DDraceNetwork - developer

send help mi map is broken

09:01

ChillBlock5_exp.map

496.83 KB

09:01

crashes my ddnet client if i try to join or open in editor

09:02

works in vanilla 0.6.4 tho

09:04

ah no it crashes my ddnet server on join and client on editor open

09:09

k saved it with vanilla client and fixed speedups/front and tele manually. Cant remember exactly but i think i did some experiments with tml back then. So i got the map back. But still a crash bug that ddnet has and vanilla doesnt. So if somebody is bored pls fix xd

Crashing servers and client with maps is extremely easy, I tried fuzzing it once

ya

all the accesses seemed to be reads instead of writes so it's probably not THAT bad

but even vanilla 0.6.4 can handle this map

basically we trust the map to report its data correctly and then just read over borders

didnt some guy reported is as security vuln latley in vanilla?

vanilla doesn't support many of the game layers ddnet does

09:17

oh?

09:17

link please

lemme check

09:18

https://github.com/teeworlds/teeworlds/issues/2071

Integer overflow and null pointer dereference in CMap::Load() in e...

Hello Teeworlds dev team, There is an integer overflow in CMap::Load() which can lead to a buffer overflow. CTile *pTiles = static_cast<CTile *>(mem_alloc(pTilemap->m_Width * pTilemap->...

[quakenet] <littlething> hey guys can someone help me a quide ? about in network.py I think you can only add whole 32 bit ints at a time. right?

09:33

[quakenet] <deen> I guess

09:33

[quakenet] <deen> There's a NetIntRange, but that still seems to put a full int in

@ChillerDragon sounds good, want to cherry-pick that over to ddnet?

do i look like a git wizard?

09:35

xd but i can try

you can start learning

09:36

and you can also test if that fixes your issue

[quakenet] <littlething> so can i add more ?

dont think so tho @deen because their fix is like a few days old and i tested in 0.6.4 but ye lez see

@ChillerDragon yea, it was our tml test, wasnt it?

no even way older think it was my first test

@deen i guess it is easier to copy over the few lines manually instead of using git magic

10:23

but we loose the real authors of the commits then ._.

i dont think ddnet is affected by this anways. Because vanilla added way more code to the Load() function . Left ddnet, right vanilla

ChillerDragon

[ddnet/ddnet] Pull request opened: #1591 Resolve conflicts ( vanilla cherry pick )

aaaah fuck

10:57

-.-

10:58

wtf where did the commits go? I did cherry pick solved the conflicts and git said i should commit. And now im the only commit there and the one i cherrypicked disapeared

ChillerDragon

[ddnet/ddnet] Pull request closed: #1591 Resolve conflicts ( vanilla cherry pick )

ChillerDragon

[ddnet/ddnet] Pull request opened: #1592 Add bound check in datafile.cpp anywhere m_ppDataPtrs is accessed as …

…an array. Should fix https://github.com/teeworlds/teeworlds/issues/2073 (cherry picked from commit e086f4b35b1adf7edc35b4ad332dc7ed1edc5988)

call me git wizard! Bitch pls

@ChillerDragon git wizard

2

def-

[ddnet-web:master] 1 new commit

feea1a0 Add donation by cris - def-

bors[bot]

[ddnet:master] 2 new commits

1e0cc51 Add bound check in datafile.cpp anywhere m_ppDataPtrs is accessed as an array. Should fix #2073 - Dune-jr 79a6ae6 Merge #1592 - bors[bot]

11:39

bors[bot]

[ddnet/ddnet] Pull request closed: #1592 Add bound check in datafile.cpp anywhere m_ppDataPtrs is accessed as …

bors[bot]

[ddnet:master] 2 new commits

e27679e Remove some code duplication in CSaveTee::LoadString - def- 4931d00 Merge #1589 - bors[bot]

12:13

bors[bot]

[ddnet/ddnet] Pull request closed: #1589 Remove some code duplication in CSaveTee::LoadString

ChillerDragon

[ddnet/ddnet] Pull request opened: #1593 Add emacs style ctrl-a ctrl-e navigation

Im sure im not the only one who keeps pressing ctrl-a and ctrl-e in the teewoods console and nothing happens .-.

does anyone know a fix for tabbing back into ddnet client on debian10 ? The window is not centered anymore

14:12

epic

14:30

sharing in a GNOME group.

