esoteric-python

[pure dew]

it what

huh

[floral meteor]

different exception code

[pure dew]

you are on windows, yes?

oh yea

[floral meteor]

and another one

[pure dew]

hrm

[pure dew]

how did you get the advanced debug details?

[floral meteor]

control panel > System and Security > Security and Maintenance > reliablity report

you can see blue-screens through here as well

[pure dew]

cool

i wonder what the handling looks like for the oserror tho

that seems strange

definitely surprised me

[pure dew]

oh yes... "hardware problems"

i really hate windows for anything beyond gaming

[floral meteor]

some system apps segfaulting :/

[pure dew]

hmm

[floral meteor]

more dead python

[pure dew]

i wonder if this is a python feature and if its recent

i feel like i remember getting a real "segfault" on windows

but that could have been WSL too

[floral meteor]

in pycharm you can get process finished with exit code (0xc0000005)

negative a billion and something

[pure dew]

i dont use pycharm

[floral meteor]

graff

[pure dew]

why so many application failures

[glass coyote]

!projects

[night quarry]

Kindling Projects

The Kindling projects page on Ned Batchelder's website contains a list of projects and ideas programmers can tackle to build their skills and knowledge.

[snow beacon]

This channel is for writing terrible Python code. I'd suggest going to one of the help channels #❓｜how-to-get-help for good Python help.

[woven bridge]

@rugged sparrow https://www.python.org/dev/peps/pep-0659/ sounds like a lot of bytecode hacks will need to be rewritten to account for this...

assuming theres even a straightforward way to detect quickening

The only way that users will be able to detect the presence of the new interpreter is through timing execution, the use of debugging tools, or measuring memory use.
hope the part about the use of debugging tools implies there'll be some implementation details exposed

[crystal urchin]

anyone help me please ```python

e = globals()
p = 95
h = 116
u = 101
y = 108
a = e[chr(u)]
q = 97
l = [p,p,98,117,105,108,h]
l = l + e["".join([chr(i) for i in l]) + "\x69\x6e\x73\x5f\x5f"].list([105,110,115,p,p])
e = e["".join([chr(i) for i in l])]
k = getattr(a[chr(q)][chr(u)],"".join([chr(i) for i in [ 103,u,h,q,h,h,114] ]))
l = [98,68,69,67,79,68,69,70,85,83,67,65,84,79,82,56,49,54,50,51,54,55,56,48,57]
a[e.chr(97)]["".join([chr(i) for i in (l + [54])])] = a["k"]
a["".join([chr(i) for i in (a["l"] + [57,49,53])])] = a[chr(u)]
l = str()
bDECODEFUSCATOR81623678096(bDECODEFUSCATOR8162367809915,"".join(chr(i) for i in [112,114]) + "".join(chr(i) for i in ([105]+ [110,116])) )((chr(a[e.chr(97)]["y"]) + chr(int("1"*3)) + chr(a[chr(121)] )))

idk how this works

my friend sent me

it prints lol

[gritty mesa]

Anything in particular? Because I mean, it's really just a big mess that can't exactly be explained well

[crystal urchin]

but like

[cloud fossil]

i like how there is pp

[crystal urchin]

how can i make it into readable form

[gritty mesa]

You don't

Or rather, shouldn't

No matter what you do here it's still going to be a mess

Well actually

heh

[acoustic gust]

At first glance it looks like malware because malware uses methods like that to hide that it's really malware lmao

[gritty mesa]

print("lol") technically

[eager sphinx]

Make it into readable form print("lol")

Jack this is literally the second time in a row

[gritty mesa]

lol

[eager sphinx]

You're stealing my thunder here

[gritty mesa]

!e

e = globals()
p = 95
h = 116
u = 101
y = 108
a = e[chr(u)]
q = 97
l = [p,p,98,117,105,108,h]
l = l + e["".join([chr(i) for i in l]) + "\x69\x6e\x73\x5f\x5f"].list([105,110,115,p,p])
e = e["".join([chr(i) for i in l])]
k = getattr(a[chr(q)][chr(u)],"".join([chr(i) for i in [ 103,u,h,q,h,h,114] ]))
l = [98,68,69,67,79,68,69,70,85,83,67,65,84,79,82,56,49,54,50,51,54,55,56,48,57]
a[e.chr(97)]["".join([chr(i) for i in (l + [54])])] = a["k"]
a["".join([chr(i) for i in (a["l"] + [57,49,53])])] = a[chr(u)]
l = str()
bDECODEFUSCATOR81623678096(bDECODEFUSCATOR8162367809915,"".join(chr(i) for i in [112,114]) + "".join(chr(i) for i in ([105]+ [110,116])) )((chr(a[e.chr(97)]["y"]) + chr(int("1"*3)) + chr(a[chr(121)] )))

[night quarry]

@gritty mesa :white_check_mark: Your eval job has completed with return code 0.

lol

[crystal urchin]

e

but how did he do this

did u use like an obfuscator

he*

[gritty mesa]

You can make anything into a mess if you try hard enough

[crystal urchin]

uhmm

[gritty mesa]

You don't need an obfuscator, just a need to torture yourself

[acoustic gust]

Pin that sentence

[crystal urchin]

🍭

i was meant to say

:lol:

but my nitro ran off

[gritty mesa]

Take for example

!e

(lambda _, __, ___, ____, _____, ______, _______, ________:
    getattr(
        __import__(True.__class__.__name__[_] + [].__class__.__name__[__]),
        ().__class__.__eq__.__class__.__name__[:__] +
        ().__iter__().__class__.__name__[_____:________]
    )(
        _, (lambda _, __, ___: _(_, __, ___))(
            lambda _, __, ___:
                chr(___ % __) + _(_, __, ___ // __) if ___ else
                (lambda: _).func_code.co_lnotab,
            _ << ________,
            (((_____ << ____) + _) << ((___ << _____) - ___)) + (((((___ << __)
            - _) << ___) + _) << ((_____ << ____) + (_ << _))) + (((_______ <<
            __) - _) << (((((_ << ___) + _)) << ___) + (_ << _))) + (((_______
            << ___) + _) << ((_ << ______) + _)) + (((_______ << ____) - _) <<
            ((_______ << ___))) + (((_ << ____) - _) << ((((___ << __) + _) <<
            __) - _)) - (_______ << ((((___ << __) - _) << __) + _)) + (_______
            << (((((_ << ___) + _)) << __))) - ((((((_ << ___) + _)) << __) +
            _) << ((((___ << __) + _) << _))) + (((_______ << __) - _) <<
            (((((_ << ___) + _)) << _))) + (((___ << ___) + _) << ((_____ <<
            _))) + (_____ << ______) + (_ << ___)
        )
    )
)(
    *(lambda _, __, ___: _(_, __, ___))(
        (lambda _, __, ___:
            [__(___[(lambda: _).func_code.co_nlocals])] +
            _(_, __, ___[(lambda _: _).func_code.co_nlocals:]) if ___ else []
        ),
        lambda _: _.func_code.co_argcount,
        (
            lambda _: _,
            lambda _, __: _,
            lambda _, __, ___: _,
            lambda _, __, ___, ____: _,
            lambda _, __, ___, ____, _____: _,
            lambda _, __, ___, ____, _____, ______: _,
            lambda _, __, ___, ____, _____, ______, _______: _,
            lambda _, __, ___, ____, _____, ______, _______, ________: _
        )
    )
)

[night quarry]

@gritty mesa :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 25, in <module>
003 |   File "<string>", line 25, in <lambda>
004 |   File "<string>", line 27, in <lambda>
005 | AttributeError: 'function' object has no attribute 'func_code'

[gritty mesa]

nice

[crystal urchin]

lol

[crystal urchin]

so he did this all manually?

like my friend

[acoustic gust]

There's a 10000 lines code to just print hello world

[gritty mesa]

Yep

[crystal urchin]

he said it's an obfuscator but alr i see i see

[paper bloom]

omg

[crystal urchin]

so hes lying

[gritty mesa]

That's possible

[paper bloom]

I was watching the bot logs since we just merged a change to the help channels

[gritty mesa]

But unlikely

[paper bloom]

and I see this pop up

[gritty mesa]

lmao

!e

class _(metaclass=(type("__", (type,), {"__init__": lambda cls, name, bases, dct: print(f"""a{setattr(cls, '__del__', lambda self: print(f"hello world {type(self)()}"[0:11]))}{cls()}"""[0].replace('a', ''), end="")}))):...

[night quarry]

@gritty mesa :x: Your eval job has completed with return code 143 (SIGTERM).

001 | hello world
002 | hello world
003 | hello world
004 | hello world
005 | hello world
006 | hello world
007 | hello world
008 | hello world
009 | hello world
010 | hello world
011 | hello world
... (truncated - too many lines)

Full output: too long to upload

[gritty mesa]

enjoy

[crystal urchin]

o lol

[gritty mesa]

!e

from __future__ import annotations


@lambda c:c()
class __annotations__:
    def __setitem__(self, name, value):
        globals()[value] = globals()[name]


class Number:
    def __init__(self, value):
        self.value = value

    def __pos__(self):
        self.value += 0.5
        return self
    
    def __str__(self):
        return str(self.value)


let: i = Number(5)

++i
print(i)

[night quarry]

@gritty mesa :white_check_mark: Your eval job has completed with return code 0.

6.0

[gritty mesa]

This is the channel where nightmares come from

[crystal urchin]

do you know a way to do it like him?

like try to understand how it works

[acoustic gust]

!e

if 1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1-1+1 : print("Hello world")

[night quarry]

@acoustic gust :white_check_mark: Your eval job has completed with return code 0.

