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 2025-03-06 00:00 and 2025-03-07 00:00

no there is not

david loves Jellycar Worlds

is there currently any way to get the client to load a demo file from a server? im wondering if loading a evil demo file would be a attack surfires worth looking into

mods often open demo files

wdym?

00:21

also how do i build the tools?

00:21

wait nvm they automaticly build

in #

?

00:22

whats #?

sorry my internet has died for some reason in #✉-create-a-ticket if someones blocking u might upload a demo to prove it

why client doesnt keep the number of players but always loops max_clients instead?

ah ok good to know

The players are usually held in an array that uses client ids as indecies

00:53

The highest used id does not always match the player count

00:53

It would require new data structures

counting players from m_apPlayerInfos?

I don’t know client code to well I am more a server dev

01:00

But I am pretty sure it has the same max clients arrays

01:01

Not sure what you mean with this

Replying to @kebs counting players from m_apPlayerInfos?

so if client receives all player infos every snap, why not count players and loop it instead of max clients

Could you link the code?

nvm i forgot its not sorted

01:04

but these are

Image attachment

yea that’s what I mean with the data structures. It’s mostly arrays where the elements are spread across the array and do not fill up the beginning or the end

[ddnet/ddnet] Issue opened: #9811 demo list shows incorect sizes

when you look at a demo in the demo list on the side it shows its size. ![Image](https://github.com/user-attachments/assets/a however this can be incorect as it is in this case. heres the demo file and you can see it is not 92.68kb like it shows in the image. its about 1.1 megs. JustRocket_239.340_David Mathlover.zip 476c153-4d06-4a14-a4cf-f5e36f054a51) tested on the lastest full steam releace on ...

Depending on how players reconnect

01:06

Yea the sorted ones could exit the loop early I suppose

01:07

There might be room for improvement not sure. I wouldn’t be surprised. We do other things that are way more expensive than 128 continue checks I assume

01:09

Question is is keeping count even worth it. It’s a code complexity vs runtime trade off.

01:15

i had to add it so it updates the player count for this

01:15

it should be simple to implement elsewhere

ummmmm lol 200%

Image attachment

01:22

500%

Image attachment

01:22

and counting

01:22

ok it stoped

[ddnet/ddnet] Pull request opened: #9812 Update Discord Rich Presence

The timer will not reset when switching servers
It will now display the server name, map name, and the number of players
You can join and invite others to your server. However, joining without Steam sometimes does not automatically open the game, I'm not sure why. When game is open it will automatically connect

!image ![image](https://github.com/user-attachments/assets/c73b11c5-fce4-498d-aba2-4ab55e...

i like it

[ddnet/ddnet] Pull request opened: #9813 Updated Chinese translation

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...

[ddnet/ddnet] Pull request closed: #9813 Updated Chinese translation

kebs

Click to see original message

What's your plan with sending invites? Is that some new discord feature?

it was a feature already in discord

06:02

no one ever uses it tho

[ddnet/ddnet] Pull request reopened: #9813 Updated Chinese translation

[ddnet/ddnet-web] Pull request opened: #318 News: Moderator Call

Leeeet's do this!

Image attachment

5

1

justatest

2

fixed (edited)

Time to switch browsers again

08:42

LibreWolf will surely keep their promise, right?

As long they don't want money

Robyt3

You could also fuzz using the demo_extract_chat tool which doesn't use the map, or write a fuzzing test driver based on that

6 hours and counting and no crashes yet

Image attachment

09:10

ill tell you if it found any when i get back up

@kebs Thank you for the "Test map locally" feature, this is amazing!

brownbear

1

[ddnet/ddnet] Issue opened: #9814 "Test map locally" feature ignores rcon from autoexec

"Test map locally" feature ignores rcon from autoexec. I don't see, that the autoexec is fully ignored. It contains "password xyz" for example and the server started in the background actually asks for the password. I believe that the rcon is simply overwritten.

