DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2021-06-07 00:00:00Z and 2021-06-08 00:00:00Z

Learath2

oh, is there an "idiomatic" way to do global application state in rust?

The idiomatic way is to avoid it, unless u are doing microcontroller stuff

06:01

I personally rly like thiserror

06:01

I dont write many binaries so i havent used anyhow much

06:04

https://docs.rs/dashmap/4.0.2/dashmap/

API documentation for the Rust dashmap crate.

^ "Blazingly fast concurrent map in Rust."

[ddnet:master] 3 new commits

403803d Give mouse settings a maximum - Jupeyy cd68b46 Make calculation which sets mouse pos safer - Jupeyy 8bd552d Merge #3884 - bors[bot]

08:37

[ddnet/ddnet] Pull request closed: #3884 Make mouse pos calculation more robust against NaNs

@Deleted User @Chairn i answered on github but i just proposed to create teams faster

08:56

idk if we must create a new issue for what u asked or if we can put it in the same one

if it's related to the same issue comment on it

i did i did bluekitty

bluekitty

nice

https://discord.com/channels/252358080522747904/293493549758939136/850380658261622834

[ddnet-web:master] 1 new commit

4788511 Remove Tom & sirius from CHN mods - def-

Ryozuki

The idiomatic way is to avoid it, unless u are doing microcontroller stuff

I'm an old man. I like my global state no matter why you kids call it unsafe :P

Learath2

I'm an old man. I like my global state no matter why you kids call it unsafe :P

well in rust ur forced to handle the unsafeness

10:24

as u have seen

10:25

monkalaugh

10:25

thats why its a chore

10:25

to use global state

(I also ise it an an extremely specific way that where I ensure the safety)

hmm i dont understand what u saying here

10:25

u worded it a bit weirdly

10:25

xd

Heinrichs wrapping of it is quite nice but I ended up passing the state along to functions at the end

10:26

Much cleaner, very expensive but much cleanrr :P

have u checked the concurrent hashmap i sent

10:26

maybe its useful

Ryozuki

hmm i dont understand what u saying here

I use global state in a way that specifically avoids the issues that cause people to call global state unsafe

ah

I ended up not using a hashmap either. I'll deal with dynamic configuration when I get to it. Just using a struct for now

im using reactjs again, seems like the idiomatic way is to use functional components for everything now

10:28

feels weird but its way less boilerplate code

10:28

so i can see why it is like this

Very expensive to pass a struct full of strings but whatever. It looks pretty

@Learath2 what are you trying to do?

Well I have some configuration that must be shared amongst threads

look at this monkalaugh

monkalaugh

Image attachment

The configuration becomes immutable way before the threads are launched, and the threads only ever read the configuration

10:31

So in theory it should be very safe and very cheap. Heinrichs wrapping of a cell looks like it would compile down to what I expect

@Learath2 for my config i usually use https://docs.rs/config/0.11.0/config/ and it maps everything to a struct

Config organizes hierarchical or layered configurations for Rust applications.

10:32

which is usually safer and faster than a hashmap after

Do you have one instance of the config and how does it handle concurrency?

10:33

I currently have a struct that I just clone to each thread but if performance starts being an issue I was thinking of putting it in an Arc<RwLock<>>

hmm im using actix-web which provides a rly nice way

10:34

i think it wraps it in a arc

10:34

i load it

Image attachment

10:34

wrap it

Image attachment

10:34

and pass it here

Image attachment

10:34

and with some magick, i get it on my endpoint

Image attachment

10:34

xd

I'm also doing another smaller project in rust where the config is so small and threads so infrequent that I actually don't care about the clone overhead

since i dont need to modify my data i dont need a lock iirc

Ryozuki

and pass it here

Yeah I'd guess it's an Arc if that clone is cheap and safe.

