*schedule
#Nyaux
edgy pilot
flat nymph
which is 11 minutes long
surreal path
nvm
its not working??
seems like i need something for virtual
for it to actually work
it isnt actually doing any of the calls for some reason
nvm?
seems to be ub caused by not making write virtual
oopsie
edgy pilot
seems like i need something for virtual
base class has virtual
child has method
surreal path
yep
surreal path
ret?
edgy pilot
your return type bruh
edgy pilot
that doesn't really do anything, it'll just throw an error if you're having skill issues during typing
just some good practice
I remember it helping me once
flat nymph
mark them with ``override`` in the derived class
it's implicit on the virtual functions (?)
edgy pilot
it is
flat nymph
but that way the compiler is allowed to complain
edgy pilot
but if you mark one, you need to mark all of them
flat nymph
Bruh, I didn't read the message above
edgy pilot
and there needs to be a virtual method with the same type in the base class
flat nymph
I need to go to sleep
But clang complains if you don't have override
At least with my flags
edgy pilot
what flags would those be
flat nymph
I think GCC is more lenient?
cinder plinth
Don't think concepts can be used for thst
How I do it is just using a bunch of templates
edgy pilot
Don't think concepts can be used for thst
for what?
flat nymph
-Wall -Wextra
edgy pilot
clang doesn't complain with that
flat nymph
Don't think concepts can be used for thst
It can help with type deduction?
flat nymph
Maybe
edgy pilot
you must have a method marked as override somewhere else
edgy pilot
why not?
flat nymph
I'm doing this for my IPC
edgy pilot
those memes must be true though
edgy pilot
*potential starship launch
surreal path
flat nymph
I'm doing this for my IPC
Now that I think of it, why the optional parameter is not std::option?
cinder plinth
But I don't think I really use them for formatting specifically, I abuse them in other parts tho
flat nymph
Now that I think of it, why the optional parameter is not std::option?
Oh, it was because you can't have references in that
C++ stl decided to not have specification for that
surreal path
@tender gorge 
flat nymph
wow a language is actually good
I think several people on this server were saying that for a long time
surreal path
love you too /p
flat nymph
I think several people on this server were saying that for a long time
But Rust blah blah blah
surreal path
I think several people on this server were saying that for a long time
im actually liking c++ a lot tho
seriously
cinder plinth
I really like C++ but there's definitely a lot of stuff to know about
flat nymph
Yeah, every day you learn something
surreal path
i like how easy everything went with classes
cinder plinth
There are lengthy wikipedia pages dedicated to very small aspects of the language
surreal path
tho there is a lot to learn
cinder plinth
Like SFINAE
flat nymph
Like SFINAE
SFINAE is easy
flat nymph
CRTP is cool
cinder plinth
Also I still don't really know what && is and what std::forward does
I just use it 
surreal path
no idea what yall talking about but oki
flat nymph
CRTP is cool
And they've made it much better in C++23? (or 26)
cinder plinth
CRTP is cool
Yeah that's a good example
flat nymph
Also I still don't really know what && is and what std::forward does
it moves objects
cinder plinth
No that's std::move!
edgy pilot
I just use it <:meme:575445094589661215>
me too lmao
I get interested, I read about it, learn, forget a few days later
cinder plinth
Also don't forget std::launder
edgy pilot
that's for telling the compiler to not assume some sane assumptions about modification of data
like modifying a const variable
daring juniper
Clearly I do not know that much C++ lol, I'm barely understanding any of this
edgy pilot
Clearly I do not know that much C++ lol, I'm barely understanding any of this
there's always std::get_money
flat nymph
inline is not inline
cinder plinth
That one I learned like last year
edgy pilot
Like `mutable`
for modifying member variables from a const method and allowing the modification of captured references from a lambda
flat nymph
yeah
edgy pilot
well you could
flat nymph
like I would have a class, where I have a container and a spinlock
edgy pilot
it minimises the need for putting mutable everywhere
flat nymph
like I would have a class, where I have a container and a spinlock
And I would make locks mutable for const getters
which find things in containers
edgy pilot
why would you need a lock in a const method though?
cinder plinth
for modifying member variables from a const method and allowing the modification...
https://github.com/nyx-os/atlas/blob/master/include%2Fatlas%2Fiter.hpp look at this code
This is probably the best piece of C++ madness I've written
edgy pilot
std::move(*this)
flat nymph
It's not that bad
cinder plinth
Yes
flat nymph
std::move(*this)
what's wrong about it?
cinder plinth
It's not that bad
Yeah that's why I'm saying it's my best one
edgy pilot
what's wrong about it? <:trl:829316641191886888>
it's strange
flat nymph
edgy pilot
<https://github.com/nyx-os/atlas/blob/master/include%2Fatlas%2Fiter.hpp> look at...
is this from that rust+cpp code you posted a bit ago?
flat nymph
some of my recent code
cinder plinth
Yea it has a pretty rust-like interface
flat nymph
I'm waiting for pattern matching to be accepted into C++
edgy pilot
why not just call the operator=
flat nymph
This should maybe be std::forward now that I think of it?
edgy pilot
{ operator=(std::move(r)); }
flat nymph
why not just call the operator=
what's the difference?
edgy pilot
what it does is more clear
cinder plinth
This should maybe be std::forward now that I think of it?
