DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.org/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 2024-04-07 00:00:00Z and 2024-04-08 00:00:00Z

did you already try just deleting your build folder and retrying? iirc this could happen using the wrong rust toolchain - even tho im not 100% sure - what compiler are you running? - if the error persists you could try swapping to gnu or msvc respectivly: rustup default stable-x86_64-pc-windows-gnu rustup default stable-x86_64-pc-windows-msvc

00:31

keep in mind i might be talkin bullshit here :D

i don't think he even has rust installed

justatest

i don't blame him really

From the official ddnet repo fof building on Windows:

Download and install some version of Microsoft Visual Studio (as of writing, MSVS Community 2017) with C++ support, install Python 3 for all users and install CMake. You also need to install Rust.

ddnet 16.5

01:14

troll

meloƞ

From the official ddnet repo fof building on Windows:

Download and install some version of Microsoft Visual Studio (as of writing, MSVS Community 2017) with C++ support, install Python 3 for all users and install CMake. You also need to install Rust.

yes please uninstall all your existing build tools so you can do it the way the outdated build instructions say to

Mazty

mingw32-make[2]: *** No rule to make target 'release/libddnet_engine_shared.a', needed by 'DDNet.exe'.  Stop.
mingw32-make[1]: *** [CMakeFiles\Makefile2:499: CMakeFiles/game-client.dir/all] Error 2
mingw32-make: *** [makefile:155: all] Error 2

(edited)

just install rust through msys

greatest achievement of my life

Image attachment

[ddnet/ddnet] Pull request opened: #8189 More details in snapshot debug dump

The method CSnapShot::DebugDump() used to print just the raw snap items. Without item names or field names: ``` 2024-04-07 09:10:01 I snapshot: data_size=356 num_items=10 2024-04-07 09:10:01 I snapshot: type=0 id=32767 2024-04-07 09:10:01 I snapshot: 0 -1824658838 933dea6a 2024-04-07 09:10:01 I snapshot: 1 -629591830 da7930ea 2024-04-07 09:10:01 I snapshot: 2 -1450210576 a98f8af0 2024-04-07 09:10:0...

[ddnet-maps:master] 1 new commit

55e7fbd A Holidays 2 - ddnet-maps

why china automatcly checks in as i open the game?

are u chinese?

not

and what do you mean then by "checks in"? chinese servers showing? or chinese language?

showing

08:52

and selects the flag

Bors Matyas

and selects the flag

did u check if your filters are correct?

Image attachment

ye

08:54

thats the problem

If you are having trouble where the flag automatically re-enable itself. Try closing the game normally instead of force closing it using keyboard shortcut

08:55

It is possible that your config is not saving properly

08:57

scratch that

08:57

it is bugged

08:57

chinese filter does automatically re-enable itself. I'll check with github issues

08:57

sorry

okay ty

china has infiltrated ddnet

what a bug to just annoy specifically russian players

kek

1

09:12

the ping of chinese servers are low enough to be annoying to russians and not others.

09:12

kek

[ddnet/ddnet] Pull request closed: #7521 Add /mode allowing teams to behave like team 0

09:16

github-merge-queue[bot]

[ddnet:master] 4 new commits

672cfb6 add /mode allowing teams to behave like team 0 - dobrykafe 4c40e6c change /mode -> /flock - dobrykafe 3a75f83 Address reviewer comment - def- ff991af Merge pull request #7521 from dobrykafe/pr-team-0-mode - def-

09:16

[ddnet/ddnet] Issue closed: #7380 Improve support for t0 runs

Teero

china has infiltrated ddnet

oh damn. I'm trying to debug this, and it turn magic

Ewan

yes please uninstall all your existing build tools so you can do it the way the outdated build instructions say to

WE Had this discussion before, and also i Said that to remind that Rust is needed. Why dont you Change it If you hate it so much

[ddnet/ddnet] Pull request opened: #8190 Fix broken rebase of ConFlock, rename previous remains of team 0 mode

The rebase of #7521 made the /flock command not be added. I've also changed any previous mentions of team 0 mode to flock mode and also added an explanation of what flock mode does.

