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

[ddnet-maps:master] 1 new commit

d93da7b A Lighthouse - ddnet-maps

[ddnet/ddnet] Issue closed: #3883 Bullets affected by alpha

I released I don't know whether teeworlds messages are garenteed to be in order.

13:57

realized*

some are i guess

14:05

the ones that send acks

14:05

vital packages

https://github.com/ddnet/ddnet/issues/3091#issuecomment-855398889

Client Side translation of Server Messages · Issue #3091 · ddnet/dd...

The server would send a packet instead of a chat message with a id and arguments, then the client would display the translated messages with those values. We would then need our own printf parser a...

Did I tell you I absolutely hate autocompletion... Why can't it just work or not work?

15:32

This inbetween garbage annoys me so much

you mean from mobile keyboards?

15:35

or coding? xd

Coding

15:35

@Ryozuki you have RLS working in your vim setup, right?

@TsFreddie did you do smth with mini screen btw? i am curious xd

not really, haven't got time

ah ok

rust-analyzer works better, now I only need a 24 core machine so it's remotely performant

15:48

amazing stuff

Learath2

@Ryozuki you have RLS working in your vim setup, right?

RLS?

16:19

hell no

16:19

i always used rust analyzer

16:19

RLS is on the path of being replaced entirely by rust-analyzer

16:19

its abandoned i would say

16:19

everybody uses rust-analyzer

16:19

btw, rust-analyzer is considered to be one of the best implementations of a language server

16:20

among all language servers of any language

yeah, it works much better, funny how the official extension on vscode uses rls by default and has abysmal rust-analyzer support

use the rust-analyzer extensio

yeah, that works much better, it's rather slow but that's my 4c/8t processor being not enough

16:21

need 24 threads to run modern autocompleters fluently

well for me the only slow thing is when u load the first file

16:21

it has to index all

16:21

but then its ifne

16:21

https://twitter.com/rust_analyzer/status/1399695658903293957

rust-analyzer (@rust_analyzer)

It seems like we're not just a good LSP for Rust, but a good LSP overall: We're now the highest-rated VS Code plugin for programming languages! If you'd like to help us become even better, consider sponsoring us on GitHub Sponsors: https://t.co/HtltXjLWL8

Likes

351

@Learath2 vscode still has to harness the power of treesitter highlighting, on the left normie vscode, on the right neovim with treesitter

Image attachment

16:29

the lens stuff on neovim are also way less invasive

16:30

xd

[ddnet/ddnet] Issue closed: #3821 bullets mapped from tiles are transparent in solo part

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

1. Since the config didn't have a limit set, they could be negative, which directly caused bugs 2. If you set a min mouse distance, lets say 5, It can mathematically happen that you drag your mouse exactly so much that you hit the (0, 0) point and the if(.. < MinDistance) branch normalizes that point, which causes non restorable NaNs. It's now safe, the min distance is later in code checked to be atleast 1, i "even" just made it safe enough, so there shouldn't be any change in behaviour: h...

[ddnet-discordbot:master] 1 new commit

f657f97 log global unbans - 12pm

how can i translate player color to RGBA (edited)

isn't it output as hex with an additional 2 hex digits for transparency

Alzheimer

22:43

thx

Depends on where you are looking for it, in the console if you print numbers it will tell you r g and b values

22:44

@heinrich5991 here?

22:44

or @Ryozuki, or anyone else with rust experience really

Learath2

Depends on where you are looking for it, in the console if you print numbers it will tell you r g and b values

oh wow didnt notice this was added (edited)

Image attachment

@Anime.pdf try to be conservative with your edits in this channel, they sort of spam irc

sry

@Learath2 here

@heinrich5991 How would you go about wrapping multiple error types? I want my InternalError to be able to hold either a std::env::VarError or a reqwest::Error

what are you writing? application or library?

application

22:48

but it's a module that really should be just propagating it's errors upwards

they say https://github.com/dtolnay/anyhow for applications and https://github.com/dtolnay/thiserror for libraries

Flexible concrete Error type built on std::error::Error - dtolnay/anyhow

dtolnay/thiserror

derive(Error) for struct and enum error types. Contribute to dtolnay/thiserror development by creating an account on GitHub.

so well do i need to enter hex code in ColorRGBA constructor

you can also emulate thiserror yourself with a few From implementations and an enum

cuz r, g and b as args are not working

@Anime.pdf no, ColorRGBA has multiple constructors, I don't think it has a hex one

i just get white instead of blue

22:50

whenwwseeachild

@heinrich5991 I have enum Error { InternalError, BackendError(String) } not really sure how to shove either error into an InternalError

do you want to keep the representation or not?

22:51

you probably want to wrap it, right?

I could do enum Error { VarError(std::env::VarError), ReqwestError(reqwest::Error), BackendError(String)} but I'd much prefer if I could have just one InternalError

22:51

I want to wrap it, so yeah, keep the representation

do you want to avoid a dependency on anyhow?

