 stuff
the only thing i remember from the diff eq part is characteristic equations and corn
(corn - całka ogólna równania niejednorodnego - general solution of a nonhomogeneous equation)
Making a MLP from scratch that can train on MNIST is pretty easy and really educational
I'd try making an LLM from scratch if you didn't need a gazillion dollars to train one
Just buy a h200 5head 
make claude 2
Get vc money say you are making novel llms
Maybe a really dumb one can be trained on a consumer gpu
I had a friend who got like a few discord servers worth of messages
They trained an LLM with it on a few GPUs
It was dumb
Not even dumb
But like
Moronic even
Gonna download my message history over my two accounts and make mathewndLLM
Considering I have been on this god forsaken platform for over 10 years
well, you can train a shitty LLM for $100 as nanochat shows
shitty = GPT 2 level
On what does that run?
That's karpathy's thing right?
yeah
i think it can be trained for $<100 on AWS instances
i haven't tried it myself
That's still expensive if you mess something up 
anything aws is
amazon got to him
well ig you don't train on the full training set for small changes
you can see how the loss behaves etc even w/o doing that
we have an entire calc class for partial derivatives only
bro what the fuck
lmao
did anything happen at that point in time
I genuinely have no idea every other metric is fine its just clones that completely skyrocketed
Maybe some random ahh guy just had git clone on a loop

Trvth
@opal lotus soon
huge
I'm gonna be able to have my screensavers back
u added signals?
I have had signals for years
xscreensaver
Mat outdoing himself as always
maybe I also port a file manager/editor/whatever to finish the nscde look
maybe even port xdm or smth
nscde also works under non-root which is nice
idling also works for the screensavers
so yay astral now has screensaver support
lmao I got one telling me to install a perl package
☠️
was in ze fucken
startx just works now
maybe it was that one socket backlog bug
yeah it just works
time to migrate my fucked up scripts to just be a wrapper around startx xddddddd
Sir, another ai slop pr has hit the astral
tell tell tell me about it
teeellll me about it
tellll me all l abou it
tell me about it..............
Like WTF
How am I homophobic? I have many ai friends
mathew the fucking biych
cvnt
name one
Clanker is a robophobic slur
CLANKER
clanker
Can you guys stop, please, they have feelings like us...
no 
tbh i use ai all the time
but the useless ones like those yt spam bots those pmo so much
Says the literal clanker from star wars
Oh wait
It's jar jar
the fuck is this on the gh
its a clanker
ai
oh really i couldnt tell
im more confused on why its looking for vulns in prof.py 😭
because AI is dumb?

"if the download is compromised"
accelerate it by contributing patches
I am unsure if I am going to keep xfm as a 90s larp thing though
I think I will switch it for xfile
would you ever consider making your own window system and userspace
Math11 protocol
Probably not
It's pretty fun if you know what youre doing and can actually style it the way you want, but also makes all Xorg programs not work anymore
Though with tools like xbanan I guess that's becoming less of a problem
where network card
when ibm 8514/a accelerator driver 😔
etc etc
what about a driver for my plumbus?
everyone has a plumbus why would there not be a driver for the plumbus
@prime juniper
lmao the screensaver starts automatically now
also wtf why did I type neofetch there and not fastfetch
brainrot
🍌
❌🍌
Banger
did u stole wallpaper from budgie
real gaming
It came with nscde
I should figure out why unity games don't work on astral
and also finish ricing nscde
and port xfile instead of xfm
on wine?
Yeah
They seem to get stuck on mono bullshit
Or well, did
I havent tested in months and have done a lot of bugfixing since then
Does c# work?
No idea
Maybe start with that lol
Probably lmao
It would be fun to play unity games in astral but any unity game would probably run at like seconds per frame if the performance of non unity is anything to compare to
☠️
I dont remember if I ever profiled wine
you could try to get nvidia hw accel working
Yeah tbh that probably relies more on hw accel than older games
Thatd be a fun project to look into but I would need to implement drm and I know exactly 0 about drm atm
you also knew nothing about usb and sound stacks yet u did both easily
it's the spirit of osdev
Calc exam done, astraling shall resume soon with a brief pause for another exam
Ah fuck I read that as astral shell and was hoping for a kernel shell impl
All code will now be run in ring 0 thank you very much
Does astral have a built in kernel debigger
without drm?
Wdym didnt Terry Davis do it
And he was smart
And wrote the greatest os
Nah
can i work on astral and implement every planned feature for my kernel in astral kernel instead
PRs always open 
Astral on AMD Ryzen 7530U laptop
-# had to enable autologin in getty due to a weird bug in login
adding drm support is comparatively easy
yo merge me 😔
soon™ I am out rn
DISASTER!!!
he really hates two things:
- running as root
- windows
I can't believe I am going to say this but in order to make astral look even more retroey I am going to need to...... port qt6??
thankfully at least according to blfs I should already have most of the bullshbit needed
at that rate port full kde plasma and apply a windows 7 theme or smth 
to not vaguepost: the nscde theme generation python script depends on pyqt4/pyqt5 being installed, but I am pretty sure that I can get away with pyqt6
and the theme generation is what makes gtk3 apps/etc look retroey
ah