I saw that Test map locally does not connect to the correct server if a local server on 8303 is already running xd

well this makes sense, doesn't it? xD

Assa

well this makes sense, doesn't it? xD

Well it cannot log into rcon then

10:02

I guess the fact that we cannot start a server inside the client kinda limits our abilities

It just calls connect localhost so it'll connect to the first port with local server

[ddnet/ddnet-web] Pull request closed: #318 News: Moderator Call

10:08

[ddnet-web:master] 3 new commits

ba28929 News: Moderator Call - murpii 3ffc4e2 Update staff page - murpii 8bd9bbe Merge pull request #318 from murpii/moderator-call - def-

Jupstar ✪

Well it cannot log into rcon then

I never was able to login into rcon with this feature, as see my ticket

10:12

log into, login to, login into what is english

[ddnet/ddnet] Issue closed: #9814 "Test map locally" feature ignores rcon from autoexec

^this was a duplicate

github-merge-queue[bot]

[ddnet:master] 2 new commits

b7a9301 Updated Chinese translation - XCWQW1 cd6a437 Merge pull request #9813 from XCWQW1/master - def-

10:15

[ddnet/ddnet] Pull request closed: #9813 Updated Chinese translation

interesting, the local server doesn't log me in even with Client()->RconAuth("", aRandomPass, g_Config.m_ClDummy); called

Maybe you've changed login

10:31

Username

that's not in my autoexec. If I start my local server normally, I can simply login with my password defined there. If I start it with the in-editor-option, the password is overwritten

Assa

that's not in my autoexec. If I start my local server normally, I can simply login with my password defined there. If I start it with the in-editor-option, the password is overwritten

and it doesnt auto log you in?

nope

10:41

and If I press the button again, it even tells me i am not authorized or connected (while I am connected):

Image attachment

weird

@kebs we could use (pStorage->ReadFile)[https://github.com/ddnet/ddnet/blob/cd6a4373aff7d525bbd053ccb27ac71edaf57923/src/engine/shared/storage.cpp#L584] and simply check if an rcon exists already. There are parsing options in the console, but none of them seem to match our usecase

better to do what roby's issue says than scan pc autoexec for password

10:56

this doesnt sound like a secure idea

server scans for autoexec anyways but okay

client doesnt, the client needs to know the password to auth

how does auth_add work? It takes an ident, a level and a password, but what are ident and level options?

havent looked into it yet

"level can be one of {\"admin\", \"mod(erator)\", \"helper\"}"

ident = identificator/key

so to summarize: auth_add editor-local-server admin password (edited)

yea i guess

and use RconAuth("editor-local-server", ...) are you fine if I do the PR or do you want to do it?

you can do the pr, im doing the discord rp thing rn

1

Does anyone know what's total width and height of CUIRect?

ForgottenCat

Does anyone know what's total width and height of CUIRect?

CUIRect has w and h values

I mean, if the rect is as big as your screen

There is a Screen() function that returns a rect of your screen

ok, thx

@kebs this also doesn't log me in. The normal rcon now works for me, but I have a 1/30 wrong passwords

debug it

Assa

and use RconAuth("editor-local-server", ...) are you fine if I do the PR or do you want to do it?

maybe the ident isnt username

how do I login with user password in rcon?

11:22

I can only enter a password there

yea then theres no user

11:22

just RconAuth("", pass)

I used "add_user" instead of "auth_user"

11:23

i need more coffee

2025-03-06 12:32:04 I chatresp: No such command: auth_user.

11:32

huh?

11:33

its fucking auth_add .....

11:37

2025-03-06 12:36:30 I server: ClientId=0 authed with key=editor-local-server (admin) Now I am authed, but rcon still asks for username password?

