DDraceNetwork - developer

github codespaces aren't half bad when you need to edit quickly on a computer you don't usually code on

deleting a branch doesn't require a force push

ddnet-maps

[ddnet-maps:master] 1 new commit

29a0d97 A 2 Days in the back - ddnet-maps

Voxel

so i read both this and got the plugin, but when i build my client and try to test my changes in a map the protocol isnt detected as TW (edited)

i got this fixed, but now i dont think my packet is sending

05:52

i'm trying to send 1 sv_kill_msg_teams in here, but its not going through

05:52

i'm in a public server because LANs are not detected

Jupstar ✪

If you struggle with something specific just name it

nevermind i finally got it working... somewhat

https://blog.rust-lang.org/2022/12/15/Rust-1.66.0.html

Announcing Rust 1.66.0 | Rust Blog

Empowering everyone to build reliable and efficient software.

ryozoozki

headbot

https://bytecodealliance.org/articles/cranelift-progress-2022

Cranelift Progress in 2022

Continuing the tradition of a year-end progress report from last year, we are excited to report that this year has also been quite a productive one for the Cranelift project! Cranelift is our optimizing compiler backend that provides the...

@heinrich5991 how does fix_style.py only check style of changes? It doesn’t even seem to check git

Headshot

nevermind i finally got it working... somewhat

nice _D

Ah, it seems we removed that. @deen why did we stop running clang_format on just diffs?

@deen for the playground, can we start just creating a new repo and give me full access to it? we obv don't need to make it open for everyone yet however, i think it would be cool to have:

nightly builds, that i can share with friends, with whom i will probably also design away some annoying limits of vanilla we always disliked. would also be cool if i can trigger the builds somehow, or simply copy them from some github CI ^^
maybe a dns alias to my VPS so i can host some servers my own, or some vps i can connect to

def-

[ddnet-web:pr-2023-funding] 1 new commit

8166b1c Remove GER10 - def-

@murpi Could you please send me the GitHub link for the file which contains the code for the new #reports discord bot replies?

Headshot

nevermind i finally got it working... somewhat

wadafuuuuk

Learath2

Ah, it seems we removed that. @deen why did we stop running clang_format on just diffs?

did we ever do that? I don't think so

Jupstar ✪

@deen for the playground, can we start just creating a new repo and give me full access to it? we obv don't need to make it open for everyone yet however, i think it would be cool to have:

nightly builds, that i can share with friends, with whom i will probably also design away some annoying limits of vanilla we always disliked. would also be cool if i can trigger the builds somehow, or simply copy them from some github CI ^^
maybe a dns alias to my VPS so i can host some servers my own, or some vps i can connect to

Why new repo? You want to start from scratch? You can have a branch/fork instead

09:15

Rest sounds fine

deen

Why new repo? You want to start from scratch? You can have a branch/fork instead

no but i guess its easier to manage(e.g. if i make prs to "myself") and silence the bot?

Just make a branch on your own ddnet fork, I can build nightlies from that

09:19

as long as you are the only one working on it

ok, i can also simply use my master branch, and can make prs to that

You can make PRs to any branch btw

yeah i just want to make sure its automatically selected

09:21

since im not reviewing my prs, just so i have overview

09:21

but can probs change default branch

For the server, should I rent you a https://www.hetzner.com/cloud CPX11? Probably cleaner to have experiments on a separate VPS. Not connected to official ranks for now or should already be connected? (Then I'd have to add the flag to tell the ranks apart)

https://github.com/Jupeyy/ddnet/tree/playground

GitHub - Jupeyy/ddnet at playground

DDraceNetwork, a mod of Teeworlds. Contribute to Jupeyy/ddnet development by creating an account on GitHub.

deen

For the server, should I rent you a https://www.hetzner.com/cloud CPX11? Probably cleaner to have experiments on a separate VPS. Not connected to official ranks for now or should already be connected? (Then I'd have to add the flag to tell the ranks apart)

doesnt need to be connected yet. i can also pay it, doesnt need to be run by ddnet's costs

If you want cheaper, OVH has for 2 € / month

i just want a ddnet subdomain, bcs i am also playing to try some new stuff out

09:23

planing

deen

If you want cheaper, OVH has for 2 € / month

yeah should be fine for my needs

Actually 1 € / month for a year now for new customers: https://www.ovhcloud.com/de/vps/

VPS

do you want to buy it for me? should i send u money or simply donate to ddnet?

I can't get the new customer thing though

09:28

and I prefer hetzner

09:28

so if you want to get the cheap one, you have to do it yourself

i am also ok with paying more

deen

If you want cheaper, OVH has for 2 € / month

or he can run instances using free credits at gcp, aws or any other

deen

did we ever do that? I don't think so

Apparently we ran on entire files, but only changed ones in the past

Oracle Cloud gives you a free 4 core 24 GB ARM VPS permanently

ReiTW

or he can run instances using free credits at gcp, aws or any other

I would not recommend those, super easy to get an attack which causes outgoing traffic, and then after 1 night with 10 Gbit/s outgoing you are charged thousands of dollars

tbh idc i just want a starting point

09:55

if i can get a subdomain and the nightly builds i can also do all stuff my own

deen

I would not recommend those, super easy to get an attack which causes outgoing traffic, and then after 1 night with 10 Gbit/s outgoing you are charged thousands of dollars

well if I understood well isn't it for testing purpose?