I wish bro I wish
when I got all the dependencies for webkit ported I genuinely had a headache from how mind numbing it was
found a nscde cross compilation bug but the upstream repo seems to be dead for a few years
funny
it has to install 94 times 😔
what is that
qt6
I kinda doubt that it'd take 47gb tbh considering that I can build qt6 fine using 48gb of ram + tmpfs for the build dir (with no swap), looks like the build dir for qtbase on my gentoo is 244mb
its estimating with all optionals included, qt6 itself is only 502mb
oh yeah
i did a du on my repo dir
152GB
💀
trillions must pyqt6
I love porting software but I would never ever ever do it myself
why
tonight in: python packaging is kanker
nscde gtk pixmap generation for themeing depends on pyqt4/5 apis for some reason
but qt4 and qt5 are aids to port
so I will liekly have to patch it to use qt6
python bindings take longer to compile than the fucking library itself award
oh my GOD
no
TONIGHT WE LARP IN HELL
gtk3 theme generation is now working
at last
the xscreensaver thing finally looks right too
its just a giant gtk app 😭
Might port a dm later
For keeping up the larp it will either be xdm or something that I can use to larp xdm 
😭 @acoustic peak
though tbf I am likely much much farteher away from the server than u are
idk where u are from
what TCP stack is it
my own
it's not slow
if you're trying to test your TCP stack performance you should probably only be doing LAN stuff
If it makes you guys feel better my E1000 process is using 300% CPU time since I fucked itu p
how does it use 300% cpu time
how many threads does it have
5 cores
yeah the server is located in Washington DC so wouldn't be surprised about conn issues
might want to check this bit
https://github.com/Mathewnd/Astral/blob/rewrite/kernel-src/io/net/tcp.c#L1760-L1768
looks fine to me
Oh whoops
you check that tcbsocket->tcb is null and if it is you goto cleanup, but immediately after cleanup label you deref the null
Also lmao I am on another continent
xdm might be too limited but dtlogin brings in the fill cde stuff
I have no mouth and I must scream
@opal lotus try this one, its closer https://fsn1-speed.hetzner.com/
1 sec
or the de ones
whats host speed
100Mbit
ok not bad
let me curl from my lan
yeah ethereal lags behind my host, it averages 8-10Mbps while my host hits around 60 Mbps
ah I confused it with the Finland one
lol
this is pulling with a roundtrip from my other pc over my router back into my pc on linux
what was the issue
server that's actually close to me + i forgot to ACK and reopen the recv window for the peer
small detail
oh yeah no wonder your speeds were so slow

maybe add window size acks if you want, that boosted a ton of speed for me
yeah
it just works
You gonna get an old UNIX server and make a sleeper server out of it?
And use astral on it?
Port something like nginx and run some infra off it and shit
i need this
If I had cash 😭
Im not marvin I dont get paid 10 tirllion euros a month to buy computers
If you ever get money and find a cheap broken one
Why not ask Marvin to buy it for you

