DDraceNetwork - developer

so hardcoded magic

So I can't connect to ddnet server using default tw?

you can

00:01

it still hooks on vanilla

but no prediction

00:01

No prediction? I thought it's only ddnet client feature

00:02

yes

00:02

Now I'm completely confused

00:02

no prediction = it does not predict freeze

00:02

prediction = see into the future xD

00:02

but no prediction

@Deleted User maybe you meant here "but with prediction"?

00:02

u asked for default tw

00:03

it has no prediction

00:03

Yep

default tw has prediction tho..

00:03

ddnet don't glitch because it disabled prediction for hook

00:03

wording is hard..

But it does glitch

00:03

On fng

00:03

but fng is something different

00:03

its a different mod

00:03

u can add support for it in ddnet code

00:03

they will add it and it wont glitch

sounds like ddnet clients has hardcoded check to disable prediction for ddnet servers

Hmmmm

so you can't fix it in server, is what he's saying.

ddnet is in first place made for ddrace... during the time other modders added mod support too

00:05

Okay I'll look for it in sources

use the fng method, you can prevent glitch for a certain distance on all clients i think.

It would be not that easy. I've seen folder predictions, what does code inside do?

if you only want to make server mods

Jupeyy

[ddnet/ddnet] Pull request opened: #2838 Only show new icons, when client is offline

Jupeyy

[ddnet/ddnet] Issue opened: #2839 Ingame server browser tab change, does not change outgame server browser tab

Change the tab of the ingame server browser, disconnect.. you get a different tab there, which causes loading wrong server browser. Since new changes

Jupeyy

[ddnet/ddnet] Pull request opened: #2840 [Idea] Add some decent hover affects and colors

Generally, would be nice to have very decent hover effects and decent colors. I just tried for fun adding hover effects to menu tabs(red for the quit button), hover effects for the server browser items and color for search/exclude I'm not an artist, it doesnt look good, if someone wants to overtake, he/she would be very welcome todo so !image ![image](https://user-images.gi...

two things i think would be pretty cool to implement: 1. being able to start a lan server inside the client without having to open an external window + program 2. "texture pack files" where you can just have one folder with game, particles, entities, etc. files so you can hotswap different game, particles, emotes, and other visuals without having to rename each individual one (and have texture packs available to change + select inside the client)

02:39

ofc it would take some time but these are just two things that would be cool

@louis isnt 1. in the newest client

oh is it

version 15

is it a beta build

it will be released in the next days

oh neat

if you have github https://github.com/ddnet/ddnet/runs/1119831512 you can download artifacts top right

ddnet/ddnet

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/ddnet

neat it works nicely

bors[bot]

[ddnet:master] 2 new commits

1cef3bf Only show new icons, when client is offline - Jupeyy 59961a2 Merge #2838 - bors[bot]

05:32

bors[bot]

[ddnet/ddnet] Pull request closed: #2838 Only show new icons, when client is offline

05:37

def-

[ddnet/ddnet] Pull request opened: #2841 Fix ingame browser tab change (fixes #2839)

@Soreu it is useful when something epic happens but u weren't recording you can still use replays (if u have them enabled) to save a replay of that epic moment

05:48

save_replay 30 saves the last 30 seconds of ur game

1

d3fault

[ddnet/ddnet] Issue opened: #2842 Turn on Anti-Ping by default

We should turn on Anti-Ping by default for our new Steam playerbase. Needing to "lead" hooks/etc by a certain amount of time, especially when that amount of time is variable (dependent on your ping), gives a poor gaming experience. I see perfectly competent gamers missing hooks/etc by a consistent amount: the amount they should be leading (but aren't). I wouldn't be surprised at all if many of the new Steam players have rq'd (permanently) simply because the "leading" is dumb. Honestly on...

No no no no

@Deleted User do github actions eat any money? Did you even meet any limitations with it?

No limitations for public repos

I'm certain they will talk to you if you go over some limit, same as with repository size

There is a limit to the builds you can run per month

@Learath2 @heinrich5991 btw, I just checked and xz was about same speed for compression and smaller file sizes than zstd for teehistorian files for me (tested with lowest and highest compression levels). Was just wondering if I could easily save some space by recompressing

09:02

I'm wondering what the proper approach would be to get a teehistorian replayer for creating proper demo files

09:02

I guess I need to reimplement teehistorian format reader in C++, run a modified server that just uses the inputs at the correct ticks?