No I think this is right
std::forward would be used for passing it
flat nymph
but the operator takes a &&?
cinder plinth
Move is clearer anyway since you're saying "this won't be used anymore"
edgy pilot
that's why I though move(*this) was strange lol
cinder plinth
Nah look at the remove_reference
cinder plinth
that's why I though move(*this) was strange lol
*this is a copy
edgy pilot
is it?
cinder plinth
Well yea you're moving from the copy
edgy pilot
Nah look at the remove_reference
wouldn't T still be Right&&
edgy pilot
Well yea you're moving from the copy
*this is a reference
cinder plinth
Yea but the remove ref would bring it to Right
edgy pilot
return type doesn't have remove_reference
cinder plinth
So it'd be back to Right&&
cinder plinth
Yes
edgy pilot
T&& = Right&&&&
cinder plinth
std::forward would return right&&
edgy pilot
T&& = Right&&&&
^?
flat nymph
T&& = Right&&&&
jumpscare
edgy pilot
lol
cinder plinth
*this is a reference
Also yea you're right
I forgot why I even move(*this) but that isn't an uncommon pattern afaik
cinder plinth
Maybe both are right in this case?
edgy pilot
hmm well I believe both would do the same thing
cinder plinth
Average c++
edgy pilot
lol
surreal path
5 minutes launch
flat nymph
they're holding it (?)
edgy pilot
yes
thorn bramble
may i introduce you to managarm/frigg
supple robin
wow a language is actually good
be fr
flat nymph
The solution to all C++ problems is just reimplementing the world 
thorn bramble
the biggest cancer of c++ are the stl streams
simple solution, don't use them
containers are fine, except for lack of hash of pair/tuple
flat nymph
the containers are not fine because they throw
thorn bramble
they are okay for userspace programming
like if you're writing a regular ass userspace program then they are more than okay
flat nymph
I agree
But you can't use them in kernel
(I mean I had exceptions in my kernel)
(but still)
flat nymph
simple solution, don't use them
Also there's std::print now
surreal path
rip booster
flat nymph
Rip ship
thorn bramble
Also there's std::print now
i am well aware
i was more so talking about using them for file io
surreal path
crazy
i was working on fbdev but uhhh now ive been up for 24 hours
and i need sleep
yet i cannot sleep weirdly?
edgy pilot
may i introduce you to managarm/frigg
nah, this follows the standard
my coding style was a bit "ew" when I wrote it though
cinder plinth
but frigg.. π₯Ί
atlas better β’οΈ
surreal path
what are we talking about
maiden ridge
what are we talking about
frigid cliff
Felt
surreal path
oh god what is that
ebon needle
i was working on fbdev but uhhh now ive been up for 24 hours
isnt fbdev is simple
surreal path
no not the info structs
supple robin
?
surreal path
they are kinda complicated idk how they work very well
surreal path
like idk what a pixel clock is or what sync is
supple robin
not relevant to you
ebon needle
like idk what a pixel clock is or what sync is
put dummy values
surreal path
where are they??
maiden ridge
I remember using this
https://docs.kernel.org/fb/api.html
surreal path
found them
surreal path
im guessing this explains how the red green and blue bits are formatted?
how is the limine framebuffer formatted
idrm
surreal path
huh?
supple robin
what huh
surreal path
whats not compatible with linux
supple robin
the limine fb struct
surreal path
ik
supple robin
what do you mean then
surreal path
i meant
like how are the bits formatted
for each pixel
ive never really did much framebuffer shit
ive slapped flanterm onto nyaux and called it a day
supple robin
vro
thorn bramble
for masking and shifting the respective color components
supple robin
exactly
surreal path
okay makes sense
supple robin
(pixel & mask) >> shift = red_color
surreal path
so if bits were formatted like RRRGGGBBB
it would be
111000000 for red
etc etc
and shift would be like 6
thorn bramble
sure
surreal path
k makes sense
thorn bramble
u don't need to worry about these values yourself
you just pass them to userspace
and xorg does the magic
surreal path
okay
thorn bramble
or whatever else you're porting
median goblet
where is this from
maiden ridge
where is this from
The code or the sticker?