09:55

if yes why not just register 0

you can also do a AA record to my domain name, then i can change servers as i want

also free credits doesn't charge you, on gcp you use your free creds then when you run out of those they just ask you to switch to a normal acc, so they won't charge you for that

or ANAME

09:57

CNAME

09:57

was it

09:57

right

So on the server you want the official maps etc?

09:59

Then I'll set up a small Hetzner server for you now as pg.ddnet.org

deen

So on the server you want the official maps etc?

I am also planning with other mods in mind So would generally be easier if I've access to it

why don't you set up the server yourself?

10:06

saves deen some work

I can

def-

[ddnet-scripts:master] 1 new commit

f18f7a9 Add playground build support - def-

I just want nightly build and sub domain.^^

10:07

So it's kinda official

nightly build is up, subdomain I can only set when you have an IP

So no cname?

Why do you want a cname?

Else i buy vps later . When at home 2-3h

deen

Why do you want a cname?

Easier to change vps later

10:08

But idc

What should I point it to?

pg.aliveclan.de

10:09

I'll set the server up to there then

Done

Thanks

Playground is also a release channel in steam, I used a PW just to make sure people don't select it accidentally, password is Playground23 (edited)

Oh great

If you want official server setup/database access etc then I should set up the server. If not, then you can go ahead

def-

[ddnet-scripts:master] 1 new commit

24da651 Fixes - def-

Yeah for now sounds better to not connect to official db

10:15

Maybe in few months

Alright, easier from my side then

pog

10:36

based linux distros got ddnet updated

10:36

well liguros uses gentoo guru which its the one i update

Now get it into Arch and Gentoo official packages, then it takes a bit longer

def-

[ddnet-scripts:master] 1 new commit

9b77e44 typo - def-

def-

[ddnet-web:master] 1 new commit

9b75d0d Add playground links - def-

@Jupstar ✪ first playground build is live. Links at the bottom of https://ddnet.org/news/nightly-build-revived/

whats the playground about?

it's a place where kids play

Ryozuki

whats the playground about?

breaking compability

deen

@Jupstar ✪ first playground build is live. Links at the bottom of https://ddnet.org/news/nightly-build-revived/

thanks a lot^^

def-

[ddnet/ddnet] Pull request opened: #6146 Instruct to use rustc-mozilla on Debian/Ubuntu (fixes #6145)

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-ad...

Jupstar ✪

breaking compability

Make sure to break it a lot

Ryozuki

Make sure to break it a lot

yes i am trying to create a complete new structure ^^

13:47

just physics stay the same xdd

@Jupstar ✪ changing the protocol?

13:55

a good thing would be to merge ex messages or smth

13:56

and make them extendable without hacks

yeah

I might not have much time to help, but if you need any further infrastructure etc for these experiments, just ping me. I'm glad to provide this kind of playground to get ddnet to next level.

13:58

@Jupstar ✪ anything in Rust?

deen

@Jupstar ✪ anything in Rust?

i'd like to play around with that too yeah, maybe for modules u can cleanly split from game logic but no guarantes yet

Cool

Just rewrite ddnet in rust, go

go for it

14:04

honestly the interop is a bit meh cuz u cant harness the power of rust on the c++ side

14:04

so it should all be rust

even if i would i still use the c++ vk backend

14:04

bcs rewriting that is lot of work

work is

14:04

but maybe u can make it more "correct"

You know what? The gfx backend would actually be a good candidate for a rust rewrite

since concurrency is easier in rust

Ryozuki

but maybe u can make it more "correct"

there are no "real" vk bindings for rust

14:05

only abstractions

https://vulkano.rs/

It's a relatively well split portion of the code

Jupstar ✪

there are no "real" vk bindings for rust

wdym

Another good candidate is the small auth_manager. I actually want to do that sometime so we can have a rust module in the code see how well our integration went

Ryozuki

wdym

well its not vulkan what u sent there

Ryozuki

https://vulkano.rs/

check this

its a higher level abstraction

14:06

designed for rust

well

14:06

i dont think u can have vulkan in rust as is

14:06

u mean t he sdk?