Checklist

[x] 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 ou...

average ddnet dev

Image attachment

1

Ryozuki

average ddnet dev

average OpenGL 1 enjoyer

Image attachment

10:49

ddnet station

the table looks a bit thin given what you have on top plus your own arms (edited)

true lol

Ewan

Click to see attachment 🖼️

WHAAAT, no way xD

11:03

are u praying there

11:03

or sleeping under ur table? xd

owo

community filter broke

11:04

the region filter breaks when a community only serve servers in one region

Ewan

Click to see attachment 🖼️

but bro, u must have high mouse sense with that few space lol

lol

~~diverse community only~~

TsFreddie

the region filter breaks when a community only serve servers in one region

@Robyt3

Jupstar ✪

@Robyt3

already updated issues so he knows

1

Jupstar ✪

but bro, u must have high mouse sense with that few space lol

lol

he just thinks about the mouse moving and it moves xDD

ikr, else it's impossible xd

https://vxtwitter.com/tercio_miranda/status/1776643807523524940

Jupstar ✪

or sleeping under ur table? xd

foot rest.

Chairn

the table looks a bit thin given what you have on top plus your own arms (edited)

table is fine

Jupstar ✪

but bro, u must have high mouse sense with that few space lol

lol

i don’t play a lot of games

11:22

high sens is economical

Jupstar ✪

but bro, u must have high mouse sense with that few space lol

lol

not as high as a chinese player constantly doing one tick snap across the screen in fng and triggering noby's autoban

xDD

11:42

chinese players play max mouse distance 10 xD

eww

not as high as the ppl that whisper me to use a certain client

Jupstar ✪

not as high as the ppl that whisper me to use a certain client

bobo use teeor client

11:51

https://www.youtube.com/watch?v=dQw4w9WgXcQ here is preview

Hey step bro, wanna see my new client brownbear

brownbear

Jupstar ✪

Hey step bro, wanna see my new client brownbear

brownbear

ah man finishing extremes with grenade automatically is so fun brownbear

brownbear

Teero

https://www.youtube.com/watch?v=dQw4w9WgXcQ here is preview

how did u post without preview

11:53

or did u just delete so quickly

xDDD

11:53

do <>

11:53

around the link xd

Jupstar ✪

how did u post without preview

you fell for it xDDD

11:54

nub

https://www.google.com/

11:54

interesting

11:54

https://www.google.com/

Search the world's information, including webpages, images, videos and more. Google has many special features to help you find exactly what you're looking for.

11:54

it's more blue apparently

:D

Search the world's information, including webpages, images, videos and more. Google has many special features to help you find exactly what you're looking for.

11:54

11:54

mh

11:54

too bad

11:54

xd

11:54

11:54

oh

11:54

that works

11:54

lmao

ye sadly you cant do https://kek.com

oh no

Image attachment

11:55

jupstar has learned the art of trolling

i invented troll

monkaStop

2

Jupstar ✪

i invented troll

you have ruined so many good runs

11:56

put him to the stake!!!! (edited)

https:\ddnet.org

11:57

lmao

11:57

looks so wrong for the brain

11:57

https:||ddnet.org

11:57

xD

⑊

12:00

https:⫽ddnet.org/ (edited)

12:00

https://ddnet.org

teach

12:02

what is happening

Jupstar ✪

XcQ

12:04

XcQ everywhere

Chairn

XcQ

does that mean anything, i'm old boomer and not into the new hot teenage stuff

Jupstar ✪

@Robyt3

omg, why is it so hard to fix without refactoring

12:06

or change in UX logic

TsFreddie

omg, why is it so hard to fix without refactoring

i feel that

12:06

i guess it's simply bcs 2 ppl have different ways of doing things

I rarely have motivation to PR

12:06

now I have none (edited)

rip

12:07

tsfreddie contributor of the year 2024 was already reserved for you

hi

https://github.com/ddnet/ddnet/blob/ff991af6dd3dd50ad9c4313c09bf4017b7c086ab/src/engine/client/serverbrowser.cpp#L1973-L1977 Since "midfun" community only has CHN servers, excluding CHN will cause CleanFilter() (which is called upon getting ddnet-info.json it seems) to call the above routine and reset the filter.

