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 2018-08-29 00:00:00Z and 2018-08-30 00:00:00Z

[quakenet] <heinrich5991> Learath2: it has that

06:09

[quakenet] <heinrich5991> see the downloader client (it's maybe a bit unwieldy :/ )

[quakenet] <deen> heinrich5991: I mean just a client that gives you chat on the command line and lets you use console and rcon

[quakenet] <Learath2> deen: I always wanted to code one but never had the time, should be pretty easy with libtw already doing the work tho

[ddnet:master] 1 new commit

a63f210 Fix segfault when deleting lock - def-

[ddnet:test] 1 new commit

2e5f643 Test - def-

09:11

[ddnet/ddnet] New branch created: test

09:13

[ddnet:test] 1 new commit

c02c0a4 Forbid copying classes in threading - def-

[ddnet:master] 2 new commits

925f7b6 Fix weird Windows compilation - def- c9f28ce Forbid copying classes in threading - def-

@deen ath has such a future. You only see the console, if i am right. Probably not what you are looking for but i thought it was worth mentioning.

[quakenet] <deen> onby also has it, so no need to rewrite it 3 times over

[ddnet/ddnet] Issue opened: #1280 Why can't I add the header here?

https://github.com/ddnet/ddnet/commit/925f7b6325ad542f5f7747d11fc771dae3d5aac8 Windows compilation fails: https://ci.appveyor.com/project/def-/ddnet/build/1.0.1450

10:57

[ddnet/ddnet] New branch created: test2

10:57

[ddnet:test2] 1 new commit

e8cac83 Test (fix #1280) - def-

@deen onby doesnt want to share his source thats why was creating my own version. And i successfully kicked the graphics and it also is able to connect but it somehow crashes after 2 secs and i have no idea how to debug that

[ddnet/ddnet] Pull request opened: #1281 Test (fix #1280)

11:50

[ddnet:test2] 1 new commit

9423c1a Test (fix #1280) - def-

why does base/system.c defines functions that are already defined?

11:54

void mem_copy(void *dest, const void *source, unsigned size)
{
    memcpy(dest, source, size);
}

11:54

is it for style convention? compatibility with something?

@ChillerDragon gdb

11:56

convention and being able to add something in there later on

the good old gdb advice .-.

12:58

any spot where i can learn how to master gdb that i finally understand the output?

@noby start sharing already thonkery

thonkery

[quakenet] <heinrich5991> ChillerDragon: you need very few commands to use gdb successfully

13:05

[quakenet] <heinrich5991> gdb ./DDNet

13:05

[quakenet] <heinrich5991> to start a debugging session for the client

13:05

[quakenet] <heinrich5991> then in the gdb prompt:

13:05

[quakenet] <heinrich5991> run

13:05

[quakenet] <heinrich5991> then make it crash

13:06

[quakenet] <heinrich5991> then run bt

13:06

[quakenet] <heinrich5991> this shows you where it crashed (and all the functions that led to this crash)

13:09

[quakenet] <heinrich5991> this is basically it. more commands: print variable: tells you the value of variable (can also be an expression), up, down goes up and down in the call stack (e.g. to print variables from different functions)

break File:LineNo or break FuncName sets breakpoints

[quakenet] <heinrich5991> oh yea, that one is also useful

13:19

[quakenet] <Learath2> c continues after hitting a breakpoint, n steps one line, s steps one line but if it can go deeper it'll go deeper, r starts the program again

13:19

[quakenet] <Learath2> that's about all you need :)

13:19

[quakenet] <heinrich5991> (continue, next, step, run; run starts the program from the start again)

[quakenet] <Learath2> oh also very useful while stepping is if you just hit enter it repeats the last command

13:34

[quakenet] <heinrich5991> l(ist) shows you the surrounding code lines

[quakenet] <Learath2> One last thing if you launch gdb with gdb -tui you get code view on top :)

13:54

[quakenet] <deen> no need to use -tui, just use ctrl-x a to toggle the TUI

13:54

[quakenet] <heinrich5991> (the point, however, is that you need very little to actually debug a segfault: gdb ./DDNet, run, bt

13:54

[quakenet] <heinrich5991> )

13:55

[quakenet] <deen> And you have to understand what a backtrace is

14:01

[quakenet] <deen> And if something is unclear, the documentation is good: https://sourceware.org/gdb/current/onlinedocs/gdb/

Debugging with GDB: Top

Debugging with GDB: Top

14:02

[quakenet] <deen> A few months ago I discovered that Python also has pdb: https://docs.python.org/3/library/pdb.html

[quakenet] <heinrich5991> nice

14:14

[quakenet] <heinrich5991> I've never looked far enough to find a python debugger, it was too easy using printf-debugging ^^

14:14

[quakenet] <heinrich5991> thanks for that hint :)

@heinrich5991 connection packets are not supposed to be compressed right?