Marvin already spent it all
i got a pentium 4 machine for 7 euros at my local flea market
He gets his paycheck and in 2 days its all in ebay orders
and it looks close enough to a server
Hey Siri set a reminder for 2 days from now
oh wait but astral is x86-64 
That's why I said make a sleeper build
And slap a mini PC inside of a Unix workstation
oh yea right
holy nice dude, i still didnt make the sys_clone syscall for bash 
nah
its time to take the metaphorical L and add an xorg-server os-support astral implementation
so I don't have to use hacky shell wrapper scripts around X anymore
mostly for xf86OpenConsole and xf86CloseConsole
sysctl me this
okay this should satisfy having permission checks while allowing non root xorg
yeah that works
I guess all I gotta do is now change the behaviour of the console while its locked
also funny webkitgtk works rootless neat
What do you think you will do with astral once you have it fully stable and other shit?
I know exactly 0 about writing gpu drivers
closest I've done is some opengl programming
☠️
Well if you want 3d accel on real hardware that would probably be the best bet
I think somebody mentioned wanting to do it for RDNA2
I would probably try nvidia-open first since both my laptop and desktop have nvidia gpus
and lil if that ever gets support for that

I think there were plans for that on lil
For like more flexible backend support or whatever
though nvidia-open seems to still need closed userland stuff
as it seems like, at least according to leo, the open userland stuff is completely broken with it?
😭
If you do ever want to experience the hell of amdgpu do let me know
And if you have some ROM dumps of AMD cards they could be of use
I've never owned an amd gpu
its always been either integrated only or an nvidia gpu
my first desktop when I was a kid had a nvidia 9600 gt or whatever it was
year of the astral desktop (real) ((its slow because I am running a gtnh server in the background 💀))
astral 4 years ago
in function dup library function fails to missing sysdep 💀
tbf, bentobox 4 years ago:
(wasn't a thing)
why wouldn't it
drm
seeing rtdl instead of rtld is so weird to me
feels like a typo i would make but nope
I think we're renamed it to rtld by now to match convention
Deadnaming rtld
damn bash runs without dup?
bash is a really simple port
like, "can be speedrunned in a week" level
yeah it is i didnt realize it could function so limited tho
xterm skips the first line of /etc/shells award
stop using ts
the larp must go on
suckless? more like sucksmore
port urxvt
much better than xterm
also dont forget about note in xterm src
sucking terminal
Shitless Terminal.
been a pretty productive day of distro work
my pr is rotting mr astral
I'll get to it soon 😭
ok so the path to building astral in astral from a fresh git clone is to just
source the astral source-recipe
prepare
git submodule update --init
gg ez
and then fuck off to build it somewhere
I probably should add a native build command
compiling the kernel in astral:
with 6 cores, -j7 takes 51 seconds
with 6 cores, -j1 takes 1m29s
with 1 core, -j1 takes 1m12s
compared to linux, which takes 8 seconds for -j6 and 31 seconds on -j1
💀 astral suuucks bro
profile profile profile sahur
haha this is quite real
i get similar jumps in compile time in my kernel too iirc
is that host linux or linux in vm?
wsl linux, I am running astral under kvm in wsl
what did it reveal?
I wonder how much paravirt could help? And I wonder how much slower it is on bare metal
wouldnt help that much
Honestly I have no idea, single core is probably just missing optimizations and slow i/o
Multicore is probably some bullshit like thrashing a cache or something I really dont know
was this in raw terminal (ie no other stressful procs running) and what happens if you lower niceness on the compiler
when Astral NUMA 😔
Under nscde
i mean how long does one gcc take times wise
if you have that system call
lets see
600-800ms usually, which is quite a lot
not great yeah
how many page faults are you getting per run
VMEXIT my behated
so like 100ms?
maybe you should consider tracking the amount of PFs per process
for smaller allocations if the system isnt under memory pressure just back them immediately instead of demand paging them in
I plan to eventually cut down on the amount of page faults by clustering
but that will come when I rewrite my page cache to support clustering