cinder plinth
im porting doom
You don't really need an advanced fbdev for that
maiden ridge
^
median goblet
The code
maiden ridge
The code
Polaris back in start of 2022
Iirc
I figured out the issue later
Neptune was doing some bit manip which returned the incorrect result at times
So fixing that it worked
median goblet
nice
surreal path
using kde is better
usable + works
and ram usage isnt too crazy
6gb with firefox discord and alacritty open and screensharing with audio on walyand
edgy pilot
and ram usage isnt too crazy
mine uses 1.9gib on fresh boot
(with no startup programs of course)
edgy pilot
me everyday:
surreal path
cooked fried boiled
literarly cannot do shit
computer has become so unusable
if installing artix no workie
going back to windows cant deal with this shit no more chat
quick quiver
Just install fedora, debian, ubuntu, whatever
surreal path
nothing works
arch trolling me
so im reinstalling with artix
i dont like fedora, debian is debian, ubuntu is ubuntu
daring juniper
I wanna try arch at some point
tawdry mirage
install gentoo
supple robin
bro is doing anything but trying opensuse π
supple robin
"what is a stable system"
tawdry mirage
supple robin
https://yiff.forsale/gentoo.gif
i saw that fucking url
tawdry mirage
i don't own that domain for the record
just a gif i have in my gif picker
(and inb4 yes the url is a part of the joke)
surreal path
marvin just told me
symbols have size
too
so we can make nyaux's stacktrace wayyyy bettterrrr
ready to boot into artix
median goblet
https://yiff.forsale/gentoo.gif
gotta love the steak logo
ebon needle
<:nooo:750466870993223750>
mossy belfry
https://yiff.forsale/gentoo.gif
i can stand by this propaganda
surreal path
better nyaux stacktraces coming ur way
thanks to @supple robin for telling me st_size exists
gray hill
thanks to <@173337436989030401> for telling me st_size exists
st_size?
edgy pilot
better nyaux stacktraces coming ur way
my boi is mixing up c and c++ and is doing stack trace
surreal path
yep
surreal path
ik
ebon needle
c++ mangles names
the sigma solution is add extern "C to every function !!!
surreal path
no
edgy pilot
the sigma solution is add `extern "C` to every function !!!
no, there is a reason why names are mangled
it's not just because they felt like it
you can't use many features that depend on it with exten "C"
surreal path
i want kitty
edgy pilot
i want kitty
I have kitty
surreal path
i want cat
edgy pilot
i want cat
you won't believe this
surreal path
crazy
ebon needle
i want kitty
At least you don't have them, but mine ran away 
ebon needle
https://www.youtube.com/watch?v=hTgb2izmPOA&pp=ygUUZ3JlZ29yeSBmZW50YXlsIGZuYWY%3...
did you play fnaf security breach
tawdry mirage
you won't believe this
actually that's a bat
is that an alias or does bat drop all the waterworks if output is not a tty
edgy pilot
is that an alias or does `bat` drop all the waterworks if output is not a tty
It's an alias
thorn bramble
it does drop all the fancy stuff if output isn't a tty too
thorn bramble
i don't get it
thorn bramble
nah thanks
surreal path
missing out
thorn bramble
if you say so
surreal path
implementing kvm clock for nyaux even tho im sick lesgo
mossy belfry
implementing kvm clock for nyaux even tho im sick lesgo
woah you're writing a kvm?
incredible
mossy belfry
oh
surreal path
for better timer shit
mossy belfry
oh that boring π
surreal path
not really
mossy belfry
good luck π
mossy belfry
why not use rtc
surreal path
ew
mossy belfry
π’
surreal path
isnt that slow as balls
mossy belfry
isnt that slow as balls
it is usable π I have it like taped together to provide timestamps for ext2 things
supple robin
imagine using the rtc for timing
mossy belfry
would be wild ngl
surreal path
FINALLY
GOT IT WORKING
HOLY SHIT
IM SUCH AN IDIOT SOMETIMES
i lit did (hpet_virt_addr + 0xF0)
but it was a fucking uint64_t*
so its adding 0xF0 * sizeof(uint64_t)
now we have a generic timer thing
cool beans
surreal path
static inline bool checkkvmvendor() {
uint32_t ecx, edx, ebx, eax;
cpuid(0x40000000, 0, &eax, &ebx, &ecx, &edx);
uint32_t reg[3] = {ebx, edx, ecx};
uint32_t matching[3] = {0x4b4d564b, 0x4d, 0x564b4d56};
for (int i = 0; i < 3; i++) {
kprintf("comparring 0x%x with 0x%x\r\n", balls[i], matching[i]);
if (reg[i] == matching[i]) {
continue;
}
// else {
// return false;
// }
}
return true;
}
does anyonr know why cpuid is returning garbo data?
comparring 0x0 with 0x4b4d564b
comparring 0xffffffff with 0x4d
comparring 0x80035220 with 0x564b4d56
im running it in kvm obviously
ebon needle
```C
static inline bool checkkvmvendor() {
uint32_t ecx, edx, ebx, eax;
cpui...