[quakenet] <heinrich5991> yes, control packets are never compressed anywhere, but vanilla supports decompressing control packets

I tried to use CNetClient from outside tw, it seems to compress the packets

15:42

what prevents their comression normally?\

[quakenet] <heinrich5991> huh

15:45

[quakenet] <heinrich5991> it seems compression is enabled

15:45

[quakenet] <heinrich5991> let me check vanilla

I'm fairly certain the packets we receive on the servers are not compressed, we can them out at the firewall

[quakenet] <heinrich5991> okay, it does allow compression

15:46

[quakenet] <heinrich5991> it's just that the compressed packets are usually longer, I think each compressed packet has at least 3 bytes of data

15:46

[quakenet] <heinrich5991> whereas the normal control messages have 1 byte of data, uncompressed

Well somehow mine "compresses" it to 1 byte which is insane, so I'm guessing I need to initialize something about huffman

[quakenet] <heinrich5991> yes, huffman needs to be initialized

15:49

[quakenet] <heinrich5991> call CNetBase::Init()

15:49

[quakenet] <heinrich5991> (or construct a CEngine)

15:51

[quakenet] <heinrich5991> Learath2: here's the compression table: https://github.com/heinrich5991/libtw2/blob/master/doc/huffman.md#appendix

heinrich5991/libtw2

Some Teeworlds stuff in Rust. Contribute to heinrich5991/libtw2 development by creating an account on GitHub.

yay, that fixes some of it

15:51

heinrich5991: How long is the securitytoken? 4 bytes?

[quakenet] <heinrich5991> yes

15:52

[quakenet] <heinrich5991> +TKEN

15:52

[quakenet] <heinrich5991> (another 4 bytes for the CONNECT message)

15:52

[quakenet] <heinrich5991> note that there is no support for that message in libtw2 yet

huffman compression is used on the whole payload?

[quakenet] <heinrich5991> yes

15:54

[quakenet] <heinrich5991> on everything except the three bytes of teeworlds packet header

hmm heinrich5991 could dbg_lognetwork not be updated to log properly with security tokens?

[quakenet] <heinrich5991> mh?

00 00 00 00 0C 00 00 00 10 00 00 02 54 4B 45 4E 02 CE E9 05

15:59

00 indicates that it's raw, 00 00 00 0C indicates it's 12 bytes, there are more then 12 bytes remaining

16:00

however this is insane, 00 00 00 0C is printed by io_write(Size, sizeof Size); and io_write(Data, Size);

[quakenet] <heinrich5991> b'\x00\x00\x00\x00\x0c\x00\x00\x00\x10\x00\x00\x02TKEN\x02\xce\xe9\x05'

16:01

[quakenet] <heinrich5991> wait what?

16:01

[quakenet] <heinrich5991> not &Size?

it' &Size, I miswrote

[quakenet] <heinrich5991> so what's insane?

a 12 is written, and the next command can only possibly write 12 bytes

16:03

so where do the rest of the bytes (which obviously belong to the packet and are the security token bytes) come from?

[quakenet] <heinrich5991> I only see 12 bytes

16:03

[quakenet] <heinrich5991> 00 00 00 00 (type)

16:03

[quakenet] <heinrich5991> 0c 00 00 00 (size)

Oh I am a moron

[quakenet] <heinrich5991> little endian? :)

endianness

16:07

heinrich5991: 10 00 00 00 02 CE E9 05 do you happen to know which packet this is?

16:07

10 00 00 00

[quakenet] <heinrich5991> that looks like a keepalive

16:07

[quakenet] <heinrich5991> sent when nothing is happening

16:07

[quakenet] <heinrich5991> I think every half a second or so

so if I start getting keepalives is it safe to assume I have connected successfully?

