development

[marble perch]

The bugs are the same every year with new os releases. Bad battery, memory leaks, etc

Just QC issues

People just love to cry about things

[grim sparrow]

idc about those, I more mean things like buttons just not working in settings

mostly ui bugs

(in mostly swiftui apps but again, not my point)

[marble perch]

Ahh swiftui moments

[grim sparrow]

I think SwiftUI will be a viable option when you can open an app and not immediately tell that its written with SwiftUI

because it just has so many tells, like slightly stuttering lists or buttons not being perfectly aligned

[gentle grove]

Upgrade to C++23 so you can get the new slop but I forgot what specifically

[grim sparrow]

Yeah I would always use the latest C++ version when I’m starting a project

Clean slate so no code to fix

[magic hazel]

It’s not in the public beta?

[grim sparrow]

It was added in the most recent dev beta

I don’t think it’s in public

[magic hazel]

Ah alr

I should not have updated

[grim sparrow]

Typically public is the last dev beta with critical fixes

[magic hazel]

This beta is so much slower and looks so much worse and is so much buggier

[floral notch]

update: dont know if anyone cares (im mostly just writing this so i can reference it in the future) but the 3rd one is the best, and super easy to vendorize due to the codebase structure and that it is MIT licensed. there hasnt been any development in over a year but i would consider OCMethodTrace to be a complete project more or less, on iOS at least. it doesn't really need maintenance

my only complaint is that it is heavily tied to iOS. OCMethodTrace.m, which is the meat of the code, has #import <UIKit/UIKit.h> at the top and you cannot simply comment that out since it hard codes support for UIEdgeInsets, UIOffset, CGRect, etc. however, that file is only 1KLOC so i imagine it will be pretty straightforward to remove the hard dependency on UIKit (probably would take a day or so) and eventually port to other platforms like GNUstep / Linux (which would probably take a week).

in a pre-ChatGPT era i would have complained about the docs/comments being in chinese but that can be one-shot in the fucking head with a prompt like "translate the comments in this code to english" or "translate this markdown to english"

[tepid nacelle]

its made by the chinese its chinese spy propaganda /j

[pearl sail]

Until

[floral notch]

i am chinese so its ok

[tepid olive]

And you still couldn’t understand Chinese docs?

[floral notch]

我是中国人

我爱习近平

taiwan is china

[tepid olive]

DeepSeek gave me some weird shit about harmless responses when I gave it this to translate

[grim sparrow]

Just found this site on Twitter, thought it was really neat https://whatsnewinwwdc.com

[magic hazel]

Awesome!

That's actually going to be incredibly useful for my backports

All the way back to 2015 as well!

[hasty ruin]

lmao i knew it

[magic hazel]

I mean it's useful either way

But yeah it was definitely screaming react 💀

[wooden yarrow]

yeah no wonder it glitches so much when i scroll

also if you scroll too far up you get flashbanged

[native dune]

you can tell its ai by the gradients

[hexed knot]

It's the colors man

[magic hazel]

what's the best mach-o decompiler?

[light owl]

@native dune Does this seem similar to something else ?

[sonic totem]

IDA probably

[magic hazel]

thats what im trying rn

surprisingly it's quite excellent

UIVisualEffectView__whatsWrongWithThisEffect```

LOL

[kind herald]

Nocturne

[harsh junco]

*cracked ida)

[frank fossil]

Who’s interested in getting amfid/launchd task port on iOS 17.0?

[trail nimbus]

What would that do

[frank fossil]

jelbrek

[trail nimbus]

Jelbrek always good

[cinder frigate]

With "jelbrek" you mean semi-jb like Nathanlr or serotonin?

[frank fossil]

Might be a bit more than semi jb if platformizing via amfid is still feasible

[cinder frigate]

Would that be 17.0 only or 16.0/1 16.6/16.6.1 would be supported too?

[cinder frigate]

When you have some time I'd like to know what that "a bit more" is. A link to a web page would be appreciated too of course

[frank fossil]

Chance that it might be possible to inject system wide, but in a limited form since there is no PPL bypass to enable JIT (or that all binaries needs to be resigned to enable JIT)
-# JIT in this context is meant for tweaks putting in-memory hooks that would otherwise trip codesigning

[cinder frigate]

@frank fossil Sorry I meant 16.6/16.6.1

[frank fossil]

should work

[cinder frigate]

Then, if it will be better than actual semi-jb, I'd be interested

[quaint rain]

Would it be possible in some way to use trollstores JIT to enable JIT?

[frank fossil]

no, binary must have get-task-allow entitlement

[granite frigate]

if resigning is still needed whats the benefit over js replacing with LC_LOAD_DYLIB

[gentle grove]

https://x.com/tritlo/status/1983665843310620936?t=fM90nbWefBSOLBGLAVqiFQ&s=19

Matti Palli 🧙‍♂️ (@tritlo)

sometimes I forget that Python is a scripting language with dynamic imports.

then I remember, and remember that you can use subprocess from WITHIN Python to install a package and then LOAD that package in the same interpreter

absolutely cursed

X · Oct 29, 2025, 10:42 PM

Wait until this guy finds out about dlopen()

[granite frigate]

im kind of impressed the interpreter like rescans the venv for packages

[gentle grove]

I think it just scans at import time but I could be wrong

[cloud yacht]

idk how python works but js and lua just lookup when you import and then cache the import. No need to lookup everything if your not going to import eveyrthing

[wooden yarrow]

i'm sure some code out there checks for import exception then proceeds to auto download from pip if not satisfied

[timid furnace]

it's not very intensive

iirc it's just direct children of everything in sys.path

that's not that many locations to check

[frank fossil]

What does that mean?

[magic hazel]

anyone know how to resolve this

My target is macOS 15 so I'm incredibly dumbfounded as to why this is occuring

[grim sparrow]

Usually this means that in one of your Package.swift files you haven’t specified the minimum version so it reverts to the default

[magic hazel]

Ugh so it means forking time

I gave up on it anyways now I'm banging my head against a wall about how to manage these freaking data structures

[frank fossil]

I found a method for TrollStore app to elevate to root without root helper. Enjoy.
Dunno if this survives 17.6 thing

[magic hazel]

Turns out just switching from structs to classes literally saved me

[grim sparrow]

what issue were you facing

[magic hazel]

still kinda struggling to move my client past dms to servers

and i was trying to make member objects

you can't inherit from a struct so i had to make a protocol which is messy since i didn't really want that plus it meant i had to add a user sub object to compare since members aren't users they're user protocols etc

whereas if i use classes, i can have a base class of a user, then a member class that is a user but has extra stuff, a client user class that is also a user but has extra stuff

might end up going back to structs though

[cinder frigate]

I imagine this has already been discussed, but I don't know how to search for it, so I'm asking.
Since we spawn a custom launchd, are there technical limitations that prevent semi-jb like serotonin or Nathanlr from using a technique similar to the one used by dopamine to put all executables in the trust cache, to instead signing them on the fly? Perhaps by checking that they are not already signed.

For "the dopamine technique" I mean the one explained in this video (link with timestamp included)

[sonic totem]

Trustcache is in PPL memory, Dopamine has a PPL bypass and Serotonin does not

[cinder frigate]

Ok, dopamine puts the CDHash in the trust cache on the fly, why can't serotonin sign with ldid (giving all the possible entitlements) on the fly, does this require a PPL bypass too?

[sonic totem]

Because the CD hash being put in the trustcache is what makes the code signing system believe the binary is trusted

If it wasn’t in the trustcache the system would just reject it, because it would have no way of knowing if it was trusted or not

[cinder frigate]

So the CT check has already been done (and failed)?