Hello world

[gritty mesa]

I mean, you basically just try your hardest to add as many unnecessary steps as possible

And make it an absolute mess

One sec lemme get an example

[crystal urchin]

uhmm but like

i dont see any print functions in the code

[gritty mesa]

class While:
    def __init__(self, condition, code):
        self.condition = condition
        self.code = code

    def __del__(self):
        if eval(self.condition):
            exec(self.code)
            globals().update(locals())
            type(self)(self.condition, self.code)

x = 3

While("x < 50", "x += 3")
print(x)
```For example, this is already cursed, but somewhat readable

[tropic night]

its because its not visalbe

[crystal urchin]

so how does it prints then?

[tropic night]

e = globals()
p = 95
h = 116
u = 101
y = 108
a = e[chr(u)]
q = 97
l = [p,p,98,117,105,108,h]
l = l + e["".join([chr(i) for i in l]) + "\x69\x6e\x73\x5f\x5f"].list([105,110,115,p,p])
e = e["".join([chr(i) for i in l])]
k = getattr(a[chr(q)][chr(u)],"".join([chr(i) for i in [ 103,u,h,q,h,h,114] ]))
l = [98,68,69,67,79,68,69,70,85,83,67,65,84,79,82,56,49,54,50,51,54,55,56,48,57]
a[e.chr(97)]["".join([chr(i) for i in (l + [54])])] = a["k"]
a["".join([chr(i) for i in (a["l"] + [57,49,53])])] = a[chr(u)]
l = str()

getattr(builtins, "print")('lol')

[gritty mesa]

class While:
    ___ = lambda self: ((exec(self.__), globals().update(locals()), 
          type(self)(self.___, self.__)
        ) if eval(self.___) else ...)
    def __init__(self, __, ___):
      self.___ = __; self.__ = ___

      ((_ := globals()['__builtins__'].setattr), (
          x := getattr(self.__class__, 
         (_(self.__class__, (t := "__del__"), lambda ________: (...,...)), 
       t)[1]), _(x, "__code__", While.___.__code__)))

[tropic night]

i changed the last line into what it really is

[gritty mesa]

Or you could do that

Same thing, make it a mess

[crystal urchin]

ohh i see

[tropic night]

getattr(builtins, "print")('lol') -> print('lol')

[crystal urchin]

thanks

[tropic night]

bDECODEFUSCATOR81623678096(bDECODEFUSCATOR8162367809915,"".join(chr(i) for i in [112,114]) + "".join(chr(i) for i in ([105]+ [110,116])) )((chr(a[e.chr(97)]["y"]) + chr(int("1"*3)) + chr(a[chr(121)] ))) is just convoluted for the sake of being so

it does not take many seconds to decode

[crystal urchin]

oh xd

so bDECODEFUSCATOR81623678096 = getattr?

uhmm

[tropic night]

yes

[crystal urchin]

i dont really understnad why is that so

like

i dont see the declaration of it

maybe it's hidden somewhere?

[gritty mesa]

It is

[crystal urchin]

oh

[tropic night]

thats the hidden part of it

[crystal urchin]

oh

is there a way to find it?

uhm it's kinda hard is it?

idk either

[gritty mesa]

It's meant to be

The whole purpose is to be as hard to read as possible

[crystal urchin]

hmm

[gritty mesa]

a["".join([chr(i) for i in (a["l"] + [57,49,53])])] = a[chr(u)] seems to be doign that

[crystal urchin]

oh?

[gritty mesa]

Defining the variables that is, and the line above that

a = e[chr(u)]

chr(u) -> chr(101) -> e -> the globals object

[crystal urchin]

oh

[gritty mesa]

So he's writing to globals

[crystal urchin]

uhmm so that's a way to register a variable

[gritty mesa]

!e

globals()["var"] = 5

print(var)

[night quarry]

@gritty mesa :white_check_mark: Your eval job has completed with return code 0.

5

[gritty mesa]

yep

[crystal urchin]

do you know how this work? ```python
"".join([chr(i) for i in (a["l"] + [57,49,53])])

it's kinda

messy i think

[gritty mesa]

It's just making a bunch of characters

[crystal urchin]

uhmm

[gritty mesa]

l = [97,97,98,117,105,108,116]

print("".join([chr(i) for i in (l + [57,49,53])]))

!e

l = [97,97,98,117,105,108,116]

print("".join([chr(i) for i in (l + [57,49,53])]))

[night quarry]

@gritty mesa :white_check_mark: Your eval job has completed with return code 0.

aabuilt915

[crystal urchin]

oh?

[gritty mesa]

Ah wait it's re-defined

!e

l = [98,68,69,67,79,68,69,70,85,83,67,65,84,79,82,56,49,54,50,51,54,55,56,48,57]

print("".join([chr(i) for i in (l + [57,49,53])]))

[night quarry]

@gritty mesa :white_check_mark: Your eval job has completed with return code 0.

bDECODEFUSCATOR8162367809915

[gritty mesa]

So there's that

[crystal urchin]

ohh

idk how did u even know how to do this

lol

but really cool

!e ```python
e = globals()
p = 95
h = 116
u = 101
y = 108
a = e[chr(u)]
q = 97
l = [p,p,98,117,105,108,h]
l = l + e["".join([chr(i) for i in l]) + "\x69\x6e\x73\x5f\x5f"].list([105,110,115,p,p])
e = e["".join([chr(i) for i in l])]
k = getattr(a[chr(q)][chr(u)],"".join([chr(i) for i in [ 103,u,h,q,h,h,114] ]))
l = [98,68,69,67,79,68,69,70,85,83,67,65,84,79,82,56,49,54,50,51,54,55,56,48,57]
a[e.chr(97)]["".join([chr(i) for i in (l + [54])])] = a["k"]
a["".join([chr(i) for i in (a["l"] + [57,49,53])])] = a[chr(u)]
l = str()
bDECODEFUSCATOR81623678096(bDECODEFUSCATOR8162367809915,"".join(chr(i) for i in [112,114]) + "".join(chr(i) for i in ([105]+ [110,116])) )((chr(a[e.chr(97)]["y"]) + chr(int("1"*3)) + chr(a[chr(121)] )))

[night quarry]

@crystal urchin :white_check_mark: Your eval job has completed with return code 0.

lol

[crystal urchin]

wow it actually works

[tropic night]

we use some repl tricks to figure it out or we use python debugging. or both i guess 😄

[prime surge]

lol

[midnight maple]

Here's a funny thing I discovered

You can use multiple nots in a statement

It's hard to make code blocks on mobile but

if not not not not not not not True:
     print("something")```

Pretty good for obfuscating

[outer dust]

!e

from typing import Callable, Union
class if_:
    def __init__(self, condition: bool):
        self.condition = condition
        self.done = False
    def do(self, action: Callable):
        if self.condition:
            self.done = True
            action()
class elif_:
    def __init__(self, parent: Union[if_, 'elif_'], condition: bool):
        self.parent = parent
        self.condition = condition
    def do(self, action: Callable):
        if self.condition and not self.parent.condition and not self.parent.done:
            self.parent.done = True
            action()
class else_:
    def __init__(self, parent: if_):
        self.parent = parent
    def do(self, action: Callable):
        if not self.parent.condition and not self.parent.done:
            self.parent.done = True
            action()
i = 7
a = if_((i == 5))
a.do(lambda: print("equals five"))
b = elif_(a, (i < 3))
b.do(lambda: print("is less than three"))
b = elif_(a, (i > 7))
b.do(lambda: print("is over seven"))
else_(a).do(lambda: print("didn't pass"))

[night quarry]

@outer dust :white_check_mark: Your eval job has completed with return code 0.

didn't pass

[outer dust]

https://github.com/dontmindmehere/python-if_statements/blob/main/if_statements.py

[terse mortar]

Ehh

[sick hound]

hello

[formal sandal]

in Smalltalk, booleans are objects that have methods like ifThen and there's no if statement in the language, so it's not that far off

[wanton kite]

print( "".join( [chr(i) for i in [105, 32, 108, 111, 118, 101, 32, 121, 111, 117]]))```

[outer dust]

lol

[sick hound]

!e ```python
print( "".join( [chr(i) for i in [105, 32, 108, 111, 118, 101, 32, 121, 111, 117]]))

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

i love you

[sick hound]

yea me too

[frail token]

not not x is also faster than bool(x)

[midnight maple]

oh cool

[golden finch]

This channel is nightmare fuel.

[golden finch]

How can I learn the ctypes nightmare fuel?

[rugged sparrow]

experiment with stuff

[golden finch]

Any advice on what to experiment with or where?

[rugged sparrow]

start with some of the code you see in here

and try to mess with it

and ask questions

[golden finch]

alright I'll search for ctypes

!e

import ctypes;ctypes.string_at(0)

interesting

!e

__import__('ctypes').string_at(5000)

!e

__import__('ctypes').string_at(id(1))

Interesting!

!e

print(__import__('ctypes').string_at(id(1)))

[night quarry]

@golden finch :white_check_mark: Your eval job has completed with return code 0.

b'x'

[golden finch]

Okay - where does the x come from?

Or am I just reading random chunks of memory?

Furthermore, it gives different strings on different machines!

!e

print([__import__('ctypes').string_at(id(i))for i in range(10)])

[night quarry]

@golden finch :white_check_mark: Your eval job has completed with return code 0.

[b'\xf9', b'v', b'Z', b'!', b'/', b' ', b'\x0f', b'\x0e', b'\x1f', b'\r']

[golden finch]

I think I'm reading random memory

this could be a good rng

@rugged sparrow What does any of this do?

[tribal moon]

You're reading some weird stuff in memory

[golden finch]

I was under the impression it should be reading 1

Or is it reading a PyObjectsomething

[woven bridge]

not random, the address at id(i) is the start of the memory for the int object i.

[tribal moon]

!e ```py
import ctypes
import sys

some_string_that_is_going_to_be_in_memory = "string"
size = sys.getsizeof(some_string_that_is_going_to_be_in_memory)
original_spot = id(some_string_that_is_going_to_be_in_memory)
length = len(some_string_that_is_going_to_be_in_memory)

for i in range(original_spot+size, original_spot+size-length, -1):
print(ctypes.string_at(i-2), hex(i-2))```

[woven bridge]

!e

import ctypes

for i in range(10):
    array = (ctypes.c_byte * i.__sizeof__()).from_address(id(i))
    print(list(bytes(array))[-4:])

[night quarry]

@woven bridge :white_check_mark: Your eval job has completed with return code 0.

001 | [0, 0, 0, 0]
002 | [1, 0, 0, 0]
003 | [2, 0, 0, 0]
004 | [3, 0, 0, 0]
005 | [4, 0, 0, 0]
006 | [5, 0, 0, 0]
007 | [6, 0, 0, 0]
008 | [7, 0, 0, 0]
009 | [8, 0, 0, 0]
010 | [9, 0, 0, 0]

@tribal moon :white_check_mark: Your eval job has completed with return code 0.

001 | b'g' 0x7f8fc0633f25
002 | b'ng' 0x7f8fc0633f24
003 | b'ing' 0x7f8fc0633f23
004 | b'ring' 0x7f8fc0633f22
005 | b'tring' 0x7f8fc0633f21
006 | b'string' 0x7f8fc0633f20

[tribal moon]

also now that you're here

about the operator thing

any luck?

[woven bridge]

ive been busy with other things but its still in the works. i think it's definitely doable with a smart enough injector like gdb but dunno if itll be possibleworthwhile to do it with ctypes alone.

and it would likely break for any new python version since the injection addresses would likely need to be hard coded

[tribal moon]

I still think it can be done with just ctypes

[woven bridge]

right, think we had this discussion before

possible for sure but perhaps difficult and brittle beyond the point of being worthwhile

[tribal moon]

What did you learn though?

[woven bridge]

like overall? hmm.. learned more about the internal structure of the cpython parser, a good bit about injection, how it's done on different OS's, learned about some interesting pyhon injection projects like pyrasite. also had never heard of code caves before.

i mean i'm still going though, but dont hold your breath.

[white verge]

what is esoteric python

[cloud fossil]

Golfing, Python VM languages, obfuscation, code gore and other general Python weirdness

(lambda: "esoteric python")()

[pastel ibex]

What exactly is code gore?

[earnest wing]

really ugly code

[sick hound]

ugly in being hard to read or just a mess but somehow works?

[tribal moon]

