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 2018-10-08 00:00:00Z and 2018-10-09 00:00:00Z

@n000b the error message should be improved. I think most players' "you're on the dnsbl list" actually means "the server hasn't received a dnsbl answer yet" (edited)

[ddnet/ddnet] Issue opened: #1338 DNSBL is not perfect

See for example https://forum.ddnet.tw/viewtopic.php?t=6646 Some players will have bad luck just because someone used to send spam mails from their assigned dynamic ip address years ago. Not sure what the best solution is here.

06:38

[ddnet/ddnet] Issue closed: #1338 DNSBL is not perfect

[ddnet/ddnet] Issue opened: #1339 Prevent fake low ping

Is it possible to prevent fake low ping on our servers?

lmao

im going to guess never based on how ping works but would be cool if there was an actual patch for this

Image attachment

Well, we could change how ping works

wouldnt it require people to use a new client

@deen i fucking needed to change the whole configuration of my router to have like max 80ms in GER

@Cellegen | HU ok, so?

hmmm

10:14

but you see

am i wrong that changing how ping is reportd would also require the client to be changed

i cant manage to bring myself less than 80ms ;-;

@noby I'd be interested if there is a way to do it without client changes

well the way i do pingfake is by changing the ackgametick thats sent along with inputs

doesn't that have adverse side effects?

10:16

like potential packet loss

im not sure how a server would be able to tell whether a player actually takes a long time to acknowledge the inputs or is faking

10:16

mm

10:16

it might make it a little laggier but not really i think

10:18

it isnt precise afaik u can only get precision of ±1 tick so ~20ms?

10:19

if ur ping goes "below 0" it causes problems

yeah, feels so

10:19

everyone has 8, 24-28, 40-48 ping etc

yeah so given this im not sure if theres any way to detect fakeping without making everyone use an updated client unlss im missing something

if ackgametick is based on the real game tick it would be probably impossible to fix without changing the client, but if it is based on the last packet received by the server, it might be possible to do something about it

why is the spoof protection on password protected servers turned off?

[quakenet] <deen> doesn't work I think

ok just tried it, your right ^^

maybe somebody else could have a look at this and state his opinion https://github.com/ddnet/ddnet/pull/1320

Implement ISDDNET server message to decouple version expectation f...

This adds a new network message sent from the server to the client when a client connects. It tells the client that the server is a DDNet server and it is used in addition to the gametype to determ...

Yeah need a tiebreaker on that one

[ddnet/ddnet] Pull request closed: #1301 Continue work on Docker

@Dev im on ddnet++ (based on 9.0.2) linux server. my score (finishtime) is only getting displayed after i renamed, how can i fix it so its like on ddnet or kog where the time is there on client connect? i already had a look into the code but didnt find why it shouldnt be working.

I have no idea. it could be a billion things. What is a ddnet++ and why do you need it?

https://github.com/ChillerDragon/DDNetPP

ChillerDragon/DDNetPP

Based on DDraceNetwork by deen & contributors (www.ddnet.tw) which is based on the great game teeworlds (www.teeworlds.com). - ChillerDragon/DDNetPP

Oh, hadn't seen this yet. I'd suggest upgrading the ddnet version but I doubt that's an option after this many patches. Soooo just debug it?

14:59

I'd start from looking at how the client decides when to display a time.

good point

15:02

@Learath2 could you help me a bit?

tomorrow maybe, kinda busy today

mh

[ddnet/ddnet] Pull request opened: #1340 Editor fixes

Address #1331 and also fix the scrollbar enabling when there isn't enough elements.

@heinrich5991 are you the one with AppVeyor access? If so when you have a moment can you rebuild my PR and/or set-up access so all members can rebuild

16:22

?

I think for some reason, both deen and I have an appveyor build

16:22

let's see if it's mine that's failing

16:23

wait, it's travis that's failing

appveyor was also failing but it suddenly succeeded

well, then it's the race condition between deen's and my appveyor

16:24

whoever finishes last sets the status, I think

16:24

which is nice, because a failing build probably finishes first

@Learath2 maybe you remember the drop weapon code from me (press f3 to drop the current weapon). Is it possible to remove that weapon if its falling into a shield?