[sonic totem]

No, the trustcache check comes first

If that fails, it moves to CoreTrust

[cinder frigate]

Ok, so why did you say that it would have no way of knowing if it was trusted or not?

[sonic totem]

Well, do you understand what the trustcache is?

It’s essentially meant to be a list of CD hashes for Apple binaries on the system

That the kernel knows it can implicitly trust to run

Dopamine uses its PPL bypass to add CD hashes to this list, that’s why it can bypass codesigning

If the binary wasn’t in trustcache, the kernel would get its CD hash and not find it in the trusted list, so would reject it

[cinder frigate]

Ok, dopamine bypasses code signing, and gets arbitrary code execution, my point was that on serotonin, (where you can't bypass code signing, and I'm aware of that) you have to manually sign every binary you build, or download, and this is painful.
So I was wondering if this signing process could be automated in the same way dopamine automate the "trust cashing"

Sorry if my first question wasn't clear

[sonic totem]

Ah, then no

Well, Serotonin does bypass codesigning, but with a CoreTrust bypass

There used to be a way to attach a signature purely in memory and not be re-signing the file

But now that required at least a kPAC bypass

[grim sparrow]

And now we need MTE proof exploits

[wooden yarrow]

the PAC PPL TXM SPTM MTE bypass

[cinder frigate]

And re-sign (actually just sign, since I hope to have some way to determine if a bin is already signed, and in general I would assume that an already signed bin, has already all the entitlement that it needs) the file is somehow out of discussion? (Is it too slow?)

[magic hazel]

Atp it’s just the Apple bypass 💀

[wooden yarrow]

i mean that's really the core of what a jailbreak is

[magic hazel]

Fr

[cinder frigate]

Just impersonate an enough democratic government and buy it from a cyber security company

[magic hazel]

How get Apple off my system

[sonic totem]

What do you mean?

Dopamine won’t sign anything

It just adds CD hash to trustcache if the binary is signed

[cinder frigate]

You said

Well, Serotonin does bypass codesigning, but with a CoreTrust bypass
There used to be a way to attach a signature purely in memory and not be re-signing the file
So my question was, is re-signing the file bad / impossible (for serotonin)? (taking into account the caveats I had put in parentheses)

[frank fossil]

• TPRO bypass

[sonic totem]

Serotonin applies the CoreTrust bypass to every file, which is kind of ‘dirty’ because it modifies the original files

[cinder frigate]

The point is that it doesn't (at least not the latest release). If I build a bin with clang and I try to run it, it gets killed

[sonic totem]

You need to ad-hoc sign it

ldid -S <file>

[cinder frigate]

Yes, I knwo, the whole point of my first question, is if this sign process could be automated patching launchd

[cinder frigate]

(you also have to run fastPathSign since it's roothide it is what applies ct_bypass)

[sonic totem]

You shouldn’t have to

RootHide is RootHide

[sonic totem]

In theory yes but in practice probably not

Like, you’d have to sign with no entitlements

And you’d have to depend on ldid being installed on the device

It should be the user’s responsibility

[cinder frigate]

To run fastPathSign? it's what applies ct_bypass. You have to run it (on serotonin, never tried Nathanlr)

[cinder frigate]

ldid is a basebin, it will always be installed on the device

[cinder frigate]

You should also sign with all the entitlements, otherwise the bin could miss some and get killed

[cinder frigate]

(Yes sorry, I confused symredirect with fastPathSign, but you have to call fastPathSign)

[naive kraken]

only in roothide

and I'm pretty sure that breaches it's license

[sonic totem]

The concept of “all the entitlements” doesn’t exist

There isn’t a predefined list

And you shouldn’t give all entitlements to all binaries because then all binaries get permissions they shouldn’t have

And yeah as opa said, ldid should not be preinstalled 😛

[sonic totem]

The jailbreak should handle that itself

If not then it’s bad design

[cinder frigate]

Doesn't this happens with Dopamine or binaries in the trust cache are still checked for entitlements?

[sonic totem]

Dopamine doesn’t sign anything

It just adds CD hashes to trustcache

Signing and entitlements is the user’s responsibility

But in regards to signing with all entitlements by default, I more meant that giving every binary as many entitlements as possible is very very insecure

[rocky oriole]

Idk what you mean, I love giving my random calculator app I found camera and microphone access

[cinder frigate]

Yes, I know

[grim sparrow]

We love saurikware

[cinder frigate]

Well, I think it is... ('cause you have to)

[sonic totem]

Yea that’s bad design then

It should just re-sign with the bypass before spawning

[cinder frigate]

Sorry but, Trollstore uses Procursus ldid to sign IPAs, doesn't it automatically install it too?
Edit: (It wasn't meant to be an accusation, just a doubt.)

[hasty ruin]

it downloads it at runtime

it doesn't get shipped with trollstore

[sonic totem]

The license just prohibits distribution

[grim sparrow]

https://github.com/ProcursusTeam/ldid/blob/master/COPYING in case you want enrichment in shit license™

[pearl sail]

But it is FOSS or you can email the patch to bingner who may or may not email saurik

[grim sparrow]

Procursus ldid is the only version that matters anymore

[pearl sail]

Let me fork it to make a version out of rust that two people may use but I will update via vibe code every 2 years

[magic hazel]

I thought if you don't modify source you're safe?

[grim sparrow]

anything you bundle it with has to be the same license

its a virus

if you are just providing a dl link ur probably fine

as long as you also include a cop y of the license

[vague parrot]

is there same as trollstore app, but can work on ip16pm?

[rocky oriole]

stop crossposting everything you ask please

[vague parrot]

aight

[rocky oriole]

asks a question in #jailbreak
gets an answer from 3 people
makes genius bar thread
asks in #development

[vague parrot]

okay

[naive kraken]

I mean my reasoning is that if a non GPL browser is allowed to download the file and a non GPL terminal is allowed to run it, then so is TrollStore

[grim sparrow]

i guess

im an engineer not a lawyer

and besides, I doubt Saurik would care atp

[rocky oriole]

does Saurik even do jailbreak related things anymore

[grim sparrow]

not to my knowledge

the last I heard he was hacking ethereum

[pearl sail]

methereum

[hollow oar]

Meth!

[rocky oriole]

https://mwrevamped.github.io/Methamphetamine/

[grim sparrow]

METH!

[magic hazel]

Why is CAFilter still a private API even when it’s remained unchanged for so long and is still quite useful?

Is there any specific reasoning behind it?

[hasty ruin]

https://github.com/MethDevelopment/AmyWasHere

[grim sparrow]

Real

[hexed knot]

Lmaoo I forgot about meth development

https://github.com/MethDevelopment/russ1anroulette

[tropic sail]

Anybody knows anything about X-Apple-ActionSignature and how it works?

I mean, how it is being generated. We need info about it, because there is an issue with logging in Apple ID account in iTunes Store/App Store on iOS 5.1.1 since spring 2025. After researchers, it was discovered that iOS 5 and below iTunes Store clients were using GET method to log in, and starting from iOS 6, it started using POST method to log in. And in spring 2025 Apple killed GET api, so it is not possible to log in iTunes Store/App Store unless you convert GET request to POST using proxy, but the thing is that it requires X-Apple-ActionSitnature, that iOS 5 and below clients were not generating. So now as we can see the only way to fix it is too convert GET request to POST and try to “teach” iOS 5 generate that signature, or something like that.

