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 2024-02-23 00:00:00Z and 2024-02-24 00:00:00Z

Yea idk sorry i only translated a snap successfully so far client side. I build a snap server side like 2 days ago and it is not working :c

00:31

But keep me posted maybe i finally realize whats wrong with my snap

00:32

Oh wait i didnt fully read the quesiton lol

00:32

yes so obj_character_core is not sent i think thats a base class. So is event_common. The others just inherit from that.

00:32

Sound global i would have to check

00:33

So 8 is not sent but it is included in 9 https://chillerdragon.github.io/teeworlds-protocol/06/snap_items.html#obj_character

chillerdragon

Sound global i would have to check

from what I can see, the sound global is sent as a SV chunk and then mocked to the netobject by the client

chillerdragon

But keep me posted maybe i finally realize whats wrong with my snap

I just wrote the documentation about snaps

03:17

check: https://protocol-docs.teeframe.com/packets/default-packets.html https://protocol-docs.teeframe.com/snap/snap-structure.html

chillerdragon

Yea idk sorry i only translated a snap successfully so far client side. I build a snap server side like 2 days ago and it is not working :c

the thing is that everything in your snap header must be right, check the infos about CRC, Size, Removed Items Count and Delta Items Count

03:19

the only thing I'm unsure is about the Removed Items Count (I'm not really sure if its related to the last snap sent or something else) and Delta Items Count (idk if its just the current items count in the snap)

03:20

however, you can try just spamming a snap single chunk with obj_game_info and obj_player_info with 0 removed items and 2 delta chunks (and everything else correct)

03:20

after the initial handshake, the client will understand you as a server and will move you to the game screen

03:21

I'm doing this after inital handshake

Image attachment

03:21

moves to this screen

Image attachment

Miguilim

check: https://protocol-docs.teeframe.com/packets/default-packets.html https://protocol-docs.teeframe.com/snap/snap-structure.html

also give some feedbacks if you find something wrong or unclear owo

owo

chillerdragon

Sound global i would have to check

yes, but from a point of view of the protocol, number 8 is unused. The message id never touches the network and only exists on the teeworlds official code (edited)

03:26

I marked 0, 1, 8, 13, 18 as unused on my listing (https://protocol-docs.teeframe.com/snap/snap-items.html)

Snap Items | Protocool Docs - TeeFrame

A VitePress Site

Another thing I'm trying to figure out is why there is a MAX_SNAPSHOT_PACKSIZE (900) value. Why we just dont use the 1400 maximum packet size - (things we send) to decide if we need to split the snap? (edited)

03:55

The thing is that value is also used on client. So I cant just not use it

Nice one good job

Replying to @Miguilim 5913e568-0285-458b-b5ad-8b96ae9d9c4b.png

05:58

Amazing thanks

Replying to @Miguilim I just wrote the documentation about snaps

chillerdragon my first merged commit, even tho it's not code, let's goooo

lgtm

<ChillerDragon>

06:22

<ChillerDragon> @MilkeeyCat ur ukrainian?

Yes ._.

<ChillerDragon> pog

Id never learn it as a second language

<ChillerDragon> how much diff to rus?

Few different letters and ' character

<ChillerDragon> woah much bigger diff than i expected

06:24

<ChillerDragon> dude even hello is changed Привіт

morning

1

Wazzupers

09:12

FFR

Jupstar ✪

FFR

Hmmmm

Image attachment

09:18

Can you fix my bug problem?

Teero

Can you fix my bug problem?

What is your bug problem

Teero

Hmmmm

Fridays for rust

Jupstar ✪

What is your bug problem

It's illegal robot talk so I can't ;(

Use asan

09:20

And tsan

It's logic bug...

What's hard then

09:21

And some logging

09:21

See where the value gets wrong

My head can't find a way to implement something some way that doesn't require a threadripper

U have a bottleneck

09:21

I see

09:22

Then you are at the right address

09:22

I make your code the fastest

Well no xD I just can't think of a way to solve my problem efficiently.

09:23

Has do with rating organisms while simultaneously having multiple goals.

09:28

My rating is fucked and I have no clue how to get it right. I could implement a bruteforce solution but that would be a 25x slowdown which I cannot afford

Teero

Well no xD I just can't think of a way to solve my problem efficiently.