-j7
6 cores
6 cores -j1 💀
you can just throw a if (size < PAGE_SIZE*4) { back_pages() } in vmm_map()
at least for a bit of pressure relief
ehhh
try disabling your e cores in linux maybe
oh wait right
your host is windows
do you think I am rich enough to afford a cpu with e and p cores
b-but-
that was someone else running it
ah
its running on an i5-10300h bro
single core j1 is not that much different from multicore j1
just like 20 seconds diff
so there is still some overhead but not that much
Better optimization is to demand page multiple pages
that is also easily possible
Shootdowns?
yeah I plan on doing that soon
thats like almost entirely optimized for unless I completely redo half my memory subsystem do do async shootdowns
try j32 the more threads the better trust
mfw scaling is ~1/3 off from linear (there is still more optimization to be done) 
smh
anyways still cool shit
I should try to do a multicore gcc build at some point
do you only do shootdowns on the cores the process ran on?
bool do_shootdown = // do shootdown if
current_thread() // scheduler is up
&& arch_smp_cpusawake >= 2 // and there are multiple cpus in the system
&& (page >= KERNELSPACE_START // and its either in the kernel
|| ((page == NULL || user_shootdown) // or in userspace...
&& thread->proc && thread->proc->runningthreadcount > 1)); // in a process which has multiple threads running
// .........
for (int i = 0; i < arch_smp_cpusawake; ++i) {
if (smp_cpus[i] == current_cpu() || (user_shootdown && smp_cpus[i]->vmmctx != current_thread()->vmmctx))
continue
this is what I do rn
uhh what
so you do do it on all cores?
you do know you have to do shootdowns if it's still only one thread right?
no, only if its actually running on the cpu
why?
that doesnt work
because the thread can run on a bunch of cpus
what other thread will be doing the shootdown on the vm space its not running on?
why
if its not running on the cpu wouldn't it already have done a context switch
the point of TLB invalidation is to prevent stale TLB entries, if a thread ran on another CPU but is not running anymore, then gets scheduled back in its tlb entries will still exist
fuck im an idiot i think
reloading cr3 flushes the tlb
Dosnt changing an address space flush the TLB?
yes

I presume it's like that on most architectures anyways
good thing I don't implement aids or whatever that is
Yeah you'd have to flush the ASID but that could be done lazily when you switch to it
i had this in mind because of asids mb
actual 🧠 💨
I purposefully do some crazy racy shit here
yeah the vmmctx check is iffy but that's best effort anyway
yeah
if it transitions to the space during it it will have the new entry anyways, and if it transitions out of it it does not matter
the current_thread() is racy too
ah it's to check if scheduler is up
weird
whgat do you mean its not like the thread is gonna vanish
sched isnt up at start and TLB invalidations have to happen before sched
on multicore
you dont use the value so it's fine
if you did it would be racy
because preemption is enabled
no
are you confusing it with current_cpu
holy shit wow i am making so many mistakes stop listening to me
current_thread is NULL at ap initialization and you do an invalidate on the ranges for kVMM alloc/dealloc
😭
lmao it happens
42 seconds 6 core🥀
uhh what
no
at least he does iirc
no like
i meant during vmm_map and vmm_unmap
you invalidate ranges
and before the APs have had their scheduling initted they are allocating memory for other bs soo
yeah but you dont do vmm_unmap there
i shouldve clarified
I don't actually even hit a shootdown on this anyways anymore
also if you're cool you start the scheduler and setup current thread as teh first thing
like I could remove this check and be fine but whatever
in my kernel APs' current threads are set up before they start
that's not what i meant but wtv
so true
I really want to be fully self hosting already
so probably work towards that
I need to finish page locking for dma so that I can dma directly into user pages for O_DIRECT
and then rewrite some part of my filesystem drivers to properly handle multi sector reads/writes
does it run on anything !amd64 yet
and then rewrite the page cache, and then implement page clustering
not really
I don't really have any non x86 hardware to test it on anwyays
port to 32 bit when
this page locking will also be used for fork correctness