11:38

also I auth before I join the server

11:39

I auth before I join the server, because the server asks for password, there are multiple issues on top of each other

david loves Jellycar Worlds

6 hours and counting and no crashes yet

Are you using persistent mode and tmpfs already? The exec speed still looks rather slow

Assa

I auth before I join the server, because the server asks for password, there are multiple issues on top of each other

how does it connect you when theres a password

kebs

how does it connect you when theres a password

it asks for the password

11:42

the server password in autoexec

11:43

I added a fourth command now witch sets password ""

11:43

because with sv_register 0 this is useless anyway

Can you use the rcon auth to pass the server password or does that only work for reserved slots?

The server respons with a successful auth, but the client doesn't know (is my guess)

Robyt3

Can you use the rcon auth to pass the server password or does that only work for reserved slots?

So I should be able to pass the password through this, but we don't know the password at this stage

You don't need the password if you can use the existing rcon auth instead

11:47

You can join reserved slots with password "username:password" but I don't know if that works without a reserved slot

do you see any benefit in letting the password in for your local server? I mean technically another person from your network could join

[ddnet/ddnet] Pull request opened: #9815 use local user for local test server

Add a local user for the local test server (which you start with "Test map locally" button in Editor). This allows you to still have a global rcon config in a server config. Also this PR removes the server password. This prevents trying to rcon into the server without accessing it before.

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit ...

Assa

do you see any benefit in letting the password in for your local server? I mean technically another person from your network could join

yea chillerdragon plays on public networks

Assa

do you see any benefit in letting the password in for your local server? I mean technically another person from your network could join

If it works with that feature I mentioned then using the rcon auth to bypass the server password would be cleaner

Two questions:

How do we add a reserved slot
How do we join it out of the client?

Is the reserved slot even necessary or does rcon auth work for server password in general?

yes, you can login with rcon even with a server password enabled, also for other users (edited)

11:53

I see, easy fix: we can generate a server password and simply connect with Connect(IP, Password)

Then setting password username:password on the client side before joining should work regardless of the server password using the editor auth

Assa

yes, you can login with rcon even with a server password enabled, also for other users (edited)

I mean do you need the server password still?

I like my global autoexec. If I test multiple clients or servers I don't want to reconfigure them individually

11:55

In this case I don't need a persistent server password, so no ig?

Robyt3

Then setting password username:password on the client side before joining should work regardless of the server password using the editor auth

whats the benefit of doing password username:password instead of password something_generated ?

Then you don't need to change the existing password

dont set random password, itll be same issue as preset rcon password

Doesn't look like rcon auth works for server password without reserved slot. Maybe we should just add that as a feature as well.

Assa

whats the benefit of doing password username:password instead of password something_generated ?

For us mods were just leaking the generated password to some of their friends

12:06

So we moved to their personal rcon password

if you are logged in, just type "password" into the console and leak whatever the password is

12:07

checkout #9815 if this is what you want

https://github.com/ddnet/ddnet/pull/9815

use local user for local test server by AssassinTee · Pull Request ...

Add a local user for the local test server (which you start with "Test map locally" button in Editor). This allows you to still have a global rcon config in a server config. Also ...

Maybe we should rename it to "local server security hardening" xD

<ChillerDragon> @Teero firefox L time to switch to Ladybird

ws-client

<ChillerDragon> @Teero firefox L time to switch to Ladybird

sadge

12:33

Been using firefox since birth

<ChillerDragon> I still feel new to firefox im not attached. I used a lot of safari and chrome in the past

Can't you just disable ads and then it's ok? xd

<ChillerDragon> i guess its the mindset to me firefox sometimes feels a bit more sluggish than chromium i just use it because it feels more free

13:09

<ChillerDragon> OOP enjoyers can i somehow make this print fng twice? https://paste.zillyhuhn.com/OQ

Isn't it already printing twice?

<ChillerDragon> in the comments you can see what it prints

13:20

<ChillerDragon> but i think i am good now. Got bigger problems than that one i might scrap the idea fully

ohh

13:21

I didn't understand your msg

13:21

Move the set config from the parent class to the child class? xd

ChillerDragon: https://stackoverflow.com/questions/962132/why-is-a-call-to-a-virtual-member-function-in-the-constructor-a-non-virtual-call

Why is a call to a virtual member function in the constructor a non...