Efficiency is not the key. Throw your 32 threads on your problem

09:31

And 64 GB RAM

no money (edited)

[ddnet-web:master] 1 new commit

9408451 Add 58€ donation by Zerol Acqua - def-

7

@Zerol Acqua Thanks!

Miguilim

moves to this screen

@Miguilim don't like phpstorm?

@Miguilim PM me please? (Your settings forbid me from starting it)

@Kaffeine I don't want to merge my own pr -.- could you add https://github.com/teeworlds-community/teeworlds/pull/61 to the merge queue pls. Ideally before I go to sleep

Fix CMP0148 #3212 by teeworlds-mirror · Pull Request #61 · teeworld...

upstream: https://www.github.com/teeworlds/teeworlds/pull/3212

<ChillerDragon> or jopsti u go if you have 1min to look at the change :3

ChillerDragon

@Kaffeine I don't want to merge my own pr -.- could you add https://github.com/teeworlds-community/teeworlds/pull/61 to the merge queue pls. Ideally before I go to sleep

Is it possible to merge a MR with CI failed? I thought we'd have CI fixed before merging something.

<ChillerDragon> no its not possible

PR

️

the best approach ever

<ChillerDragon> @Kaffeine but only the jobs i activated to be required to pass

11:52

<ChillerDragon> since half of the jobs fail due to the cmake issue its only 2 jobs that need to pass for now. Thats why im pussing #61 so much

11:52

<ChillerDragon> thanks babe

I found out it's also possible to re-run only individual jobs that failed, so you don't need to force push to refresh if only macOS CI is acting up again

<ChillerDragon> ye ik thats epic

11:53

<ChillerDragon> i use rerun more than i would like to admit

11:54

<ChillerDragon> robsti wats your opinion on teeworlds-community?

11:55

<ChillerDragon> https://zillyhuhn.com/cs/.1708689299.png watafak is going on now

Embedded image

I don't think I have the time and motivation to contribute code or large reviews to teeworlds-community. Seems good if you keep the scope low any only fix build issues and bugs. I'd consider most refactoring out of scope though.

12:01

For build/CI failures I'd look at the difference between teeworlds and DDNet and then cherry-pick the relevant commits

Robyte is already team dd pg

what is teeworlds community

Ryozuki

what is teeworlds community

0.7 maintained by the community xd

xd

12:23

just use ddnet 4head

If we'd have website etc would be cool. But as is, I also find it useless xdd

12:23

Official website and steam

https://factorio.com/blog/post/fff-399

Friday Facts #399 - Trash to Treasure | Factorio

Hello, It's Earendel back for another electric adventure. You got your first look at Fulgora in FFF-398. (If you haven't read that already please read that first.) Now let's take a look at the new planet's mechanics.

12:24

@Jupstar ✪ ah its just chiller xd

Ryozuki

https://factorio.com/blog/post/fff-399

Have you ever ran into performance issues playing that game, eg with a lot of entities on the map or smth

Jupstar ✪

Have you ever ran into performance issues playing that game, eg with a lot of entities on the map or smth

some ppl do with megabases but its hard to hit

12:27

people make crazy stuff since there is no limit

12:27

i never had personally

12:28

this is from a video

Image attachment

12:28

50 fps on a mega base

wtf

hard to grasp how big is this cuz this is the map look

Ryozuki

hard to grasp how big is this cuz this is the map look

Interesting, does it then use all cores?

probs yes

12:30

idk

12:30

read their blogs

12:30

xd

Ok

they rly care about perf

12:30

and nice code

12:31

i think belts are indeed threaded

12:31

@Jupstar ✪ > Factorio is primarily limited by cache misses - which is why the (otherwise rather mediocre) 5800X3D and its enormous L3 cache dominates your linked benchmark.

12:32

lmao

12:32

Image attachment

12:32

@Jupstar ✪ there is a amd epyc cpu with 1gb of l3 cache

12:32

imagine that

12:32

https://www.servethehome.com/amd-genoa-x-the-1-1gb-l3-cache-cpu-launched/

Patrick Kennedy

AMD Genoa-X The 1.1GB L3 Cache CPU Launched

AMD EPYC Genoa-X is a 1.1GB L3 cache part designed for high-performance computing with massive performance gains in some applications

Ryozuki

@Jupstar ✪ there is a amd epyc cpu with 1gb of l3 cache

Just too epyc

12:37

I wish future would be today

<ChillerDragon> @Robyt3 yea thats my plan to cherry pick ddnet fixes to get the build passing. Some of the more motivated ppl might have more ambitious plans lets see how long until i rq from democracy. Yea i assumed youd be busy but ye if you ever feel like approving a pr that would be pog.