12:08

the easiest fix is just to tell midfun to host servers outside of CHN as well TrollDespair

TrollDespair

bluekitty

Xalt

hi

hi wazzup, how are you, hows your day so far

Jupstar ✪

hi wazzup, how are you, hows your day so far

hi

12:10

Image attachment

12:10

this is my cat

ah

Xalt

Click to see attachment 🖼️

gigachad cat xD

12:10

nice

this is my cat

Image attachment

Teero

gigachad cat xD

shes disgusted by me

xDDD

Jupstar ✪

this is my cat

thats a hammer

meine katze ist der hammer

why do i understand that

maybe i had enough discord for today maybe maybe maybe

having a hammer as a cat isn't healthy

xD

Jupstar ✪

maybe i had enough discord for today maybe maybe maybe

don't leave

mixing tee eyes looks beyond cursed lol

lol

Image attachment

justatest

12:54

btw @Jupstar ✪ did you do the splitscreen gaming thing?

12:55

i remember that you posted a vid where you played split screen xD

12:55

gg kog servers

Image attachment

Teero

btw @Jupstar ✪ did you do the splitscreen gaming thing?

you mean dis?

Image attachment

12:59

it aint finished yet

lol xD

12:59

infinite dummies cool

yeah

Jupstar ✪

i added a dummy counter now, wanted to see how many entities until i reach 50 FPS

@Teero

Yoyo, if any of you likes doing crackmes, could you do this one https://crackmes.one/crackme/65df05c5199e6a5d372a3d25 and show how you dided it kek

kek

faceless dummies creepy

xd

13:02

nowadays they have faces

13:02

in the picture they have mixed eyes

13:02

still creeps me out

opensource?

13:04

what kind of structure are you using for the tees btw?

13:04

array, vector?

Teero

opensource?

an outdated version is open sourced

13:06

but it will be open source ofc in future

Teero

what kind of structure are you using for the tees btw?

linked hashmap

Jupstar ✪

linked hashmap

didnt think of that

13:06

i did infinite tees serverside once

it uses a vec or smth internally i think

just using a vec was somehow very slow

13:07

maybe my pc was just shit

it defs isnt very fast

i could connect ~250 tees until server went below 50tps

but with WASM in mind i also don't see a chance to make it multi threaded maybe a better algorithm can improve it a bit

Teero

i could connect ~250 tees until server went below 50tps

i think i reached 700 tees

Jupstar ✪

but with WASM in mind i also don't see a chance to make it multi threaded maybe a better algorithm can improve it a bit

would destroy the physics sadly

but on desktop pc ofc

Jupstar ✪

i think i reached 700 tees

xDD

13:08

new video

13:08

aipgores with all the players

Teero

would destroy the physics sadly

i think you can reform it a bit without destroying physics the most expensive checks are simply the distance checks

13:10

but i'd say it defs makes the code harder to read

13:10

so dunno if worth it

are there squareroots?

13:10

i dont think there are

afaik yes

hmm

but even without

tee to tee collision is heavy

n² bcs 64 ^ 64 checks

13:10

lol

13:10

64^2

collision is the most performace heavy out of all i think

13:11

atleast that would be the bottleneck imo

well 99% of the time they dont collide i guess?

13:11

but yes in a worst case it's even worse

i mean tile collisions

ah

all those intersect functions

well u could thread out all tees that likely wont collide with each other

13:12

and do these checks in parallel

13:12

but again, it will make the code harder to read

13:12

if u really really need 500+ players ok

13:12

else just buy better hardware

13:12

lol

tile collision could be optimized insanely with chunking and actual line collision instead of instead of for-looping over all the blocks xd

Teero

tile collision could be optimized insanely with chunking and actual line collision instead of instead of for-looping over all the blocks xd

wait what?