hard to read and ugly

for example look at JSON source code: ```py
#json.py

if (not skipkeys and ensure_ascii and
check_circular and allow_nan and
cls is None and indent is None and separators is None and
default is None and not sort_keys and not kw):
iterable = _default_encoder.iterencode(obj)
else:
if cls is None:
cls = JSONEncoder
iterable = cls(skipkeys=skipkeys, ensure_ascii=ensure_ascii,
check_circular=check_circular, allow_nan=allow_nan, indent=indent,
separators=separators,
default=default, sort_keys=sort_keys, **kw).iterencode(obj)```

[pure dew]

thats not really codegore tho

code gore is more like cursed code imo

[floral meteor]

The names are too long, that's what makes pep8 compliant code ugly.
As a mathematician, i like my variables are single letter, and following conventions for single letter variables

[snow beacon]

Have you ever wondered why only mathematicians read mathematical research?

[floral meteor]

Cos most current education systems are a joke, except where the mathematicians study

¯\_(ツ)_/¯

I'm still trying to find the integral of x^x

[floral meteor]

Ooh I got it, you gotta speak mathematics to read mathematics, and in order to speak mathematics you need to study mathematics

And in order to study mathematics, you need to write a lot of equations, and that's more paper friendly with single symbol operations and references

I use python like a mathematical language

It has the right notation for sets, it sorta has matrices

Tuple pairs

Arithmetic

Integrals and other symbolic mathematics takes a bit more work

[heavy ermine]

trivia: what will cause a fatal error?
1: while True: time.sleep(1)
2: def foo(): while True: print('hi') threading. Thread(target=foo).start() foo()
3: print('idk')

[floral meteor]

Pls code block the code

[heavy ermine]

1: ```py
while True:
print('hi')

2: ```py
def foo():
while True:
print('hi')
threading.Thread(target=foo).start()
foo

3: ```py
print('idk')

what will cause fatal error

[floral meteor]

And neither, if you want a fatal error you do this:

from threading import Thread as T
def morte():
  try:morte()
  except BaseException:morte()
while 1:T(target=morte, daemon=True).start()

[heavy ermine]

lel

[slim sonnet]

): formatting is bad

[heavy ermine]

🔡

⬅️

[snow beacon]

What do you mean?

[heavy ermine]

🔢 👎 0123 👍

#869634288977584158 message

finally my bot lagged

[floral meteor]

!e And if you want to fatally not raise an error on Windows,

Death = lambda*a:type(*a)()
class Obitus(metaclass=Death):
  def __del__(self):
    try:[Death(self) for _ in range(13)]
    except:self.__class__()

print("memento mori")

[night quarry]

@floral meteor :x: Your eval job timed out or ran out of memory.

memento mori

[heavy ermine]

bro why does this server have 3 PATRON ROLES? !!?

!e ```py
for i in range(40):
print('a' * i)

!e ```py
for i in range(100000):
print('a' * i)

[last locust]

Because we have 4 different levels (3 of which give a role) https://www.patreon.com/python_discord

[heavy ermine]

!e ```py
for i in range(100):
print('a' * i)

[heavy ermine]

thank you

a mountain of a

[floral meteor]

Try obfuscating the process

[heavy ermine]

!e
for i in range(100):
print('a' * i-1+2-3+2-1+3-2)

!e ```py
for i in range(100):
print('a' * (i-1+2-3+2-1+3-2))

[gusty island]

what you trying to do?

[heavy ermine]

!e
for i in range(100):
print('a' * (i-1+2-3+2-1+3-2))

[night quarry]

@heavy ermine :white_check_mark: Your eval job has completed with return code 0.

001 | 
002 | a
003 | aa
004 | aaa
005 | aaaa
006 | aaaaa
007 | aaaaaa
008 | aaaaaaa
009 | aaaaaaaa
010 | aaaaaaaaa
011 | aaaaaaaaaa
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/mayeganaya.txt?noredirect

[heavy ermine]

a mountain if a

[floral meteor]

!e ```py
def p(a, *b):
if b:
h = b.pop(0)
if b:x, y, (z,) = a, h, b
else:x,(y,),z = a,b,1
else:x,y,z = 0,a,1
while x < y:
yield x
x += z
def q(a): return'a' *a
r=(1<<7)-28
for i in p(r):
print (q(i))

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

001 | 
002 | a
003 | aa
004 | aaa
005 | aaaa
006 | aaaaa
007 | aaaaaa
008 | aaaaaaa
009 | aaaaaaaa
010 | aaaaaaaaa
011 | aaaaaaaaaa
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/uboyevibis.txt?noredirect

[floral meteor]

It needs to be valid python syntax

Newlines between statements

Or run it through a proper onelinerizor

[gusty island]

!e

from threading import Thread as T
def morte():
  try:morte()
  except BaseException:morte()
while 1:T(target=morte, daemon=True).start()

[night quarry]

@gusty island :x: Your eval job has completed with return code 139 (SIGSEGV).

001 | Fatal Python error: _Py_CheckRecursiveCall: Cannot recover from stack overflow.
002 | Python runtime state: initialized
003 | 
004 | Current thread 0x00007facc5eb8700 (most recent call first):
005 |   File "<string>", line 3 in morte
006 |   File "<string>", line 3 in morte
007 |   File "<string>", line 3 in morte
008 |   File "<string>", line 3 in morte
009 |   File "<string>", line 3 in morte
010 |   File "<string>", line 3 in morte
011 |   File "<string>", line 3 in morte
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/vebufehuye.txt?noredirect

[gusty island]

!e```

!e

Death = lambda*a:type(*a)()
class Obitus(metaclass=Death):
  def __del__(self):
    try:[Death(self) for _ in range(13)]
    except:self.__class__()

print("memento mori")

[night quarry]

@gusty island :x: Your eval job timed out or ran out of memory.

memento mori

[fickle scarab]

TIL

[gusty island]

not too surprizing

[fickle scarab]

what does this mean?

[floral meteor]

Try running it on windows

it'll take about 10 minutes

depending on the system

something small will blue screen

your graphics drivers are usually the first to go

not recommended on your personal or work pc

completely useless on Unix systems

[pure dew]

it's a singleton. None does the same

[floral meteor]

I'm still working on a kernel-panic-induction-protocol as a Linux version of blue-screen-induction-protocol

it's a bit harder to whack linux, as unlike windows it's actually an alright OS

[fickle scarab]

huh. can i create singletons or are they specially handled by the compiler interpreter?

[tribal moon]

!e print((lambda i,c:c({~0XBis(0o2)&0xFfFfor(0b01, ):-0.9j,0:i@9.7e+1} [0.e+0jand(0o00)])+c({~0XBis(0o2)&0xFfFfor(0b01, ):-0.9j,0:i@9.7e+1} [0.e+0jand(0o00)] -(0x02<<0x04))+c(~(~((0x42<<i@42.e-24) &{.2:0b0is{0.e-1j:...}}[2e-1]) <<( 0x06&0o06)>>1)*3 +0b101)+c(~(~((0x42<<i@42.e-24) &{.2:0b0is{0.e-1j:...}}[2e-1]) <<( 0x06&0o06)>>1)*3 +0b101 -(0x02<<0x04)))(type("",(object,),{"__matmul__":lambda _,__:int(__)})(),chr)) make a full on obfuscator with this design

[night quarry]

@tribal moon :white_check_mark: Your eval job has completed with return code 0.

001 | <string>:1: SyntaxWarning: "is" with a literal. Did you mean "=="?
002 | <string>:1: SyntaxWarning: "is" with a literal. Did you mean "=="?
003 | <string>:1: SyntaxWarning: "is" with a literal. Did you mean "=="?
004 | <string>:1: SyntaxWarning: "is" with a literal. Did you mean "=="?
005 | aAbB

[tribal moon]

ignore the stupid warnings

[fickle scarab]

christ

[pure dew]

sure you can make them, but enforcing the use may not be possible

[night quarry]

Hey @slim sonnet!

It looks like you tried to attach a Python file - please use a code-pasting service such as https://paste.pythondiscord.com

[heavy ermine]

!e
for i in range(100):
print('a' * (i-1+2-3+2-1+3-2))

[night quarry]

@heavy ermine :white_check_mark: Your eval job has completed with return code 0.

001 | 
002 | a
003 | aa
004 | aaa
005 | aaaa
006 | aaaaa
007 | aaaaaa
008 | aaaaaaa
009 | aaaaaaaa
010 | aaaaaaaaa
011 | aaaaaaaaaa
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/ucosilaror.txt?noredirect

[heavy ermine]

!e
for i in range(1000):
print('a' * (i-1+2-3+2-1+3-2))

[night quarry]

@heavy ermine :white_check_mark: Your eval job has completed with return code 0.

001 | 
002 | a
003 | aa
004 | aaa
005 | aaaa
006 | aaaaa
007 | aaaaaa
008 | aaaaaaa
009 | aaaaaaaa
010 | aaaaaaaaa
011 | aaaaaaaaaa
... (truncated - too many lines)

Full output: too long to upload

[heavy ermine]

!e
for i in range(999):
print('a' * (i-1+2-3+2-1+3-2))

[night quarry]

@heavy ermine :white_check_mark: Your eval job has completed with return code 0.

001 | 
002 | a
003 | aa
004 | aaa
005 | aaaa
006 | aaaaa
007 | aaaaaa
008 | aaaaaaa
009 | aaaaaaaa
010 | aaaaaaaaa
011 | aaaaaaaaaa
... (truncated - too many lines)

Full output: too long to upload

[slim sonnet]

!e

globals()[''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x5f', 16), int('0x5f', 16), int('0x62', 16), int('0x75', 16), int('0x69', 16), int('0x6c', 16), int('0x74', 16), int('0x69', 16), int('0x6e', 16), int('0x73', 16), int('0x5f', 16), int('0x5f', 16)]])].__dict__[''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x70', 16), int('0x72', 16), int('0x69', 16), int('0x6e', 16), int('0x74', 16)]])](''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x68', 16), int('0x65', 16), int('0x6c', 16), int('0x6c', 16), int('0x6f', 16), int('0x20', 16), int('0x77', 16), int('0x6f', 16), int('0x72', 16), int('0x6c', 16), int('0x64', 16)]]))

[night quarry]

@slim sonnet :white_check_mark: Your eval job has completed with return code 0.

hello world

[slim sonnet]

!e

getattr(globals()[''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x5f', 16), int('0x5f', 16), int('0x62', 16), int('0x75', 16), int('0x69', 16), int('0x6c', 16), int('0x74', 16), int('0x69', 16), int('0x6e', 16), int('0x73', 16), int('0x5f', 16), int('0x5f', 16)]])].__dict__[''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x70', 16), int('0x72', 16), int('0x69', 16), int('0x6e', 16), int('0x74', 16)]])], ''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x5f', 16), int('0x5f', 16), int('0x63', 16), int('0x61', 16), int('0x6c', 16), int('0x6c', 16), int('0x5f', 16), int('0x5f', 16)]]))(getattr(getattr(__import__, ''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x5f', 16), int('0x5f', 16), int('0x63', 16), int('0x61', 16), int('0x6c', 16), int('0x6c', 16), int('0x5f', 16), int('0x5f', 16)]]))(''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x73', 16), int('0x79', 16), int('0x73', 16)]])), ''.join([chr(x) for x in []]).join([chr(x) for x in [int('0x70', 16), int('0x61', 16), int('0x74', 16), int('0x68', 16)]])))

[night quarry]

@slim sonnet :white_check_mark: Your eval job has completed with return code 0.

['', '/snekbox/user_base/lib/python3.9/site-packages', '/usr/local/lib/python39.zip', '/usr/local/lib/python3.9', '/usr/local/lib/python3.9/lib-dynload']

[slim sonnet]

An elegant solution for printing the path
https://pastebin.com/Bxrs2dnQ

hastebin wouldn't fit it

[fickle scarab]

how so?

and how is it done for Ellipsis? or is it an interpreter / runtime env thing

[floral meteor]

!e print(....new(type(...))is...)

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

True

[floral meteor]

returns the already existing...

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

001 | None
002 | <class 'NoneType'> Ellipsis

[floral meteor]

!e ```py
class Cell:
_reg = {}
_PyObject = import('ctypes').py_object
def new(cls, victim):
if id(victim)in cls._reg:return cls._reg[id(victim)]
else:return super().new(cls)
def init(self, prisoner):
type(self)._reg |= {id(prisoner):self}
self.cell_contents = prisoner