16:33

I mean I cant get it to work :D

I don't remember your code, does it spawn a CPickup?

it spawns a CWeapon, made by me

16:34

https://github.com/ChillerDragon/DDNetPP/blob/master/src/game/server/entities/weapon.cpp

ChillerDragon/DDNetPP

Based on DDraceNetwork by deen & contributors (www.ddnet.tw) which is based on the great game teeworlds (www.teeworlds.com). - ChillerDragon/DDNetPP

16:36

should be possible huh?

It is possible but I must say that looks very, uuhhm dirty

ik :D

You just need to check that like IsCharacterNear()

ye

16:39

but how xd

16:39

i couldnt do it for shields :(

You look for nearby ENTTYPE_PICKUPS and check that their type is POWERUP_ARMOR

how do i check the type of it?

16:40

ah

16:40

nvm xD

16:41

what do i have to use for the MAX size?

Who/what is a max size?

?

16:41

MAX_CLIENTS?

I mean does it even matter? Think about it

no

You just need one armor to make it disappear

ye

So i'd just look for 1

oo

I'd add a new function to CPickup int Type(); that returns m_Type rather then making m_Type public btw

Ye

16:45

you mean like for example GetActiveWeapon from Character?

Yes, but you need to stop thinking this way

16:46

It's a very simple thing I said, just trust your instincts instead of trying to make it "like" something

Okay :)

what's the masterserver of ddnet?

bool CWeapon::IsShieldNear()
{
    CPickup *apEnts[2];
    int Num = GameWorld()->FindEntities(m_Pos, 20.0f, (CEntity**)apEnts, 2, CGameWorld::ENTTYPE_PICKUP);

    for (int i = 0; i < Num; i++)
    {
        CPickup *pShield = apEnts[i];

        if (pShield->GetType() == POWERUP_ARMOR)
        {
            return true;
        }
    }

    return false;
}

@Learath2 this works, is it right?

@fokkonaut I'd do CPickup *pEnt; and pass (CEntity **)&pEnt instead

how to pass it?

and replace that 2 with a 1 obv

17:45

I told you how to pass it in that sentence

but it didnt work then

17:46

tryed that already

Are you sure?

let me see

Oh actually in this case you don't even need a pointer to the entity

17:49

Oh actually you do nvm

17:49

anyways

17:49

That should still work

cannot convert 'CPickup' to 'CPickup*' in initialization CPickup *pShield = pEnt[i];

Well DUH

xd

17:49

sory

17:49

xDD

17:50

im blind

Actually I'm stupid

17:50

You need more then 2 though

17:51

You need 9 to be sure

?

With 2 if the weapon is falling between 2 hearts and has an armor below it, it is possible that the 2 hearts get found and the armor doesn't get found

ahh

So to be safe look for 9 pickups, one you could be on and 8 could be around you

ok^^

17:53

thanks!

Uhm @heinrich5991 I think I broke something as I pushed right after you bors r+'d

oh, awesome, you also did that thing

Can you eye over it to make sure I didn't introduce any bugs

mhk

18:13

@Learath2 you changed behavior in https://github.com/ddnet/ddnet/pull/1337/files#diff-f60b98c9084662cc2855d8c1a78e458bR390, previously the port wasn't recorded. probably won't break anything. wanna change it to the old behavior or do you think it's ok?

Ignore DNSBL on "empty" servers. by Learath2 · Pull Request #1337...

DNSBL is to prevent annoyances. There is no one to annoy on an empty server.

18:13

also, that function used to modify the pointer that looks read-only…

18:14

good job, function

Yeah those functions look horrible

18:14

it takes a pResult it never uses aswell

18:14

VoteMute and Mute I mean

18:15

I'll restore the old behaviour just in case

maybe don't modify the pointer that's passed in

18:16

that sounds outright horrible

Would you prefer it like it is or if I reverted to the old behaviour?