Suppose I have two C++ classes: class A { public: A() { fn(); } virtual void fn() { _n = 1; } int getn() { return _n; } protected: int _n; }; class B : public A { public: B() : A() {}...

14:09

I guess don't use a constructor but a separate Init function

<ChillerDragon> thanks

DDNet

https://github.com/ddnet/ddnet/pull/9815

@Learath2 should we print the rcon/password in the client console when the local server is started? After you disconnect from it you have no way back in (edited)

why do you want to set password on a local server

15:22

someone can do it in autoexec if they want to

15:22

if then auto rcon doesnt work edit code so it works when connected

2 reasons:

to prevent others in my local network from joining
to overwrite the existing password from my autoexec which is automatically loaded

idk how you gonna do it

15:23

if you just print passwd into console, what if i dont check console

15:23

or exit game? the server will still run

We should just add the feature to auth with rcon username and password to join a server

doesn't help you to join back, because you are never informed about your credentials

Not sure I created an issue for that already, but basically the automatic auth should also extend to the normal start server button and not only the editor

15:24

So whenever you start a local server via the client it should be able to auth and join automatically

yea you have it in your sv_rcon_password issue

15:25

but hes saying to lock local server entirely by password, not just rcon

The client should remember the credentials so you can join transparently I guess

i dont like the idea of it being password locked when i dont want to

15:27

im restarting the client so many times during testing

There's no need to set the password on the server, but the client should always use the automatic username:password to join the local server so it works in any case

kebs

im restarting the client so many times during testing

If you restart the client that also quits the local server though

not with the restart button in vs

I assume because it just kills the client?

yes

Then your config also isn't saved

Robyt3

There's no need to set the password on the server, but the client should always use the automatic username:password to join the local server so it works in any case

how do I configure this? It currently doesn't

15:32

also if you kill the client @kebs the server also closes

Assa

how do I configure this? It currently doesn't

We would have to first add the feature to allow authing with rcon credentials instead of general server password

Assa

also if you kill the client @kebs the server also closes

Only with a clean exit, the server doesn't detect if the client stopped running (except on Android)

Robyt3

Only with a clean exit, the server doesn't detect if the client stopped running (except on Android)

doesn't the OS kill all the child processes if the parent dies?

15:34

at least on windows I can't do that easily, not even if I kill the client over task manager

Assa

doesn't the OS kill all the child processes if the parent dies?

From a quick search this would require admin permission on Windows

So far I didn't find a way to kill the client without killing the local server

I assume if you crash it it does not close the server

there is a subissue in this ticket: If I rcon into a server, and then connect into it afterwards, the rcon console isn't unlocked (edited)

this is only issue that you need to fix i think

I literally can't login into rcon without setting and joining with server password automatically or preventing a server password to be set in the first place

this will also fix if someone has password set up in autoexec

I also asked if we need a password for the local server. I would be fine to overwrite ith with "", empty

15:41

but "testing a map locally" shouldn't mean "Test your map with friends"

just add rcon auth on local server join if it was set by the test map locally

15:42

maybe in client.cpp ProcessServerPacket

@MilkeeyCat got some enums + generic enums + impl with enums working

Image attachment

Ryozuki

@MilkeeyCat got some enums + generic enums + impl with enums working

epyc

Ryozuki

@MilkeeyCat got some enums + generic enums + impl with enums working

is # a special syntax to access enum members?