[quakenet] <heinrich5991> yes

16:08

[quakenet] <heinrich5991> however in a normal connection, you'll never receive KEEPALIVEs

xD

[quakenet] <heinrich5991> because there's always an outstanding conversation (log on) or you're already in game so you're receiving snaps every 0.04 seconds

16:09

[quakenet] <heinrich5991> Learath2: what are you doing? :)

coding a chat only client

[quakenet] <heinrich5991> why you not using libtw2? :)

deen wanted one that is a target, and I wanted to see how spaghetti the tw code is

[quakenet] <heinrich5991> ah, like a part of ddnet?

and my rust is pretty rusty (ha)

16:12

yeah make text_client

with rust?

16:12

ayyy

not with rust, with C

16:13

well C+(+?)

16:13

@heinrich5991 so after connecting, the client is supposed to send 10 00 00 01?

[quakenet] <heinrich5991> no, that's the first thing the client sends, that's the CONNECT message

So wait, I am getting keepalives, the server is definitely expecting me to send sth

16:15

is there a protocol doc somewhere that outlines the process?

16:15

NETMSG_INFO?

[quakenet] <heinrich5991> lol

16:15

[quakenet] <heinrich5991> but yes, I wrote a bit on that

16:15

[quakenet] <heinrich5991> only a tiny bit though

16:15

[quakenet] <heinrich5991> but lol docs

16:16

[quakenet] <heinrich5991> https://github.com/heinrich5991/libtw2/blob/master/doc/connection.md

heinrich5991/libtw2

Some Teeworlds stuff in Rust. Contribute to heinrich5991/libtw2 development by creating an account on GitHub.

16:18

[quakenet] <heinrich5991> these are the system messages: https://github.com/heinrich5991/libtw2/blob/97c8d8291f81f4e742bd458e84052b009bef7206/gamenet/generate/msg_system.py#L3-L25

heinrich5991/libtw2

Some Teeworlds stuff in Rust. Contribute to heinrich5991/libtw2 development by creating an account on GitHub.

16:18

[quakenet] <heinrich5991> the game messages can be found in datasrc/network.py in your ddnet source code

16:20

[quakenet] <heinrich5991> Learath2: questions?

can u help me with the first 8 bytes? xD

16:33

Image attachment

16:33

9th byte is client_id from the writer

16:33

then the message follows

16:33

and last 4 bytes are some footer

[quakenet] <heinrich5991> I should really finish my wireshark dissector some time

xD

[quakenet] <heinrich5991> https://github.com/heinrich5991/libtw2/blob/master/doc/packet.md gives information about the first 6 bytes

heinrich5991/libtw2

Some Teeworlds stuff in Rust. Contribute to heinrich5991/libtw2 development by creating an account on GitHub.

thx

[quakenet] <heinrich5991> next byte is message ID

16:38

[quakenet] <heinrich5991> that's 6/2=3, i.e. the third item in the list https://github.com/ddnet/ddnet/blob/c9f28ce0941ef1727242005ff1cd18c4b741aef3/datasrc/network.py#L230

DDraceNetwork, a mod of Teeworlds. Contribute to ddnet/ddnet development by creating an account on GitHub.

16:38

[quakenet] <heinrich5991> so it's a Sv_Chat message

16:38

[quakenet] <heinrich5991> containing two integers and a string

why /2 though

[quakenet] <heinrich5991> (also, capturing bytes as picture, really? :P)

16:39

[quakenet] <heinrich5991> the number is either even or odd

16:39

[quakenet] <heinrich5991> if it's even, it's a game message and /2 is the ID

16:39

[quakenet] <heinrich5991> if it's odd, then it's a system message and /2 (rounded down) is the ID

hm ok

hmm, I really should have just used libtw

16:40

why does network,cpp depend on a config being available? :/

16:40

atleast it doesn't request it from a kernel

16:41

at that point I might aswell work backwards and remove stuff from the actual client

lol

16:42

what is libtw2 btw

[quakenet] <heinrich5991> some building parts for teeworlds, rewritten from scratch, in rust

16:42

[quakenet] <heinrich5991> can read maps, demos, teehistorian, can decode/encode messages, has a working map downloader and a bare-minimum server

heinrich5991 any tip on getting config working?