18:18

I don't think it makes any difference though, as _noport is also used to check the ban itself

18:21

eh, I'll just clean up those functions and make sure

awesome

18:22

inb4 #1337 expands to refactor the whole ddnet code base

18:23

do stop though (with refactoring these functions) if it becomes too much work

@heinrich5991 not that bad apparently, cleaned it up and made sure the behaviour doesn't change

18:31

Also is bors k?

@Learath2 fixed the score on client enter thing myslef

\o/

18:33

What was it?

just added this line

m_apPlayers[ClientID]->m_Score = (Score()->PlayerData(ClientID)->m_BestTime) ? Score()->PlayerData(ClientID)->m_BestTime : -9999;

from changeinfo to OnClientEnter()

Things not to do: stop gdb session while saving a .map file

wew..

[ddnet:master] 7 new commits

1073b72 Ignore DNSBL on "empty" servers. - Learath2 f1867e0 Grammar fix "then->than" - Learath2 2e137e6 net_addr_comp_ip -> net_addr_comp_noport - Learath2 254ff3d Oversight - Learath2 5b95edd DRY with net_addr_comp_noport - Learath2 571b069 Small cleanup of *Mute functions - Learath2 1f4585c Merge #1337 - bors[bot]

19:03

[ddnet/ddnet] Pull request closed: #1337 Ignore DNSBL on "empty" servers.

@jxsl13 we use the regular TW master servers but ddnet.tw is one of them

1

1337!

and @heinrich5991 still hasn't done http masters

salt

2

feelsbadman

@heinrich5991 extended map items work great

Nice job (iguess) feelsgoodman

feelsgoodman

@heinrich5991 I can't really follow the map items thing, here to give me a hand?

yes, I guess

So, what is an external item type and what is an internal item type?

ok, so we have item kinds and extended item kinds (edited)

19:57

non-extended item kinds are referred to by a small integer, both in code and in the actual datafile

19:58

code is "external" (because it's the API the datafile exposes) and the actual file is "internal" (because only the datafile class has to concern itself with how it saves stuff in there) (edited)

19:59

UUIDs also happen to be referred to by integers, although bigger ones, I think starting from 2**16

20:00

so when the datafile class gets a item kind > 2**16, it sees that it's an extended item kind that is associated with a UUID

20:01

it has to store that some way in the datafile, which only supports item kinds < 2**16, and certainly no UUIDs

20:02

so I reserve the item kind (2**16 - 1) to provide a mapping from internal item kind IDs (in the datafile) to external ones (referred to by UUIDs, to be translated to the corresponding integers by the datafile reader)

20:02

so when you read a item kind that's likely to be extended (>=2**15 I think)

so you use 2**16 - 1 like we used NETMSG_NULL for the protocol extension?

then you check whether we have an item with type ID 2**16-1, and the type ID you're looking for as item ID

20:03

yes

20:03

because 0 was already taken

20:04

well

20:04

no

20:04

type ID 2**16-1 only provides the mapping of internal (in the datafile) type IDs to UUIDs

20:05

i.e. if I add one item with extended type, it'll get type ID 2**16-2

20:05

next one will get 2**16-3

20:05

but to know what it's actually dealing with 2**16-2 doesn't map to a type, the reading code has to check the type map

sooo 2**16 - 1 actually is a type-map?

yes

20:08

I probably should have written all that into the commit message…

20:11

@Learath2 questions?

I'm trying to give it a re-read with my newly acquired knowledge

[quakenet] <Learath2> heinrich5991: do asserts still kill the client/server in release builds?

20:16

[quakenet] <heinrich5991> I fear they do not

20:16

[quakenet] <heinrich5991> yup, they don't :(

20:17

[quakenet] <Learath2> Well assert in CDataFileWriter::GetExtendedItemTypeIndex could allow an attqcker to craft a map that crashes the client.

20:17

[quakenet] <heinrich5991> lol

20:17

[quakenet] <Learath2> Not a particularly useful crash but still

20:18

[quakenet] <heinrich5991> have you read the datafile code? :P

20:18

[quakenet] <heinrich5991> it's full of crashes

20:18

[quakenet] <heinrich5991> but you're right, I shouldn't make it worse

20:18

[quakenet] <heinrich5991> Learath2: no, that's a datafile writer

20:18

[quakenet] <Learath2> Oh yep mb

20:24

[quakenet] <Learath2> I can't seem to read this at all :/

20:30

[quakenet] <Learath2> heinrich5991: did you have a document on datafiles and their structure?