16:24

i suppose it is used to avoid name collisions with methods

16:24

or is it just temporary

its to name the variant to use

16:28

because we using context free grammars

16:28

otherwise its ambiguous, but maybe there is a way, but i dont know yet

16:28

rust grammar is not context free

Image attachment

Ryozuki

Click to see attachment 🖼️

how people who never wrote rust see rust code

kebs

maybe in client.cpp ProcessServerPacket

problem with all of this is how stateless this is: You start the server

send a connect message, hope for the best
send an rcon message, hope for the best There is no verification in any of this steps and I don't see how to add them

test map>save rcon to gameclient>when joined server check if local server is runnning and ip is local>rcon auth (edited)

can still be the wrong server, as an other issue suggested
you don't know if you can successfully connect or successfully auth

17:21

I bet there is some obscure ubuntu local network setup where this fails

you dont need to check if sucessfully connected

17:21

just run auth when connected (edited)

but you don't do that

17:22

you send a connect message and immediately an auth message

yea bcs the feature was already complicated to do from ground up

Ryozuki

otherwise its ambiguous, but maybe there is a way, but i dont know yet

it will be ambiguous but it will look the same as rust!

17:22

owo

if your connection takes any longer for whatever reason your rcon will be too early

if i was to improve it rn i would do that

kebs

yea bcs the feature was already complicated to do from ground up

yes, I don't want to kill your feature, as I said, it's amazing and I'm glad to have it! owo

owo

you can just add the auth_add for now i think

17:24

doing the rcon auth with checks would be more complicated

17:24

and password thing especially

the auth add part is not the issue, this is working perfectly ^^

17:26

the auth feature just breaks with server password. That's my issue

17:26

at least with auth_add i can still login with my usual rcon

17:27

what does the f2 rcon expect for a username for the root user? root or nothing?

why do you run a password on local

17:33

unless you're opening it to the public?

local servers are still findable in the LAN network

ah like within apartment networks etc.?

also this way I only put sv_register ipv4 into the console to make it public for testing

louis

ah like within apartment networks etc.?

yes or if you life in teeworlds households

[ddnet/ddnet] Pull request opened: #9816 Fix unpractice when locked

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 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...

def faster in merging, than I can review (edited)

Assa

def faster in merging, than I can review (edited)

same

kekw

Assa

def faster in merging, than I can review (edited)

bcs its a gramebreaking bug

18:08

and its copy paste from other part of code

can't we skip merge queue for these things then?

18:09

tho ngl, just make a follow up PR tbh, if you push it now it gets removed out of the merge queue

18:09

oh the check fails °-°

[ddnet/ddnet] Issue closed: #9037 Map downloaded with incorrect expected SHA256

i guess reviews blocked merge?

18:15

idk what happened

Reviews shouldn't cancel the jobs, maybe a github issue

18:16

Internal server error. Correlation ID

will push random fix to rerun jobs

kebs

i guess reviews blocked merge?

sorry, maybe I caused it then. I thought def- stopped the jobs

i changed to your printing function

18:35

but theyre exactly the same it doesnt matter

18:35

when pr is fixing bug thats used for cheating ranks

<ChillerDragon> I spent one day refactoring code and the result looks way worse than before. I regret my life decisions.

19:07

<ChillerDragon> https://zillyhuhn.com/cs/.1741287993.png

Embedded image

19:09

<ChillerDragon> i feel like i need the struct as parameter robyte keeps talking about

passing half-initialized structs is so C-coded

I love it

kebs

when pr is fixing bug thats used for cheating ranks

Thanks for reading my mind justatest

justatest

, looks like you already refreshed while I was typing

God GitHub is really annoying today isn't it

Image attachment

20:39

cc @kebs

monkalaugh

monkalaugh

its in mergeq

20:42

pr checks passed

20:42

@Robyt3

[ FAILED ] Net.Ipv4AndIpv6Work

20:43

did the runner somehow lose internet?

github-merge-queue[bot]

[ddnet:master] 2 new commits

6a36358 Fix unpractice when locked - KebsCS a83f5b2 Merge pull request #9816 from KebsCS/pr-fix-unpractice - Robyt3