self._content_type = self._PyObject.from_address(id(prisoner)+8).value
def hack(self, new:type):
victim = self._PyObject.from_address(id(self.cell_contents)+8)
old, victim.value = victim.value, new
return old
def del(self):
self.hack(self._content_type)
del self._reg[id(self.cell_contents)]
def repr(self):
return f"<cell({self.cell_contents!r}) at {hex(id(self.cell_contents))}>"
def hash(self):return id(self.cell_contents)

EllipsisType = Cell(...).hack(type(None))
print(...)
NoneType = Cell(...).hack(EllipsisType)
print(NoneType, ...)
print(Cell(...) is Cell(...), Cell(True)is Cell(False))
T, O = Cell(True), Cell(1)
O.hack(T.hack(int))
if 1: print(1 == True, 1)

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

001 | None
002 | <class 'NoneType'> Ellipsis
003 | True False
004 | 1 False

[floral meteor]

why tf is 1 look like False but is still truthy?

this is confusing af

[gusty island]

e! import subprocess

[floral meteor]

~

[gusty island]

e!

bashCommand = "ls"
import subprocess
process = subprocess.Popen(bashCommand.split(), stdout=subprocess.PIPE)
output, error = process.communicate()

[tribal moon]

[floral meteor]

attempting to break the bot is allowed in #bot-commands

not here

here we break python itself

!e ```py
_0='echo 69';
from subprocess import Popen as _1;
from subprocess import PIPE as _2;
_3 = _1(_0, stdout=_2);
_4 = _3.communicate()
print(*_4)

[tribal moon]

It doesn't allow that

[floral meteor]

stripped terminal?

[tribal moon]

Can't access the terminal in any way I think

[floral meteor]

!e ```py
_0='/usr/local/lib/python3.9';
from subprocess import Popen as _1;
from subprocess import PIPE as _2;
_3 = _1(_0, stdout=_2);
_4 = _3.communicate()
print(*_4)

[night quarry]

@floral meteor :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 4, in <module>
003 |   File "/usr/local/lib/python3.9/subprocess.py", line 951, in __init__
004 |     self._execute_child(args, executable, preexec_fn, close_fds,
005 |   File "/usr/local/lib/python3.9/subprocess.py", line 1821, in _execute_child
006 |     raise child_exception_type(errno_num, err_msg, err_filename)
007 | PermissionError: [Errno 13] Permission denied: '/usr/local/lib/python3.9'

[gusty island]

!e

import subprocess
subprocess.call('python')

[floral meteor]

bruh you have the ! and the e the wrong way around

and it will error 2

32512

[tribal moon]

alright then

[floral meteor]

im going to use that to obfuscate something aha

[gusty island]

!e

import subprocess
subprocess.call('python')

[floral meteor]

error 2

!e ```py
import('os').abort()

[night quarry]

@floral meteor :warning: Your eval job has completed with return code 139 (SIGSEGV).

[No output]

[gusty island]

!e

[floral meteor]

!e ```py
from signal import*
from os import abort
for i in range(2,22):
try:signal(i, lambda*a:print(*a)or 0)
except Exception:pass
abort()

[night quarry]

@floral meteor :warning: Your eval job timed out or ran out of memory.

[No output]

[dawn kayak]

!e ```py
(s := lambda: import('sys').setrecursionlimit(2**24) or s())()

[night quarry]

@dawn kayak :warning: Your eval job has completed with return code 139 (SIGSEGV).

[No output]

[gusty island]

!e

import weakref
ref = None
class Target:
  def __del__(self):
    global ref
    ref = weakref.ref(self)

def g():
  w = Target()
  w = None
  print(ref)

g()

[night quarry]

@gusty island :white_check_mark: Your eval job has completed with return code 0.

<weakref at 0x7f7195d8e3b0; dead>

[floral meteor]

that's a large id

I'm still trying to makes sense of the opcodes for just an annotation

!e not e!

okay the LOAD_CONST is actually two lines after

damn stack based bytecode

[night quarry]

@gusty island :warning: Your eval job has completed with return code 0.

[No output]

[floral meteor]

!e ```py
def Mortis():
try:Decessus
except:Mortis()
Mortis()

[night quarry]

@floral meteor :x: Your eval job has completed with return code 139 (SIGSEGV).

001 | Fatal Python error: _Py_CheckRecursiveCall: Cannot recover from stack overflow.
002 | Python runtime state: initialized
003 | 
004 | Current thread 0x00007f90cfc28740 (most recent call first):
005 |   File "<string>", line 2 in Mortis
006 |   File "<string>", line 3 in Mortis
007 |   File "<string>", line 3 in Mortis
008 |   File "<string>", line 3 in Mortis
009 |   File "<string>", line 3 in Mortis
010 |   File "<string>", line 3 in Mortis
011 |   File "<string>", line 3 in Mortis
... (truncated - too many lines)

Full output: https://paste.pythondiscord.com/jepijomizo.txt?noredirect

[floral meteor]

that's how you death it

[gusty island]

@floral meteor would there be a way to lock a resource that the python interpreter needs to access and then kill the program with the lock still in place?

!e

import os
os.system('ping kekex.cc')

[night quarry]

@gusty island :warning: Your eval job has completed with return code 0.

[No output]

[golden finch]

facts, bro - a,b,c,x,y,z,i,j,(k)

[gusty island]

!e

import os
while True:
    os.system('ping kekex.cc')

[night quarry]

@gusty island :warning: Your eval job timed out or ran out of memory.

[No output]

[golden finch]

oh lord what is this

[golden finch]

the box is sealed

[gusty island]

😭

[golden finch]

It was made by those who are Dead, and the Dead keep it, until the time comes. The box is sealed.

[gusty island]

there must be a way

[floral meteor]

print the return code

oh and it will be -1

[gusty island]

!e

import os
while True:
    os.system('ping 27.0.0.1')

[golden finch]

There is no known way. The boxes have been exhaustively tested.

[golden finch]

not gonna work - it has no network connection

[gusty island]

it must have some network device

maybe it's sealed but maybe it can be crashed

[golden finch]

I've had some experience breaking into and out of boxes before. I did not succeed.

I learnt that generally the only way to get data in is via the way it inputs code, and the only way out is how it outputs output.

I haven't looked into snekbox, but I'm guessing it's a virtual machine.

[gusty island]

well, I guess I will continue hacking the box before I make another atempt

[golden finch]

suggestion: track network traffic and see if anything goes in/out

[gusty island]

i just did that

but I guess that's too easy

[rugged sparrow]

it has no network

you also cannot start new processes (so os.system fails immediately)

[gusty island]

no sockets no os.system?

[rugged sparrow]

correct

[gusty island]

well, kinda makes sense to disable those

[rugged sparrow]

you can use os.execv however

because that does not call fork()

[tribal moon]

You can't escape the isolated environment or do anything really bad

I think Google made it and tested it

[gusty island]

thonk
there must be a better way

[golden finch]

I really doubt there's a way out.

[tribal moon]

There's always a way out

[golden finch]

Yes, there is a way out. However, it probably involves incredible amounts of voodoo and access to the source of the box and the host os.

Read: Unless you're a nation-state, it's impossible

[rugged sparrow]

Even if you root the box you'd need an 0day to get out of the vm

[golden finch]

that's my point

wait, if the box is rooted would it still filter ingoing/outgoing net requests/

my guess is yes

[rugged sparrow]

The box has no network connection at all

afaik

[golden finch]

I don't doubt you. How is input/output sent?

could you explain?

anyone know how to use __build_class__?

[pine edge]

lmao how long ago was that

uhh ping me again tomorrow i'm going to sleep

[golden finch]

sure

when is tomorrow for you?

I got a spicy error

better than a segfault imo

[pine edge]

let's say 13 hours

[golden finch]

okay - that's about the same time as tomorrow for me

What are the dunders:

__class__
__base__
__subclasses__
__name__
__module__
__builtins__

?

Specifically, the code is this:

(([x for x in (1).__class__.__base__.__subclasses__() if x.__name__ == 'catch_warnings'][0]()._module.__builtins__))['__import__']('os').system('ls -la')

I can't find documentation on object

seems to just list subclasses or smth

so then we get a catch_warnings thing

[earnest wing]

looks like a "missing __builtins__" escape

fetches the catch_warnings object, recovers __builtins__ from its _module, then does standard sandbox escape stuff

[golden finch]

yeah it is

but what is object?

[earnest wing]

It's just a plain featureless type.

The ancestor of every* object.

[golden finch]

so just type('',(),{})?

[earnest wing]

No.

[golden finch]

Or am I misunderstanding ancestor?

[earnest wing]

That would create a subclass of object.

[golden finch]

okay

so all objects inherit from object

[earnest wing]

yeah

[golden finch]

ok

and then subclasses just lists all objects

[golden finch]

the asterisk intrigues me

[earnest wing]

You can manually alter the mro and __class__

[golden finch]

mro? that's method resolution order, but I don't know what that means

and class is just what it was an instance of, right?

[earnest wing]

yes and yes

[golden finch]

so what does mro do?

[earnest wing]

method resolution order
the order that methods are resolved

[golden finch]

do you have a quick demo?

[earnest wing]

which class's methods are resolved first

in the bases

[golden finch]

bases are those things in class x(bases): which it inherits from right

[earnest wing]

no I don't have a memo I'd like to post something else in fact but it's a bit inconvenient right now

[golden finch]

but I'm guessing resolve has a definition I don't know

[earnest wing]

Unrelated, here's a challenge:
It's possible to perform a sandbox escape with the following conditions:

• Your code will be passed to exec with __builtins__ set to {}
• Your code cannot contain the " or ' characters
  What's the shortest possible program to spawn a shell and execute ls? (Assuming the latest 3.9 version)

[golden finch]

just use [*map(chr,[array of ints])]

[earnest wing]

Map and chr are missing with no builtins.

[golden finch]

oh

oh

that is not fun

!e

print((1).__class__.__base__.__subclasses__())

!e

print(len((1).__class__.__base__.__subclasses__()))

[night quarry]

@golden finch :white_check_mark: Your eval job has completed with return code 0.

114

[golden finch]

hmm

[rugged sparrow]

||(s:=().class.base.subclasses())[84].load_module(s6.decode()).system(s6)|| this works, but offsets need to be adjusted in some cases

[earnest wing]

Yeah that's pretty much as expected

I wonder if it's possible to force obtaining the str class through other means (||repr(), doc||, etc))

[rugged sparrow]

a full __builtins__ rebuild is also fun

i mean getting str is fairly easy

[earnest wing]

Another idea would be to restrict what kinds of literals are allowed :>

like, ellipsis only

I'm sure you could enforce that, too

[rugged sparrow]

parse it with ast

[earnest wing]

just visiting the ast nodes- yeah

[rugged sparrow]

_del = delattr
_imp = __import__
for i in dir(__builtins__):
    _del(__builtins__, i)

sys = _imp('sys')
del _del, _imp
sys.meta_path.clear()
sys.modules.clear()``` write some code to recover from this

[earnest wing]

well step 1 is to define __builtins__ lol

oh that was just repl

inconvenient

[golden finch]

How can I change readonly attributes? I'm sick of Python telling me I can't break things

[golden finch]