That is why I am asking

[grim sparrow]

I don't know personally but just searching for that header key on GitHub gave a lot of results

[wooden yarrow]

@torn oriole another one for the filters

[fluid lintel]

https://github.com/StephenDev0/StikDebug?tab=readme-ov-file

Hi, I was analyzing this tool thinking I could enable JIT even on apps downloaded from the store, but unfortunately it seems that it only works with apps that are being debugged or sideloaded. My idea was to be able to analyze an app without having to re-sign it, because re-signing the app often causes the loss of the original entitlements (keychain, iCloud, etc.) and often requires a binary patch to avoid crashes. Do you have any advice?

[granite frigate]

there’s ellekit jitless but it’s only limited to 8 breaks i think

oh 6*

https://github.com/tealbathingsuit/ellekit/blob/main/ellekitc/JITLess.c

[fluid lintel]

But doesn't it need to be jailbroken to work?

[slim bramble]

No

[fluid lintel]

However, it only works for an app that has this code inside it.

[slim bramble]

What device

[fluid lintel]

Iphone 11 iOS 26

[tepid olive]

JIT only works on specific apps on iOS 26

[fluid lintel]

What do you mean? Has iOS 26 introduced further restrictions?

[slim bramble]

To simplify, JIT was killed by apple in iOS 26

Here is JITLess ellekit

[fluid lintel]

Ok but only for jeilbroken device

[fading shell]

<@&355174844205367317>

[kind herald]

Machine

[magic hazel]

Isn’t there another method

[grim sparrow]

<@&355174844205367317>

[north lance]

That was quick

[grim sparrow]

third ones the charm

[magic hazel]

crazy

[magic hazel]

any way of recieving touch events outside the bounds of a view, clip to bounds wont work

[hasty ruin]

hitTest maybe

[grim sparrow]

My immediate reaction would be no, the event would go to whatever view you actually pressed

What’s your use case here, just from that description it sounds unintuitive

[frank fossil]

are you okay with having the view completely hidden instead

[magic hazel]

Profile picture hangs off message bubble i don’t want the bubble to expand

Only the part hanging in gets touch events

[grim sparrow]

Just give the view a container view that’s invisible

And that container view can listen for events

[thorn ether]

anyone know how I can set custom paths for Shortcuts?

currently it uses /var/jb/usr/bin/ but i want to add another path

nvm figured it out. put paths in /private/var/root/.bashrc

[grim sparrow]

Gm

[marble perch]

[magic hazel]

is limneos down again :/

i should try backing it up

[torn oriole]

shocker, annoying repo outage

[lean forge]

found something called tightbeam_stub inside my iphones filesystem

its a unix executable

anyone know what it does?

[wooden yarrow]

https://arxiv.org/abs/2510.09272 read the paper

specifically 7.10

[lean forge]

thanks

was just confused about it cuz I couldnt find anything about it

[untold drift]

hi guys, is there any alternative MSHookFunction for sideloaded app's tweak? (livecontainer)

[indigo kraken]

You can’t on sideloaded apps

[untold drift]

i just wanna call a function in app's framework :/ is it impossible on sideloaded? i can get address with dlsym

[grim sparrow]

Calling a function doesn’t need a hook

If you can get the address then you can just call it

Hooking is for overriding what a function does

[untold drift]

yea i know but i dont want to more change on my code 😄 im trying to pointer to cast a func now.

[radiant idol]

may you provide some code as to what you're doing? that may help

[grim sparrow]

Or what your goal is. If the framework you’re calling is open source or publicly documented you may just be able to call it normally

[untold drift]

long long (*orig_getData)(const char *str);

thats the function i wanna call. i hooked and logged some data when jailbroken but im not jailbroken now. just wanna call sometimes. its enough for now

when i get symbol with
void *dataFunc = dlsym(RTLD_DEFAULT, "getData");

its returns an address (pointer)

just i will cast it and work i think 😄

[radiant idol]

long long (*orig_getData)(const char *str);

// ...
void *dataFunc = dlsym(RTLD_DEFAULT, "getData");
orig_getData = (typeof(orig_getData))dataFunc;

I think this should do it

[untold drift]

i cant see crash logs but crashing. i will try your code o/

[radiant idol]

make sure the pointer is valid though

[grim sparrow]

Crashing would usually mean the pointer is nil or one of your function signature is incorrect

[untold drift]

yea, maybe i need to debug it

[grim sparrow]

As a first step I’d just print the value of dataFunc

See if it’s 0x0 or it actually found it

[untold drift]

no, i did it. return an address with 0xdeadbeef etc

i just upgrade my ios for whatsapp :/ i love jb community but bad times.

[untold drift]

its work! thank you ❤️

[radiant idol]

nice!

[magic hazel]

does ios crash out if you lipo together 2 different binaries

as in one built by xcode 13 and one built by xcode 14

[grim sparrow]

I’m pretty sure someone figured out the way to do it so it does work

[light owl]

Developers my buddy @native dune has a question

[native dune]

Nah its working

[grim sparrow]

@native dune what’s your question

[native dune]

Nothing i was wondering how to inject into a platform binary but i ended up disabling amfi

[kind herald]

/dev/apple_mfi

— 8:15 PM
Developers my buddy @#maxinedemoted has a question
#maxinedemoted

— 8:16 PM
Nah its working
amler #promotemaxine

— 8:25 PM
@#maxinedemoted what’s your question
#maxinedemoted

— 8:26 PM
Nothing i was wondering how to inject into a platform binary but i ended up disabling amfi

[light owl]

Youre wasting your time on that btw

[native dune]

Absolutely

[light owl]

Precious time that you could be spending on The Method instead

[native dune]

I just wanna see it work

[visual meadow]

Anyone know why this happens? I can't ssh in on 17.0

nathan@Nathans-MBP ~ % ssh root@192.168.1.133
Warning: Permanently added '192.168.1.133' (ED25519) to the list of known hosts.
root@192.168.1.133's password:
shell request failed on channel 0
nathan@Nathans-MBP ~ %```

Same stuff works on 16

[velvet path]

I’m going to be real I don’t think many people would have an idea as to why since this is a relatively new thing

[indigo kraken]

is sshd alive?

[indigo kraken]

nvm it must be if the client is being rejected

[visual meadow]

spawns when i ssh in exits when i enter password

17.0 really changed a lot lol

[timid furnace]

run with -v...

and check sshd logs

[cosmic briar]

does anyone know any reasons why SpringBoard crash reports wouldn’t be generated after crashing into safe mode? someone is facing a safe mode issue when using my tweak but they have no crash reports neither in Settings nor in Cr4shed and this has always been an issue for them apparently anytime they have a crash problem

I was thinking maybe they disabled crash reporting in icleaner or have too many logs but idk if there are other reasons

[sonic totem]

They got hacked by a nation state and the spyware disabled crash logs 🗣️

[acoustic imp]

Yea I’ve had issues like this it’s really annoying. Try having them clear logs, and restart user space. Then if that doesn’t fix it, just do a trusty old disable things then re enable group/one by one and see if it crashes

[cosmic briar]

ok thanks!

[radiant idol]

Also doing a full reboot and rejailbreaking helps sometimes

[visual meadow]

Nothing helpful here and nothing in StandardOutPath/StandardErrorPath

sshd is exiting with signal 2

[grim sparrow]

Did palera1n have any specific issues / patches?

[timid furnace]

run sshd manually with verbose flags

[visual meadow]

Issue is when I run it like that it gets further

Like it doesn't even generate the keys when its ran by launchd

But when i run keygen wrapper manually it generates the keys

[timid furnace]

edit the launchdaemon plist to add verbose flags, edit the sshd config to add verbose flags, try to get as much logging as possible

[visual meadow]

I set LogLevel to verbose and even then nothing

Tf

is roothides forkfix broken on 17

[unkempt magnet]

So you have issues running ssh because sshd running abnormaly, right?