[quakenet] <heinrich5991> no, not really

16:45

[quakenet] <heinrich5991> src/engine/config.h seems to be the interface

16:46

[quakenet] <heinrich5991> haven't really worked with that a lot

eh, just created a kernel and registered config in it

16:51

kernel itself depends on storage, yeah this is all pretty spaghetti

well got it connected and sending messages, now just need to set up a send buffer

wow thanks a lot for the gdb instructions ima test them later c: !!! ❤ @Learath2 @heinrich5991 @deen

bt full

17:34

for more detail

17:36

 Thread 1 "zcatch_srv_d" received signal SIGSEGV, Segmentation fault.
0x000055555558ca66 in CRingBufferBase::Allocate (this=0x7ffff6ca73b8, Size=71) at src/engine/shared/ringbuffer.cpp:109
109                pNewItem->m_pNext->m_pPrev = pNewItem;
#0  0x000055555558ca66 in CRingBufferBase::Allocate (this=0x7ffff6ca73b8, Size=71) at src/engine/shared/ringbuffer.cpp:109
#1  0x0000555555573b6a in TStaticRingBuffer<CNetChunkResend, 32768, 0>::Allocate (this=0x7ffff6ca73b8, Size=71) at src/engine/shared/ringbuffer.h:59
#2  0x0000555555572b49 in CNetConnection::QueueChunkEx (this=0x7ffff6ca7398, Flags=1, DataSize=31, pData=0x7fffffffcf00, Sequence=416) at src/engine/shared/network_conn.cpp:128
#3  0x0000555555572c4b in CNetConnection::QueueChunk (this=0x7ffff6ca7398, Flags=1, DataSize=31, pData=0x7fffffffcf00) at src/engine/shared/network_conn.cpp:154
#4  0x000055555558ec66 in CNetServer::Send (this=0x7ffff6c8de08, pChunk=0x7fffffffce80) at src/engine/shared/network_server.cpp:647
#5  0x0000555555591754 in CServer::SendMsgEx (this=0x7ffff6a47010, pMsg=0x7fffffffcf00, Flags=1, ClientID=3, System=true) at src/engine/server/server.cpp:683
#6  0x000055555559252e in CServer::SendRconLine (this=0x7ffff6a47010, ClientID=3, pLine=0x7fffffffd780 "[game]: leave player='7:iheb'") at src/engine/server/server.cpp:927
#7  0x00005555555925d4 in CServer::SendRconLineAuthed (pLine=0x7fffffffd780 "[game]: leave player='7:iheb'", pUser=0x7ffff6a47010) at src/engine/server/server.cpp:942
#8  0x000055555557c618 in CConsole::Print (this=0x555555981720, Level=0, pFrom=0x5555556b5e33 "game", pStr=0x7fffffffdbb0 "leave player='7:iheb'")
    at src/engine/shared/console.cpp:204
#9  0x00005555555d1fc5 in CPlayer::OnDisconnect (this=0x555555958450 <ms_PoolDataCPlayer+8176>, pReason=0x7ffff6cd10a0 "") at src/game/server/player.cpp:331

17:36

anyone :/?

baut the console client. Why i used the ddnet source is because i want a client that can do more than chatting. I just love consoles and using some ncurses stuff or what ever to enable keypresses would be awesome. And rendering a poor image out of text just shwing tees and collision tiles would be awesome. If you just want to chill on a server vote, chat. move around see a bit of the gameplay etc. I would love to see such thing.

@jxsl13 pretty hard without seeing some code

https://git.johnbehm.de/jxsl13/zcatch

[quakenet] <heinrich5991> print pNewItem->m_pNext

17:41

[quakenet] <heinrich5991> print pNewItem

how ._. this occurs quite often, but not in the way I can reproduce it.. and the debugger is turned off atm ?

[quakenet] <heinrich5991> you should write these lines in the debugger ^^

17:43

[quakenet] <heinrich5991> (if you can get the stack trace, you should be able to print these as well, from inside gdb)

then gotta wait for it to crash again, @ChillerDragon mind flooding my server johnbehm.de:8303 ?

[quakenet] <heinrich5991> you can save coredumps or so

heinrich5991: do you know how CGameClient populates its m_aClients?