my settings*

10:36

actually idk

10:36

i should look into this

10:37

i dont think it clones the data

10:37

since my struct is not Clone

Ryozuki

since i dont need to modify my data i dont need a lock iirc

You don't multiple readers is safe

yeah

10:37

if i wanted to modify it i would need to do this

Image attachment

10:37

but i dont so

But you can only have immutable borrows of your struct iirc. Otherwise rust will complain

ye

10:39

btw calling a clone on an Arc<T> is not cloning the internal data

10:39

so its always cheap iirc

Yeah. Just an atomic increment

10:39

Fairly expensive but much cheaper than memcpying an entire struct full of string data

yeah, it only clones it initially when starting the threads

10:40

i think

Ryozuki

yeah, it only clones it initially when starting the threads

I don't think so, doubt things live in thread local storage by default

10:41

An Arc should probably never cause a clone

i mean the arc clone

10:42

idk

10:42

what actix-web does is spun up x threads for each cpu core initially, thats why u clone the settings_data there

10:43

this runs 12 times for me

Image attachment

10:43

monkaS

I'm using tokio for the small discord bot I'm making

10:44

I'll probably only ever need to run it single threaded but still need to keep it thread safe

@Learath2 can u show me how u added tokio to ur toml

10:45

if u added full it uses a multi threaded runtime btw

@Deleted User (or anyone else who might know this): do you happen to know the scale of the ingame emoticons?

so u are on a multithreaded enviroment already

Hm, the docs said the full version still uses the single threaded scheduler, you sure?

where do u see this?

10:47

i used tokio full with warp and it definitly shwoed me 12 threads

10:47

i have 12 virtual cores

10:48

Image attachment

ah, I misread, yeah I guess it does use the mt scheduler

10:49

Oh btw, if any of you want to make a discord bot in rust I've already went through all the available libraries and I like twilight the best

i wanted to do one but i was always lazy

serenity is waay too idiomatic with it's frameworks

you mean it forces you to use the frameworks for commands e.g?

10:51

twilight logo looks better poggers

poggers

The way the framwork is set up with an abundance of macros is just unwieldy imo

10:52

You can indeed use it without a framework, but eeeh. I liked twilight better

yeah i definitly see why twilight looks better

10:53

starred

10:53

HN is so funny, i always see a wikipedia link there

Image attachment

10:54

This particular drive model was reported to have unusually high failure rates, approximately 5.7 times higher fail rates in comparison to other 3 TB drives.

10:54

ah

You can only skip a docker update on mac if you pay for the PRO version, LOL

13:13

I'm so glad we all embraced this obviously corporate company into our ecosystem

13:13

s/(obviously|company)//g

13:14

sudden brain damage leading to loss of english skills

Patiga

@Deleted User (or anyone else who might know this): do you happen to know the scale of the ingame emoticons?

should be same as the body

Learath2

You can only skip a docker update on mac if you pay for the PRO version, LOL

lel

any electrician here?

Heh, I asked this a couple weeks ago, so nope

14:28

But I can try to answer simple things

you can make entities invisible if you try to recolor game layers using multi selection

Image attachment

I need to fix alot about multi selection :/

but at least if you do this then everything in entities is invisible

ah thx for the reminder,tsfreddie also found out you can do this(not in editor), guess we'll just not allow it before it gets abused on some mods xd

14:33

except this is wanted

14:33

bcs even front layer etc. respect these values, kinda strange

Learath2

But I can try to answer simple things

It's about connecting my router to the upper room via an old telephone cable

14:35

It works in the one room, even gigabit connection (we have 150mbit by ISP, so our full strength gets up there), but not in the room next to it...