20:30

[quakenet] <heinrich5991> yes, but only little

20:30

[quakenet] <heinrich5991> https://github.com/heinrich5991/libtw2/blob/baf355f8f07ce0b58818598930e769121fcaa3c1/doc/datafile.md

heinrich5991/libtw2

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

20:30

[quakenet] <Learath2> Well an overview would be enough

20:31

[quakenet] <heinrich5991> might very well be that this doesn't help

hm, blockers/stoppers don't work when flipped. Known bug?

hum. no, not known I think

20:48

they should stop in the direction you can see in the editor

known

[quakenet] <heinrich5991> Learath2: updated the PR

21:08

[quakenet] <Learath2> heinrich5991: how do datafiles handle endianness issues? do we save them all little endian and swap based on architecture?

21:09

[quakenet] <heinrich5991> yes, something like that

21:09

[quakenet] <heinrich5991> in the real world, the implementation for big endian is broken, and no one seems to have noticed in years

21:09

[quakenet] <Learath2> :D

21:09

[quakenet] <Learath2> Well not many big endian machines around

21:09

[quakenet] <heinrich5991> but for the PR you just have to know that integers get saved correctly

21:10

[quakenet] <heinrich5991> (because they work in other map items=

21:10

[quakenet] <heinrich5991> and I serialize the UUIDs into integers, in a native-independent way

21:10

[quakenet] <heinrich5991> *endian-independent

21:10

[quakenet] <heinrich5991> so it doesn't matter how we save them, as long as integers work

21:11

[quakenet] <Learath2> Great, I'm merging then

21:11

[quakenet] <Learath2> than*

21:12

[quakenet] <Learath2> heinrich5991: doesn't sizeof(CUuid) / 4 assume sizeof (int) == 4?

21:12

[quakenet] <Learath2> Is that a fair assumption to make?

21:13

[quakenet] <Learath2> I guess it is given sizeof (int) is guaranteed to be >= 4

21:14

[quakenet] <heinrich5991> if it's not 4, the whole serialization code breaks

21:14

[quakenet] <heinrich5991> like

21:14

[quakenet] <heinrich5991> the datafile code

21:15

[quakenet] <heinrich5991> hm

21:15

[quakenet] <heinrich5991> if one were to refactor the datafile code, one would replace all ints with int32_ts

21:16

[quakenet] <heinrich5991> (obvious spot for rust promotion) you know that in rust, you usually know the integer sizes you work with? :P

21:16

[quakenet] <heinrich5991> hm

21:17

[quakenet] <Learath2> One last thing before I r+, should you maybe test for the uuid being saved correctly aswell in your testcase?

21:20

[quakenet] <heinrich5991> hm. I considered that private to the datafile classes

21:21

[quakenet] <heinrich5991> should I add a test regardless? I don't really have the structures for that around, because they're in the .cpp file

21:22

[quakenet] <Learath2> I guess it's not really mandatory under the assumptions that storing ints work and that sizeof(int) >= 4

21:23

[quakenet] <Learath2> first one is true because, well things work :D and the second one is guaranteed by the standard so I guess we are golden :)

21:25

[quakenet] <heinrich5991> do you feel that I'm being more strict with you on PRs than you are with me?

21:25

[quakenet] <heinrich5991> because I currently have that feeling

Well with this one it's not really my area so I can't really be sure if the things I worry about are superficial or not

21:34

Also you've always been more strict then me anyways :D

good night

hm does anyone know if something got changed in the protocol or sth on 7th. october?

23:53

since I can't seem to retrieve the stats from most servers suddenly

23:53

https://i.imgur.com/M9rXcF8.png

Embedded image

23:56

just started debugging but it ran completely fine for multiple months and now test and local environment showed this behaviour so thought I'd ask if there was something recently

master servers ddosed

Exported 234 message(s)