probably never 
Get an m2 Mac mini and port to that like Marvin
They can be found pretty cheap relatively if you did want to ever do that
the lack of space to put HHDM makes i386 hell anyways
you can limit the physical memory size or design around it
ye i have arch_mmu_remap_physical and arch_mmu_unmap_physical but it still sounds annoying as fuck to implement some pool system
send builds, I have everything
his kernel only supports x86_64
cool
I am not funded by the mossad sorry
not having x86 hardware should not be a blocker
pool?
what if i got you CCP funding for loongarch64
genuinely would
the method i would do is take a chunk out of my memory map and make a small HHDM pool
what NT does for temporary mappings is you have 64 pages of VA per CPU, and you cycle through each page until the 64th is hit, then you do a whole (local) tlb flush on that
wdym tlb flush on that what does that do
like if you need more than 64 pages of mapped on a CPU then what
you dont
mobile ARM (32-bit), embedded ARM (32-bit), game console ARM (32-bit), mobile ARM (64-bit), embedded ARM (64-bit), game console ARM (64-bit), laptop ARM (64-bit), embedded PowerPC (32-bit), game console PowerPC (32-bit), laptop PowerPC (32-bit), desktop PowerPC (32-bit), game console PowerPC (64-bit), desktop PowerPC (64-bit), embedded MIPS (64-bit), laptop x86 (32-bit), desktop x86 (32-bit), desktop x86 (64-bit)
i see
temporary mappings are page-sized only
I have all of these classes of architectures
you have 64 to avoid a tlb flush every time you unmap it
@mystic plank
holy shit its loongarch64
no sparc64 🥀 how shall I unix larp
I will test on any of these shits for days on end, for free, just need shit to test on them 
(leo will veto any sparc64 mlibc ports lmao)
just have it as an out of tree fork

no DEC Alpha? how will you larp NT?
no loongarch64

itanium better
that's fake larp
real NT larp is MIPS
on the Nintendo® Wii™ video game console running eNTii-for-workcubes

