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 2020-03-13 00:00:00Z and 2020-03-14 00:00:00Z

@heinrich5991 why did you use #define GET_INT(integer) (integer) = str_toint(Up.GetString()) instead of Up.GetInt()?

because I figured it would be easier to implement for third-party software

14:06

feel free to change it, I think in 0.7 they changed it for the normal protocol as well @Learath2

heinrich5991: I have a very tough issue with fixing the protocol compatibility in 0.7, do you have a second to brainstorm?

Learath2: what's the issue?

Old versions trust m_NumClient to be the amount of clients we send. But we cant know how many clients fit in the packet before we craft it

14:10

I had a very meh idea of crafting the packet until we add num clients, craft the rest of the packet and drop clients until the extra int for numclient fits, however you know that's very ugly due to the way the packer works

14:12

The only two ways I know of to pop off clients is either finding a way to iterate the packer backwards or keep note of the offsets of every client I add, both are ;(

returning different server info depending on the requesting client is not an option?

@heinrich5991 problem is how many clients we send to an oldclient

ah

14:15

remember offset of last client

14:15

after each client check that the int + header + clients still fit

14:15

if not, return previous packet

14:15

(use CVariableInt directly)

Oh I'm stupid, I tried handling it after I filled the packet

14:17

Ofc I should handle it during filling it. Thanks @heinrich5991

you're welcome

Two heads are better then one :P

my uni postponed the next semester by one week, corona

14:18

or rather my state (berlin) did that

A month for italian unis

ah

14:19

do you have other effects?

Well I'm completely stuck in istanbul

oh

But friends who are stuck on the other side in italy report that the streets are completely empty and the vibe is pretty meh

got a sensible computer at least?

@heinrich5991 sensible enough for working on some coding

14:21

I don't deserve to play games anyway so I don't mind

~~

@heinrich5991 how is berlin? Italy is supposedly the worst in europe but idk

berlin looks fine

14:23

people go about their day mostly as usual

14:24

large events are cancelled, at least the uni tries to encourage people to work from home

14:24

btw, are you into coding challenges?

14:25

I found some nice ones: https://cryptopals.com/

@heinrich5991 r u doing the ctf that liveoverflow announced?

no, link?

liveoverflow is a fun watch, I thought the guy was @heinrich5991 for a while

nowhere near that skilled

Wtf @Learath2

14:27

https://youtu.be/kLZdnvvQRKY

We are Organizing a CTF! - CSCG Announcement

oh

14:27

cscg

14:27

I considered participating this year

I tried but too lazy

doit

The Perl challenge is so frustrating

@heinrich5991 he might be more acquainted with ctfs but dont undersell yourself :P

you're flattering me, thanks

14:31

I've never really done any serious RE unfortunately

14:31

I should kick myself a little more. I'd love to do some hardware disassembly/RE

I'm actually also really interested in some hardware work, especially "security" devices

I'd be fine with non-security devices for starters

That ring buffer attack on most garage doors was really interesting e.g.

given that I have no experience

14:37

ring buffer?

Not a ring buffer, soz. A shift register

shift register?

Turns out many garage doors out there use a shift register to store the transmitted code, which allows you to reduce the amount of transmissions needed by a whole lot which made all of them very easy to bruteforce

I don't know what a shift register ist

14:43

*is

Ah, it's a type of register you can shift bits into, and out of one at a time. Like a queue of bits

ah

Problem was that the cpu was checking the code every time a bit is pushed and not after getting all 12 bits

14:46

You generate a string of all possible 12 bit combinations with overlap and transmit that a couple times instead

14:47

Turns out to be a lot less bits then without overlap. The guy figured it out with one of those logic analyzers

14:47

I think there was a youtube video where he describes how he devised the attack

[ddnet/ddnet] Pull request opened: #2083 Add Discord mod log

Logs usage of ban, kick and muteid to a discord channel with a webhook !image

@Learath2 happen to have the link?

15:01

also @Learath2 mind merging the antibot stuff? ^^

@heinrich5991 I'll find you the link when I get home, I'll review the antibot stuff as soon as I'm done with this extended serverinfo patch for 0.7

15:02

@heinrich5991 is there a way to estimate the size of a packed int cheaper then Packing it?

just pack it. should be optimized out

15:03

(if you pack it into a buffer that is not used elsewhere)

You sure? i'd need to pack it and then subtract the buffer from the end of the packed data, is the compiler clever enough?

15:04

well I guess if any compiler is smart enough it's a C/C++ compiler, those are so aggressive with optimisations

cant u assume max size (4 bytes per?) if its only an estimate

https://godbolt.org/z/LYWgCt

Compiler Explorer - C++ (x86-64 gcc 9.2)

unsigned char *CVariableInt_Pack(unsigned char *pDst, int i) { *pDst = (i>>25)&0x40; // set sign bit if i<0 i = i^(i>>31); // if(i<0) i = ~i *pDst |= i&0x3F; // pack 6bit into dst i >>= 6; // discard 6 bits if(i) { *pDst |= 0x80; // set extend bit while(1) { pD...