14:35

cables are both the same

14:36

patched it all together the same way, in the second room it caps at 93 mbit, router interface says that connection is a 100MBit/s connection and the one thats working is a 1 Gbit/s connection

14:37

And no, I will not use dlan, i am trying to get away from it D:

gbit over phone cable? I'm guessing you mean ethernet

yes, but over a telephone cable

14:42

its an old cable in the walls of the house

then you're lucky it's even getting 100 mbit/s

^^

you need some isolation, use cat5 or higher rated cables

deen

then you're lucky it's even getting 100 mbit/s

well, the other room gets 150, which is our ISP limit and i dont know why it wouldnt work in the other room

14:43

its the same cable

14:44

i know about the isolation

Ravie

but at least if you do this then everything in entities is invisible

❌

14:45

then how will i play your maps

Uhm, can you show me the cable or atleast the jack? I don't think it's possible to run gbit over the 2 pairs inside a telephone cable

louis

❌

oh yeah I wasn't able to single out front layer tho

its 8 veins

14:47

its also correctly applied to the "Netzwerkdose"?

14:47

xD

maybe cause it was the same map file that was already messed up

There are larger telephone cables with multiple pairs inside, we have the same in the multi-apartment building I'm living in

Image attachment

14:47

such a thing

so 4 pairs, rj14 I guess

it has no pairs

14:48

4 white veins and 4 colored ones, i am pretty sure the white veins arent set correctly, but i have tested that already... so idk

14:49

weird enough: fast ethernet (100 mbit) needs 4 veins (1,2,3 and 6)

14:49

for me, it just worked with 3 veins

14:49

and thats super weird

14:49

I really have no idea, i think i will need an eletrician for that

14:49

and gbit ethernet only works in the other room

oh the bugs continue xD

Image attachment

as i said, router interface shows connection between pc1 and router as 1 gbit/s and the otherroom as 100mbit/s, so that means it detects the cable strength or whatever, but its weird because the cable is identical

Anyway, @fokkonaut as far as I'm aware the devices will do autonegotiation to determine the speed to run at. iirc part of autonegotiation is also checking the line. If they don't like the SNR they might be falling back to fast ethernet

found something about that too, but the laptop i tested with couldnt select 1gbit full duplex

14:51

it wasnt listed, so i left it on auto negotiation

14:52

in the other room i got 150 also with the laptop, and it is a gigabit network adapter built in, so thats not the proiblem

fokkonaut

for me, it just worked with 3 veins

SUPER WEIRD!!

14:53

fast ethernet requires 4 veins usually

i get 250mbits with 1 vein

thats not one vein xD

1 vein download one upload

250 mbits over a single twisted pair?

yes

what is this sorcery?

didnt i post my router results once

what standard even supports it?

its this new super vectoring

Ah, VDSL with super vectoring

14:55

We are talking about ethernet here though

yeah but in the end its same physics xd

14:55

u just need the components to build that

Well very different modes of operation, very different characteristics

DSL and ethernet work quite differently

you could indeed have a DSLAM in your house

they work differently but you just need to translate it for the computers using it

14:57

in the end they can send the same data

every method of communication can send the same data "in the end"

so he needs a programmer that implements his hack xd

14:58

and ofc the hardware

Ravie

oh the bugs continue xD

did you try to give it a color animation?

@fokkonaut I couldn't really think of how to get 100baseT working over just 3 wires, maybe you have an odd device that supports 100BASE-T1 and can do 100mbit over a single pair

no just different stuff happens either on a new file or different layer combos, not sure (edited)

ahh too bad

15:02

but i think its even possible to animate the game layers xD

15:02

u just need to hack an envelope id into it

Learath2

@fokkonaut I couldn't really think of how to get 100baseT working over just 3 wires, maybe you have an odd device that supports 100BASE-T1 and can do 100mbit over a single pair

i dont think so, i connected it as the standards say

well if you connected it as the standards say you need 2 pairs to do 100base-tx so that's 4 physical cables

15:05

which one did you omit to test with 3 cables?

mh?