Yes I think that about covers it

09:03

You might also want to run the server at a higher tickspeed so it doesnt take hours

And then ideally we throw a DDNet client to watch at it and make it record the video to a stream that gets proxied to the user watching the teehistorian?

09:04

well, or implement a demo viewer in teewebs javascript client

09:05

@Learath2 I thought we could remove the waits between ticks totally and just run through the ticks instantly

One problem I can think of is the rather massive teehistorian files

09:05

These are sometimes like 20 hours of server uptime, all in one file

There are teehistorian files with > 100 hours I think

So when a tee requests a demo, you need to chop the massive demo up

and nearly full server for much of the time, so they are easily multi-GB

09:06

well, can't find out where to cut it

If only we were saving race starts

09:08

I was thinking of skipping until the race start, then running it backwards until I have enough for a full snap and starting there

Well, we have the finish time

09:12

we could go back from that and try to find when the players that finished started

09:12

but it's a bit hacky

But it's not very easy to figure out when you have enough for a full snap

09:13

I wonder if we can just run it through in the server, if that's cheap enough, no need to bother with hacking around teehistorian files

@Ryozuki im such a github influencer i wonder if i can get sponsored and make some github ads xd https://zillyhuhn.com/cs/.1600260808.png

12:55

btw join #olive on freenode

QingGo

[ddnet/ddnet] Issue opened: #2843 maybe add bundle lib of libwebsockets 4.1 for windows

It can make it easier to build a windows DDNet client with websocket feature

ChillerDragon i found it curious cuz i was searching for a video wditor

@heinrich5991 can you give me a hand with some rust?

13:36

just a handful of lines

or @Ryozuki ^^

if its not too complicated i could try my best @Learath2 :)

It's trivial, I have a program that takes a filename as an argument, I need it to be able to get the file from stdin too

14:02

https://paste.pr0.tips/xV

ah, ive never actually worked with stdin, only ever used the command arguments

Meh

shouldnt be that complicated tho, no?

14:04

have you tried https://doc.rust-lang.org/std/io/struct.Stdin.html?

std::io::Stdin - Rust

API documentation for the Rust Stdin struct in crate std.

process takes a Path though, hm

14:05

I can break it and rewire it to just work with stdin

wait so your problem is that you want to allow either filename as argument or via stdin?

Smth with pathbuf irc

14:06

Not home rn

I was thinking of wiring it like cat, treat - filename as stdin

What the problem ezactly

14:08

Taking arguments from stdin

14:08

Or converting it to a path?

14:08

https://doc.rust-lang.org/std/path/struct.PathBuf.html

std::path::PathBuf - Rust

API documentation for the Rust PathBuf struct in crate std.

stdin isn't a path

14:09

but process takes a path

14:09

I think I should move the opening of the file out of process, so I can just pass a teehistorian reader to process

Whats process here?

14:14

Oh u sent a paste

14:14

Didnt see

14:18

I think u should create the path with pathbuf the diference is that pathbuf is owned

14:18

Also u dont need extern crate

14:18

Pathbuf::from()

@Learath2 make sure to use a BufReader

14:51

Read isnt buffered

14:51

std::io::Read

14:51

https://doc.rust-lang.org/std/io/struct.BufReader.html

std::io::BufReader - Rust

API documentation for the Rust BufReader struct in crate std.

@Learath2 cna't think of something simple except pass /dev/stdin to the process

14:56

I decided against doing everything generically in the teehistorian crate, so the only thing I accept is a File (or a Path) that will be opened

QingGo

[ddnet/ddnet] Pull request opened: #2844 update cmake file to use libwebsockets bundle

15:22

12pm

[ddnet-web:master] 1 new commit

9fdb4cc remove lady saavik - 12pm

@heinrich5991 why would you tie parsing to the opening?