15:07

seems so

15:07

that it's optimized out

@heinrich5991 one more little bug and I can review yours, do you know of a way to append a packer to a packer?

15:18

I tried adding the second packer as raw, but that doesn't seem to work :/

AddRaw(.Data(), .Size()) (edited)

15:18

should work

Ah I forgot to reset the packer

the liveoverflow is apologizing for the name

15:20

rightfully so

NOOO, old clients don't trust the m_NumClients so the server looks as full as the amount of clients we send

15:21

@heinrich5991 is that an acceptable shortcoming? I don't think I can fix it

I guess

wasnt this a problem during ddos attacks

15:22

when the server wasnt sending players

15:22

and then not showing up

they "fixed" it in 0.7

15:22

so you can't send short packets anymore

eh

Well given in the past the server wouldn't even show up if the clients didn't fit in the packet I think this is acceptable

when in reality they only helped mitigate dos attacks

15:22

@Learath2 yes

i guess im thinking of a different problem then

15:23

but the one i describe is where the client tries to determine playercount by counting the players the server sends

15:23

which doesnt work when the server omits player list under heavy load

@heinrich5991 https://github.com/teeworlds/teeworlds/pull/2492 I'm done

Extend serverinfo by Learath2 · Pull Request #2492 · teeworlds/teew...

This isn't the same extended serverinfo protocol heinrich defines. However the only feasible way to get the entire thing in would be a NETVERSION bump or breaking compatibility. I vote we l...

15:30

Take a look at mine, I'll take a look at yours

1

I won't merge yours though(?)

15:31

oy does that

15:33

@Learath2 why do you drop the more-players-packet if it arrives before the main packet?

@heinrich5991 because the entry doesn't exist by then and I have no idea how to handle that

15:35

and yeah I know you won't merge, just looking for a review

I think™ ddnet handles that

15:35

maybe you can take a look at that

15:36

for the protocol extension, I'd add a UUID, and scan the whole packet for it

15:36

just so that you're compatible with other (possible, unlikely to exist) modifications

@heinrich5991 this isn't really a whole extension, I'm not sure if we can extend the protocol without breaking compatibility

15:39

I kept the initial packet the same so the old clients can also use it, I have "giex" for requests from the serverbrowser but that's not really a thing for ingame serverinfo updates, we'd need some way to keep track of the version of the client and whether it supports the new protocol, is that a thing vanilla wants? I'm not sure if oy would merge an extension like that

15:40

@heinrich5991 why do you have a separate ANTIBOT_MAX_CLIENTS? Shouldn't you tie that into the engines MAX_CLIENTS so there is no chance it falls out of sync?

I have a separate ANTIBOT_MAX_CLIENTS so that the API doesn't accidentally become incompatible

15:40

hm

15:41

maybe I should add a test that they're the same and add a comment to recompile all anticheat modules if you change it

15:44

given that there are probably no protocol extensions that you could be incompatible with, I guess this is fine

15:44

(re your PR)

Your pr LGTM, the last concern I have is how we'll build the official servers

how do we build official servers?

Currently we just git pull and build, I think we need a new (private)repo for the antibot code

nah

15:45

antibot can be built sporadically and the .so can just be copied in after the build (edited)

Does this generate an antibot.so?

the build? yes

15:46

if you add ANTIBOT=ON to the command line

so we just don't copy the one built along with the server and use @noby's module instead

yes

Great, I don't see any issues then

\o/

:D

hi ping

hi

15:47

cool to see this geting implemented

We still should have a private repo with the antibot code we can all work on

currently it's an almost-not-modified version of noby's

gitlab?

just includes

15:49

and I believe no one except noby really wants to change that code

@ @noby

I can merge this tonight and we can try deploying it. Will add a new script to ddnet-scripts to pull from a repository and build the antibot too

i plan on making updates

15:49

ive been constantly updating it for the past couple years rly lol

I do trust @noby but I'd still not want to run code we can't see on ddnet servers

yes sure

15:50

just a note regarding modification (edited)

ye if we make a private repo for the 3 of us or whichever other admins then its fine as long as the code never leaves that repo

Github private repo?

would work up to a couple of members, not as a subrepo of ddnet though

i use gitlab since free private repos

Yep, I could own the repo, ddnet org doesn't have private repo allowance

oh, you even have a subscription (edited)

nah, I'm a student

15:54

we could host the repo on gitlab aswell if you prefer that

15:55

(I'm a student for atleast 2 more years so that'd be the lifetime)

its about the same right

15:56

other than offering free private. repos

yes, github too, for up to three (four?) collaborators

is that new

15:57

i only moved from gitlab to github because of the private repos thing

15:57

if yall can do it on github then fine by me

Okay, I'll create the repo on github, add y'all as contributors

:D

15:58

so when i make fixes to the anti bot code

15:58

i will push them there? and theyll be added whenever is next convienent time

exactly

sounds good

can we push our glue code there, as well?

glue code?