there's problem in your cpuid function implementation
i getting normal values from cpuid
surreal path
static inline bool cpuid(uint32_t leaf, uint32_t subleaf, uint32_t *eax,
uint32_t *ebx, uint32_t *ecx, uint32_t *edx) {
uint32_t cpuid_max;
asm volatile("cpuid"
: "=a"(cpuid_max)
: "a"(leaf & 0x80000000)
: "rbx", "rcx", "rdx");
if (leaf > cpuid_max) {
return false;
}
asm volatile("cpuid"
: "=a"(*eax), "=b"(*ebx), "=c"(*ecx), "=d"(*edx)
: "a"(leaf), "c"(subleaf));
return true;
}
this is what it looks like
ebon needle
try this
inline void __cpuid(int code, int code2, uint32_t *a, uint32_t *b, uint32_t *c , uint32_t *d) {
__asm__ volatile("cpuid":"=a"(*a),"=b"(*b),"=c"(*c),"=d"(*d):"a"(code),"c"(code2));
}
wait
why you printing balls[i]
when you need to print reg[i]
surreal path
i changed it to be
more um
less stupid varialbe name
its not actually that in the code
it works???
weirdly
lol
thanks
thorn bramble
```C
static inline bool cpuid(uint32_t leaf, uint32_t subleaf, uint32_t *eax,
...
this function wont work with the 0x40000000 leaves
it only works with with the regular leaves
brisk zenith
```C
static inline bool cpuid(uint32_t leaf, uint32_t subleaf, uint32_t *eax,
...
this will unconditionally fail for 0x4xxxxxxx leaves because it checks for existence using the maximum for 0x0xxxxxxx leaves
and because you don't check for failure the code goes ahead with the garbage that was in the uninitialized locals
crystal scarab
cant you just use the __get_cpuid from cpuid.h?
brisk zenith
that has the same issue
edgy pilot
this function wont work with the 0x40000000 leaves
what are those
ebon needle
when nyaux hyfetch port 
thorn bramble
what are those
hypervisor defined leaves or smth
edgy pilot
hypervisor defined leaves or smth
kvm and stuff?
because I have very similar code and it seems to work
I think (about the similarity) (because it definitely works)
thorn bramble
i believe kvm, hyperv and others use that for identification and hypervisor specific stuff
and his code work wont because he calls cpuid with eax=0x4000β0000&0x8000β0000
i actually dont know what that returns exactly
edgy pilot
ah
I just do 0x8000etc for max leaf check
ebon needle
real sigmas don't do checks and blame userspace for all errors
edgy pilot
real sigmas don't do checks and blame userspace for all errors
even before userspace?
ebon needle
even before userspace?
yes 
surreal path
hi i woke uo
surreal path
and his code work wont because he calls cpuid with eax=0x4000β0000&0x8000β0000
makes sense
surreal path
working on pvclock for nyaux
which is just kvmclock v2
lol
whats crazy is that pmm init isnt taking forever suddenly
nyaux boots in like 3 seconds??? from kernel to userspace????
wtf????????????
idk what i did but
nice
im still sick
so
yea
:c
ebon needle
nyaux boots in like 3 seconds??? from kernel to userspace????
with kvm or not
surreal path
with kvm or not
kvm
ebon needle
what is your cpu
surreal path
thats stupid whoops
epic
uacpi: time source backwards drift detected, this may cause problems?????
oh wait im opid
okay somethings wrong
my code
pvclock::pvclock() {
// as pvclock is a x86_64 only feature we can assume shit here so its fine :3
// verify we actually HAVE tsc
uint32_t eax, edx, ebx, ecx;
cpuid(1, 0, &eax, &ebx, &ecx, &edx);
if (!(edx & (1 << 4))) {
panic("okay so no tsc = die");
}
// pvclock is per cpu :c
struct per_cpu_data *vcpu = arch_get_per_cpu_data();
// listen, kvm wants a full page. because of the way it handles it. so heres
// what we do, directly call pmm alloc
uint64_t addr_of_page = (uint64_t)pmm_alloc();
memset((void *)(addr_of_page), 0, 4096);
vcpu->arch_data.pvclock = (struct pvclock_vcpu_time_info *)(addr_of_page);
wrmsr(0x4b564d01,
((addr_of_page - (uint64_t)hhdm_request.response->offset) | 1));
kprintf_log(TRACE, "pvclock has now inited lol\r\n");
kprintf_log(TRACE, "pvclock version %u\r\n",
vcpu->arch_data.pvclock->version);
}
size_t pvclock::get_ns() {
struct per_cpu_data *vcpu = arch_get_per_cpu_data();
struct pvclock_vcpu_time_info *info = vcpu->arch_data.pvclock;
size_t time = rdtsc() - info->tsc_timestamp;
if (info->tsc_shift >= 0) {
time = time << info->tsc_shift;
} else {
time = time >> -info->tsc_shift;
}
time = (time * info->tsc_to_system_mul) >> 32;
time = time + info->system_time;
return time;
}
size_t pvclock::get_us() { return get_ns() / 1000; }
int pvclock::stall_poll_ms(size_t ms) {
size_t nanosecondsrnlol = get_ns();
kprintf_log(TRACE, "nanoseconds lol %lu\r\n", nanosecondsrnlol);
return -1;
}
looks good to me, the only thing i can think of is that the info is getting updated in the middle of a read and because you don't check version you're using stale values
reading the values and the rdtsc is all supposed to be done in a seqlock on version
brisk zenith
this is what i do
brisk zenith
if you don't use atomics things can get reordered out of the seqlock
or theoretically the new_version load can get reordered before the version load
like the docs don't describe it as such but this is essentially a lock and you have to treat it like one
surreal path
or theoretically the new_version load can get reordered before the version load
like in the compiled machine code, the compiler might reorder the loads
is that what u mean?