Fast Ethernet (100BASE-TX) runs over 4 wires, terminated on pins 1, 2, 3, 6. If you ran it over 3 wires instead you need to have omited one

fast ethernet uses one RX pair and one TX pair = 4 wires, but both wires in one pair carry the same signal (one is inverted and the second one is not), so technically it can "work" with 3 wires only

15:12

by the way, fast ethernet works just fine over telephone cable if it's not too long and there's nothing else around

it could, but I'm not sure many receivers would compensate for that well

yes

15:13

but gigabit ethernet over telephone cable.... you know, those wires are twisted for some reason

(edited)

(fwiw I had a similar thing back in my parents house, I ended up pulling a cat6 cable through by tying it to the phone cable, very very risky but it worked out with a bit of soap to lubricate

1

@Deleted User I guess if you had some tool to mess with map files you could make really funky stuff that editor doesn't allow?

yes, Patigas map tool reports, if you hit "undefined" behavior in a map, and some of these things can be abused

telephone cables must be removed from this world because they create ~~problems~~workarounds like xDSL

bcs the client doesnt check for correctness of the map

@Comrade well they give us better range than ethernet, so if they are going away we need fibre optics everywhere to replace it

myy router's telephone cable litterally goes outside so it has the shortest path from underground to router xd

15:19

for short distances inside the house it could probs still handle quite alot of bandwith

Learath2

@Comrade well they give us better range than ethernet, so if they are going away we need fibre optics everywhere to replace it

yes, xDSL should be considered as something wrong and disgusting, otherwise people get used to it and there will be no reason to put fibers everywhere

15:22

cable television is quite common in some countries, so there must be a way to do one more upgrade to fiber

in our country it was prevented bcs the CEO from Sat.1 was good friend with a politican of that time, and Sat.1 is a television company xd

15:26

"Die sozialliberale Koalition unter Helmut Schmidt hatte bereits 1981 Pläne für einen bundesweiten Glasfaserausbau beschlossen. Ein Jahr später kam Helmut Kohl an die Macht, legte die Pläne aufs Eis und förderte lieber das Kabelfernsehen." Means, we had a plan in 1981 from the SPD leader but then corrupt CDU politican joined the chat xd

15:27

welcome to 2021 the future

@Ryozuki are you familiar with macros in rust?

quite a bit

15:47

https://github.com/edg-l/formy

A rust derive macro to generate HTML forms from structs. - edg-l/formy

15:47

here is something i made

15:47

dont ask me why

15:47

xd

Can you think of a way to have non-local effects? I need to generate code elsewhere for an annotated function

elsewhere?

15:49

derive macro?

15:49

sry im in the middle of a doto ranked

15:49

you probs can do it

15:49

macros can even connect to a database to check schemas

Like I annotate a function with #[command] I need to generate a line inside another function fn register_commands to register it

uhm idk rn

15:51

u can make command create code that automatically registers tyhis function

15:51

somehow

15:51

u know like python decorations

I probably can, but something needs to call all these registerers we generated

15:52

and that something needs to know what to call, so a list of all the functions I've annotated

when i made a discord api in python i had a array of event subscribers and a decorator to easily add a function there

15:53

you can make the decorator like this

15:53

impl_event!({ pub fn your_function() {} });

15:53

brb 5 min

@Learath2 you probs need a derive macro, derive macros must be in their own crate and can only export derive macros, you probs can achieve this looking at how https://github.com/dtolnay/no-panic is implemented

dtolnay/no-panic

Attribute macro to require that the compiler prove a function can't ever panic - dtolnay/no-panic

this exact guy has a crate that does what I'm looking for using a rather nasty hack, it abuses ctor as the list of things to run

its a rly simple crate that shows how to modify a function

16:08

ah

16:09

this guy is a rust mastermind btw

16:09

Image attachment

yeah he has very interesting repos

he literally made the most famous crates

I don't think what I'm looking for is possible :/

@Learath2 why do u need to generate code inside a function rather than call a function to register that command?