@noby

15:59

yea, glue code from the other side

15:59

the code was fucking entangled

like the code that connects the

15:59

lol

well it'd be nice if you could use the .so on your servers aswell @noby

yes heinrich made stuff a lot cleaner in the interface

16:00

lol

@Learath2 private repo when

and at some point i wana use wat he did

16:00

on blocker at least, fng is a different repo cus laser stuff

I mean I could try to help you with that code as well, if you want

16:01

so we can have one unified codebase

16:01

❤️

@noby what is your github name?

heartw

16:01

https://github.com/nobody-mb

nobody-mb - Overview

nobody-mb has 13 repositories available. Follow their code on GitHub.

@heinrich5991 you can push whenever

16:03

I have horrible rebound rhinitis and it's giving me headaches :/

16:06

@heinrich5991 how could ddnet handle the main packet not arriving first?

16:06

Do you queue up the packets based on NETADDR or sth?

storing the players for that IP address

16:06

no

I see no easy way to do that

[ddnet:master] 3 new commits

3c64bed .gitignore shared libraries - heinrich5991 a3163c5 Add support for dynamically loading an antibot module - heinrich5991 fdf7e4e Merge #2081 - bors[bot]

17:14

[ddnet/ddnet] Pull request closed: #2081 Add support for dynamically loading an antibot module

feelsamazingman

1

@heinrich5991 let me try building this on the server

17:16

see if it loads antibot null

@Learath2 I already modified stuff on ddnet.tw

17:16

go there plx

17:17

(only added flags, maybe to git clean if you want a clean slate)

17:17

ah, be careful with git clean I guess

17:17

17:17

git checkout -- build.sh

I'm trying it out on ddnet.tw btw

17:18

just there for now

k

17:19

food

@heinrich5991 ddnet.tw:8370 now runs antibot null

@heinrich5991 all the servers are going to be running antibot null in a second, but there is an issue

17:45

the server binary links to libantibot.so in the src folder, seems it is linked with absolute path, you need to do sth there to not do absolute paths but I'm no dynamic linkingwizard

17:47

ddnetger ldd DDRace64-Server_sql | grep antibot
    libantibot.so => /home/teeworlds/src/master-build/libantibot.so 
(0x00007f174f4ff000)

@heinrich5991 other servers won't load the antibot or the debug message is somehow getting lost, I need to go

17:58

please take a look

uh oh xd

It compiled fine and I checked that it also linked fine and it can find the dynamic lib, checked it through ldd

18:03

And if I run the server manually it does load the antibot, however if i let the server restart it wont

18:03

Couldnt figure it out and had to leave

Is antibot also on 0.7 servers?

@ChillerDragon no. antibot isn't really on 0.6 either, it's just a test phase right now

18:55

the changes needed can be seen here: https://github.com/ddnet/ddnet/pull/2081

Add support for dynamically loading an antibot module by heinrich59...

Support for this feature is turned off by default. This feature allows to integrate with noby's anticheat feature.

@heinrich5991 I think it broke something and I dont have my laptop with me, can you please take a look?

I only have a couple of minutes

19:01

@Learath2 restart doesn't work, you say?

Restart doesnt work but more importantly @nexusrf says votes are broken

~~

I think you should just roll it back if you dont have the time to diagnose

what to do to rollback?

19:03

./build after checking out the commit?

Yeah votes are broken, i can't change map rn

@Gwendal server, port?

@heinrich5991 I think so, am not completely sure

ddnet ger 8309

@Learath2 change is applied everywhere?

jao tested with admin command and it didn't work aswell

looks like a crash on map change

19:05

will roll back

I think I have a build-fixup.sh there for this occasion

looks good, ty

19:08

@Learath2 what do I need to do after executing this script?

19:10

nothing, I guess

Can you check the script, im not a 100% sure if i do the copying

you do

19:11

checked the script

Thanks, had to leave urgently

@Gwendal ger2 should be fine now, can you check? maybe try twice

ok

19:12

1st time didn't work

19:13

ok

19:13

2nd worked

19:13

thanks!

nice, ty

19:13

k, empty servers restarted, should be good now

19:14

classic "didn't test the map change" bug from me

19:14

sorry

19:14

I should add that to the test suite

I wonder what the crash is, you dont seem to change anything relevant

the server crashes after the map change

19:14

probably some data structure in the antibot that thinks some player is still there

19:16

will take a look tonight, have to leave now

ye it just go back on the map that was here before voting

19:21

and it says u got timeout

Exported 267 message(s)