 #0  0x000055555558ca66 in CRingBufferBase::Allocate (this=0x7ffff6ca73b8, Size=71) at src/engine/shared/ringbuffer.cpp:109
        pNewItem = 0x7ffff6ca74a8
        WantedSize = 96
        pBlock = 0x7ffff6ca7448
#1  0x0000555555573b6a in TStaticRingBuffer<CNetChunkResend, 32768, 0>::Allocate (this=0x7ffff6ca73b8, Size=71) at src/engine/shared/ringbuffer.h:59
No locals.
#2  0x0000555555572b49 in CNetConnection::QueueChunkEx (this=0x7ffff6ca7398, Flags=1, DataSize=31, pData=0x7fffffffcf00, Sequence=416) at src/engine/shared/network_conn.cpp:128
        pResend = 0x5007fffffffcec0
        pChunkData = 0x7ffff6caf58d ""
        Header = {m_Flags = 1, m_Size = 31, m_Sequence = 416}

[quakenet] <heinrich5991> then you can investigate the crash even if no debugger was attached at the time

bt full info

[quakenet] <heinrich5991> print pNewItem->m_pNext

17:45

[quakenet] <heinrich5991> Learath2: in Reset or so and in the snapshot handling

how do I save the coredump?

[quakenet] <heinrich5991> I don't know

17:45

[quakenet] <heinrich5991> there's an option in linux to save them automatically

damn, I have to handle snapshot stuff to get client names

[quakenet] <heinrich5991> yes

17:46

[quakenet] <heinrich5991> there you'll also find one obvious flaw in the tw protocol

17:46

[quakenet] <heinrich5991> :

17:46

[quakenet] <heinrich5991> nothing synchronizes game messages with snapshots

17:46

[quakenet] <heinrich5991> i.e. you can get chat messages for people who have left/whose slot has been reassigned

17:46

[quakenet] <heinrich5991> then you get the wrong sender in the client

in gdb this seems to do the job generate-core-file ... let's wait xD

fk it i give up

18:06

writing python program to chat in tw 2 hard

@ZimE Use libtw2 and learn yourself some rust for great good

lol thats all ur trying to do??

18:18

Image attachment

bypassing ur protection was too ez

u tried so many times lmao

nub apple

[quakenet] <heinrich5991> maybe I should write a python wrapper for libtw2

18:33

[quakenet] <heinrich5991> might come handy for some easy scripting

18:34

[quakenet] <heinrich5991> https://github.com/heinrich5991/libtw2/issues/13

Wishlist, the next big thing™, aka never gonna happen · Issue #...

Python bindings

https://github.com/Learath2/ddnet/commit/d069f7dcd

Add chat client · Learath2/ddnet@d069f7d

18:46

Such a horrible experience

18:47

Will finish it up when I can see again

[quakenet] <heinrich5991> receive! not recieve ^^

fixeth

18:54

https://github.com/Learath2/ddnet/tree/dd_pr_twchat

DDraceNetwork, a mod of Teeworlds. Contribute to Learath2/ddnet development by creating an account on GitHub.

18:59

now the question is do I introduce sth like readline or do I just do a non-blocking read of stdin until I get a \n in my buffer and call it a day?

[quakenet] <heinrich5991> if I had to guess, I'd say you'd settle for the latter

18:59

[quakenet] <heinrich5991> :P

1

I really don't feel like implementing a whole input buffer and a send queue :/

19:01

i'll check how fstd does it in icat, maybe I can steal sth

[quakenet] <heinrich5991> icat just reads from stdin

19:05

[quakenet] <fstd> actually the process is quite a bit involved

19:05

[quakenet] <fstd> iirc

19:06

[quakenet] <heinrich5991> mhk

19:06

[quakenet] <heinrich5991> select()?

19:08

[quakenet] <fstd> on the icat level, it first checks whether it can read (which ends up as a select, yes)

19:08

[quakenet] <fstd> and confusingly already fills the read buffer as part of the check

19:09

[quakenet] <fstd> whereas the actual "readline" function just considers that buffer and never does any i/o by itself

does the chat client allows to interact with your character

19:18

inb4 machine learning bot (edited)

It can't even chat yet

19:20

waiting for it

19:20

i always wanted try stuff with a headless client

meh fok windows

C:\MinGW\bin\gdb.exe: TUI mode is not supported

hi, is there a flag to allow editing a config variable only in the cfg file?

?

21:57

client or server side

21:58

i think sv_test_cmds can only be set in config

1

1

1

1

tear

1

what

1

21:59

just load the config var value into a variable after loading the config and use this variable instead

srv side

idont think there is a flag

nah test cmds can be edited

but it has no effect

22:00

u cant change a ddnet server into a testddnet server

22:00

ingame

it has only CFGFLAG_SERVER as flag

22:00

it can be edited

yes

22:01

maybe done in a similar way i suggested

22:01

it has no effects

22:01

but it could be edited

Exported 248 message(s)