DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/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 2020-08-03 00:00:00Z and 2020-08-04 00:00:00Z

I really need help, I cannot seem to get the Secruity Token working for my 0.7 mod... I am even at that point where I cause client crashes while I try to connect :(

01:18

I would really appreciate if anyone could help me implement it, I know the logic behind it, heinrich told me the other day already, but I just cant get it to work...

01:24

also clients dont get inserted in network_conn::Feed instead in DirectInit? I just dont get how and why all is so different between the versions :/

[ddnet-maps:master] 1 new commit

d3bdf1e M Multimap - ddnet-maps

@fokkonaut which part does already work? do you see the CONNECT messages of the client in the server? the CONNECTACCEPT in the client? or even the ACCEPT in the server?

[ddnet:master] 2 new commits

07f18d4 Client: Support .map file drag-and-drop to open map in editor - def- abb7e7b Merge #2531 - bors[bot]

10:45

[ddnet/ddnet] Pull request closed: #2531 Client: Support .map file drag-and-drop to open map in editor

10:52

[ddnet:master] 2 new commits

bf4008c Provide sqlite3.lib file - def- 6e6e51e Merge #2530 - bors[bot]

10:52

[ddnet/ddnet] Pull request closed: #2530 Provide sqlite3.lib file

@heinrich5991 my best result was that I got the connect and the accept from the client, tokens matching, but the weird part is that in ddnet the server tries to accept the client in the accept msg, and in 0.7 in the Connect msg which comes earlier from ddnet. So no matter what I tried, the best I could get was a "connecting" tee, but the client never rendered anything. I even caused client crashes xd. When i try to accept the client on the connect msg it doesnt work. maybe its because ddnets and 0.7 Recv() part is completely different, 0.7 doesnt have these OnTokenCtrlMsg etc. or TryAcceptClient... Also, I dont understand the part in network_conn.cpp::Feed, where the server is checking for the token (dbg msg: "generate token %d"). It doesnt get called at all? I am just lost trying to do this

again there are too many maps in maptesting, so new maps can't be added, idk if i should post that here or soemwhere else (edited)

@jao I guess?

@fokkonaut try to debug the connection process step by step. try to send a token back with the CONNECTACCEPT, and check whether you get a response back from the client

I obviously get, because the tokens match

then continue with the next message. do you get the next message?

which one is the next?

how would one go about finding that out?

i guess its the netmsg_info?

I think so too

no wait

15:01

in ddnet its the clientver thing

15:01

in preauth state

oh yea, in new ddnet, true

so i need to take care of that too?

first you'd probably check whether the message even makes it to the relevant part of the server code

15:02

before trying anything further

15:03

always try small steps of things that definitely have to work

yea but I think I will not finish this working anyways... And if I get it the code is really ugly

15:04

What do you use DirectInit for? why not the normal path using Feed(

15:04

)

15:04

I have too many questions

you oculd treat it as a learning opportunity and try to understand the code

15:05

idk what DirectInit is either

it initializes the netconnection witout going through Feed to check the msgs again etc

15:06

I should maybe structure the Recv() part of network_server the same as in ddnet

15:06

with all the functions and Slot

the general code structure is relatively irrelevant if you understand what is sent over the network

It's rather difficult to work with when the code structure doesn't match what's send over the network

15:17

sent*

I disagree here

15:18

the protocol is trivial to describe, it should fit fine into any code base

15:18

it's just three messages and only the last one causes state to change

eh, look at the mess that is the 0.7 compatibility

15:20

The token protocol of 0.7 doesn't match 0.6 at all and it sticks out like a sore thumb

example?

15:22

and can you explain the difference between the 0.6 and the 0.7 protocol here?