Zoozti

[ddnet/ddnet] Issue opened: #1594 hookcollcolour is not working correctly when in solopart and watch other people using it

When you are in a solopart and watch people play, their hookcoll is not getting yellow if they aim at a tee. Even when you are in spectator mode. You have to unsolo first to see it working correctly and it works also for spectator mode then. Should not be too hard to fix i guess?

2

zoozti

1

14:31

xd

14:33

remove solotiles -> problem solved

@ᶰ°Konͧsti "Should not be too hard to fix i guess?" go fix it then

mean

no, he is the mean one

how does he come to this conclusion

2

i guess cus it already works correctly in some places so it could be a simple fix that doesnt require alot of new code

where does anything related to solo prediction work?

probably in gameclient.cpp, players.cpp, maybe gamecore.cpp

don't think client knows whether another player is in solo or not

does /showothers hide people who are in solo

15:56

if so then doesnt it know somehow

[quakenet] <heinrich5991> all / commands are handled by the server

15:57

[quakenet] <heinrich5991> this only indicates that the server knows that ^^

oo right

1

def-

[ddnet/ddnet] Issue closed: #1594 hookcollcolour is not working correctly when in solopart and watch other people using it

@ᶰ°Konͧsti https://media.discordapp.net/attachments/314776480880132097/564492483791290399/fixed-0.jpg?width=1216&height=684

16:51

is this what u wanted

This is what i mean Ye xd

16:55

Or idk

ok now pay me 10€

its just when ur in solopart and watch people playing u will not see the hookcoll yellow when they aim a tee

ryozooz

17:03

wtf 3000 fps

ryozoOoz (edited)

@noby linux + good pc

17:08

xd

linux + potato pc + no design + overclock

it's funny cuz Gameclient on a method named OnRender is sending packets to the server, gg

also in vanilla ._.

17:17

(src/game/clinet/components/chat.cpp, CChat::OnRender)

vanilla no lag

HADOKEN!!!

17:21

weeb code

what's hadoken?

https://www.urbandictionary.com/define.php?term=hadoken

Urban Dictionary: hadoken

1. From Japanese Ha Dou (Wave or Surge) and Ken (Fist, Technique, or Attack); a blastwave or fireball formed from a martial artist's life energy (or ki/ch'i). The hands are brought together, heels of the palms touching, and moved to beside the fighter's waist (which is as clo...

its the thing that the guy that fights people spams

17:38

alpha for solo players!

17:38

yay

17:38

now i just have to fix the colision bug and the hook moving a little

17:41

fixed nameplate alpha too xd

1

what does it change

17:42

from /showothers

17:42

@Ryozuki

?

17:43

on solo players

17:43

before it was u see him or not

17:43

now if u have showothers it appliues the alpha

17:43

from teams

17:43

idk if i explain good

no

17:43

xd

17:43

before u could see fully or not at all solo players

17:43

u get that

ah

17:43

i get it

gud

Ryozuki

[ddnet/ddnet] Issue opened: #1595 Projectiles don't have m_ClShowOthersAlpha applied.

Looks like client doesn't know who owns the projectile when rendering it.

im wondering where does the client checks for colisions i want to fix the bug where if u are not in solo and other is u see a weird collision

18:00

intersectcharacter doesnt do this

Well, how would you know that the other player is in solo=?

18:00

That information is not sent to client yet

i added it

ok

i got m_IsSolo on cnet CNetObj_Character

that sounds like it would break compatibility with vanilla?

hmm

I thought we should use the new protocol extension thing

if you explain me how i can try

18:02

what is it

and maybe send a single uint64 where each bit says whether a tee is in solo

18:02

I don't know

the network character by hmh

18:02

was closed

see protocol_ex_msgs.h

:o

the @DDNet.tw things I thin

what domain i put? ddnet.tw

yes

18:05

if your goal is ddnet server/client

y

18:09

how to convert a unsigned char * to uin64 xD

(uint64_t)

int SoloPlayers = (uint64_t)Unpacker.GetRaw(64);

18:10

this easy?