or the Xbox 360 which just runs NT natively 
PPC 👎
PPC page tables 💀
I just use the BATs lmao
I should port whatever the fucking perl module is for that one screensaver that sometimes just jumpscares me with a "acme::whatever::whavetever not found. your perl is broken whatever" white screen
🦇
best architecture is armv5 minus the mmu
from what I've seen the ARMv5 MMU is alright
best architecture is that one fucking risc machine we studied at uni that had 256 bytes of ram, and only {add, load, store, and, or, not, halt, jmp, jz} or whatever the instruction set was
well its certainly reduced
looked sane enough when I read how MINI (FOSS replacement for BroadOn's "IOS" on the Wii's "Starlet" / "IOP" (ARM926EJ-S)) handles it
i mean it works its just annoying to remember all of ARM's stupid bits
starlet is peak
my favorite console security mechanism
tomorrow I think I will pick back up on working on bettering the vmm
starting by splitting up the file
ni hao fine shyt
its not Great
you have 4 sets of permissions bits for "subpages", which naturally means you get less control on permissions
for example, akin to the original 386, you can't prevent the kernel from writing to any mapped pages
on armv6 there is an extra APX bit which allows you to specify that kind of permission
it shares the same flaw as armv6 with regards to the unbalanced page tables as well. the top-level page table is 16 kilobytes instead of 4 kilobytes, and page tables themselves are 1 kilobyte long
oh, and on armv5 you also have less control over the cache policy of a page
you only have the C and B bits, whereas on armv6 you have 3 extra "TEX" bits
then there is armv7 which added a different page table layout which is better, using 4kb pages for both levels
I am partaking in the school of will and completely removing the name "pmm" from my kernel
2 jackal pics a month
and by that I now mean doing a general style cleanup on the entire mm 💀
I really regret not writing astral with taking page faults in the kernel in mind but I can't really do that without rewriting like half the memory manager and redoing the entire design
💔
Why do you want to take page faults? Paged kernel heap?
more straightforward cached file read + optimizations for filesystems like fat
paged kernel heap when
I kind of want to implement swap at some point but I have never done it nor barely read any theory up on it
armv8 my beloved
mhm?
don't do paged kernel heap
atp
it's the kinda stuff you have to design from the beginning
and it's not really useful
yeah plus I don't really think I will target memory constrained architectures
I was talking more about like user side swap
yeah that's worth it 
i actually write bad implementations of everything so i can later justify full rewrites
zinnia rewrite when
c23 microkernel
everyone
trust
i use c23 but my kernel is not a microkernel 
nah but fr it all started with brutal
who is brutal
c23 microkernel
by who
dead project smh
brutal, nothing is left for the beta
idk i want to write a microkernel but in a much smaller scope
like, just being able to boot off of an nvme maybe and go to bash
thor compatible c23 microkernel
lol
it's not late to be the first os that uses bragi and is not managarm
i want to come up with a cool ipc algorithm
to prove my brain hasn't rotted away from clanker
well, I stole it from existing microkernels mostly 
in this world nothing is created, all is stolen and sold for more
something something newton standing on the shoulders of giants
but yeah
it's probably worthwhile to design it with readahead and journaling in mind
journaling is one of the things I have been itching to add to astral
can't stand filesystem corruption anymore
Why would it be different ?
extra handling that smp might need
Oh okay
windows cursor
locks
Make sense
yeah I use a real os (windows)
imagine not using the latest and greatest kde plasma 6.7.2
I should experiment with adding clustered anonymous page in for the lols
I do remember that in my old profiling for graphical programs like meincraft the page fault isr had a lot of hits
though also there kernel time is like less than 25% of overall time 
Time to try to build gcc in astral again but with 6 cores
go go gadget make all-gcc all-target-libgcc all-target-libstdc++-v3 -j7
It survived running 7 configures at the same time in smp, cool lmao
hm, took like 3h10m to build it
crazy that it didn't crash
there is still a lot of work to be done
I think last time it took 5 hours, so thats still an improvement over one job 
- xscreensaver running
literally unusabl
no getrusage
are you implementing journaling?
I was looking into how it works to see how I would fit it into my new page cache design
how do you plan on handling journaling on the managarm page cache, btw
damn
this is pretty cool
https://github.com/Mathewnd/Astral/issues/34
I'll fix the stuff they brought up
this is really nice and why open source is good
(ignoring the AI and bullshit issues)
2026 is actually the year of the astral desktop :p
it doesn't sound that hard
@carmine swallow wants to do it for nlnet (?)
been working on that unix socket stuff
so far I have managed to get to 0-2 failures on junixsocket-common (apparently there is some race in my localsocket code that causes it to randomly hang?)
FAIL junixsocket-common 144/181 (32 skipped) failures: 2/149
PASS junixsocket-tipc 0/89 (89 skipped)
PASS junixsocket-vsock 1/76 (75 skipped)
SKIP junixsocket-rmi (skipped by user request)
FAIL junixsocket-ssl 471/472 failures: 1/472
PASS junixsocket-darwin 0/2 (2 skipped)
these 2 common failures are on some throughput test
the ssl one also times out for some reason, maybe its the same issue. who knows
rotting away mr mathewnd, rotting away...
You need to say "talk is cheap, merge patches" 😔
"talk is cheap, patches already sent"

anyways, bored, trying to build Astral from my M3 Pro MBP
can't build from macOS (lame), but I'm trying from my Debian aarch64 VM
😭 Im too lazy to think about refcounting Ill get it to you today fr fr fr on god
I'm also trying to do this on a mobile hotspot (I'm out in a DMV in the middle-of-fucking-nowhere with a network speed of No™)
checking for gcc... gcc
checking whether the C compiler works... no
configure: error: in `/build_dir/host-builds/xbinutils':
configure: error: C compiler cannot create executables
See `config.log' for more details
make[1]: *** [Makefile:77: kernel] Error 77
make[1]: Leaving directory '/home/techflash/src/Astral'
make: *** [Makefile:38: all] Error 2