13:13

why should it loop

13:13

over all blocks

13:13

or do you mean if u dont do step by step moving

it goes pixel for pixel when doing line collision

but doesnt it do that also to do tee bounces

if all tiles were represented by lines you could just do line collision which is simple math i think

u'd need to track moving objects

Jupstar ✪

but doesnt it do that also to do tee bounces

im talking about tile collision still bro xD

Tile collision could be optimized insanely with bucketing

TILE

Teero

im talking about tile collision still bro xD

yeah but they happen at the same time

13:15

"time"

@Jupstar ✪ these things

Image attachment

dude

Jupstar ✪

yeah but they happen at the same time

no they dont xD

u not even reading

Teero

no they dont xD

then u mean ddrace tiles

13:16

ok

tile collision and tee to tee collision is something different right? (edited)

it happens before

13:16

yeah yeah fine

xd

then the physics sucks anyway lmao

wot

Learath2

Tile collision could be optimized insanely with bucketing

whats that?

with elasticity you would bunce into walls that way

Tee to tee collision is actual n^2, horrible

bounce*

Teero

whats that?

I meant tee to tee collision, sorry, what is the issue with tile collision? It should already be fairly optimal

13:18

Constant time per tee iirc

i think he means that the check happens in a ray shooting approach where u have a step count and go step by step

1

poggers2

Tee to tee collision is n^2 which can be improved by splitting tees into buckets so you don’t ever check against tees that can’t possibly be colliding with you

Jupstar ✪

i think he means that the check happens in a ray shooting approach where u have a step count and go step by step

the heavy thing are the array accesses to the m_pTiles array. they take a lot of time

13:24

i thought something like this. i think this is flawed too

Image attachment

Teero

the heavy thing are the array accesses to the m_pTiles array. they take a lot of time

well i cannot say last time i checked some sqrt was defs in top3 heaviest things

Jupstar ✪

i think he means that the check happens in a ray shooting approach where u have a step count and go step by step

Am I misremembering how we do it? I think we just slowly move you along from one position to the other in steps, at each step we check for collisions at the corners of the box

check CCollision::GetMapIndices

i know that code

13:25

but it's mostly some integer calcs

13:25

if u benchmarked it, i believe you

Teero

the heavy thing are the array accesses to the m_pTiles array. they take a lot of time

Do they? That access should be all cache hits if I think about it

i am just surprised that this is the biggest bottleneck

13:25

the distance check sounds scarier

this

1

13:26

btw the distance still uses sqrt

13:26

since u asked

Learath2

Do they? That access should be all cache hits if I think about it

i did some tests with perf and basically while(1) Predict(); (edited)

ok i should check in newest ddnet code

13:26

but sqrt² will also change physics sadly

13:26

even if it's faster

13:27

distance squared

i dont see a sqrt that isnt speeder related or other things

Image attachment

with zwelfs test cases we could check how much we can change physics without breaking edge cases maybe

Teero

i dont see a sqrt that isnt speeder related or other things

look in gamecore.cpp

13:28

float D = distance(Pos, pCharCore->m_Pos)

13:28

now open distance

ah

13:28

i forgot distance existed

Jupstar ✪

does that mean anything, i'm old boomer and not into the new hot teenage stuff

that's old boomer shit, and rick roll end of url

Chairn

that's old boomer shit, and rick roll end of url

ah that's the boomer way to check if the URL is rick roll? xD

Learath2

Tee to tee collision is n^2 which can be improved by splitting tees into buckets so you don’t ever check against tees that can’t possibly be colliding with you

with a quad tree or what?

Peter0x44

with a quad tree or what?

Even constant sized chunks would help and be very simple. But you could also do a quadtree yes

Jupstar ✪

float D = distance(Pos, pCharCore->m_Pos)

Could be replaced with dot product though, the actual distance value is never used

Though tbf it is just fast enough, it becomes more of a concern as playercount increases, e.g. 128/256p servers could probably benefit more

Vagina?

Bizarre behaviour

how dare he say such words

12yo behaviour

wait i thought we're all like in our mid 20's here, who invited him!