13:12

<ChillerDragon> @Ryozuki yea i did not invite you or deen since i assumed you have no interest at all in the teeworlds code base. Let me know if you want write access.

holy hell. it doesn't look as cursed as before poggers

poggers

Image attachment

ws-client

<ChillerDragon> @Ryozuki yea i did not invite you or deen since i assumed you have no interest at all in the teeworlds code base. Let me know if you want write access.

u can invite me if u want just in case

<ChillerDragon> ok cool will do

MilkeeyCat

holy hell. it doesn't look as cursed as before poggers

poggers

Hmmmm, looks like I saw smth like this before thonk

thonk

in ddnet code base?

16:17

time to add textures, woo... im pretty sure it will be fun

@MilkeeyCat: you are really close to jsfiddle for tw graphics :3

yea, i also have thought about that

Pls do

i would need to make ui

16:19

sounds not fun at all

16:20

but i think when i make Graphics class it wont be that bad after

You mean an editor?

16:20

Use monaco editor it’s vscode as js lib

16:20

The rest of the ui is just the canvas

i mean.. basically im rewriting teeworlds graphics in js

Yes

16:21

That’s what I almost did

16:21

Just to build a fiddle

16:21

Now you did it. Pls do fiddle

js the best pepeW

pepeW

Image attachment

You such pro coder

Matodor

@Miguilim don't like phpstorm?

nah, I use vscode for everything xd

yes i am, it took only one hour to put true in setFloat32 (edited)

MilkeeyCat

js the best pepeW

pepeW

are these floats?

17:52

does cpp has fused multiply add

Ryozuki

are these floats?

yes

https://doc.rust-lang.org/std/primitive.f64.html#method.mul_add

A 64-bit floating point type (specifically, the “binary64” type defined in IEEE 754-2008).

MilkeeyCat

yes

i mean size * i + 4

17:53

https://en.cppreference.com/w/cpp/numeric/math/fma

cpp/numeric/math/fma

17:53

some cpus have a fma instruction

17:53

it can yield more accurate

17:53

and faster probs

rotated square btw

Image attachment

Due to its infinite intermediate precision, std::fma is a common building block of other correctly-rounded mathematical operations, such as std::sqrt or even the division (where not provided by the CPU, e.g. Itanium).

not rly a square, not rly rotated tho

Ryozuki

i mean size * i + 4

it's byte offset

nvm then

chillerdragon

You such pro coder

You say in your documentation that snap items have sizes. However, we send ints inside the items, and ints does not have fixed sizes

[ddnet/ddnet] Pull request opened: #8010 Improve envelope evaluation and validation

Call EnvelopeEval functions directly instead of passing them and their arguments to CRenderTools::RenderTilemap and CRenderTools::RenderTileRectangle. Only evaluate color envelopes for tiles layers once instead of separately for the opaque and transparent passes. Only evaluate relevant number of envelope channels instead of always evaluating all channels. Avoid unnecessary calculations by only evaluating position envelopes for quads which are not fully transparent. Fully ign...

Ex. obj_projectile size is 6, however m_X can use 2,3,4 bytes to encode it value

Image attachment

18:40

Or not? O.o Idk if I'm missing something here

Miguilim

You say in your documentation that snap items have sizes. However, we send ints inside the items, and ints does not have fixed sizes

All ints are 4 bytes in snapshots

Learath2

All ints are 4 bytes in snapshots

hmmmmmmmmmm I didnt confirmed but from what I tested this is actually not true

I really doubt we used packed integers for the contents of the snapobjects but there is a slight possibility that I'm misremembering

This is the ChillerDragon lib example

Image attachment

20:05

You see game_start_tick is 00 -> 0, x is 90 0c -> 784 and so on

20:06

tick is 9b 13 -> 1243

20:06

thonk

You'll see everywhere in the code that we use sizeof(int32_t) to iterate through the snapshot though :/

yes, thats why is so confusing

20:08

whoops

20:08

just found

Maybe we do pack them but then unpack them very early and I'm just forgetting that?

when the server sends the snapshot, it compress

Image attachment

Ah, yep that would make sense

the aCompData is passed by reference and altered inside the function

20:09

Image attachment

20:09

correct if im wrong XD

Soo when you get a snap, first thing you should probably do is to unpack so you can work with nice 4 byte ints all over

Learath2

Soo when you get a snap, first thing you should probably do is to unpack so you can work with nice 4 byte ints all over

yeauhuhok

20:11

So the snap having a fixed size is a wrong assumption from a protocol perspective right?