[visual meadow]

yes

It works fine on 16 though which is odd

[velvet path]

it wouldn’t surprise me if it was

(either that or another set of entitlements are required)

[visual meadow]

bash seems to crash with this when i run it in newterm

Thread 0 name:   Dispatch queue: com.apple.main-thread
Thread 0 Crashed:
0   libsystem_kernel.dylib                   0x1f7c4075c __kill + 8
1   bash                                     0x104d9ba08 0x104d58000 + 277000
2   bash                                     0x104d9bc74 termsig_handler + 444
3   bash                                     0x104d9b56c termsig_sighandler + 144
4   libsystem_platform.dylib                 0x21a1a1888 _sigtramp + 56
5   libreadline.8.2.dylib                    0x104f2a4fc _rl_init_locale + 228
6   libreadline.8.2.dylib                    0x104f11d18 _rl_init_eightbit + 24
7   libreadline.8.2.dylib                    0x104f13f34 rl_initialize + 244
8   bash                                     0x104da52d0 initialize_readline + 920
9   bash                                     0x104dd40f0 0x104d58000 + 508144
10  bash                                     0x104dd971c 0x104d58000 + 530204
11  bash                                     0x104dd6f28 0x104d58000 + 519976
12  bash                                     0x104d98a08 yyparse + 544
13  bash                                     0x104d98744 parse_command + 372
14  bash                                     0x104d981d0 read_command + 268
15  bash                                     0x104d9841c reader_loop + 504
16  bash                                     0x104dbebf4 main + 5180
17  dyld                                     0x1d3877d44 start + 2104

but this happens in both 16 and 17

[unkempt magnet]

Try running bash with environment LC_ALL=C.UTF-8

or just C

[languid sonnet]

check DM pls

[acoustic imp]

~~i typed this and deleted it 🤦‍♂️ ~~

[magic hazel]

is jsonencoder/decoder faster than jsonserialisation

[ocean raptor]

I think bash is crashing

[ocean raptor]

roothide is broken? Who could have every possibly seen this coming

[grim sparrow]

They use the same core

If you need performance use https://github.com/michaeleisel/ZippyJSON

It’s incredible

[magic hazel]

requires new ios

already looked at it

[grim sparrow]

You can back port it so easily

It’s just a wrapper of a c library

[magic hazel]

why would it specifically be iOS 11+ though

before that it was iOS 10+

Which is rather odd

Usually it goes 8+ 9+ then 12+ or 13+

[grim sparrow]

actually idk if you could cause it uses simd

[magic hazel]

yeah right thats why

oh no

bruh

[grim sparrow]

iirc that was the minimum version that spm let you target

[magic hazel]

simd is swift

[grim sparrow]

[grim sparrow]

What no

[magic hazel]

https://developer.apple.com/documentation/swift/simd

lmao

[grim sparrow]

my brother in Christ simd refers to a technology, that’s a wrapper of the c library

[magic hazel]

its also built into the stdlib

ive used it before on ios 6

[grim sparrow]

I’m pretty sure you could get simdjson to compile for iOS 6

it’s a c library

used by fucking everything but apples jsondecoder (except in newer iOS)

[magic hazel]

oh

[grim sparrow]

jsondecoder was baked in objc for the longest time and was a lot slower than any c or c++ alternative

[magic hazel]

nvm

yeah i see what you mean

[grim sparrow]

Because objc is objectively slow as balls in comparison to c++ or c

[magic hazel]

yeah that's probably not gonna backport

1.21mb header file

[grim sparrow]

ZippyJSON is just a Swift wrapper of simdjson, which is designed to be a universal C based json library

[magic hazel]

ill try

its really really big though

they re implement json serialisation

667kb file

if that uses a bunch of ios 11+ stuff it might be cooked

[grim sparrow]

Yeah because apples is slow as balls

[magic hazel]

mm

[grim sparrow]

I don’t think it would, I’m more inclined to believe it’s because that’s the minimum you could target with spm5

[magic hazel]

compiling this will be slow as balls..

[magic hazel]

i'm sitting here targeting 7.0

all my packages do

you just get a warning

who cares about that

[grim sparrow]

Not everyone likes that

[magic hazel]

¯_(ツ)_/¯

i mean if you're doing it on purpose then you don't really care to begin with

1.2mb of text to compile is really rough

sorry 1.8

[grim sparrow]

Good thing you’re not doing it by hand

And you have a thing called a computer which can compute incredibly quickly

[magic hazel]

a 2013 mac pro

[grim sparrow]

1.8MB of source is really not a lot for most projects

[magic hazel]

i mean my project is well beyond that

but it's 1 file

[grim sparrow]

the compiler doesn’t particularly care how many files it’s in

[magic hazel]

it seems to

hangs on huge files

even with the built-in xcode compiler

wait shite its targeting 5.7

why is it targeting 5.7

grr

[grim sparrow]

It optimises on a per module basis so if you’re using it as an spm package it will treat it as a separate module and it will take longer

Compiling is a great excuse to go and do something else with your life

I’ll hit go, know it’s gonna take 5-10 minutes and go and buy a monster from the store 💀

[magic hazel]

hmmmm

[grim sparrow]

Oh I might know why

[magic hazel]

y

[grim sparrow]

Oh nvm I was gonna say so it can add visionOS as a target but it just doesn’t

lame

[magic hazel]

lmao it builds

thing is i dont actually use jsondecoder rn

i use jsonserialisation

time to see if merely having the library causes a crash

[grim sparrow]

I used to be all in on serialisation but I ended up going back to decoder

It’s just generally a lot nicer to work with in the long run

And can be more performant

(In certain situations)

[magic hazel]

Its a lot more cumbersome and it’s slower however I’m hoping this library will fix the later

It takes twice as long to get into the app when using decoder

(Unless you’re on iOS 17+)

[grim sparrow]

This should help that a lot

The funny thing is JSONDecoder is backed by JSONSerialisation, the difference is JSONDecoder will only decode the keys you’re asking for whereas JSONSerialisation will just get everything all at once

[magic hazel]

It can’t just be as good as serialisation though it has to be much better to justify it

[grim sparrow]

it’s not using objc what more do you need

[magic hazel]

Since conforming to codable is a royal pain in the ass

[slim bramble]

LOL

I didn't know that

[magic hazel]

Codable inits for big objects is like rewriting the entire class

[magic hazel]

Yeah it’s kinda goofy

I looked at the open source foundation

Pretty funny

[slim bramble]

I'd love to know how much of Swift's foundation is backed by objc

[magic hazel]

On Apple platforms? Most of it

On Linux however most of it is entirely Swift

And works mostly the same

Kinda funny

[slim bramble]

Could prolly be checked easily by analyzing the dsc

but cba to write a script

[magic hazel]

You can check by looking at swifts source straight up

The Stdlib is just an overlay of objc outside core types

Hence why it works on iOS 6…

[slim bramble]

swift is a joke

[magic hazel]

Swift is great

I would’ve quit all programming ventures on iOS if it weren’t for it

[slim bramble]

prolly

[grim sparrow]

Yeah

For objc interoperability on Apple platforms Swift uses the objc foundation types under the hood

But then on Linux it used the Swift version which has the same api and runs considerably faster

But they’ve done hella optimisation to make the Swift objc foundation work really well

If you don’t need the objc interoperability it mostly won’t use it

It will only invoke it when you try and pass it into an objc method

[slim bramble]

You know swift is a joke when only 5% Foundation's imports are actually swift

[magic hazel]

Yeah swift is held back by objc interoperability on Apple platforms

[slim bramble]

the rest is well, objc's Foundation or CF

[grim sparrow]

Why rewrite the wheel though

[magic hazel]

Swift the language exists standalone with foundation

[grim sparrow]

It works

[magic hazel]

Only on apples platforms does it use objc

[slim bramble]

it's slower

[magic hazel]

And that’s for this reason

[slim bramble]

🤷‍♂️

[magic hazel]

Then again… they already have??