you didn't delete sys - you can use that to find the builtins

[golden finch]

what is the ().class.base.subclasses())[84] on your machine?

[rugged sparrow]

_frozen_importlib.BuiltinImporter

[glass coyote]

new logo

nive

nice*

[golden finch]

this is impressive algorithmically too

>>> s_assign('a',0,'b')
>>> x='a'
>>> x
'b'

however it opens the path for significantly more voodoo than I expected

[woven bridge]

strings use different internal layouts depending on their contents. e.g. 'uh oh: \U0002a6a5'

[golden finch]

>>> s_assign('foo',0,'bar')
>>> 'foo'
'foo'
>>> x='foo'
>>> x
'foo'

:(

How can I change builtin methods of int as a class?

[woven bridge]

forbiddenfruit module

[golden finch]

https://github.com/clarete/forbiddenfruit
this?

so how does it work?

oh lord

bit too much

[rugged sparrow]

@golden finch try fishhook it's more stable

!pypi fishhook

[night quarry]

fishhook v0.1.1

Allows for runtime hooking of static class functions

[golden finch]

what if I want to write my own?

[rugged sparrow]

Go for it

I wrote fishhook the same way

But fishhook can handle exceptions in hooks while forbiddenfruit crashes

[golden finch]

alr

[rugged sparrow]

Feel free to ping me with questions

[golden finch]

I have no idea where to start @rugged sparrow .

[rugged sparrow]

So generate_slotmap does the dynamic work of figuring out where each dunders slot is

Start there

[golden finch]

okay

also

[exec('from '+i+' import *',globals())for i in[*__import__('sys').modules]]

for no reason

oh lord it's all one file

@rugged sparrow what's a slotmap?

[rugged sparrow]

It's a mapping that says where different slot pointers need to go

Slot pointers are where c functions for Dunders go

[golden finch]

What's a mapping, and how can you make a dunder point to a c function?

oh wait

Is mapping similar to in math?

[rugged sparrow]

Oh it's just a dictionary, dunder names to locations

[golden finch]

locations being memory addresses of c functions?

[rugged sparrow]

The way that fishhook works is it takes the slot functions that exist on normal python classes and sticks them into builtin classes

Well data that can be used calculate the memory address

[golden finch]

okay

wait - normal python classes have slot functins?

[rugged sparrow]

Yea

All they do is look up the corresponding python function and call it

So by placing them in a builtin class, you can just place a python function in the class dictionary and it works

[golden finch]

so if I have the class

class foo:
  def bar(self,x):
    return x+1

what is the slotmap?

@rugged sparrow ?

[rugged sparrow]

The slot map isn't dependent on the class

It's based on how type objects look in memory

[golden finch]

type objects?

@rugged sparrow ?

Do you mean subclasses of type?

or things like int, str, etc (are those the same thing though)

[rugged sparrow]

So all classes are type objects (including type itself)

[golden finch]

okay

so type objects are things those metaclass is type?

do you have an example slotmap?

[rugged sparrow]

Slotmap shows every dunder name to its location

It's kinda big

[golden finch]

could you paste it?

(furthermore, how do you access it?)

@rugged sparrow you there?

[rugged sparrow]

You can import generate_slotmap from fishhook and call it

[golden finch]

returns a thing

looks like a dict of 3-item tuples keyed with method names

[rugged sparrow]

Yea that's the slot map

It's size of struct, base offset and secondary offset

[golden finch]

alright time to learn a bit more c

structs are just grouped variables, right?

like "classes" without methods

[rugged sparrow]

It's a set of values arranged specifically in memory

[golden finch]

it's a variable that holds several data items?

[rugged sparrow]

Yes

Sortof

But that description works for understanding fishhook

[golden finch]

ok

now base offset

[tribal moon]

It's a thing that holds some variables in some part in memory that you have to set later on, and all those variables can be accessed through the name of the structure variable you defined

[golden finch]

alright how do you set it later on?

[tribal moon]

You define it, and then set it by name, you could also set it like an array too

[golden finch]

Do you have an example?

[tribal moon]

struct Points {
  int x;
  int y;
}

struct Points points;

points.x = 4;
points.y = 3;``` something like that

[golden finch]

Okay - looks a lot like a methodless class.

[tribal moon]

in Python there's also ctypes.Structure that you can use

and I think it works if you pass it into an actual function that requires a struct

[golden finch]

ok

How would you set it like an array?

[tribal moon]

struct Points points = {4, 3};```

you could also initialize it all to 0 by just putting a 0 inside the curly braces

[golden finch]

okay

could the type of x and y differ?

[rugged sparrow]

Yes

[golden finch]

any restrictions?

[rugged sparrow]

No

[golden finch]

ok

so now what are base and secondary offsets?

[rugged sparrow]

Base specifies the offset of the struct on the type object struct, secondary specifies the location of the slot pointer on that base struct

[golden finch]

the what

offset being distance in memory

code examples might be useful if you have some

?

@rugged sparrow

[rugged sparrow]

Yea offset is distance in memory

Sorry I don't haven't code examples rn I don't have my computer

[tribal moon]

Read up on a stackoverflow article or something it explains it really well

things are usually referenced by an offset

Like for example local variable and function arguments are referenced by offset

they're referenced relative to the base pointer

[golden finch]

base pointer being?

the address of the first byte?

[rugged sparrow]

Yes

[golden finch]

So base offset is?

[rugged sparrow]

The base pointer + offset is an address inside the struct of a specific value

[golden finch]

I'm not sure what you're referencing

my c knowledge is - variables and structs exist, and pointers

code examples would really help if you could give any

[tribal moon]

Okay wait just a second

if you have something like ```c
#include <stdio.h>

struct Points {
int x;
int y;
};

int main() {
struct Points points;

points.x = 4;
points.y = 3;

printf("Address of `points`: %p, size of `points`: %d, value of `points`: %d\n", &points, sizeof(points), points);
printf("Address of `points.x` %p, size of `points.x`: %d, value of `points.x`: %d\n", &points.x, sizeof(points.x), points.x);
printf("Address of `points.x` %p, size of `points.x`: %d, value of `points.y`: %d\n", &points.y, sizeof(points.y), points.y);

}and run this this will output:
Address of points: 000000000061FE18, size of points: 8, value of points: 4
Address of points.x 000000000061FE18, size of points.x: 4, value of points.x: 4
Address of points.x 000000000061FE1C, size of points.x: 4, value of points.y: 3``` notice the addresses? See how the address of points is 0x000000000061FE18, and the value of it itself is x, that's because it points to the first starting value of the struct, where x is stored. Since x is an int, the size of it is 4 bytes, so if you wanted to access y for example, you'd go to base address which is 0x000000000061FE18, add the offset to it 000000000061FE1C (0x000000000061FE18 + 0x4) and then deference it to get the actual value. This is a really bad and rough explanation cause I'm eating right now but I'm sure you can find a better one on youtube or something

[golden finch]

reading now

[golden finch]

alright I think that makes sense

understood

[golden finch]

@rugged sparrow slotmaps?

[golden finch]

later...

[rugged sparrow]

Slotmap is just a term I coined for fishhook, because the dictionary is a mapping of the slot pointers

[golden finch]

okay

and so it is pointers to structs

maybe addr, size, val?

[pine edge]

i'm gonna be honest i don't fully know what i was on about

i THINK

i was referring to how, using the gc module, you can get all the referrers to an object

[golden finch]

referrers? things that reference an object, right?

[pine edge]

yeah

so if you and your buddy share a secret object (or constant, since the parser makes all literal constants in the same block of code the same)

[golden finch]

okay

[pine edge]

then you can use gc to find your buddy and then call methods on it

that was probably it

so hypothetically if you write two things that use the literal constant 12353823672

in the same source file

one can find the other

[golden finch]

ah

okay

gc.get_referers or something

[pine edge]

(assuming nothing has changed about the cpython's treatment of literal constants since the original post)

yeah

[golden finch]

my knowledge of gc is rusty

[pine edge]

i once did some fucked up shit where the gc module formed part of my program's logic

nobody told me weak references existed

so i like

made them from scratch

[golden finch]

weak references?

[golden finch]

well now I want to see

[pine edge]

yeah, uh, things don't get cleaned up until their reference count drops to zero OR they're in a cycle that can't be accessed from globals()/locals()/etc scopes

[golden finch]

okay

[pine edge]

so what if you want to hold a reference to something, but you want it to get deallocated if you're the only thing holding a reference

you can make a "weak reference" that doesn't count

https://docs.python.org/3/library/weakref.html

[golden finch]

Okay - why would you need that?

[pine edge]

i was implementing something SORT OF like a language

and that language needed something like a garbage collector

[golden finch]

rebuilding a lang inside python?

[pine edge]

but why should i implement a garbage collector when python can garbage collect for me

so by using weakrefs when appropriate python can clean up some of my messes automatically

[golden finch]

oh no

oh no

[pine edge]

but yeah i can't think of a good reason to use them except "i need python to free up some resources for performance reasons but i also need to know if it's still getting used"

[golden finch]

print_exception threw an exception and crashed
the cache died with it
print_exception somehow crashed again
and again
and then when calling exit() exit crashed
and then it segfaulted

we need a thread for spicy errors

[golden finch]

this is art - but I reckon you should communicate through integers in range(-5,255) to make it truly esoteric

[floral meteor]

I use [*{*some_massive_string}] and pop a character when it's needed

[golden finch]

I knew I wasn't the only one to use [*{*}]!

[golden finch]

What do you use it for?

[floral meteor]

Importless random

But it's like a runtime hash

It's random between runtimes, it's consistent in the same code block

[golden finch]

ah

that is truly awful

[floral meteor]

I guess you could probably find the seed or figure out the seeding algorithm using this

[golden finch]

Myself, I read id() or random memory

[floral meteor]

id(1) also works

But the nested stars is a convenient, short, obfuscated shuffle

But some days I prefer to boogie rather than shuffle ;)

[golden finch]

"shuffle"

Furthermore, isn't there a limited amount?

[thin trout]

I am looking for a way to remove every attribute from new classes, because that's funny IG. I thought about using type.__prepare__ = lambda *_: {} but you can't assign to type. Any idea?