Yes that sounds wrong, most snap items however do have fixed size

Learath2

Yes that sounds wrong, most snap items however do have fixed size

thonk

but they only have fixed size in the code, not in the encoded packet that is sent

Yes, so more exactly, most snap items have fixed size after they get unpacked

fddrace, @fokkonaut switches arent shown, bug or intended?

Image attachment

Learath2

Yes, so more exactly, most snap items have fixed size after they get unpacked

pepeglasses

on the level of what ends up on the wire, you can't really say anything about sizes

cyberFighter

fddrace, @fokkonaut switches arent shown, bug or intended?

they should, if not maybe a client update fucked something up

Now everything is more clear, thanks for the help mate

i dont know about recent client history

im in a kinda newish version

cyberFighter

fddrace, @fokkonaut switches arent shown, bug or intended?

Does it work again after you disconnect and rejoin, or after you restart the client entirely?

disconnect rejoin doesnt fix it

20:17

its never shown

20:17

it used to be shown before on other client versions

20:19

ok i updated to latest nightly and its still broken

Likely caused by #7979

https://github.com/ddnet/ddnet/pull/7979

Fix entities textures not being built for non-DDNet/DDRace types, f...

Closes #7965. 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...

is there nightly before that pull i can download to check? or i dont need to

Seems like switch layer is not rendered anymore when m_GameInfo.m_DontMaskEntities is true

20:21

@fokkonaut I assume you send this flag on your servers then?

Ye, its required in order to show tiles which are non Default

20:22

It was added by Heinrich Back then

20:22

for this purpose

Shouldn't we just always render the switch layer if it exists? Kinda weird to render it only when DDNet and DDrace types are set

ye

[ddnet/ddnet] Pull request opened: #8011 Always build switch entities layer when map has switch layer

Instead of only building the switch entities when the server uses the DDNet/DDrace type. Fix switch entities not being shown anymore on servers which do not mask entities. Regression from #7979.

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 a...

[ddnet/ddnet] Issue opened: #8012 add error handling when renaming fails using auto_replay

slicing a demo using save_replay and specifing an invalid filename will dump the entire replay but will still output: -- Successfully saved the replay! along with the invalid filename !image

my brain - i cant even correctly name my issues

20:52

feelsbadman

[ddnet/ddnet] Issue closed: #7900 Editor - HD-off only Layers

[ddnet/ddnet] Issue closed: #7051 Envelope scaling is too fast when zoomed in (e.g. with default color envelope)

https://vxtwitter.com/EliasDaler/status/1761090706800537610

Elias Daler (@EliasDaler)

Vulkan official art goes hard. In the era of soulless corporate logos and art, it makes me happy that this exists.

21:39

I have to agree it goes hard

Hey, I'm trying to cross-compile DDNet to MacOS from Linux. I got osxcross installed with MacOS 11.1 SDK. My issue is that CMake complains about SQLite3 not being installed, so I'm wondering if anyone knows how to do this?

cmake -DPREFER_BUNDLED_LIBS=ON -DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/darwin-x86_64.toolchain -DCMAKE_OSX_SYSROOT=../../osxcross/target/SDK/MacOSX11.1.sdk/ ..

CMake Error at CMakeLists.txt:638 (message):
  You must install SQLite3 to compile DDNet

furo

Hey, I'm trying to cross-compile DDNet to MacOS from Linux. I got osxcross installed with MacOS 11.1 SDK. My issue is that CMake complains about SQLite3 not being installed, so I'm wondering if anyone knows how to do this?

cmake -DPREFER_BUNDLED_LIBS=ON -DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/darwin-x86_64.toolchain -DCMAKE_OSX_SYSROOT=../../osxcross/target/SDK/MacOSX11.1.sdk/ ..