14:15

heresy! - TREASON!

Except me, I’m simultaneously 12 and 80

12 in your mind, 80 in your back

14:16

(i have severe back pain)

Robyt3

Could be replaced with dot product though, the actual distance value is never used

what do you mean never used?

14:17

it's used one line later

14:17

the question is how accurate is it changing dist to dist_squared

He means the actual value is not used except for ordering, which sqrt preserves because it’s increasing

Jupstar ✪

the question is how accurate is it changing dist to dist_squared

yeah, I wonder if it could change physics justatest

justatest

It might

hard to say, since distance must be rather small anyway (physical size)

14:19

it might ™️ work

But perhaps we are close enough to 0.0 that it doesn’t matter

on ARM it might already perform differently

I think we just might have enough precision to get away with it

or on windows

14:20

or on different compiler flags

14:20

i dunno if squar root is standard

do u think adding hands to tees makes the game more fun?

Milf

do u think adding hands to tees makes the game more fun?

they have hands

14:20

hook smth and u see the hand

not really

14:20

bruh

14:21

thats not it

Image attachment

14:21

2 hands

Jupstar ✪

it might ™️ work

actually TAS comes in handy here. i can just replay the same inputs and check if there are any diffrences

there just as short as my forehead

14:21

lol

Milf

there just as short as my forehead

should be like real hands man

14:22

not like ice cream

1

Teero

actually TAS comes in handy here. i can just replay the same inputs and check if there are any diffrences

@zwe

bro do u know how to make a channel like u can creat some kinda chat

@Zwelf might have a solution to these already

or sug

Learath2

@Zwelf might have a solution to these already

did he write tests?

He reimplemented the entirety of teeworlds physics in rust with tests to make sure they are the exact same

Milf

should be like real hands man

fr make this a feature if u can im leaving here

It’s part of his work to replay teehistorian files iirc

Learath2

He reimplemented the entirety of teeworlds physics in rust with tests to make sure they are the exact same

woah cool

14:24

is it faster? troll

troll

Teero

is it faster? troll

troll

not faster not slower probably

14:25

it's almost 1:1 copy at least physics wise

14:25

but u can see the code

14:25

https://gitlab.com/ddnet-rs/twgame/-/tree/master/ddnet/src?ref_type=heads

ddnet/src · master · ddnet-rs / TwGame · GitLab

GitLab.com

@Teero i replaced all distance with dist_squared except for the one in // check character collision it made 5 fps difference 45 vs 50 fps xdd

14:39

i guess sqrt is already too optimized

14:39

with 400 dummies btw

14:39

these fat asses

Image attachment

waoh

lol

well it could always be the rendering code that is slow too

could be

15:00

#off-topic stop spam

:c

[ddnet/ddnet-web] Pull request closed: #291 Fix syntax of news atom feed

15:03

[ddnet-web:master] 2 new commits

2252ebb Fix syntax of news atom feed - furo321 37768b8 Merge pull request #291 from furo321/fix-atom-feed - def-

@Jupstar ✪ i managed to recreate the situation in the demo renderer where it sleeps insanely long

Image attachment

15:19

SleepTimeInNanoSeconds is 8218685

mh 8 milliseconds sounds ok

15:19

xd

yea weird

other threads busy?

nop

15:20

all sleeping tight

how big is the resulting video

Jupstar ✪

how big is the resulting video

i didnt render it even. i wanted to change smth so i just quick via the menu

mh very weird

15:21

does it go into the render call

15:21

of gameclient

15:22

or OnRender

15:22

or whatever it's called

it does render it but i only made 1 frame sooo

15:22

since i did pause on start and then quit

i mean your client

15:23

is it still actively calling render calls

15:23

maybe some other vairable is broken

15:24

ok asking differently: does it exit the sleep

15:24

or is the sleep itself the problem

it exits sleep

15:26

but the next time it sleeps again

15:26

its like having fever

15:26

poor client

i tried to analyze using perf... i still can't use that tool xD it's just so bad.. why is there no tool like valgrind's callgrind but without being the slowest shit ever xD (edited)

Jupstar ✪

these fat asses

is it 512 slot server ?

that's a 6400 slot server (i added two 0's to the 64)

Peakies

is it 512 slot server ?

but it's 400 dummies connected

is it still get vanished or you fix it ?