GetRaw(64)?

const unsigned char *CUnpacker::GetRaw(int Size)

18:10

i can only get this

18:10

or a int

18:11

(or a string)

I'd assume Size is number of bytes, not bits

oh

18:11

so 8

the way we usually do it is not that pretty

18:12

unsigned GhostMapCrc = (m_Header.m_aCrc[0] << 24) | (m_Header.m_aCrc[1] << 16) | (m_Header.m_aCrc[2] << 8) | (m_Header.m_aCrc[3]); is for 4 bytes for example

uhm

18:13

uint64_t SoloPlayers = (uint64_t)Unpacker.GetRaw(8);

18:13

xD

so you get out each byte and shift them so that they all fit together

18:13

what you're doing casts the pointer

ok

not the actual content

i get what u mean

You could cast the content straight to uint64_t and probably it would mostly work, but it's probably not standard compliant

18:16

and about collision, I'd look for occurences of the player_collision tune

18:16

that leads to gamecore.cpp:417

18:17

sounds good: // handle player <-> player collision

thanks

18:18

iff i shift << 56 clang tidy says it has more width than a unsigned char, i guess i should cast every array memeber to uint64? xD

18:19

or it isnt needed

18:22

The behavior is undefined if the right operand is negative, or greater than or equal to the length in bits of the promoted left operand.

18:22

so yes i have to cast

yeah

ugly code (tm)

told you

not even a unicode ™

@heinrich5991 do you remember if copy_mem for the same thing would be fine on all kinds of platforms?

18:34

@heinrich5991 also, want the walkthrough now?

no, because of big-endian vs little-endian

ah, right

yea, let me look at the PR again

I think we still have locations that do it with mem_copy though

18:35

might have to fix them then

yes

18:35

we have faulty code for big-endian machines, at least in the map reading code

weird, no one noticed on debian?

18:36

i thought on debian we're built for dozens of platforms or something like that

possibly elsewhere as well, if no one tests it, it doesn't work

18:36

I don't know if anyone actually executes the stuff or whether they just build it

18:36

we don't have a test suite for datafiles yet, that might uncover it

true enough

18:37

probably no one tests it

18:37

https://packages.debian.org/sid/ddnet see ports at the bottom

mips is le

18:39

@deen okay, so the idea behind OnCompletion is that it can close the file and thus check whether we wrote correctly?

18:39

if that fails, we just go to the generic error path?

def-

[ddnet/ddnet] Issue opened: #1596 Fix little endian

copy_mem might be causing some dangeorous situations on little endian, should go through and fix a few.

eh, big endian, sorry

18:40

we're on little endian by default (x86, arm, …)

I changed OnCompletion?

yes, you did. and you're also claling it in the error path

whats the reason behind putting everything on a interface and on a class ? IServer CServer Iwathever and Cwathever

18:42

double work

so that they have a cleanly defined interface, in theory

@heinrich5991 oops, that seems unintentional, let me take another look

doesnt look that clean twintri

(edited)

it's kinda clean in vanilla, ddnet kinda walks over it due to casting it to CServer anyway

ah, the idea is that if you don't call OnCompletion then the updater is just stuck

18:43

and nothing will be shown

18:43

so you need to call it and OnCompletion can then do something depending on whether m_State is HTTP_DONE/ERROR/whatever

18:45

and before it was called in most error paths, just not all consistently

okay

bors[bot]

[ddnet:master] 2 new commits

e539bba Handle FS failures while updating (fixes #1560) - def- 52c00dc Merge #1573 - bors[bot]

19:01

bors[bot]

[ddnet/ddnet] Pull request closed: #1573 Handle FS failures while updating (fixes #1560)

void CServer::GetSolos(unsigned char *aPlayerSolos)
{
    for(int i = 8; i > 0; i--) {
        unsigned char x = 0;
        for(int j = 8; j >= 0; j--) {
            x |= m_aClients[64 - (i * 8 + j)].m_Solo << j;
        }
        aPlayerSolos[i] = x;
    }
}

is this clever, or really stupid

19:10

sounds much more expensive than the previous solution

wich one is it?

19:11

which

19:11

the cast?

yeah

xD

19:12

i spend 10 mins to do this