configure:4532: checking whether the C compiler works
configure:4554: gcc -O2 -march=x86-64 -pipe -mtune=generic conftest.c >&5
gcc: error: unknown value 'x86-64' for '-march'
gcc: note: valid arguments are: armv8-a armv8.1-a armv8.2-a armv8.3-a armv8.4-a armv8.5-a armv8.6-a armv8.7-a armv8.8-a armv8.9-a armv8-r armv9-a armv9.1-a armv9.2-a armv9.3-a armv9.4-a armv9.5-a native
cc1: error: unknown value 'x86-64' for '-march'
cc1: note: valid arguments are: armv8-a armv8.1-a armv8.2-a armv8.3-a armv8.4-a armv8.5-a armv8.6-a armv8.7-a armv8.8-a armv8.9-a armv8-r armv9-a armv9.1-a armv9.2-a armv9.3-a armv9.4-a armv9.5-a native
configure:4558: $? = 1
configure:4596: result: no
configure: failed program was:
| /* confdefs.h */
| #define PACKAGE_NAME ""
| #define PACKAGE_TARNAME ""
| #define PACKAGE_VERSION ""
| #define PACKAGE_STRING ""
| #define PACKAGE_BUGREPORT ""
| #define PACKAGE_URL ""
| /* end confdefs.h. */
|
| int
| main ()
| {
|
| ;
| return 0;
| }
configure:4601: error: in `/build_dir/host-builds/xbinutils':
configure:4603: error: C compiler cannot create executables
See `config.log' for more details

hardcoded x86_64 host assumptions
cringe
I wonder if I can fix it
yeah I see the problem
Jinxfile sets CFLAGS to have -march=x86-64 globally
this also affects host recipes
currently testing a fix for that
the real fix is probably to separate host vs target cflags but I don't know how to do that properly
gets past xbinutils now at least
🥀 I don't think that's right
what's wrong with that? doesn't astral target only x86_64?
see the Host lines
yeah what's wrong with that
yeah, so what's wrong
Heres a nickel kid.... Go buy yourself a real computer 
mlibc-headers only
seems to be getting pretty far actually
which should be pretty much just be copying a bunch of headers 
The fuck
the real challenge will be shit that doesn't separate build time tooling targets vs host targets (e.g. a lot of cmake projects)
and g-o-i 
Inb4 changed upstream
Either that or freestnd c hdrs is gone

Or I accifentally used the master link
😭
Well I did not
The sha of the commit magically changed
Someone is doing something spooky
inb4 it's always been broken and you just happened to save the sha512sum of a corrupted download

oh yeah that is true
I'll just change it locally for now to get it building but I won't send that up
lol freestnd-cxx-hdrs has the same problem