Jupstar ✪

that's a 6400 slot server (i added two 0's to the 64)

tasty

Peakies

is it still get vanished or you fix it ?

what do you mean with fix it?

I mean on your server when you get close to some point that too many tee is rendering they get invisible or something don't know how to explain it

ah i know what you mean

16:24

yeah that doesnt exist in that version

16:24

it's litterally 400 tees rendered

16:25

no network limitations

hmm well 400 is too much I was trying to make like 150 or something for my server special events like arena tournament or something

16:26

but couldn't make it possible

yep, the client has to support it too

16:26

server side only it's not possible with the ddnet cpp version

well my server have it's own client

16:28

that base on ddnet 17.1 or 17.2.1 don't remember but anyway

mh i guess then you could make it work if MAX_CLIENTS points to 128 and no other weird limit is in the client somewhere

if you have plan to public your 400 slot server and client I like to learn how to do it

Peakies

if you have plan to public your 400 slot server and client I like to learn how to do it

i do plan to, but not very soon

Jupstar ✪

i tried to analyze using perf... i still can't use that tool xD it's just so bad.. why is there no tool like valgrind's callgrind but without being the slowest shit ever xD (edited)

Gprof isn't as good as perf but easier to use

Jupstar ✪

mh i guess then you could make it work if MAX_CLIENTS points to 128 and no other weird limit is in the client somewhere

I don't know where should I find that limitations

Peakies

if you have plan to public your 400 slot server and client I like to learn how to do it

also note that it's written in rust

Jupstar ✪

i do plan to, but not very soon

fine to me I can wait until that day

Peakies

I don't know where should I find that limitations

well the client checks a lot of stuff with MAX_CLIENTS

Jupstar ✪

also note that it's written in rust

oh shit I just installed rust don't know anything of rust language

just search it

There is also a tool to convert perf data into callgrind data so you can use tools like kcachegrind on it

Learath2

There is also a tool to convert perf data into callgrind data so you can use tools like kcachegrind on it

nani

16:30

that would be awesome

16:31

but anyway my findings are: the biggest bottleneck is simply the iterating itself the two O(n²) are simply expensive

16:31

especially the prediction tick calls it every frame

16:31

on the client

16:35

maybe i rewrite the physics for my version, since it doesnt need to be ddnet compatible O(n) should be possible with small breakages

:D that would be very cool

@Robyt3 switch display screen with 3rd party application on DDNet still need attention like using DisplayFusion to use it's functions for sending focused window to next display

16:41

resolution and display number won't update

Peakies

@Robyt3 switch display screen with 3rd party application on DDNet still need attention like using DisplayFusion to use it's functions for sending focused window to next display

Please open an issue for it if you think it's something we could fix

Robyt3

Please open an issue for it if you think it's something we could fix

I think it's possible to check current screen number with gfxScreen or something

[ddnet/ddnet] Issue opened: #8191 Switch screen still need attention

switch display screen with 3rd party application on DDNet still need attention like using DisplayFusion to use it's functions for sending focused window to next display resolution and display number won't update solution: as I think maybe possible to check GetWindowScreen and g_Config.m_GfxScreen to when it's not equel call SwitchWindowScreen but someone with better knowledge of DDNet take a look at it

bandicam.com

Ryozuki

bandicam.com

*a screenshot of a bandicam recording uploaded on discord

deep

19:39

idk but this is not the channel

Ryozuki

bandicam.com

unregistered hypercam 2

Image attachment

literally O(n²)

whats intra tick

the time between the current tick and the next

20:39

or better, the ratio

20:39

if it's a float

ok, thanks :DD

@murpi could you please update twmap-tools to 0.3.2 for the ddnet discordbot? it adds compatibility to the bezier envelope points

23:48

*for

Exported 413 message(s)