"Provides a low-levelish API around Vulkan."

vulkano is low level

14:06

it just designed to not allow u to do unsafe stuff

14:07

iirc

lets say it like this if that lib would tell me "yo bro, here i have to insert a memory barrier, except you change this and that code" i'd say gg epic lib

Can you even have direct vulkan bindings for rust? It'd all have to be wrapped in unsafe

but what probably happens is, that it does this implicitly

let vertex_buffer = CpuAccessibleBuffer::from_iter(
    device.clone(),
    BufferUsage {
        vertex_buffer: true,
        ..Default::default()
    },
    false,
    vec![vertex1, vertex2, vertex3].into_iter(),
)
.unwrap();

14:08

xd

14:08

this is low level

Learath2

Can you even have direct vulkan bindings for rust? It'd all have to be wrapped in unsafe

the way it is is, you make a -sys crate, usually autogenerated with a tool looking at C headers

14:09

and then u build a safe abstraction on it

14:09

thats what https://gtk-rs.org/ does

Unlocking the GNOME stack for Rust – gtk-rs

Rust bindings for GTK and GLib-based libraries

14:09

well all

14:09

xd

Ryozuki

the way it is is, you make a -sys crate, usually autogenerated with a tool looking at C headers

I guess this would actually be about what @Jupstar ✪ would want. Completely unsafe though, so the gfx backend would all be unsafe code, unless he builds some safe abstractions of his own to build the backend on

but its stupid

14:10

the safe wrapper just disallows things u would be wrong doing

14:10

it doesnt add a higher abstraction

14:10

that hides stuff

Ryozuki

the safe wrapper just disallows things u would be wrong doing

idk what exactly it wraps, but I can see it having different opinions on concurrency safety

in the vulkano case tho i think they try to also be correct and not allow doing stuff which the vulkan api says u woudl be doing wrong

A higher level abstraction can't know that it might not be worth dropping a lock because you are about to re-use it

Learath2

idk what exactly it wraps, but I can see it having different opinions on concurrency safety

a -sys crate is like a library full of extern functions

Anyway, idk. I'm just guessing what Jupstar wants, only Jupstar knows what Jupstar really wants

xd

def-

[ddnet/ddnet] Pull request opened: #6147 Use %zu formatter for size_t, not PRIu64 (which is for uint64_t)

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-ad...

Ryozuki

in the vulkano case tho i think they try to also be correct and not allow doing stuff which the vulkan api says u woudl be doing wrong

problem is, its probably not like the rust compiler which tells u what u doing wrong. instead it implicitly adds stuff to confort to what rust can do so what i mean is, that it will not detect data races in the graphics pipeline, but instead abstracts the whole stuff so that u cannot even do that. It would be amazing if its such powerful, but that would require lot of work probably, since its like a completely new programming language

14:21

i mean their github basically says it all

hm

14:25

i see

14:25

but vulkane is p high quality i heard

14:25

vulkano

Jupstar ✪

problem is, its probably not like the rust compiler which tells u what u doing wrong. instead it implicitly adds stuff to confort to what rust can do so what i mean is, that it will not detect data races in the graphics pipeline, but instead abstracts the whole stuff so that u cannot even do that. It would be amazing if its such powerful, but that would require lot of work probably, since its like a completely new programming language

well im not so knowledgeable about the pipeline so ur probs right, but it can prevent data races on code that runs on the cpu

i dont doubt that

by design

Ryozuki

well im not so knowledgeable about the pipeline so ur probs right, but it can prevent data races on code that runs on the cpu

i dunno vulkano, i simply doubt its great enough yet xD

why not try it

14:26

xd

14:26

best way to know

yeah but currently dont want to learn another graphics api xD

i mean its not another api its more learning how the wrapper works

14:27

but yeah

for now i'd like to redesign/test some stuff in the playground i always had in my mind

the vulkan concepts u know probs are 90% of what u need to know

and i hope i dont get too lazy for that already xD

xd

ok on a vps first time rust compiling is really slow xD

14:32

do they plan for binary shipping some day?, but probs too annoying per rust version

https://github.com/cberner/redb

GitHub - cberner/redb: An embedded key-value database in pure Rust

An embedded key-value database in pure Rust. Contribute to cberner/redb development by creating an account on GitHub.

Jupstar ✪

do they plan for binary shipping some day?, but probs too annoying per rust version

binary shipping what?

14:35

the compiler is binary shipped

Ryozuki

binary shipping what?

crates stuff xd

Dynamic linking when? Binary crates when? Smh, how is this even a srs language

no, static linking

14:48

just provide ready static libs so compiling doesnt take so long xd

14:48

at least for current stable rust compiler release or smth

No, is bad for performance

do they do it different rn?

14:50

else ofc dont do it, all in for performance xd

Afaik it's all march=native compiled rn, so you get the best performance possible

ah i c

14:50

well then make it optionally

If they ship you generic binary releases they wouldn't use epic vector instructions

yeah sounds sane xDD

14:51

on the other hand

14:51

if ddnet ships ddnet release with rust, march native sounds wrong

We probably set some flag that makes it generate generic x86-64 code

nah, default gcc behavior is only sse instructions

14:53

it wont use avx if you don't ask it to

Rust isn't compiled with gcc

dammit, got tricked by the flags

Afaik rustc has march native as a default

thought you were talking about gcc

I wonder if someone actually came up with a solution to at load time swap out more optimized "modules" if e.g. avx is available

14:57

I know you can branch at runtime and the branch predictor is good enough, but you can't guarantee 0 mispredicts afaik

Learath2

Afaik it's all march=native compiled rn, so you get the best performance possible

i dont htink so

15:17

cargo install doesnt install it with march native

15:17

u gotta set the RUST_FLAGS

15:17

RUST_FLAGS="-C target-cpu=native"

I thought they changed the default sometime ago. I might be misremembering tho

oh