They already did it for Linux

And it’s completely standalone

[grim sparrow]

But they would need to rewrite literally thousands of objc binaries

[magic hazel]

Ah true

Yeah right

Because so much bridges to objc types

Hell most UIKit does

[grim sparrow]

its so much easier for them to maintain the existing objc tech and write the new tech in swift

[magic hazel]

Scratch that basically all of UIKit does

[slim bramble]

The entirety of UIKit is objc

[grim sparrow]

Except one class that is Swift only

[slim bramble]

exceptions

[grim sparrow]

(For SwiftUI compat)

[slim bramble]

Yeah

[magic hazel]

UIViewRepresentable?

[grim sparrow]

@magic hazel you know what would be a much much much funnier project

Making an objc api for SwiftUI

[slim bramble]

LOL

[grim sparrow]

one of the internal classes it uses

[slim bramble]

Using based language for utter garbage

😭

[grim sparrow]

Why is objc based

[magic hazel]

Funny ❌ impossible/awful/terrible/against design ✅

[grim sparrow]

What’s your reasoning

I like objc but I like Swift more

what’s your logic for thinking objc is better

[slim bramble]

I don't like the "neo-languages"

Personal preference

[grim sparrow]

Fair

[magic hazel]

Yeah no fair enough

[grim sparrow]

But you can’t call Swift shit just because you don’t have a personal preference for it

You just don’t like it, doesn’t mean it’s a joke or shit language

[slim bramble]

The language maybe isn't bad, but it's implementation is

the toolchain is garbage too

🤷‍♂️

[grim sparrow]

There’s problems but objc had the same problems

[slim bramble]

At least binaries aren't 150MB

[grim sparrow]

New language features not being backportable

[grim sparrow]

This is a skill issue

[magic hazel]

Also the binaries aren’t that big what

The app might be because stdlib

[grim sparrow]

Pure Swift binaries that do the same thing as an objc thing are objectively smaller

[magic hazel]

^^^

[grim sparrow]

We can’t judge a language by shit devs who bundle half the world with their apps

[slim bramble]

binaries are static, the more libraries you import, the bigger the final binary will be

[grim sparrow]

objc has the same problems

[grim sparrow]

This is optional

[magic hazel]

Idk anyone who turns that shit on

Except like

Embedded systems?

[grim sparrow]

Google

[magic hazel]

Google is ass tho

[slim bramble]

🤨

[magic hazel]

Google out here cooking up straight shit

[grim sparrow]

it’s on by default but optional, Swift has been ABI stable since iOS 12.2 making dynamic libs possible

[slim bramble]

who would run swift on embedded systems

[magic hazel]

Surprisingly a decent amount of people

[grim sparrow]

see I think that’s insane

[slim bramble]

These people are sick

[magic hazel]

I mean if it works it works

[grim sparrow]

C is perfect for embedded systems idk why people need to keep reinventing the wheel for it

[magic hazel]

I don’t agree with it

But if it works for them then go ahead

[magic hazel]

I agree

[slim bramble]

UnsafeMutablePointer usage pov

[gentle grove]

You must use swift

[magic hazel]

Nah C definitely wins on embedded systems

[grim sparrow]

Once you know how to use it it’s fine

Swift is designed to be a safe language, pointers open you up to incredibly unsafe behaviour, what’s the big deal

[slim bramble]

The code usually looks kinda bad using this

[magic hazel]

C code looks bad, low level code looks bad

[grim sparrow]

Wrong

[grim sparrow]

this is a skill issue

[magic hazel]

Never seen good looking C code in my life

Except basic stuff

[slim bramble]

You get .pointee .makeCallable() and all those apis, that wouldn't be needed if you were using C. "neo-languages" aren't made for unsafe.

[grim sparrow]

There isn’t really ever a need for this

[slim bramble]

C code doesn't look bad at all

[grim sparrow]

People over complicate it a lot of the time for no reason

[magic hazel]

I mean looks are subjective but imo it looks incredibly ugly

[grim sparrow]

9/10 if you need to use data pointers just use [UInt8] which the compiler casts to a pointer for you at compile time

[slim bramble]

There is if you work with low level things

If I used swift, I would need that for all my projects

[grim sparrow]

var eg = [UInt8](repeating: 1, count: 10)
memset(0, eg, 10)

you can use an unsafe pointer but there is 0 need to

and there’s 0 performance benefit to

[slim bramble]

The code can still be read (and understood) by anyone...

[magic hazel]

Idek where to start w this one

[slim bramble]

How would you cast an address to it tho ?

[magic hazel]

There are 80 million different styles of C code you can write in or just make up your own on the spot

[grim sparrow]

What’s ur use case

[slim bramble]

????

[grim sparrow]

9/10 the address to it is done for you

[magic hazel]

There’s no proper guidelines as to what a C program should look like

[slim bramble]

Whatever you might wanna use an uint8_t array for ?

[grim sparrow]

[UInt8] is completely interoperable with that

that’s my point

the compiler just does it for you

[slim bramble]

There is conventions, just like in most programming language

[magic hazel]

Barely any and they’re not strict

[grim sparrow]

this is objectively not true

Clang and gcc both have strong guidelines

Lookup clang tidy

[slim bramble]

clang tidy my beloved

[grim sparrow]

9/10 clang tidy can fix my obscure error

It’s magical

[magic hazel]

Why is it that every time I look at a different C repo that the codebase is structured completely differently, each file has been laid out completely differently, hell down to the way variables are handled can be vastly vastly different in the worst cases

Meanwhile C++ doesn’t have nearly the same problem

[grim sparrow]

Because people don’t follow the convention?

[magic hazel]

Yes

[grim sparrow]

The convention exists but people don’t follow it

[magic hazel]

Because the language doesn’t enforce it

[grim sparrow]

Every single damn language has that problem

what’s your point

[slim bramble]

No language enforces it???

[grim sparrow]

You could write the worst possible swift on the planet and it would still work

[gentle grove]

rust:

[grim sparrow]

rust got overtook by asm this year it’s not real

[gentle grove]

name your variables in camel case and rustc will shoot you

[magic hazel]

That’s actually something I like about rust

[slim bramble]

Is it actually ?

[gentle grove]

Wdym

[gentle grove]

It has a built in formatter and a couple lints in the compiler by default too

[grim sparrow]

In terms of usage

[magic hazel]

I don’t like rusts syntax or stdlib at all however I find myself agreeing with its philosophy a lot sometimes

[grim sparrow]

Oct 2025 over Oct 2024

[gentle grove]

as far as I'm aware rust is the closest to having a uniform style

[slim bramble]

BASED

Assembly to the moon

[gentle grove]

Rust is by far my favorite syntax

[slim bramble]

mentally deranged statement

[magic hazel]

Rust syntax is just pure awful to look at

[slim bramble]

^

[grim sparrow]

@magic hazel have you ever ran swiftlint?