[dawn kayak]

how evil of you? how evil of you that you separate attribute from their classes? what happens to their feelings?

offended feelings intensify

[woven bridge]

__builtins__.__build_class__ = ... could be an approach

[thin trout]

What is this function for?

!e

print(__builtins__.__build_class__.__doc__)

Umh

[last locust]

It's help() but it's not directly accessible

You have to monkeypatch it

[thin trout]

that's very embarrassing

Yeah, I need to manually load the site module

[night quarry]

@thin trout :white_check_mark: Your eval job has completed with return code 0.

001 | __build_class__(func, name, /, *bases, [metaclass], **kwds) -> class
002 | 
003 | Internal helper function used by the class statement.

[thin trout]

.<

!e

from dis import dis

dis(__builtins__.__build_class__) ```

[night quarry]

@thin trout :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 3, in <module>
003 |   File "/usr/local/lib/python3.9/dis.py", line 85, in dis
004 |     raise TypeError("don't know how to disassemble %s objects" %
005 | TypeError: don't know how to disassemble builtin_function_or_method objects

[woven bridge]

iirc it basically takes care of setting up a class' namespace by executing its body as a function with the locals set to the class' __dict__

the approach could then be this: you implement a new build class and pass an empty function instead of the actual function thats supposed to be used for its body

!e

build_class = __builtins__.__build_class__


def new_build_class(func, *args, **kwargs):
    return build_class(lambda *args, **kwargs: None, *args, **kwargs)

__builtins__.__build_class__ = new_build_class


class A:
    a = 1

print(A().a)

[night quarry]

@woven bridge :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 13, in <module>
003 | AttributeError: 'A' object has no attribute 'a'

[woven bridge]

@thin trout

[thin trout]

Haha niiiice

[floral meteor]

I can assign to type.
Scroll up and see either my Cell class or my hack class to reassign the class position in memory.
Don't forget to subclass the original value at the memory, which means actually fetch it first using hack(type).value

[golden finch]

That's funny! I'm working on this right now

It seems fairly simple

the basic idea is

old=__builtins__.__build_class__
__builtins__.__build_class__=lambda args:(out:=o(args),type(something with out))

[thin trout]

Interesting, is it using ctypes?

[golden finch]

@thin trout
this might work for some arcane reason

__builtins__.__build_class__=lambda*a,**b:type(a[1],(),{})

[thin trout]

That's interesting too!

[golden finch]

ah, not yet

just need to make it support kwargs

>>> __builtins__.__build_class__=lambda*a,**b:type(a[1],(),{})
>>> class a(int,foo='bar'):
    n=2
    def f(self,x):
        print(x)
        return 3

    
>>> a.n
Traceback (most recent call last):
  File "<pyshell#28>", line 1, in <module>
    a.n
AttributeError: type object 'a' has no attribute 'n'
>>> a.f(1,2)
Traceback (most recent call last):
  File "<pyshell#29>", line 1, in <module>
    a.f(1,2)
AttributeError: type object 'a' has no attribute 'f'

yeah it works

sometimes the solution without ctypes is the good solution

@thin trout

[thin trout]

Noice

[thin trout]

.bm yeet the class attribute

[golden finch]

o=__builtins__.__build_class__;__builtins__.__build_class__=lambda f,n,*bases,**kwargs:(globals().update(_=o(f,n,*bases)),type(n,bases,{i:getattr(_,i)if i[:2]=='__'else lambda*_:exec((lambda:0).__code__.replace(co_consts=()))for i in dir(_)}))[1]

class a(int,foo='bar'):
  def f(self,x):
    print(x)
    return 3

print('test')

a.f(1,2)

stdout:

test
[SIGSEGV]

[golden finch]

slightly more complex example

Is there a #esoteric-python role?

[woven bridge]

for notification purposes?

[thin trout]

Nah we don't have any

I guess a certificate of some mental illness could work

Definitely not going insane after reading this channel too much

What is this code doing though haha

[golden finch]

No, just as a general certificate of insanity using this channel.

[golden finch]

segfaults when any non-dunder methods of a class are called

[rugged sparrow]

@golden finch do you know why it segfaults?

[golden finch]

It's the ```py
exec((lambda:0).code.replace(co_consts=()))

generally, CodeObject manipulation results in an easy segfault

[rugged sparrow]

Yea but you should try to figure out why that segfaults

[golden finch]

It tries to get an element from co_consts that doesn't exist

which I presume is the source of the segfault

I stole it from codegolf stack exchange

[night quarry]

Hey @thorny shoal!

It looks like you tried to attach a Python file - please use a code-pasting service such as https://paste.pythondiscord.com

[rugged sparrow]

Yea but why does that cause a segfault? (This is a cool thing to look at because you find a way to something really cool)

[golden finch]

o=__builtins__.__build_class__
__builtins__.__build_class__=lambda a,b,*c,**d:(_:=o(a,b,*c,**d),type(b,c,{i:getattr(_,i)for i in dir(_)}|_._))[1]

I'm liking this new trick

[golden finch]

I don't know where the source for evaluating CodeObjects is.

(I haven't looked into the cpython source)

[rugged sparrow]

look at Python/ceval.c

In the cpython source

[golden finch]

it's massive

[pastel ibex]

What exactly is this trick?

[golden finch]

On class creation, it looks for the attribute _, and if so, combines it with the class' attributes on creation

i.e

!e

o=__builtins__.__build_class__
__builtins__.__build_class__=lambda a,b,*c,**d:(_:=o(a,b,*c,**d),type(b,c,{i:getattr(_,i)for i in dir(_)}|_._))[1]

class foo:
  def __init__(self,value):
    self.value=value
  _={'__repr__':lambda self:self.value}

x=foo('potato')
print(x)

[night quarry]

@golden finch :white_check_mark: Your eval job has completed with return code 0.

potato

[golden finch]

it's purely to make your code harder to debug

I am making the father of all undebuggable programs

[pastel ibex]

huh thats actually really cool

[golden finch]

I might make it segfault if you try to declare another class without it

no idea where to look in ceval.c

it's big

[rugged sparrow]

Do a find in file for TARGET(LOAD_CONST

[golden finch]

TARGET(LOAD_CONST): {
            PREDICTED(LOAD_CONST);
            PyObject *value = GETITEM(consts, oparg);
            Py_INCREF(value);
            PUSH(value);
            DISPATCH();
        }

I've been reading that bit

and looking at defs

[rugged sparrow]

Yea now look at GETITEM

[golden finch]

#ifndef Py_DEBUG
#define GETITEM(v, i) PyTuple_GET_ITEM((PyTupleObject *)(v), (i))
#else
#define GETITEM(v, i) PyTuple_GetItem((v), (i))
#endif

so

#define GETITEM(v, i) PyTuple_GetItem((v), (i))

now to PyTuple_GetItem

where is that defined?

some other file

[rugged sparrow]

Nope, Py_DEBUG isn't defined

[golden finch]

ah

🧠

ndef

missed the n

[rugged sparrow]

So you're looking for PyTuple_GET_ITEM

[golden finch]

can't find the definition either

https://github.com/python/cpython/blob/main/Include/cpython/tupleobject.h
here somewhere?

[rugged sparrow]

https://github.com/python/cpython/blob/bb3e0c240bc60fe08d332ff5955d54197f79751c/Include/cpython/tupleobject.h#L23

[night quarry]

Include/cpython/tupleobject.h line 23

#define PyTuple_GET_ITEM(op, i) (_PyTuple_CAST(op)->ob_item[i])```

[golden finch]

https://github.com/python/cpython/blob/main/Include/cpython/tupleobject.h#L23

[night quarry]

Include/cpython/tupleobject.h line 23

#define PyTuple_GET_ITEM(op, i) (_PyTuple_CAST(op)->ob_item[i])```

[golden finch]

that was almost simultaneous

[rugged sparrow]

So now we know what LOAD_CONST actually runs on the C side

[golden finch]

so now what does that c mean?

from my java background I'm guessing it's a typecast

[rugged sparrow]

It's indexing the tuples array

[golden finch]

is a tuple an array internally?

[rugged sparrow]

It has a an array inside Its struct

[golden finch]

okay

[rugged sparrow]

But the important thing is that there isn't any bounds checking happening

[golden finch]

like checking that i is in a valid range?

[rugged sparrow]

Yes

[golden finch]

mhm

[rugged sparrow]

That's why it segfaults

Because it's trying to load whatever memory is there as a python object

[golden finch]

okay - makes sense so far

now what's the cool thing?

[rugged sparrow]

It's possible to make sure the memory it tries to load is a valid python object

[golden finch]

how?

[rugged sparrow]

By altering the argument to LOAD_CONST

I challenge you to try to figure out how to do that

[golden finch]

(which is the index it takes?)

I know almost no c

[rugged sparrow]

Yes

[golden finch]

I can change the argument, sure

[rugged sparrow]

You can change the argument to a max of 255 (there are ways to get larger but I'll explain that later)

[golden finch]

I have no clue what to do, but I can change the argument

[rugged sparrow]

I can show you an example implementation with comments

[golden finch]

(lambda:0).__code__.replace(co_code='\x64\0x01')

for example

[rugged sparrow]

https://github.com/chilaxan/pysnippets/blob/main/native_ctypes/load_addr.py

[golden finch]

Would love that

no clue how to implement it myself

[rugged sparrow]

Download my implementation and mess around with it

[golden finch]

okay makes sense

so you can load_const arbitrary memory

[rugged sparrow]

Yes

Once you have figured out how this works, LOAD_FAST has the same kind of bug, try to implement this there

[golden finch]

(within the context of python's memory?)

[rugged sparrow]

Yea the os limits us to pythons address space

[golden finch]

just change the 0x64 to 0x7c, right?

[rugged sparrow]

Not quite, LOAD_FAST Is related to a different object, not the consts tuple

So look at how LOAD_FAST is implemented and try to reason what object you need to work with

[golden finch]

TARGET(LOAD_FAST): {
            PyObject *value = GETLOCAL(oparg);
            if (value == NULL) {
                format_exc_check_arg(tstate, PyExc_UnboundLocalError,
                                     UNBOUNDLOCAL_ERROR_MSG,
                                     PyTuple_GetItem(co->co_localsplusnames,
                                                     oparg));
                goto error;
            }
            Py_INCREF(value);
            PUSH(value);
            DISPATCH();
        }

uh

¯_(ツ)_/¯

[rugged sparrow]

We have an oparg getting passed to a #define again

So follow it

[golden finch]

can't find getlocal

#define GETLOCAL(i)     (localsplus[i])

[rugged sparrow]

Yup

Now find where localsplus is set

[golden finch]

PyObject **localsplus = _PyFrame_GetLocalsArray(frame);

[rugged sparrow]

We need to know where that array comes from so we get the right offset (i) from it

@golden finch I gtg for a bit, but try to figure out where localsplus is stored (and what the struct of that object looks like)

[golden finch]

okay

[pastel ibex]

I just wanna say you people are geniuses

[golden finch]

We're insane

I'm actually new here - it's fairly easy to learn

[pastel ibex]

holy cow youre new?!

[golden finch]

yeah I've only been here for maybe a week and a half

you have more messages than me in this server

[pastel ibex]

geez although im guessing you had pretty good knowledge of lower level stuff even before you joined