21:04

[ddnet/ddnet] Pull request closed: #9816 Fix unpractice when locked

@Ryozuki just look at dis GEP

instructions = vec![
    Instruction::Alloca {
        ty: Ty::Struct(vec![Ty::I8, Ty::Struct(vec![Ty::I8, Ty::I32])]),
        out: 0,
    },
    Instruction::Store {
        value: Operand::Const(ValueTree::Branch(vec![
            ValueTree::Leaf(Const::I8(1)),
            ValueTree::Branch(vec![
                ValueTree::Leaf(Const::I8(2)),
                ValueTree::Leaf(Const::I32(3)),
            ]),
        ])),
        place: Place::Register(0),
    },
    Instruction::GetElementPtr {
        ty: Ty::Struct(vec![Ty::I8, Ty::Struct(vec![Ty::I8, Ty::I32])]),
        base: Place::Register(0),
        indices: vec![
            Operand::Const(ValueTree::Leaf(Const::I8(0))),
            Operand::Const(ValueTree::Leaf(Const::I8(1))),
            Operand::Const(ValueTree::Leaf(Const::I8(1))),
        ],
        out: 1,
    },
]

gives

.section .text
.global main
main:
    push rbp
    mov rbp, rsp
    sub rsp, 16
.L0:
    mov byte ptr [rbp - 12], 1
    mov byte ptr [rbp - 8], 2
    mov dword ptr [rbp - 4], 3
    lea r15, [rbp - 4]
    jmp .Lmain_ret
.Lmain_ret:
    leave
    ret

@MilkeeyCat is this good or bad news?

Teero

@MilkeeyCat is this good or bad news?

good

MilkeeyCat

good

cool

lol

1

MilkeeyCat

@Ryozuki just look at dis GEP

instructions = vec![
    Instruction::Alloca {
        ty: Ty::Struct(vec![Ty::I8, Ty::Struct(vec![Ty::I8, Ty::I32])]),
        out: 0,
    },
    Instruction::Store {
        value: Operand::Const(ValueTree::Branch(vec![
            ValueTree::Leaf(Const::I8(1)),
            ValueTree::Branch(vec![
                ValueTree::Leaf(Const::I8(2)),
                ValueTree::Leaf(Const::I32(3)),
            ]),
        ])),
        place: Place::Register(0),
    },
    Instruction::GetElementPtr {
        ty: Ty::Struct(vec![Ty::I8, Ty::Struct(vec![Ty::I8, Ty::I32])]),
        base: Place::Register(0),
        indices: vec![
            Operand::Const(ValueTree::Leaf(Const::I8(0))),
            Operand::Const(ValueTree::Leaf(Const::I8(1))),
            Operand::Const(ValueTree::Leaf(Const::I8(1))),
        ],
        out: 1,
    },
]

gives

.section .text
.global main
main:
    push rbp
    mov rbp, rsp
    sub rsp, 16
.L0:
    mov byte ptr [rbp - 12], 1
    mov byte ptr [rbp - 8], 2
    mov dword ptr [rbp - 4], 3
    lea r15, [rbp - 4]
    jmp .Lmain_ret
.Lmain_ret:
    leave
    ret

why is a place a register

21:36

i think a place should have no notion of register, stack or otherwise

21:36

a place is like a pointer to place in memory

21:36

btw its handy if a place has a projection

21:36

so there is a local, that the place points to and the projection (store in the place struct)

21:37

the projection for example can be a field access, a dereference, a index

21:37

projection is a vec

21:38

a local is like the base pointer to the alloca

21:38

i guess it depends on how low level u want ur ir to be

ok yep still no crashes

Image attachment

Robyt3

Are you using persistent mode and tmpfs already? The exec speed still looks rather slow

not in this case but i am using asan so that slows it down some

Ryozuki

why is a place a register

because I wrote so xd

im gonna keep it running for now though because its still getting new finds

21:41

witch is surprising

Exported 287 message(s)

Timezone: UTC+0