15:22

0.6 sends CONNECT, CONNECTACCEPT, ACCEPT over the network

15:22

the 0.7 equivalent is TOKEN, TOKEN, CONNECT, CONNECTACCEPT <- okay, here's an extra step

15:23

but you only have to cut it short by one step for 0.6

15:23

if you start from 0.7

Well the fact that detecting 0.7 requires us to unpack the message first makes it even worse

15:26

but take a look at network_server.cpp where we had to unpack the packet twice

sounds like a problem with the code, not with the concept

15:27

I think I described how to do it with unpacking only once

15:28

by pattern matching unconnected messages against 0.7 TOKEN and CONNECT, and treating the rest of the unconnected messages as 0.6 (edited)

It'd take some very clever restructuring of the code to get it to look nice

not really

the point stands, the code doesn't match the protocol so it's a PITA to work with

15:28

Pattern matching is hardly cute

it's how e.g. the "ssh and https" on one port works

(which is a very ugly thing imho)

if you want to have it nice and clean you could do it on a second port

At which point the code would match the protocol we are handling but we'd get a lot of duplicated code in return, to fix that you'd need to clean up CNetBase and CNetServer a fair bit

otherwise something like pattern matching is practically required I think

15:31

what?

15:31

you can do the pattern matching very nicely

15:31

idk what you're talking about, it fits into the existing code. okay, call it unclean

Feel free to implement it, it's open source after all

why would I?

15:32

you just claimed some impossibility and I claim possibility and that means I have to implement something now?

I couldn't get it in there in a way that didn't make it look out of place, neither could timakro come up with a cleaner way to do it

I can do it in rust one day, when I wnat it there

I never claimed an impossibility, I just said "It's rather difficult" and that the "code ends up ugly"

would some pseudocode help?

Just because I couldn't get it clean obviously doesn't mean it's impossible, though I'm flattered you think that highly of me

15:35

@heinrich5991 sure, don't make it too abstract though

bonjour hello hi, is it possible to add a feature to ddnet demo playing? So that you can rewind the demo? Or a "jump back 10sec" button?

15:42

rewind= play backwards

it already exists

15:44

left and right arrow

pro

15:44

ty xD

15:44

but rewind not in the client, right?

15:45

oh is it possible to change the time that gets jumped by pressing the arrow? so I could maybe change from 5sec to 10sec

wdym

15:45

j and k for 10 sec

playing the demo backwards

playing a demo backwards is not supported

or k and l

alrighty :)) @Patiga plans to implement it?

rewinding -> going to an earlier timestamp is supported

ye

i dont think its planned

playing backwards could be a fun coding challange

@jao what is the bind for 5 or 10sec jump? or cant I bind it to another key?

you could convert the demo and play the video backwards ^^

15:48

you can render a demo to mp4 in the client and play the video file backwards if that is your goal :)

you cant rebind it

15:48

i think its j and k for 10s jumps

@Patiga ah nono I just like to rewind videos to see again what happened xD just preference over jumping to a certain point. but doesnt matter that much

15:51

okay thank you :) 5 or 10sec jumps do that job just right. tyty

i think ddnet7 is not working anymore since https://github.com/teeworlds/teeworlds/commit/293209e7227c81e92b99c89890ebfdc264374a4a :/

do a proper bounds check for types when unpacking the snapshot delt...

… closes #2649

That breaks the protocol extension

send help @redix :D

just remove ddnet7

Well that can be fixed without breaking compatibility. Oy probably didn't think/care about mods

Ofc not

17:40

Not even surprised

[ddnet-web:master] 2 new commits

68d3518 add louis - 12pm 955ff5c dont allow renames of banned names - 12pm

droguis

3

droguis

3

extremenouis

3

extremenouis

3

noouis

3

noouis

3

noouis

3

nouis

4

nouis

3

nouis

3

nouis

3

nouis

3

penouis

3

penouis

3

zoozuis

3

zoozuis

3

rect1840

3

poois

3

poois

3

poundis

3

https://ferrous-systems.com/blog/mindsets-and-expectations/

Learning Rust: Mindsets and Expectations

Ferrous Systems is a Berlin based Rust Consultancy

jao added nouis

eh i always thought a bad harddrive means slow io ops but stuff like typing should be fast. but if the os is on a slow driver ervythign is slow huh?

20:35

i have like 10 fps while typing in my gnome terminal on tha old hdd i found in my basement xd

isnt typing input

Exported 114 message(s)