[magic hazel]

No

[gentle grove]

I don't care about aesthetics

I am a utilitarian

[slim bramble]

LOL

[grim sparrow]

so how can you know you’re following the Swift guidelines

[gentle grove]

It is very uniform and logical and autistic

[magic hazel]

I ready the wiki and follow others code bases

[slim bramble]

bad idea

[grim sparrow]

How the fuck can you be on this argument that every code base looks different because nobody follows the guidelines when you don’t yourself

[slim bramble]

yeah 😭

[grim sparrow]

I’d bet a grand you have linting errors

Default swiftlint rules

[magic hazel]

I’ll bet a grand I do too probably due to laziness but eventually I try and cleanup

I am actively aware of it though

And I don’t like it

[grim sparrow]

Then stfu about code bases not following guidelines😭 😭 😭 😭

[magic hazel]

It bothers me

Eve in myself

[gentle grove]

rootfs the type of maniac to use K&R function declarations

[magic hazel]

I’d like it if Swift got angry at me

[gentle grove]

C does have a fairly wide range of accepted styles

[grim sparrow]

FWIW I disagree with a lot of swift lints rules

but a lot of people do

There’s a reason you see temporary disable comments everywhere

[magic hazel]

Is swift lint something that runs during write time or compile time

[slim bramble]

I only do if a function has like 20000000 parameters

[grim sparrow]

I don’t know how to answer that…

[gentle grove]

All linters and formatters ate run at what I call bored time: I only run them if I am bored

[grim sparrow]

Xcode is compiling frequently in the background to index, swiftlint rules can show up then

[magic hazel]

Write time

[grim sparrow]

but write time doesn’t really exist as a time?

I think?

[magic hazel]

“Write time” is shorthand for realtime compile time

[grim sparrow]

It runs before the compiler

And generates warnings

[magic hazel]

Hm it’s for Swift 6 it seems

[magic hazel]

Can those be errors without doing all warnings as errors

[grim sparrow]

yes

[magic hazel]

Perfect

If I can find it for swift 5 I’ll start using it

[slim bramble]

@gentle grove what is the name of that style

int main(int argc, char *argv[]) {
  // Code
  return 0;
}```

[grim sparrow]

Correct

[magic hazel]

Wild

[slim bramble]

Yeah, true

[gentle grove]

Mormal

Normal

[slim bramble]

GNU Style is mentally ill

[gentle grove]

mentally I'll

[slim bramble]

like why the fuck would you indent the curly brackets

[magic hazel]

Aw man it doesn’t like you using force casts?

But those are peak lazy 😔

[gentle grove]

I mean I get it

[grim sparrow]

No because Swift style guidelines tell you not too

[magic hazel]

I know I know

[grim sparrow]

Are you telling me your codebase doesn’t follow guidelines

[magic hazel]

But when I’m testing ! Is so much faster than full optional unwrapping

[grim sparrow]

Garbage

Trash

[magic hazel]

I end up replacing ofc

[gentle grove]

all the things in gnu style basically make sense but when you combine them all together, it turns into shit

[grim sparrow]

Unreadable

[magic hazel]

Bcus ! Is awful and unsafe

And gives crashes

All the time

[grim sparrow]

It only crashes if you’re wrong

[magic hazel]

Or… you’re in a network application!

Where payloads can in fact be nil

[grim sparrow]

So you’re wrong

[magic hazel]

And that’s fine

[slim bramble]

PearDB (cc. @torn oriole) taught me to never force unwrap

[magic hazel]

What do you mean by this

[grim sparrow]

Well

! Means “I’m right and trust me”

[magic hazel]

Yes it’s quite bad practice

[grim sparrow]

Everytime it fails it means it were wrong

[magic hazel]

I guess so yes

I am wrong on behalf of the server not giving me what I expected

[grim sparrow]

I adore how this conversation has shifted from C sucks because people don’t follow the style guidelines to I don’t follow the style guidelines either but it’s ok because I know I don’t

[magic hazel]

C doesn’t suck C styling sucks

[grim sparrow]

Which is literally what every dev ever says

[gentle grove]

C
"THE" style guidelines

[gentle grove]

no

[grim sparrow]

GNU Style guidelines aren’t real

Clang is the only one that matters

[slim bramble]

Fr

[gentle grove]

Idk what clang looks like

[grim sparrow]

Correct C

[gentle grove]

Is it like the normal thing rootfs sent

[slim bramble]

Normal

[grim sparrow]

Yes

[gentle grove]

that sounds good to me

[slim bramble]

Because it is good

[gentle grove]

I assume jwi meant more like the layout and structure though

[magic hazel]

Yeah

[gentle grove]

The literal syntax is fairly similar among everything

[grim sparrow]

If you’re in c++ though it prefers

auto myfunc() -> uint8_t {}
over
uint8_t myfunc() {}

[slim bramble]

So we went from styling to layout and structure

[grim sparrow]

which I find funky

[slim bramble]

What’s gonna be the next thing, place your bets

[magic hazel]

It’s a bit of both

[gentle grove]

I thought it was supposed to be layout and structure from.the start

[grim sparrow]

Isn’t that what style guidelines are though

or at least, partly

[slim bramble]

I don’t use c++ 🔥

[gentle grove]

Style guidelines are the syntax

not like how you structure your files and includes

[magic hazel]

Not always solely

[grim sparrow]

Unfortunately Margaret thatcher introduced the class system in the UK so I’m forced to use C++

[magic hazel]

There are often naming conventions for files and file structure

[slim bramble]

😔 C OOP just too good

So I just do that

[gentle grove]

"coop" rootfs is a dei nerd

[grim sparrow]

HAHAHAHAHAHA

[slim bramble]

C++ mfs trying not to use templates for a day challenge (impossible)

[grim sparrow]

I’ve never used templates in c++

But I also don’t use c++ that much

I use c mostly but with c++ libs

[gentle grove]

c++ users attempting to understand whether their object is uninitialized or not

[slim bramble]

Because it shouldn’t exist in the first place

[magic hazel]

I’ve never used templates in any language

[gentle grove]

But what about my monomorphism

[grim sparrow]

You’ve never used protocol in Swift????

[magic hazel]

Oh I thought you meant file templates

[grim sparrow]

oh

[magic hazel]

I was confused for a second

[gentle grove]

I thought protocol was just swift equivalent of interfaces

[magic hazel]

Sry it’s like 12pm

[grim sparrow]

No c++ templates are the equivalent of protocols

[magic hazel]

Oh

[grim sparrow]

Swift doesn’t have interfaces

[magic hazel]

Why is using protocols bad 😭😭

[grim sparrow]

Protocol is fhe closest to an interface

[gentle grove]

What's the difference

[grim sparrow]

It blows in C++

[magic hazel]

Ah ic

Ngl I don’t like protocols and structs very much

Idk why

Classes make my life easier

Mutability and inheritance is nice

[gentle grove]

C++ templates are the way to implement what most languages call "generics", among other things
It is implemented by literal preprocessor text substitution

[grim sparrow]

Interfaces typically refer to only one specific class, protocols can be inherited by multiple

[magic hazel]

Yikes alr

[gentle grove]

wat

[grim sparrow]

That’s what objc treats it as

[gentle grove]

Wdym interfaces only refer to one specific class

[grim sparrow]

Like

[magic hazel]

It’s interesting how objc types are mostly all classes and then swift types are all structs

[grim sparrow]

It’s where you say what methods and variables an object would have

so in objc you would say @glossy jasper MyObject: NSObject

Somewhere in a header file

Lmao

[magic hazel]

I never liked that

[gentle grove]

Idk reading the swift page for protocol sounds exactly like an interface

[granite frigate]

)
{
is >> than ) {

[gentle grove]

Oh except that you can also declare fields

[grim sparrow]

It might be. I always see them slightly different because objc has both interfaces and protocols and they’re different

but then again objc is a special being

[gentle grove]

Maybe objc is weird

This wouldn't be what a c++ template itself does, but kinda a building block to making templates (or generics) useful

Because you need to be able to generalize wanting to have an object with "some behavior" and that's all you know about it

[wooden yarrow]

huh wait so are extern templates not possible?