well i help alot in the help channels

[golden finch]

All I knew was a little bit of bytecode

I don't think anyone who spends time in #esoteric-python should touch the help channels

My bad - I've been here for ~28 days

and now I'm lost to the dark side

[golden finch]

o=__builtins__.__build_class__
__builtins__.__build_class__=lambda a,b,*c,**d:(_:=o(a,b,*c,**d),type(b,c,{i:getattr(_,i)if'_'in dir(_)else exec((lambda:0).__code__.replace(co_consts=()))for i in dir(_)}|_._))[1]

updated slightly

[pastel ibex]

Well i mean its not all that bad at least you'll know how to help with errors and complex solutions

although the massive amount of insanity may be a bit overwhelming for the people wanting help

[golden finch]

o=__builtins__.__build_class__
__builtins__.__build_class__=lambda a,b,*c,**d:(_:=o(a,b,*c,**d),type(b,c,{i:getattr(_,i)if'_'in dir(_)and type(_._)==dict else exec((lambda:0).__code__.replace(co_consts=()))for i in dir(_)}|_._))[1]

yeah just ping us if any of the help channels ask for something truly awful ig

[pastel ibex]

How did you start learning this stuff if you dont mind me asking?

[golden finch]

I really just fiddled with dir() until I found __code__, then I read the documentation and some blogs on it, and I arrived here

yeah @rugged sparrow can't find it

[rugged sparrow]

localsplus can be found on frameobjects

[golden finch]

where?

[rugged sparrow]

Inside their c struct

Look in Include/cpython/frameobject.h

[golden finch]

https://github.com/python/cpython/blob/main/Include/frameobject.h
not seeing them?

[tribal moon]

Look into the implementation

not the header file

[golden finch]

where?

[rugged sparrow]

Make sure you're looking at 3.9

[golden finch]

I have no clue

[rugged sparrow]

I can help in an hour

[golden finch]

okay

[golden finch]

@rugged sparrow

[rugged sparrow]

sorry i got swept into looking at 3.10 optimizations

[golden finch]

all good I worked on a personal project

so @rugged sparrow where is the struct?

[rugged sparrow]

the struct definition is in that file you sent a link for

but im still looking at 3.10, prob all night

i can help tomorrow

[golden finch]

ok

can't do tmrw

[rugged sparrow]

later then

[inner vector]

hi

[golden finch]

hi

[river idol]

Hello @inner vector Just wondering why you have posted 'hi' in a dozen different channels?

[inner vector]

I wan to send 50 messages

[river idol]

Ah. Please don't spam to get there.

@inner vector Why not chat in #python-discussion or #ot0-psvm’s-eternal-disapproval?

[inner vector]

oh

[river idol]

If you join VC, people do read the messages in the associated text channels.

[inner vector]

ok

thx

[golden finch]

@rugged sparrow you free now?

[rugged sparrow]

I'm about to go to bed

[golden finch]

damn

[floral meteor]

I'm not free, I am enslaved by the need to torture python until it reveals all it's secrets, and 2+2 evaluates to 5

[golden finch]

didn't we do this last week?

>>> ctypes.cast(id(4),ctypes.POINTER(ctypes.c_int))[6]=5
>>> 2+2
5

[floral meteor]

Make it a bit more inherently evaluate to 5 rather than a substitution

Make python love big brother, not just pretend to

[golden finch]

What do you mean then? Change int.__add__ somehow?

[floral meteor]

Gimme a sec I'm on phone

.

Hmm

!e ```py
class Cell:
_reg = {}
_PyObject = import('ctypes').py_object
def new(cls, victim):
if id(victim)in cls._reg:return cls._reg[id(victim)]
else:return super().new(cls)
def init(self, prisoner):
type(self)._reg |= {id(prisoner):self}
self.cell_contents = prisoner
self._content_type = self._PyObject.from_address(id(prisoner)+8).value
def hack(self, new:type):
victim = self._PyObject.from_address(id(self.cell_contents)+8)
old, victim.value = victim.value, new
return old
def del(self):
self.hack(self._content_type)
del self._reg[id(self.cell_contents)]
def repr(self):
return f"<cell({self.cell_contents!r}) at {hex(id(self.cell_contents))}>"
def hash(self):return id(self.cell_contents)

class BigBrother(int):
def add(self, other):
a=super().add
return a(a(other))-1
two = 2
Cell(two).hack(BigBrother)
print ( two + two )
Cell(two).hack(int)

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

5

[golden finch]

what the hell

so confused what that even does

also why do you name your variables like so

[snow beacon]

Probably 1984.

[golden finch]

True, but I don't recall 'hack' being mentioned anywhere in 1984. I've got a copy on my shelf, and I'm planning on rereading it, but I don't think I'll do it by hand.

Yeah, the string hack doesn't occur anywhere in the book.

[snow beacon]

I guess it's mixing its references.

No pun intended.

[golden finch]

challenge:

#your code here
try:
  raise
except:
  pass
else:
  print('Foo')

Make this code print Foo, by putting something at the top. The last line must be executed.

[snow beacon]

I'm not very good at ctypes, but the RuntimeError from having nothing to raise might be editable, or something could use the exception to hook into frames.

[sick hound]

i wonder why bare excepts are okay but bare raises arent

[woven bridge]

they are in the right context

!e

try:
  raise Exception()
except Exception:
  raise # propagates the exception

[night quarry]

@woven bridge :x: Your eval job has completed with return code 1.

001 | Traceback (most recent call last):
002 |   File "<string>", line 2, in <module>
003 | Exception

[golden finch]

but why?

[woven bridge]

sometimes you might need to do a little work after an exception occurs without making any prescriptions about what should be done with the exception

so the responsibility is passed on through propagation

[golden finch]

excellent loophole

alright you can only use official pure-python versions

[woven bridge]

!e

import dis
import sys
import ctypes

def mutable_bytes(instance: bytes):
  return memoryview(
    (ctypes.c_uint8 * len(instance)).from_address(
      id(instance) + bytes.__basicsize__ - 1
    )
  ).cast('B')

code = mutable_bytes(sys._getframe().f_code.co_code)

for i in range(0, len(code), 2):
  instruction = code[i]
  if instruction == dis.opmap['RAISE_VARARGS']:
    code[i] = dis.opmap['NOP']
      
try:
  raise
except:
  pass
else:
  print('Foo')

[night quarry]

@woven bridge :white_check_mark: Your eval job has completed with return code 0.

Foo

[golden finch]

wow

consider me impressed

that's exactly the sort of voodoo I was looking for

how does it work?

(like, you're getting the frame and editing the codeObject, but how does mutable_bytes work?)

[woven bridge]

accesses the bytecode for the module frame and rewrites the instruction for the raise to one that does nothing, aka a no op

[golden finch]

how does mutable_bytes work?

it looks like it's an iterable of references

[woven bridge]

oh, mutable_bytes

1. is taking the memory address of the bytes object, adding the boilerplate size for the bytes struct to it to get to the content memory id(instance) + bytes.__basicsize__ - 1
2. using this new address to create an array of uint8 (unsigned bytes)
3. wrapping it up in a memoryview for some extra functionality e.g. being able to slice the memory without copying
4. and finally casting the memoryview to B (unsigned 8bit int with no prescribed endianness) because otherwise you might get complaints regarding endianness when trying to assign values to its indexes.

[golden finch]

mhm

voodoo

[woven bridge]

format characters can be seen here https://docs.python.org/3/library/struct.html#format-characters

[night quarry]

@sick hound :white_check_mark: Your eval job has completed with return code 0.

Foo

[golden finch]

you could mess that up a lot more

[golden finch]

bit by bit:
what does (ctypes.c_uint8 * len(instance)) do?

okay

that's horrible syntax

list or array?

[woven bridge]

c_uint8 by itself represent the type for an individual c_uint8, whereas c_uint8 * i returns another type which represents a fixed size c_uint8 array of length i.

the right hand side is supposed to be an int, e.g. len(code) not code itself.

[golden finch]

yes, I got that

[woven bridge]

!e

import ctypes
print(ctypes.c_uint8 * 20)

[night quarry]

@woven bridge :white_check_mark: Your eval job has completed with return code 0.

<class '__main__.c_ubyte_Array_20'>

[golden finch]

so it's actually an array not a list

[woven bridge]

its a class that is used to represent a fixed size c array

[golden finch]

ok

so now what happens if you call that class's from_address on a... byteobject's memory location?

[woven bridge]

from_address is a classmethod, so it returns an instance of the class which, using usual subscript syntax, can be used to access individual items of the array starting from the given address. the class also implements the buffer protocol which is why it can be passed into memoryview.

!e

import ctypes

data = b'\x01\x02\x03\04'

array_1 = (ctypes.c_uint8 * len(data)).from_address(id(data) + bytes.__basicsize__ -1)
print(array_1[0])
print(bytes(array_1))

array_2 = (ctypes.c_uint16 * (len(data)//2)).from_address(id(data) + bytes.__basicsize__ -1)
print(array_2[0])
print(bytes(array_2))

[night quarry]

@woven bridge :white_check_mark: Your eval job has completed with return code 0.

001 | 1
002 | b'\x01\x02\x03\x04'
003 | 513
004 | b'\x01\x02\x03\x04'

[woven bridge]

ok, so here you can see the kind of abstraction these types provide when dereferencing memory. while both arrays are using the exact same data underneath, one dereferences to 8 bit ints, and the other one as 16 bit ints.

and another thing is, since theyre both sharing the same underlying memory, changes in one will apply to the other

!e

import ctypes

data = b'\x01\x02\x03\04'

array_1 = (ctypes.c_uint8 * len(data)).from_address(id(data) + bytes.__basicsize__ -1)
array_2 = (ctypes.c_uint16 * (len(data)//2)).from_address(id(data) + bytes.__basicsize__ -1)
print(data)
print(array_2[0])

array_1[0] = 255
print(data)
print(array_2[0])

array_2[0] = 0
print(data)
print(array_1[1])

[night quarry]

@woven bridge :white_check_mark: Your eval job has completed with return code 0.

001 | b'\x01\x02\x03\x04'
002 | 513
003 | b'\xff\x02\x03\x04'
004 | 767
005 | b'\x00\x00\x03\x04'
006 | 0

[woven bridge]

it's bit more involved than that

https://docs.python.org/3/c-api/typeobj.html#c.PyTypeObject.tp_basicsize

theyre closely related so its a description of both

The basic size includes the fields in the instance declared by the macro PyObject_HEAD or PyObject_VAR_HEAD (whichever is used to declare the instance struct) and this in turn includes the _ob_prev and _ob_next fields if they are present.

to try and summarize, the object's struct has some fields at the start that are always there in every instance (though strings are an exception to this). basicsize is the combined size of these always present fields

the stuff afterwards is the variable length data of the instance, each item of which takes up itemsize bytes of space

[deep heart]

hi

[woven bridge]

👋 hey

[gusty island]

people when they find this

[pastel ibex]

i might be wrong since im very new to this but it seems to be so they dont actually have to call the class

[dire yew]

nevermind, i get it now

so it basically creates the class

then initiates a class object

and discards the old class

@lambda _:_()
class A:
    pass
```is equivalent to```py
class A:
    pass
A = A()