16:10

hmmm

Well I'm looking to do something like we do in teeworlds, where just defining a macro and including a file registers all the commands

@Learath2 do u know u can declare functions inside functions right

I have a handful of commands now, so registering them one by one is no big trouble, but I just wanted to see if it was possible

16:12

yeah, I saw nested functions

idk rn

16:13

monkaS

me neither and I'm not crafty enough with the macros yet to achieve it

im sure thinking of a different way to architecture this is better

16:14

https://xkcd.com/974/

xkcd: The General Problem

16:14

monkalaugh

Well there is no avoiding a list of all the functions I want to register if there is no way to do non-local code generation, or compile time storage

16:17

ahahaha, someone worked around the fact that there is no compile time storage by using a file in /tmp

16:19

https://github.com/rust-lang/rust/issues/44034 meh

Crate local state for procedural macros? · Issue #44034 · rust-lang...

I'm tinkering a bit with procedural macros and encountered a problem that can be solved by keeping state in between proc macro invocations. Example from my real application: assume my proc-...

16:25

and serenity makes you use another macro to list all the annotated commands, if there was a way around it I guess they'd have implemented it

16:27

Oh I know a solution, running rust code through the C preprocessor

monkaS

@Learath2 https://github.com/mattsse/cargo-memex

mattsse/cargo-memex

compile rust code into memes. Contribute to mattsse/cargo-memex development by creating an account on GitHub.

hm, I thought rust would be much more flexible at compile time but it's still fairly rigid

@Learath2 why u learn rust btw(i mean motivation wise) or do you just want to test its limits

nim and zig are probably as good as it gets compile time wise

Deleted User

@Learath2 why u learn rust btw(i mean motivation wise) or do you just want to test its limits

Well I had a small project I wanted to implement and rust is all the hip right now

16:40

I wanted to see why so many people like it so much

i'd like to start too, but i'd prefer a non object orientated project as start, but dunno never really find anything interesting i want todo xd

so far I'm still on the fence, I like that it's safe and all but I find myself wasting far too much time thinking about edge error cases and borrow semantics

cuz u need to get more used to it

16:42

at the end u end up doing things directly in a way you know you wont have problems with the borrow

traits are not really a good implementation of OOP imo, so it's a good idea to start with some project that doesn't do OOP

16:42

Generics are among the best I've seen in a language though, by far

well i dont think rust claims to be oop

16:43

its data oriented

16:43

A trait tells the Rust compiler about functionality a particular type has and can share with other types. We can use traits to define shared behavior in an abstract way. We can use trait bounds to specify that a generic can be any type that has certain behavior.

16:43

you dont have inheritance

i'd love to see a language beeing more aggressive on range checks(like defining what range a variable can have, to e.g. prevent more logic bugs

16:44

so it forces you to handle all cases urself

The inability to specify fields inside traits is annoying, it pretty much forces the compiler to emit a virtual call

rust forces u to handle all enum cases

but u can still do out of bounds, it just panics instead of going undefined behavior

and e.g u can declare a let x and initialize it inside a if else

16:45

and it forces u to have it initialized on both

and i feel like most bugs i do are logic errors

thats not what statistics tell

16:45

monkalaugh

@Deleted User I saw an interesting RFC for rust a couple days ago for implementing ranged integer types like Ada

thats why i need personal programming language

Deleted User

and i feel like most bugs i do are logic errors

isnt this solved by those super logical languages

16:46

e.g prolog?

if they dont come in trade of usebility

16:46

i often feel like they have bad syntax or arent as powerful as like c++

whats not as powerfull?

i mean like, the things u can do in a matter of the language itself, e.g. templates

16:47

you can abuse so much stuff in c++, and i kinda like it many times

prolog is very declarative, not much you can do in it

main reason im interested in rust is, bcs it still tries todo zero cost stuff, even if some aren't like array index checking

16:48

but its still around as fast as c, thats nice

it's like C++ but safer

Deleted User

main reason im interested in rust is, bcs it still tries todo zero cost stuff, even if some aren't like array index checking

u can disable the index chewcking btw

16:49

checking*

16:50

iirc

in unsafe you can call non checking functions, cant u?