[gentle grove]

Wdym

[wooden yarrow]

because then the function wouldnt know what type the generic would be

[gentle grove]

Usually all templates are in a header so if you call a templated function, it can generate one with the concrete type

[wooden yarrow]

oh

but so like say a static library right

what if you expose a public template

or better yet a dynamic library

[gentle grove]

Like how

[wooden yarrow]

just by compiling as a library and exporting the function that uses the template

[gentle grove]

You can think of a generic as describing a family of functions "foo_something" where something is some arbitrary type (with some constraint like needing to implement certain methods, which are used in the function body)

[wooden yarrow]

no yes i know what a generic is

but like

how does cpp handle them when it cant decide on the concrete type upon compilation

i.e. can cpp even make a generic interface based just off ig something like traits

[gentle grove]

Like as in a trait object in rust?

[wooden yarrow]

yeah

or something

idk what cpp does

(will reply in 15 mins brb)

[gentle grove]

Idk if I quite visualize what you're saying

Neither a c++ template nor a rust generic have any form in the compiled output, like in a shared library

So if you want to handle arbitrary types then in rust you can easily just take a Box<dyn Trait> trait object, rendering a version of the "foo" designed for that box dyn trait

In c++ it doesn't really work right because usually you don't even use a trait or interface at all, you just literally call methods. If the method names don't exist, it fails to compile. If the methods are the wrong ones but happen to have the same name, your program behaves wrong or even is undefined

But I think the abstract base class, the c++ version of the interface, then would also work as your "trait object"

Something like that

The abstract base class would have all virtual methods so you'd just replace all the pointers in your implementor object

I think its something like that

[slim bramble]

rot in hell

[gentle grove]

C++ templates does it by duck typing while real languages like Java, Rust, etc are doing it through actual type theory solving

[slim bramble]

fr

[light owl]

@native dune is it working

[shrewd smelt]

yes

[gentle grove]

It might help to re-remember how an "abstract class" (just a class where all methods are virtual/dynamic dispatch) is laid out, its just a bunch of function pointers

---

So I think the actual answer is, on the low level, the library provider would create an abstract class and publish the version of the method that uses that abstract class as the concrete type

[slim bramble]

Ever since I stopped doing java, I forgot all that stuff about inheritance 😭

[gentle grove]

Instead of creating an interface/trait and having the language create the trait object for you (with special syntax), you create that trait object yourself (the abstract class with virtual methods)

This exists in C too, a struct with function pointers as the members

You can do polymorphism in C which would probably impress some people on Twitter or something

[slim bramble]

That's what most C++ libraries do in order to create a C compatibility layer

[gentle grove]

I believe it's the same as what C++ does to talk to other C++, just with different abi

[slim bramble]

Can you stop being a compiler nerd

fr

[gentle grove]

I am not compiler nerd

That requires more credentials

[slim bramble]

How many PRs have you made to compilers ?

Whatever they might be

[gentle grove]

None unfortunately

[wooden yarrow]

huh, ok yeah that answers my question then

thanks!

[slim bramble]

Aren’t c++ templates compile time too

[gentle grove]

Yes, and so are generics, I believe always

So a template or generic doesn't exist in the compiled output, only variants of it for concrete types

[slim bramble]

Yeah obviously

[light owl]

Is she working

[gentle grove]

I mean you can use templates for anything. Here I'm just focusing on the specific case of using it to implement a generic

[shrewd smelt]

yes

[pearl sail]

My only flex in software development is that I have an unofficial PR merged in a widely used package manager

[slim bramble]

[visual meadow]

Why is it when I resign an encrypted app store app with ldid, then ct_bypass, the app opens fine on 16, but not 17? It opens fine on 17 if I just ct_bypass it. The odd thing, however is if I attach lldb to the process and then call anything and then detach it the app opens fine

something like po getgid()

[orchid fulcrum]

what does the crash report say ?

[visual meadow]

Doesn't generate a crash report weird enough

[orchid fulcrum]

does it open when you attach, don't call anything and then detach ?

[visual meadow]

no

[magic hazel]

@grim sparrow

The simdjson library is designed for 64-bit processors and it seems that you are not compiling for a known 64-bit platform. All
      fast kernels will be disabled and performance may be poor. Please use a 64-bit target such as x64, 64-bit ARM or 64-bit PPC.```

[grim sparrow]

Damn I forget ur compiling for 32bit

[magic hazel]

indeed

im assuming this will make performance far far worse

[grim sparrow]

I don’t know

I would test it

/ benchmark

I would still think a c++ impl would do better than objc

[magic hazel]

objc isn't that heavy surely

[grim sparrow]

Objc is heavy as balls

[magic hazel]

yikes

i was always under the impression that swift was the heavy one

[grim sparrow]

Nah

Swift 1 was 3x faster than objc4

[magic hazel]

Fire

[magic hazel]

Swift 1 was heavily reliant on Objective-C though

[grim sparrow]

The performance gains only improved

[magic hazel]

True

Swift has gotten much faster

I can speak to that since I've tried basically every minor version atp

[orchid fulcrum]

in what area ? maybe on runtime stuff, method calls. other than that its just c no ?

pretty sure c is faster than swift

[grim sparrow]

When did I say c?

I’m talking about objective c

Not c

[magic hazel]

C code can run in Objective-C but you'd have to be comparing Objective-C code to Swift code

[orchid fulcrum]

ik, i am saying i thought other than runtime stuff objc is just c, at least in terms of execution speed

[grim sparrow]

This

[magic hazel]

Anything not Objective-C is not comparable

Because then it's just C

And yeah, no shit, C is faster

[orchid fulcrum]

so you are comparing method calls ?

[magic hazel]

If that's what Objective-C brings to the table in your testing then that's what you're comparing

Anything that Objective-C offers over C is testable

[grim sparrow]

this is mostly what benchmarks consist of for a language

[orchid fulcrum]

thats unfair ngl, objc uses dynamic dispatch while swift doesn't

[magic hazel]

I fail to see how it's unfair when one language makes one choice and another makes another and one choice leads to better performance and the other choice doesn't

[grim sparrow]

It’s comparing language A vs language B

[magic hazel]

^

[grim sparrow]

And in this case we’re comparing the implementation of a system in Language A and Language B

And I’m saying that json decoding is faster in C++ than it would be in ObjC

[magic hazel]

It's like saying "Python has all this stuff that makes it great but it also makes it slow so it's not a fair comparison"

[grim sparrow]

Ur just missing context

[magic hazel]

I honestly didn't know Objective-C was that heavy

I would have thought it would at least be close to as fast as C considering the glaze it gets

What does Objective-C actually have going for it then

[grim sparrow]

Objc method calls themselves are slow as balls but you mostly use C and C++ in those methods

so it can balance out

[magic hazel]

But like, if you're writing an iOS app, you're using Objective-C calls

[grim sparrow]

yeah mostly

[orchid fulcrum]

you are doing it again tho 😤 , its just as fast as c/++ if you don't put your code in a method

[magic hazel]

But then you're not using Objective-C are you..

[grim sparrow]

Apple did kinda fix the dynamic dispatch problem with objc direct though

[magic hazel]

You're relying on C to give you performance

[grim sparrow]

but this was in like iOS 15?