brisk zenith
or the processor itself but yes
surreal path
makes sense
brisk zenith
(and also without atomics the compiler will assume the check is always true as nothing in between writes to version)
surreal path
cooked
brisk zenith
(and without atomics the compiler is allowed to use split loads, i.e. load part of the value in one instruction and another part in another)
i use __attribute__((packed, aligned(8))) for my kvmclock_info_t
surreal path
makes sense
that works
lets see if this works now
still not working
size_t pvclock::get_ns() {
struct per_cpu_data *vcpu = arch_get_per_cpu_data();
struct pvclock_vcpu_time_info *info = vcpu->arch_data.pvclock;
uint32_t version = __atomic_load_n(&info->version, __ATOMIC_ACQUIRE);
uint64_t tsc_value, tsc_timestamp, system_time;
uint32_t tsc_to_system_mul;
int8_t shift;
for (;;) {
if ((version & 1) == 0) {
tsc_timestamp = info->tsc_timestamp;
system_time = info->system_time;
tsc_to_system_mul = info->tsc_to_system_mul;
shift = info->tsc_shift;
tsc_value = rdtsc();
}
uint32_t new_version = __atomic_load_n(&info->version, __ATOMIC_ACQUIRE);
if (((version & 1) == 0) && new_version == version) {
break;
}
version = new_version;
}
size_t time = tsc_value - tsc_timestamp;
if (shift >= 0) {
time = time << shift;
} else {
time = time >> -shift;
}
time = (time * tsc_to_system_mul) >> 32;
time = time + system_time;
return time;
}
all i changed
brisk zenith
what does your rdtsc() look like?
surreal path
static inline size_t rdtsc() {
uint32_t low = 0;
uint32_t hi = 0;
__asm__ volatile("rdtsc" : "=a"(hi), "=d"(low));
__asm__ volatile("lfence"); // wait until rdtsc completed
// explained by sdm go read the manual lol
return (size_t)((uint64_t)hi << 32);
}
wait
LMAO???
π
surreal path
HOLY
THE SPEED INCREASE IS CRAZY
THATS
1MILLION RIGHT?
yea 1.1 m ops
@kind root
1.3m ops now
its consisently getting 1.3-1.4m ops
lesgooo
anyways cool kvmclock is implemented
surreal path
kvmclock
ebon needle
and where you used it
ebon needle
i used to get 900k
can i try your iso
surreal path
k hold on lemme build it for u
ebon needle
k hold on lemme build it for u
k
surreal path
is uploading
git are u fucking braindead
it was too big to upload here
make sure u run with -M q35
and -m 512M at least
desert haven
lol I see limewire has rebranded
surreal path
the initramfs needs some memory to be loaded in
surreal path
yea
ebon needle
surreal path
guessing ur cpu is slower
ebon needle
ryzen 5 3600 
surreal path
oh yeah input is via serial
surreal path
ryzen 5 3600 <:chad:815144781842743334>
old af
ebon needle
yeah
surreal path
even my intel laptop cpu can beat it
ebon needle
even my intel laptop cpu can beat it
π
surreal path
yogurt
gurt: yo
git status
On branch master
Your branch is up to date with 'origin/master'.
Changes to be committed:
(use "git restore --staged <file>..." to unstage)
modified: .clang-format
modified: kernel/src/acpi/acpi.c
modified: kernel/src/acpi/kernelapi.c
modified: kernel/src/arch/arch.c
modified: kernel/src/arch/arch.h
modified: kernel/src/arch/x86_64/cpu/lapic.c
modified: kernel/src/arch/x86_64/cpu/structures.h
modified: kernel/src/arch/x86_64/instructions/instructions.h
modified: kernel/src/term/term.c
modified: kernel/src/term/term.h
modified: kernel/src/timers/hpet.cpp
modified: kernel/src/timers/hpet.hpp
new file: kernel/src/timers/pvclock.cpp
new file: kernel/src/timers/pvclock.hpp
modified: kernel/src/timers/timer.cpp
modified: kernel/src/timers/timer.hpp
modified: kernel/src/utils/basic.h
β― git push
Enter passphrase for key '/home/anon/.ssh/id_rsa':
Everything up-to-date
blud
ok works now
kind root
ebon needle
edgy pilot
looks good to me, the only thing i can think of is that the info is getting upda...
time easily overflows without it being uint128
tender gorge
time easily overflows without it being uint128
uint64_t is big enough to store the amount of nanoseconds it takes for the heat death of the universe
edgy pilot
uint64_t is big enough to store the amount of nanoseconds it takes for the heat ...