22:53

enum Error { Internal(anyhow::Error), Backend(String) }

Well I can use anyhow, but I'm still getting more familiar with rust so I'd prefer to also know what it does to achieve this

anyhow uses a trait object representation (like virtual base class polymorphism). we can handcraft it though

(I'm guessing it's a Box<dyn Error> underneath it all)

under anyhow, yes

22:56

enum Error { Internal(Box<dyn Error>), Backend(String) }
// or
enum Error { Internal(InternalError), Backend(String) }
enum InternalError { Var(std::env::VarError), Reqwest(reqwest::Error) }

hm, we should be able to avoid this kind of dynamic polymorphism though, since all my error types are known in advance, no?

that's the latter representation above

Ah, someone should do a rfc for anonymous enums

then do

impl From<std::env::VarError> for Error {
    fn from(e: std::env::VarError) -> Error {
        Error::Internal(InternalError::Var(e))
    }
}

22:58

and the same for the reqwest error

22:58

that enables you to use the ? operator

Yeah already had the froms there, I was just missing the second enum

22:58

Q: How costly is rusts runtime polymorphism? About around C++?

the runtime polymorphism works quite similar, except that the vtable pointer is stored next to the pointer instead of inside the pointed-to object

23:00

i.e. Box<dyn std::error::Error> is two pointers wide

23:00

This emoji doesn't work here because it's from a different server. Discord Nitro can solve all of that, check User Settings > Nitro for details

23:01

I'll have to use a Box<dyn Error> anyway, apparently reqwest errors don't implement the trait Clone

I see no reason why boxing would help with that issue

23:03

thiserror helps with the enum boilerplate btw, and anyhow with the trait object boilerplate

Hm, I guess I could drop the Clone trait from my own error type

(a boxed reqwest error is still unclonable)

If you want to propagate the errors upwards, you probably also want to match them, right?

23:05

is matching possible with anyhow?

not really (a little bit)

23:06

but in practice I never ended up matching on "what library did this error originate from" (for libtw2 at least)

It's nice that this language makes you think about errors but sometimes you just want to get on with it

Patiga

If you want to propagate the errors upwards, you probably also want to match them, right?

ah I suppose the assumption here might be the problem then ^^

@Learath2 take anyhow then

I'll do the two enums, I don't need to change my code then, just need to rename a couple things

it's really nice. it gives you the possibility to attach context

I need to match the errors outside, is that possible with anyhow?

not really

23:07

(you could downcast them, but it's ugly)

23:08

can you explain why it's interesting whether you get a VarError or a reqwest error from the outside?

A VarError I can't recover from (but I'm not really sure how to cleanly die inside an async), A Reqwest error I'll just exponentially backoff

23:13

You know what, if I knew what was good for me I'd just extract all the vars at the start

sounds good

23:14

rust led me to overcomplicate stuff before, because it's suddenly in the "possible" range

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

do it as simple as possible to get it to work. is my new motto

23:14

do you want it thread-local or global?

global

23:15

So a Arc to a hashmap is what I was thinking

I think a lazy_static can have a mutex

23:15

a normal static can't because rust relies on posix mutexes which cannot be constructed at compile time

I don't need a mutex, it won't change at runtime

23:16

(well not after startup)

https://docs.rs/once_cell/1.7.2/once_cell/sync/struct.Lazy.html

once_cell::sync::Lazy - Rust

A value which is initialized on the first access.

23:17

this then, I guess?

23:18

it'll get into the standard library at some point

23:18

ah, you probably need parameters as well

23:18

https://docs.rs/once_cell/1.7.2/once_cell/sync/struct.OnceCell.html

once_cell::sync::OnceCell - Rust

A thread-safe cell which can be written to only once.

23:18

this then

23:18

call get_or_init to initialize it once

such a complex thing, what does it provide even?

oncecell?

yeah

thread-safe initialization

23:21

that the initialization func will only get called once

sounds like something I don't need, at startup I have a single thread, I need to populate a hashmap from some config and some env, then all the threads just read from it

https://news.ycombinator.com/item?id=23920945

heinrich5991

You can do that in Rust by creating your own abstraction around UnsafeCell. Say InitializeCell: use std::cell::UnsafeCell; use std::mem::MaybeUninit; use std::ops; pub struct InitializeCell<T> { inner: UnsafeCell<MaybeUninit<T>>, } unsafe impl<T> Sync for InitializeCell<T> {} impl<T> Initiali...

23:23

does that suit you? ^^

23:24

you can have the code under any license

23:25

fun, apparently I went through the same exchange with that person as with you

Still looks quite complex, hm I'm going too fast again, concurrency is annoying enough in C++

okay, I wouldn't consider this complex

23:29

the stuff compiles down to memcpy

I mean complex for the trivial thing I'm doing, I can read it if I squint really hard

so much stuff I could suggest. but I wanna go to bed :/

go ahead, I'll figure something out

good night

@heinrich5991 if you didn't go yet what's wrong with just a Arc<HashMap<K, V>>?

that was actually my next suggestion

23:34

*that would actually have been

23:34

I think the quic library from cloudflare does it

23:34

it's an atomic instruction per clone, can be fine

23:34

probably will be fine

Ah I also understand what you did now, cute

Exported 150 message(s)