CMake Error at CMakeLists.txt:638 (message):
  You must install SQLite3 to compile DDNet

Looks like there is no sqlite3 for mac provided as a bundled lib

22:03

you could try building and installing it into the sysroot so the CMakeLists.txt of DDNet can find it

22:05

https://github.com/ddnet/ddnet-libs/tree/master/sqlite3 there's no "mac" directory here (edited)

ddnet-libs/sqlite3 at master · ddnet/ddnet-libs

Precompiled libraries and headers of the dependencies of DDNet - ddnet/ddnet-libs

22:05

personally I wonder why not just put an sqlite source file and build it? it's totally trivial

deen is always compiling nativly on mac right? i've yet to see one person actually cross compile to that

yeah it's a bit nontrivial to do it

22:08

I've never done it before

22:11

looking at the commits here I see:

Revert "Remove sqlite3 on macOS since it's part of system"

but it got deleted in the next commit:

Update libs

which seems like it's (maybe?) a mistake?

22:11

this is from 2 years ago though

22:13

could the part about "part of system" be true? since I think otherwise someone would have noticed

22:19

https://github.com/ddnet/ddnet/blob/f6fdac442112bbf1d33d2e7564371511e0c857b6/.github/workflows/build.yml#L94 mac CI doesn't download SQLite

youtube, you drunk, go home

Image attachment

[ddnet/ddnet] Pull request closed: #8011 Always build switch entities layer when map has switch layer

22:33

github-merge-queue[bot]

[ddnet:master] 2 new commits

8350010 Always build switch entities layer when map has switch layer - Robyt3 e84278d Merge pull request #8011 from Robyt3/Entities-Unmasked-Fix - archimede67

furo

Hey, I'm trying to cross-compile DDNet to MacOS from Linux. I got osxcross installed with MacOS 11.1 SDK. My issue is that CMake complains about SQLite3 not being installed, so I'm wondering if anyone knows how to do this?

cmake -DPREFER_BUNDLED_LIBS=ON -DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/darwin-x86_64.toolchain -DCMAKE_OSX_SYSROOT=../../osxcross/target/SDK/MacOSX11.1.sdk/ ..

CMake Error at CMakeLists.txt:638 (message):
  You must install SQLite3 to compile DDNet

okay, for some reason sqlite only provides an autotools configure script You might be able to cross-compile sqlite3 like this: CC=target-triple-idk-clang ./configure --prefix=<wherever darwin-x86_64.toolchain specifies> --host=x86_64-pc-linux-gnu make install though autotools is so bad and flaky this may well not work

I would try myself but it seems this osxcross requires an apple account to download build tools or whatever

22:58

I can't do that

Peter0x44

could the part about "part of system" be true? since I think otherwise someone would have noticed

Yeah, I also saw commit about sqlite3 being included since 10.4

Peter0x44

I would try myself but it seems this osxcross requires an apple account to download build tools or whatever

I downloaded a SDK from here, https://github.com/joseluisq/macosx-sdks

furo

Yeah, I also saw commit about sqlite3 being included since 10.4

I would first try seeing if you can coerce autotools to get sqlite in the correct directory so cmake can find it

23:21

autotools is horrific junk though so gl

github-merge-queue[bot]

[ddnet:master] 4 new commits

73c3636 Add color4_base::Multiply function - Robyt3 040d767 Remove dead code - Robyt3 ab3c6c4 Improve envelope evaluation and validation - Robyt3 59a734c Merge pull request #8010 from Robyt3/Map-Envelope-Render-Refactoring - def-

23:24

[ddnet/ddnet] Pull request closed: #8010 Improve envelope evaluation and validation

23:24

[ddnet/ddnet] Issue closed: #7985 Crash when trying to open old map

furo

Hey, I'm trying to cross-compile DDNet to MacOS from Linux. I got osxcross installed with MacOS 11.1 SDK. My issue is that CMake complains about SQLite3 not being installed, so I'm wondering if anyone knows how to do this?

cmake -DPREFER_BUNDLED_LIBS=ON -DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/darwin-x86_64.toolchain -DCMAKE_OSX_SYSROOT=../../osxcross/target/SDK/MacOSX11.1.sdk/ ..

CMake Error at CMakeLists.txt:638 (message):
  You must install SQLite3 to compile DDNet

i don't think PREFER_BUNDLED_LIBS is really expected to be used on any platform but windows. ddnet-libs might have a few mac/linux binaries within it but it doesn't seem complete at all (edited)

23:45

i think you're more expected to use brew or macports

Exported 236 message(s)