uint64 is not big enough for calculation
(sometimes)
tender gorge
managarm be like:
edgy pilot
(I've encountered the issue before)
edgy pilot
:P
edgy pilot
time easily overflows without it being uint128
of course the final result can fit in u64
brisk zenith
time easily overflows without it being uint128
yeah on my system one of the intermediate values (specifically the result of time_after_shift * mul) overflows a u64 within a few seconds
so i do that calculation using u128 arithmetic
gray hill
okay somethings wrong
lol, your ops went down too XD
edgy pilot
593 π
edgy pilot
@surreal path you should probably be pinged about this
surreal path
time easily overflows without it being uint128
that is fair
edgy pilot
fair?
surreal path
meaning yes i should
thorn bramble
white is for bold text
ebon needle
white is for bold text
in macos the normal text for logs is white and it looks nice
edgy pilot
I am sick too 
ebon needle
from the sickness
ebon needle
I am sick too <a:death:816178972672131083>
why all my minecraft hardcore targets is sick
snow wolf
```Cpp
size_t pvclock::get_ns() {
struct per_cpu_data *vcpu = arch_get_per_cpu...
I really dislike the size_t get_ns();. Why no nsecs_t (which can just be typedefed to unsigned long long or uint64_t)?
surreal path
I really dislike the `size_t get_ns();`. Why no `nsecs_t` (which can just be ty...
size_t is funny
plus i dont plan to support 64 bit
and size_t just seems nice
it just feels right
edgy pilot
plus i dont plan to support 64 bit
you sure?
kind root
at least make it a uint64_t
surreal path
median goblet
just make it uint64_t
surreal path
also
amazing commit message
yes ive comfirmed it
files have uint64_t offset
or well
file descriptor
handles
yk what i mean
broken depot
you ci fail reminds me that I also need to fix my ci fail lol
surreal path
you ci fail reminds me that I also need to fix my ci fail lol
fix mine for me
broken depot
surreal path
i have shitty ci
that only builds the kernel
thats it
no sysroot
im too lazy to get jinx working
edgy pilot
probably for the best
surreal path
and i do not wanna bother mint to get it working lol
edgy pilot
you'll just waste free minutes
surreal path
if it builds for my machine it should build for you 
edgy pilot
workflows aren't unlimited
edgy pilot
same
edgy pilot
real
ebon needle
we should get back to nyaux dev soon
what you will do
edgy pilot
it's just a struct
surreal path
ik
crystal scarab
Lmfao
tawdry mirage
best i can do is three fiddy
ebon needle
where a guy bought nyaux source code for 5$
i can buy nyaux license for 1$ !!!
elder shoal
how is nyaux nyaxuing
elder shoal
we got xorg before disk drivers π₯
Honestly get disk drivers itll make testing so much faster
elder shoal
no <:trl:829316641191886888>
elder shoal
make me like furries and gay somehow then maybe
sure
broken depot
make me like furries and gay somehow then maybe
transes your gender
There you go!
surreal path
bro made me into czzpek
broken depot
I have no idea what "czzpek" is π
surreal path
I have no idea what "czzpek" is π
@thorn bramble
thorn bramble
π
glossy walrus
ebon needle
lets start by getting frigg to link againist nyaux
when nyaux intel hda support
glossy walrus
when nvidia modeset support
ebon needle
when nvidia modeset support
nyaux need a drm support cuz it need to be cooler than windows
surreal path
what
glossy walrus
nyaux need a drm support cuz it need to be cooler than windows
stroke?
supple robin
nyaux needs a job
surreal path
thats crazy
maiden ridge
nyaux needs a job
Tty jobs
rich trellis
nyaux needs a job
stop scaring people like thatπ
surreal path
okay so
why does bash die
when launching doomgeneric
ive implemented fbdev
do u guys know whats going on with that??
like the bash process doesnt die
but it doesnt show a prompt nor accept input anymore
ill dbg this tmrw i feel sick in my stomach rn
coral dove
do u guys know whats going on with that??
Where are you storing the executable?
surreal path
Where are you storing the executable?
in /usr/bin????
surreal path
im triyng to
right so doomgeneric spawns another thread of itself
it does stuff
kills that thread
kills itself
but bash still runs
now how it somehow made bash stop working is crazy
no it did not kill serial put input
yea i have no idea whats going on
atexit is not calling the function cleanup somehow
thats the issue
anyways back to eeping
ebon needle
but it doesnt show a prompt nor accept input anymore
restore your ioctl state when doom exiting
ebon needle
no its atexit
then why bash is dont want input ?
surreal path
because doom doesnt retore the original tty state
meaning the tty is probs in non interactive mode or some shit
whatver doom configured it to
ebon needle
maybe restore state in syscall exit ?
surreal path
i do ofc?