I tie it to a File

15:23

i.e. I can't read from arbitrary streams

15:25

the core library doesn't do that (teehistorian/src/raw.rs), but I didn't expose that in the public API

I guess I'll pass /dev/stdin

QingGo

[ddnet/ddnet-libs] Pull request opened: #13 add libwebsockets bundle lib 4.1

build steps is as below: ``` bash git clone https://github.com/warmcat/libwebsockets.git git checkout tags/v4.1.0 mkdir windows-build64 mkdir windows-build32 cd windows-build64 cmake .. -DCMAKE_TOOLCHAIN_FILE=../contrib/cross-w64.cmake -DLWS_UNIX_SOCK=OFF -DLWS_WITH_SSL=OFF -DLWS_WITHOUT_EXTENSIONS=ON -DLWS_WITH_SYS_SMD=OFF make -j8 cd ../windows-build32 cmake .. -DCMAKE_TOOLCHAIN_FILE=../contrib/cross-w32.cmake -DLWS_UNIX_SOCK=OFF -DLWS_WITH_SSL=OFF -DLWS_WITHOUT_EXTENSIONS=ON -DLW...

can mastersrv and twping executables be renamed to something similar, but not the same? (edited)

19:29

git ignores /mastersrv and /twping, github actions can't compile everything

what does that have to do with each other?

its not very comfortable to make mod under ddnet base on fork with modname branch, it needs to be reuploaded on new repo (edited)

19:33

and when im starting, always tools or various targets are showing up (edited)

tools or various targets are showing up?

19:33

where?

19:34

I'm not quite following

the mastersrv and twping folders are in gitignore

19:34

if you create a mod and upload your initial base, then these folders dont show up

compiler can't find /mastersrv and /twping, because mastersrv and twping are written in .gitignore

in the repo

ah

19:35

we can fix the gitignore

19:35

(but you should not use create a new git repository for your mod, you can keep the history anyway, even with a different github repo)

ye

oh, need to learn about it more

It's coming together

wdyd?

Making a dissector

building on any prior work?

Nope

20:25

There is no prior work except the libtw2 dissector

20:25

I didn't feel like plunging into depression because of the rust compiler

and fstd's dissector

Ah, he has one too?

20:26

Anyway I'm using conversations to dissect 0.6 0.6.5 and 0.7 all in one

nice!

With connless flag set, numchunks doesn't mean anything, does it?

numchunks is 0xff for num_chunks I think

20:29

for connless packets*

Seems the client doesn't set it to anything

20:31

I found one with 0x56 one with 0x4f and it seems to vary between runs so, I'm guessing uninitialized memory

@Learath2 there's also some part that is used as part of the extended serverinfo protocol

20:35

maybe you saw that

20:35

?

20:35

normal (non-extended) should have it at 0xff

Oh is that where we packed the extended bit?

we made the first two bytes of the packet

20:36

xe

20:36

and then 4 bytes for free use IIRC

But the first two bytes of the packet are after numchunks, no?

no, first two bytes in the header

20:37

it happened that 'xe' contains the connless flag ^^

Oh wow

(we didn't have much space to work with, forgive us)

But that only destroys flags and ack

20:38

I guess after its detected as extended the rest is just the packet?

ye, and then next four bytes are for free use. I don't know which, but we use two of them as the token for the serverinfo

we have serverinfo tokens?

Not the handshake kind, only for one way verification iirc

In 0.6

20:39

?

20:39

ah ye

20:39

i remember

I see

20:42

I guess I'll need to keep track of ddnet too, to dissect TKEN properly

can also be detected by having exactly four bytes remaining after decoding four chunks

20:43

*all

I already keep state for the connection. Might aswell keep this there too

fair

Wireshark has a rather meh plugin system

please do publish your even unfinished end result

20:48

it'll be interesting to look at

Idk how best to publish it tbh, you are supposed to develop plugins in tree

20:49

I guess I could have an entire fork of wireshark

fstd did that IIRC

Ah I'll host the teeworlds folder

20:50

Add the folder as a submodule

I like how well plugins integrate though, when you dissect in a plugin it looks exactly like a native dissector

20:52

You have access to exactly the same features as a built in dissector too

ye

Learath2: you're going to lua that?

@Learath2 nice on the dissector

fstd: nah sticking to C

probably for the better, it makes the decompression and deserialization copypastable

21:12

would suck to reimplement that in lua I suppose

Why there is no moving platforms or dynamically changing blocks on maps?

21:25

I'm sure this idea already was in some minds

The only implementation of it was on kog and it wasnt clean enough to add to ddnet iirc

So maybe ddnet should be initiator of this change? Not static maps would be great to see

Let me take a look at kogs implementation again

21:29

@qshar can I take a look pls?