DDraceNetwork - developer

may i ask someone check this cmakeerror

message.txt

4.14 KB

00:26

i have wavpack.h

00:27

but he say

C:\Users\cen940k\source\repos\cen94ok\xpanic\build\CMakeFiles\CMakeTmp\CheckSymbolExists.c(2): fatal error C1083: Не удается открыть файл включение: wavpack.h: No such file or directory,

00:27

Не удаётся открыть файл включение is the same as could not open the include file

00:27

pls someone

https://www.youtube.com/watch?v=_RIPrZtl_vw I don't get this vid

deepfreez [TeeWorlds]

Evolution of teeworlds-DDnet (client) for 10 years

10:16

You can see me doing turkish translations in 2:44

@heinrich5991 does your wireshark dissector still work?

well at the very least packet_decode is completely unaware of these new packets

Ah I see why timakro kinda hacked around the token cache

wireshark dissector doesn't work for 0.7 packets, no

11:53

I don't implement the 0.7 protocol

I wonder if we can get the ddnet token protocol compatible with the token cache

what's the issue?

Issue is that I don't even know where I'd start that as it's two completely different approaches

11:55

e.g. the ddnet one doesn't have the concept of peer tokens

11:58

I don't even think we do tokens for our connless partners

yes, that's true, not on the protocol level (edited)

@heinrich5991 can you use a procedural macro on a binary program in rust?

yes, sure

12:09

a binary crate is not that different from a library one

but to use proc-macro

12:10

it requires [lib]

12:10

error: the #[proc_macro_derive] attribute is only usable with crates of the proc-macro crate type

12:10

that's for declaring a proc macro

i want to declare a proc macro

12:10

and use it

you have to declare it in a separate crate

meh

that cannot be done in the same crate as the library/binary that uses it

why?

I'd guess because it'd cause a two-stage building process otherwise

12:12

if it were declared in the same crate as it's used, you'd first have to compile the proc macro code to be able to read the rest of the crate

[regsixup]: WARNING: No master servers. Retrying in 60 seconds this is one part I didn't think would require any work given I didn't touch it

12:19

talked too soon

HAHAHA

12:23

okay cool

12:24

so now i am not the only one anymore having problems with 0.7 master

Mine more than likely stems from some token mess

Mine comes from the rcon login

12:28

because it sends all the rcon cmds

12:28

maybe thats a reason for u too

Nah, found it, I'm having trouble with the 50 different kinds of token that need to be juggled around

12:31

[2020-06-16 15:31:29][regsixup]: chose 'master3.teeworlds.com' as master, sending heartbeats \o/

good job @Learath2

@Learath2 wdym

def-

[ddnet/ddnet] Issue opened: #2267 Handle missing map from server-side demo better

teehistorian-created demos contain no map file, so they fail to load, but nothing is shown to the user: ``` [2020-06-16 14:38:43][client]: loading map, map=Barren wanted sha256=7f43f924a26a9a25b2c009bba48bd019fde0e8bdf36f83d6f4ecf8280b2aa000 crc=30082080 [2020-06-16 14:38:43][datafile]: loading. filename='maps/Barren.map' [2020-06-16 14:38:43][datafile]: could not open 'maps/Barren.map' [2020-06-16 14:38:43][datafile]: loading. filename='downloadedmaps/Barren_30082080_7f43f924a26a9a25b...

@fokkonaut when you Recv a message, there is the peertoken, there is your token, there is the ddnet token

12:44

I was sending the wrong one along to the register code

ok

nooooo the FWOK is lost

see

aaand idk what's wrong

13:09

great

@deen Could you please add a score limit voting option to zCatch SA servers

Lots of people are requesting it

@fokkonaut do you know how I can disable all the masters but one so I can actually debug this?

nope haha

amazing usability by matricks

13:12

fine I'll patch it out

iptables

what is the point of a masters.cfg if you'll just ignore it...

idk lol

xD

it choses a master, gets a token, sends a heartbeat, has a stroke and asks for another token

yea, I saw a similar pattern in the pcap from the guy hosting @fokkonaut's servers

13:23

but if you could also put gdb to work, we can probably find the root cause @Learath2

13:23

maybe just patch MAX_MATSERVERSERS to 1 or so

Yeah already did that and used master4 so I can tcpdump on ddnet.tw if needed

not really necessary, you could tcpdump locally as well

but I doubt my issue is related to his, the fw?? never seems to make it to CRegister for me

could you log whether messages are dropped due to invalid tokesn?

hm, I don't understand what timakro did here, but I don't think the masterservers messages token is ever verified

@timakro ^

Hm?

tbh I don't even know what token is going where anymore

@Learath2 connless packets contain two tokens

Me neither, in the beginning I tried choosing two consistent names for the client and server tokens but I gave up after a while so if it's called ReponseToken or SecurityToken or Token, it means nothing

one token for "how can you answer me"

13:31

and one token for "this authenticates that I do own the source address"

hm the fw?? never makes it out of Recv anyway

13:35

Ah

13:35

I even left a note for this, yes it's dropped for an invalid token

The way I get the tokens (don't ask me which one I don't remember) from the NetworkServer::Recv function to the CServer::PumpNetwork function is ugly

As I inherited it from you, mine is also ugly

13:38

The UnpackPacket twice isn't too pretty either

It was the only way without major changes

but none of this mess is easy to clean up, I wasted like a month trying to clean it up

13:38

took a week to get things to here as soon as I gave up on the low level networking ever being clean

13:40

Okay, the random requests for token are still there though

13:40

but it registers

@TwItCh what numbers are reasonable? 10, 20, 600, 1000?

@heinrich5991 isn't this new connless protocol very noisy? Every time someone needs serverinfo they need a 520b token message

isnt that the point

that's the idea I think

to make it harder to ddos

the request has to be >= the response

to avoid amplification attacks ye

To this day I don't understand why the master server register protocol works without this line in 0.6 https://github.com/unique-clan/unique-race/commit/160d35d1bd3e90fe0ae89a1fae64122ab7e2d17a#diff-2b61fb3875a3e845adc670e15aa20fb7R308

Serverinfo and masterservers for 0.7 · unique-clan/unique-race@160d...

13:45

The state machine is just broken without it

probably some race condition that's more often taking the correct path than not

I don't know if this fixed it but master server connections always time out and take like 3 tries to get going