[golden finch]

yes

[dire yew]

and of course, since it's esoteric, they have to use the underscore variable 😔

[golden finch]

Speaking of that - https://repl.it doesn't support decorators such as @lambda _:_)( - you've got to do ```py
c=lambda :()
class A:
pass

[dire yew]

oh wow, didn't know that!

[golden finch]

"it works on my machine"

[woven bridge]

it's a python version thing. decorator syntax was relaxed significantly in 3.9. replit only uses 3.8 at the latest iirc

if you try to use a lambda decorator in any python version below 3.9 it wont work, replit or not.

!pep 614

[night quarry]

**PEP 614 - Relaxing Grammar Restrictions On Decorators**

Link

Status

Final

Python-Version

3.9

Created

10-Feb-2020

Type

Standards Track

[golden finch]

huh - thanks for the @eval("lambda _:_()") loophole!

[proper vault]

I used @getattr(0, '', lambda x:x())

[pastel ibex]

wait so why does it discard the class?

oh wait nvm i see now

what would be an actual use of this tho?

[snow beacon]

Do you mean decorators, lambdas as decorators, or that particular lambda as a decorator?

[pastel ibex]

that one in particular

[snow beacon]

If you want to make a particular object, and don't care about needing its class again, then you might define the class and use the decorator on it.

[pastel ibex]

is there a way you can use the object tho from that?

[snow beacon]

The object gets stored in a variable with the name of the class.

!e ```py
@lambda :()
class x:
def add(self, other):
print(other)
x + "Hello, world!"

[night quarry]

@snow beacon :white_check_mark: Your eval job has completed with return code 0.

Hello, world!

[pastel ibex]

ohhhh

thats actually kinda cool

[floral meteor]

!e i usually end up trying to make a mutable int by using

n:0 = [0]
def mutate(a):
  n [0] += a
mutate(4)
print (n [0])

[night quarry]

@floral meteor :white_check_mark: Your eval job has completed with return code 0.

4

[snow beacon]

That's not very object-oriented.

[floral meteor]

I end up making a Cell that once it holds a value, it has heightened mutability

[tribal moon]

@woven bridge Any discoveries yet?

[floral meteor]

so i know that in memory an object is an array of memory starting at its id.
So far all I know about it is the 8th position after that id is a reference to it's class.
What are the other values?
i.e. 2, 4, 16, 24, 32, etc...

[rugged sparrow]

The first 8 bytes are the reference count of the object

Second set is the class

After that depends on the object

[unique heath]

!e

[night quarry]

Command Help

!eval [code]
Can also use: e

*Run Python code and get the results.

This command supports multiple lines of code, including code wrapped inside a formatted code
block. Code can be re-evaluated by editing the original message within 10 seconds and
clicking the reaction that subsequently appears.

We've done our best to make this sandboxed, but do let us know if you manage to find an
issue with it!*

[unique heath]

!e

import os

print(os.getcwd())

[night quarry]

@unique heath :white_check_mark: Your eval job has completed with return code 0.

/snekbox

[unique heath]

rIP

/snekbox

they were 10 steps ahead

import logging

logging.error('test')

[woven bridge]

yes that windbg is frustrating to use

[floral meteor]

SOME_CONSTANT = int(''.join(['0b',*{*'0123456789abcdef'*32}]),16)

a great start for a script, i guess?

[pastel ibex]

howd you get it to randomize like that?

[snow beacon]

Sets have random-ish order. the *{*...} unpacks the string into a set, then unpacks the set into the list containing it.

[pastel ibex]

so when you unpack something into a set the order becomes random?

thats pretty interesting makes no sense to me but interesting

[short crag]

How to call a function with >>? So like: my_func >> "a arg".

I seen someone do it