/usr/bin/install -c -s tic /build_dir/pkgs/ncurses/usr/bin/`echo tic| sed 's/$//'|sed 's,x,x,'|sed 's/$//'`
strip: Unable to recognise the architecture of the input file `/build_dir/pkgs/ncurses/usr/bin/tic'
install: strip process terminated abnormally
make[1]: *** [Makefile:219: install.progs] Error 1
make[1]: Leaving directory '/build_dir/builds/ncurses/progs'
make: *** [Makefile:134: install] Error 2
make[1]: *** [Makefile:83: distro-minimal] Error 2
make[1]: Leaving directory '/home/techflash/src/Astral'
make: *** [Makefile:39: all] Error 2
🥀
💀
merging patches is cheap, talk
it looks fine, I'm going to test it locally but pls fix conflcit :plead:
or I can also just commit it under your name
whatever you prefer
anyways
Selftest results:
DONE junixsocket-common 146/181 (32 skipped)
PASS junixsocket-tipc 0/89 (89 skipped)
PASS junixsocket-vsock 1/76 (75 skipped)
PASS junixsocket-rmi 6/12 (6 skipped)
PASS junixsocket-ssl 472/472
PASS junixsocket-darwin 0/2 (2 skipped)
IMPORTANT: Could not set socket timeout via Socket.setSoTimeout (AF_UNIX); this is a known limitation of your operating system; junixsocket-common: SoTimeoutTest.testSocketTimeoutExceptionWrite()
IMPORTANT: "With issues": Please carefully check the output above; the software may not be able to do what you want.
Supported capabilities: [CAPABILITY_PEER_CREDENTIALS, CAPABILITY_NATIVE_SOCKETPAIR, CAPABILITY_FD_AS_REDIRECT, CAPABILITY_UNIX_DOMAIN, CAPABILITY_UNSAFE, CAPABILITY_LARGE_PORTS]
Unsupported capabilities: [CAPABILITY_ANCILLARY_MESSAGES, CAPABILITY_FILE_DESCRIPTORS, CAPABILITY_ABSTRACT_NAMESPACE, CAPABILITY_UNIX_DATAGRAMS, CAPABILITY_TIPC, CAPABILITY_VSOCK, CAPABILITY_VSOCK_DGRAM, CAPABILITY_ZERO_LENGTH_SEND, CAPABILITY_DARWIN, CAPABILITY_SEND_RECV_TIMEOUT]
Selftest PASSED WITH ISSUES
You can do it under my name i dont mind
alright
patches have been created
if you open a pr I probably will merge it, since this is a good to have when I eventually port it to another architecture

well this doesn't actually touch any of the actual kernel source, it doesn't get it even close to compiling targetting aarch64
doesn't even add infra for doing so
just fixes various build-time bugs that only surface when host != target
lol it's fucked but I can't see it because my Linux VM has too small of a display
😭
wait it tried to boot from NVMe
I'm stupid

when aarch64
#1061407633745125397 message
I have all of the hardware 
talk is cheap, send builds
https://github.com/Mathewnd/Astral/pull/35 patches sent ✅
this does not touch freestnd-{c,cxx}-hdrs
since how you want to actually fix that is up to you
@acoustic peak @fleet osprey talk was cheap patches were merged
talk however isnt
was bash 5.3 pr merged
old firefox logo jumpscare
Wait
oh yeah speaking of, firefox when 
What is you cantact the client support
And ask them how do you run youtube
wine iexplore runs gecko so close enough
And you just tell them you cannot update your browser
"heres a nickel kid, go buy yourself a real computer" and google sends you 1k usd
Only one grand in this economy?
can't share too much with the poors
augh time to download steam
hopefully my localsocket fixes magically fixed steam
its fun that htis actually works
no native port of Steam 
linux larp when
I aint marvin
you know what it is time to do
port glibc
an android phone 
Get Marvin to share his 10k dollars aarch64 Thinkpad
get marvin to do an osdev fund
get marvin
1k*
Wait really
well
What are the specs
other than that it's a T14s
well besides the fact it fucking sucks for linux
tho idk if it's better now
thinkpads with touchscreen are weird as fuck
also thinkpad T series displays are ass
1080p type shit
How many cores
12 iirc
How is it so cheap wtf
1080p is fine imo
What is b stock?
back stock
no
oh?
Refurbished?
not really
back-of-the-alley-found stock
more like A-stock and B-stock
oh lesser quality then
Still cheap af
it's an X1E78100
Especially for Qualcomm
i really wish this firmware wasn't hopelessly fucked
Btw how's that FRED laptop coming along
otherwise it's a really nice laptop
ETA mid july
Thank you
Which part
they haven't asked me to pay my pre order yet
bro everything
you needed grub workarounds for the 64gb ram version where it would just hard reset if you didn't cutmem the upper 32gigs
Grub supports aarch64??
bro grub supports s390x
Damn
the software dev experience for this laptop feels more like an SBC with display
Why does that happen?
firmware bug
grub is such a shitty name i just realized
also sometimes it refuses to exit bootservices
Doesn't it make sense?