[orchid fulcrum]

method calls isn't the only objc feature, there are blocks

[grim sparrow]

and only works if your app targets iOS 15+

[magic hazel]

Which, in my opinion, is a pretty lame option

If your language performs so poorly you need C to save you

[orchid fulcrum]

starting to feel like ragebait 😭

[magic hazel]

Yikes

So no back deploy

[grim sparrow]

What’s the difference between a C closure and objc block

[magic hazel]

Mega yikes

[slim bramble]

simd my beloved

[grim sparrow]

I only recently learnt the armv7 spec had neon

[magic hazel]

No 32 bit

[grim sparrow]

armv7 had neon but it was optional, I don’t think Apple did it?

[harsh junco]

c++)

[magic hazel]

?

What are you saying by saying C++

C++ performs very well

Generally

[orchid fulcrum]

wdym by c closure ? { //code } this, nesting functions, or some cursed function pointer stuff

[slim bramble]

I need to figure out how to properly use simd in armv8 just to support bigger ints on my assembly fibonacci calculator

[slim bramble]

(none)

[harsh junco]

you don’t really need plain c, you can use c++

[orchid fulcrum]

int a = 4; // global scope = nesting depth 0
int b = 2;
{
  int c = 2; // nesting depth 1
  void foo() {
    int a = 1; // nesting depth 2
    printf("%d %d\n", a, b);
  }
  foo();
} 

[grim sparrow]

this is what I thought

[magic hazel]

Then you're using Objective-C++

[orchid fulcrum]

doubt

[magic hazel]

Which isn't really the same because it's not like Objective-C where it's completely inline

Iirc

You have to segment your code

[orchid fulcrum]

first of all blocks are objc objects

[slim bramble]

Which holds a pointer to the function...

[magic hazel]

lmao it's so funny how much of a language is just hiding pointers

[grim sparrow]

https://fuckingblocksyntax.com

Based website

[slim bramble]

FR

I use it all the time

[grim sparrow]

you know a language is good when someone needs to buy that domain

[slim bramble]

I don't know which psychopath designed blocks but they are mentally deranged

[magic hazel]

Ngl I prefer C's syntax which feels wild to say

[grim sparrow]

So do I

[magic hazel]

Objective-C syntax is goofy asf

[grim sparrow]

It really is

[orchid fulcrum]

by this logic everything including c++ classes is c cuz all they are is structs and function pointers

[magic hazel]

It's not quite far enough removed to forget you're using a C wrapper but it's far enough that it feels like you're in an uncanny valley

[slim bramble]

That's only true for vtables

the rest is direct function call

[magic hazel]

ICL something deeply offends me about how you call methods in Objective-C and I don't know why

[grim sparrow]

Is There actually a functional difference other than methods should take blocks and functions should take closures

[harsh junco]

whoever designed smalltalk, probably https://wiki.c2.com/?SmalltalkBlocksAndClosures

[orchid fulcrum]

just use objc_msgSend

if you hate it that much

lol

[magic hazel]

That's even worse

[grim sparrow]

See that’s insane

[orchid fulcrum]

blocks are subject to retain/release, as they are objects

[slim bramble]

Next goal, make a fully functional UI tweak using only C

[grim sparrow]

That is true

[grim sparrow]

Didn’t someone do this

[magic hazel]

"Make anything in C" -> Possible -> Fun

[slim bramble]

i mean it's easily doable

[grim sparrow]

Yeah

[orchid fulcrum]

yeah its a calculator tweak made by luki

[grim sparrow]

I also saw a full app

It really isn’t that bad

[slim bramble]

You could prolly make a preprocessor which turns m file into full c code

[grim sparrow]

That’s called clang

Someone beat you to the punch

[magic hazel]

Wait are we talking C using iOS's libraries or C using your own homemade graphics library, foundation replacement, etc

[slim bramble]

objc_msgSend casts

[magic hazel]

Oh

Well then nvm

[orchid fulcrum]

https://github.com/Luki120/Cassia

[magic hazel]

Basically just using Objective-C without directly writing Objective-C code, like you do when you use Objective-C classes in Swift

[slim bramble]

Can you even emit C code out of an objective c file

[magic hazel]

Objective-C's overlay is in C, and it sits on C, why wouldn't you be able to?

[grim sparrow]

There’s probably some archived tool on GitHub

[slim bramble]

fr

[grim sparrow]

There is also the ability to just have a think about it

[orchid fulcrum]

you can't easily, there are more to it than turning method calls into objc_msgSend. declaring and defining classes etc has to use that objc4 repo code, which is c++ not c

[slim bramble]

logos already does the whole class part

just need that extra brackets parser

[magic hazel]

I remember when I first wanted to start programming I had this idea in my mind that C++ was this great abstraction over C and was a high level language, oh how wrong I was

[orchid fulcrum]

well logos is making it so that classes are added at runtime, not sure if thats what happens when you @glossy jasper a class and compile it

[slim bramble]

Classes are just memory objects

there are apis to do everything

[grim sparrow]

But wouldn’t you need objc to import the runtime api

Not objc

C++

(It’s late)

[orchid fulcrum]

objc/runtime.h is just c afaik, not ++

[magic hazel]

I wonder if Swift will ever drop Objective-C interop

[grim sparrow]

Heavily doubt

[magic hazel]

Shame

[orchid fulcrum]

god forbid

[magic hazel]

While it's useful it holds the language back

[slim bramble]

I've been using objc runtime api in c

[grim sparrow]

Why would Apple rewrite every framework in Swift

It’s a monumental waste of time

Objc direct fixes most of the performance problems

[slim bramble]

clang handles objective c, thus if you needed the runtime api, objc compilation would just be handled by clang++ which isn't the case

[magic hazel]

I mean the issue is right now they don't have to but in the future Objective-C is only going to get older and less frequently updated and as time goes on it'll be harder and harder to move

[grim sparrow]

That’s not true

They update it every year

Objc is still frequently getting optimisations

[magic hazel]

Major updates?

When was the last major update Objective-C got

[grim sparrow]

It’s all just performance updates

[magic hazel]

Which is basically just life support lmao

[grim sparrow]

but it’s all it needs

[slim bramble]

What would you change

the goal was just to bring OOP to C

The rest is just C.

[grim sparrow]

When you’re targeting modern iOS and using the modern toolchain it’s really not that bad for performance

it’s not as good as Swift but it’s not that bad

[slim bramble]

as long as you don't use Foundation or UIKit

[grim sparrow]

it’s just really not worth the hassle to rewrite literally thousands of binaries and then force probably tens of thousands of devs to rewrite

[orchid fulcrum]

from an 2019 nshipster post

[magic hazel]

Make syntax shorter, inline block literals perchance, better weak capture rules like [weak self], type safety perchance, better importing for modules, so no more header files

[slim bramble]

Would also be a waste of disk space

[magic hazel]

Better concurrency

[grim sparrow]

But these are all benefits for future programs

[slim bramble]

^

[grim sparrow]

If your program already exists who cares

They’re not problems that couldn’t be solved before

there’s no point rewriting it for the sake of rewriting it

Also weak self is not a swift thing

[slim bramble]

What do you mean by inline block syntax?

[grim sparrow]

weak captures are an arc thing

[slim bramble]

yes

[magic hazel]

It exists in Objective-C but it's not as good iirc

[slim bramble]

Importing modules is just linking and importing a header...

[grim sparrow]

clearly you’ve never really used objc

[magic hazel]

[array map:^(id obj) => obj.name];```

[slim bramble]

and you even have @import

[grim sparrow]

@import is legendary

[magic hazel]

That I didn't know about

[slim bramble]

I never use it but it exists

[grim sparrow]

I pretty much always use it now