thats not the issue here
the issue is that atexit doesnt call the callback function
now i dont know how that works in the internals of mlibc but im sure that is related to a missing sysdep or some kind of bug in my kernel
does atexit work for orange @ebon needle u can test this by writing a small program that uses it
ebon needle
how
surreal path
you can just make a recipe for ur own program
that gets compiled then installed into ur sysroot
see ttytest recipe in nyaux
ebon needle
you can just make a recipe for ur own program
i mean what code i should use for atexit
ebon needle
honestly
codeberg moment
i cant compile my os because codeberg is died
@surreal path it doesnt works too π
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <termios.h>
struct termios orig_termios;
void disableRawMode() {
printf("got atexit\n");
tcsetattr(STDIN_FILENO,TCSAFLUSH,&orig_termios);
}
void enableRawMode() {
tcgetattr(STDIN_FILENO, &orig_termios);
atexit(disableRawMode);
struct termios raw = orig_termios;
raw.c_lflag &= ~(ICANON);
raw.c_cc[VMIN] = 1;
tcsetattr(STDIN_FILENO,TCSAFLUSH,&raw);
}
int main() {
enableRawMode();
printf("enter any key to continue\n");
char key;
read(STDIN_FILENO,&key,1);
}
weird hm?
yea so this may be a think that mlibc wants
like some sysdep
or something
now i couldnt verify this but maybe someone more knowledgable in mlibc internals could
im also gonna verify that doomgeneric actually REACHES exit
thorn bramble
atexit is not a sysdep
it should in fact already work
so if it doesn't go figure out why
__mlibc_do_finalize is implemented in options/lsb/generic/dso_exit.cpp and it's called from exit in options/ansi/generic/stdlib.cpp
the code is really simple and, as i said, it should just work as is
surreal path
it should in fact already work
yea weird
i think its not actually reaching exit
it may be stuck somewhere
median goblet
π
surreal path
weird issue hm
tawdry mirage
chex.wad
tawdry mirage
i wonder if doom-generic actually has support for chex quest
since iirc it has custom things
surreal path
i wonder if doom-generic actually has support for chex quest
have u ever made a program that is open source
random question
i might just throw it in nyaux
for shits and giggles
median goblet
doomgeneric devs are smoking
i mean its generic
they probably assume exit doesnt exist
tawdry mirage
have u ever made a program that is open source
well one util i've written is https://github.com/qookei/dir-diff
surreal path
well one util i've written is <https://github.com/qookei/dir-diff>
ill write a recipe for it
surreal path
the pixels doomgeneric are sending in are RGBA
this is hard coded
for some fucking reason
and i have no idea how to convert the pixels on the fly in a reasonably and non slow way
doomgeneric should respect the format i tell it but no
like genuinely why even print the framebuffer shit when its all fucking hard coded
even the bpp is hard coded
β οΈ
also i dont know why it didnt choose the top of the screen but
idfk
i may need help with this cause i do not know how to do this reasonably so anyone who has ported doomgeneric please tell me what i should do that wouldnt be a shit and slow as balls solution
quick quiver
you can also port qookie's mini-lspci but that requires sysfs ;D
brisk zenith
like genuinely why even print the framebuffer shit when its all fucking hard cod...
where is it hard coded?
ah in I_InitGraphics
honestly i'd just patch it to not hard code it
surreal path
honestly i'd just patch it to not hard code it
doesnt matter if i change the values to match what nyaux would use
doomgeneric just
doesnt care
π
brisk zenith
then i don't think that's the issue
surreal path
i dont think theres much of an issue with the implementation of nyaux's fbdev either, because ive checked the values it gives to userspace and they do look sane
brisk zenith
what are those values?
brisk zenith
that does look sane yeah but it also matches what doom's printing
and besides the pattern isn't really what i would expect from those parameters being wrong
surreal path
yea
for more context this is what my write and read functions for fbdev look like
size_t read(thefunny) override {
size_t end = epicfb->pitch * epicfb->height;
if (offset >= end) {
*res = EINVAL;
return -1;
}
if (size + offset > end) {
size = end - offset;
}
void *calculated_dest = reinterpret_cast<void *>(
reinterpret_cast<uint64_t>(buffer) + static_cast<uint64_t>(offset));
memcpy(
reinterpret_cast<void *>(reinterpret_cast<uint64_t>(calculated_dest) +
reinterpret_cast<uint64_t>(offset)),
epicfb->address, size);
*res = 0;
return size;
}
size_t write(thefunny) override {
size_t end = epicfb->pitch * epicfb->height;
if (offset >= end) {
*res = EINVAL;
return -1;
}
if (size + offset > end) {
size = end - offset;
}
void *calculated_dest = reinterpret_cast<void *>(
reinterpret_cast<uint64_t>(buffer) + static_cast<uint64_t>(offset));
memcpy(
reinterpret_cast<void *>((static_cast<char*>(epicfb->address)) +
static_cast<uint64_t>(offset)),
calculated_dest, size);
*res = 0;
return size;
}
int ioctl(void *data, unsigned long request, void *arg,
void *result) override {
switch (request) {
case FBIOGET_VSCREENINFO: {
*static_cast<fb_var_screeninfo*>(arg) = variablefbinfo;
*(void**)result = 0;
return 0;
break;
}
case FBIOGET_FSCREENINFO: {
*static_cast<fb_fix_screeninfo*>(arg) = fixedfbinfo;
*(void**)result = 0;
return 0;
break;
}
default:
kprintf_log(ERROR, "fbdev(): unknown ioctl\r\n");
break;
}
return ENOSYS;
}
thefunny is a marco
just this
brisk zenith
ah you're doing it with read/write not mmap
that makes things simpler
(also imo read shouldn't be allowed on fbs but that's unrelated)
surreal path
(also imo read shouldn't be allowed on fbs but that's unrelated)
why not?