ye

like get_unchecked or smth

16:50

sure but if i'd use rust i'd prefer to minimize use of unsafe

16:50

else i could also use c++ xD

yea

16:50

i prefer it having bounds checking anyway

16:51

idk how much perfomance overhead ith as

for me its not about that

but its prob worth for 99.99% of software out there

its that the logic bug isnt prevented

16:51

if you have out of bounds, u still have a bug

16:51

doesnt matter if crash or not

16:52

i mean the "dream" would be ofc to have a closed perfect system only input/output has to be checked

16:53

everything inside the system just works ™️

i think this is too hard to achieve

16:53

given how complex stuff can get

yeah, i dunno, hard to say

but hey if they ever find a way to do it i welcome it

16:53

poggers

maybe programming gets too unusable at some point

16:54

imagine the optimization potential if the compiler has a garantuee of the flow of the program

16:55

would defs be interesting, even at the cost of slower programming

@Deleted User with rust thanks to the explicit mutability you can enable optimizations, and they try, but llvm has bugs (mutable noalias) cuz nobody ever used it (hehe c++)

16:57

so rust is making llvm improve

16:57

poggers

16:58

https://github.com/rust-lang/rust/issues/84958

Regression: Miscompilation due to bug in "mutable noalias" logic · ...

The ring test suite started failing when I tried to upgrade to untrusted 0.8.0. The regression first shipped in nightly-2021-03-23: We were further able to narrow down the regression by bisecting d...

well bugs are ok, aslong as the idea behind it isnt wrong

17:00

some day they fixed and then we've smth better

17:00

O3 in kernel: https://lore.kernel.org/lkml/CAHk-=wjuoGyxDhAF8SsrTkN0-YfCx7E6jUN3ikC_tn2AKWTTsA@mail.gmail.com/

17:00

still considered "unsafe", ok kernel also does alot of weird hacks probably

17:00

but many projects already use O3 default, e.g. SDL2 (edited)

>

I personally think -O3 in general is unsafe.

17:01

xd

17:01

yeah i think they are rly paranoid about this

17:01

but i guess rightly so

i even use ofast for my graphic driver xD

17:01

never really had a bug, and in ddnet it even helped finding undefined behavior

17:01

but yeah

17:02

i understand for kernel somehow, if it crashes it's more horrible than e.g. a user space programm

well its what he says, if no one uses, bugs wont be found

17:02

so we should all use it monkalaugh

monkalaugh

yeah, well i assume the compiler teams also test it xD

https://open.library.ubc.ca/cIRcle/collections/ubctheses/831/items/1.0079748

Can you win at TETRIS?

TETRIS is a popular video game in which you try to fill rows in a rectangular well using a sequence of tetrominoes chosen by the machine. Each time you succeed in filling a row, it is deleted from the well. Your game ends when you have stacked pieces up to the top of the well. I build a model of TETRIS and analyze the worst-case scenario, in whi...

17:49

proof that u cant win tetris

17:49

monkalaugh

Sounds like the tetris player is at a slight disadvantage

Image attachment

17:54

lmao

@Deleted User there are some languages where you can write provably safe code. but I don't know if you want to use them

18:41

check e.g. idris

18:41

https://en.wikipedia.org/wiki/Idris_(programming_language)

Idris (programming language)

Idris is a purely-functional programming language with dependent types, optional lazy evaluation, and features such as a totality checker. Idris may be used as a proof assistant, but it is designed to be a general-purpose programming language similar to Haskell. The Idris type system is similar to Agda's, and proofs are similar to Coq's, includi...

but defs sounds interesting already

I was down that road too, once ^^

18:50

you can also try coq

18:50

I ended up proving some pre-first semester math in coq instead of writing useful code

such languages would require alot of hype, so we'll see how good they actually can be

you can try languages without hype, too

yeah i mean like, that languages with these concepts are invented, until they reach a point of usefullness just like rust

it's not entirely clear that they could reach that level

18:53

e.g. I'd say that something along the lines of brainfuck can't

hype

pog

poggers

19:14

i love urbandict

Image attachment

haha

poggers

5

Exported 366 message(s)