astral allows it
brisk zenith
WC memory is not architecturally guaranteed to be coherent
aka on some cpus this assertion would trip even without any race conditions ```c
*fb = 0xabcdef;
assert(*fb == 0xabcdef);
i see
brisk zenith
this is because wc was originally as implemented as uc with the buffering being completely separate from the regular cache layer
so you could get a read to go through to the actual framebuffer memory before the buffered writes are committed
surreal path
so you could get a read to go through to the actual framebuffer memory before th...
yea makes sense
brisk zenith
for more context this is what my write and read functions for fbdev look like
``...
why are you adding offset to the input buffer? it's only the destination position, not the source position
surreal path
in read?
brisk zenith
in both
surreal path
memcpy for nyaux is dest, src
brisk zenith
you're doing memcpy(fb->address + offset, buffer + offset, size)
surreal path
also good point
brisk zenith
it should be memcpy(fb->address + offset, buffer, size)
cinder plinth
i may need help with this cause i do not know how to do this reasonably so anyon...
Idk I didn't have that problem
surreal path
considering thats the entry point that rip looks like garbage @brisk zenith
either that or its in the libc
it is probs in the libc somewhere
looks like /lib/libm.so
cinder plinth
What'd you change
surreal path
lucky ig
brisk zenith
what instruction is at that address?
cinder plinth
It's not luck I'm just pretty sure it can detect the framebuffer format
surreal path
i am checking hold on
surreal path
It's not luck I'm just pretty sure it can detect the framebuffer format
no not really
its hard coded lol
brisk zenith
It's not luck I'm just pretty sure it can detect the framebuffer format
i checked, it can't
it's hard coded in I_InitGraphics
cinder plinth
Huh
brisk zenith
it just matches the format limine uses on qemu and most hardware
cinder plinth
Yeah that's what I figured
supple robin
is lib/libm not real
yes
supple robin
it's a dummy
brisk zenith
the functions libm needs to provide are provided by libc
surreal path
right
brisk zenith
libm still needs to exist though so it's just an empty shared lib
cinder plinth
It never was real, it's just a bunch of numbers on your computer!
supple robin
what?
surreal path
im looking for the instruction it faulted?
supple robin
do you know what a relocation is
surreal path
well i know it exists
brisk zenith
i don't really trust that rip value
cinder plinth
Doubt that's what faulted
surreal path
yea
surreal path
yes i know this
that is why i subtracted the rip on the address mlibc loaded "libm" at
surreal path
looks like /lib/libm.so
^
supple robin
that is why i subtracted the rip on the address mlibc loaded "libm" at
but the symbol is in libc
libm is empty
surreal path
right ill check where it loaded libc
looks better
looks better
interesting
0000000000043dc0 <_ZN5mlibc11strncasecmpEPKcS1_m>:
seems to be in strcasecmp
i can try and get the parameters it loads
thorn bramble
just look at the source code
surreal path
for strncasecmp?
surreal path
good point
surreal path
yea
interesting? printing cook in R_CheckTextureNumForName doesnt cause it to gpf ???????
it just gets stuck here
tf is happening
i removed it
and its still stuck??
nvm i know why
fixed that
printing in R_InitSpriteDefs in doomgeneric doesnt cause the general protection fault but it does the same thing getting stuck with init status bar
printing anything at all in R_InitSpriteDefs just gets it stuck
i pushed everything to github if anyone wants to also take a look
dunno why its commiting .cache even tho i have that in my gitignore but k
kind root
brisk zenith
surreal path
i was copying shit to the buffer + offset in write from buffer + offset
π
that was the issue lol
cinder plinth
does keyboard work too?
cinder plinth
how do you type commands then
surreal path
serial input
cinder plinth
well you could just use that
thorn bramble
no ps2?
surreal path
fuck
weirdly this happened
i cannot seem to get it to be reproducable
@brisk zenith could this be related to why one of ur machines couldnt boot nyaux?
brisk zenith
no idea
surreal path
fuck
it seems to be faulting with a kernel code address?
that is quite weird
also ive noticed that when launching doom it doesnt really look like its being animated?
i cannot verify this as no input
also the rip is a heckin
address that isnt code??????
no it is
nvm
and cr2 is memory
nvm
weird it died checking the us variable ?
