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 2019-10-25 00:00:00Z and 2019-10-26 00:00:00Z

hmm, gameworlds being handled by the gamecontroller turns out to be quite unwise :/

[ddnet7:d7] 1 new commit

0179f2f Add Teams - Learath2

@Ryozuki && @heinrich5991 can you both take a look at that commit? I'm not sure I like the organization with the worlds being owned by the controller

01:02

(The commit works, but you need to modify the code to put tees in separate teams and initialize the world by hand in CGameControllerDDRace )

01:02

Also I had no idea how to handle the m_Paused

01:04

I also tried my very best not to touch the vanilla gamecontroller but I need to get rid of some of the team code and I don't really know how

01:04

Vanilla really should have had a fully abstract IGameController ...

[ddnet/ddnet] Pull request opened: #1949 Add cl_dummy_resetonswitch 2

Idea by @ChillerDragon : Only resets the active tee. So if you hold right and switch your dummy keeps walking. But if you switch back you don't have to press right twice to walk again. Fix #1854

[ddnet/ddnet] Pull request closed: #1946 Don't locally render your own dummy's afk emote. Other players will still see it

01:37

[ddnet:master] 2 new commits

2943eb0 Don't locally render your own dummy's afk emote. Other players will still see it - d3fault 4fc9079 Merge #1946 - bors[bot]

[ddnet/ddnet] Issue closed: #1879 Undefined behavior when registering an unchangeable command with more flags

[ddnet-maps:master] 1 new commit

d58ca08 M Sleepy - ddnet-maps

@Learath2 paused is the thing used in start of tournaments right?

Yeah, so maybe a global paused is something i need to have

@Learath2 one thing i never liked is that character adds himself to the gameworld, and i think it would make more sense to be managed by the gameworld directly

09:40

Ala its the gameworld who respawns and adds the entity

09:40

Not character itself

09:40

It kinda breaks the hierarhy

09:40

But its unrelsted to this tho

09:41

Damn mobile

@Ryozuki sadly all entities add themselves, I dont really think its a good idea to make so many changes just to get better hierarchy

I guess

09:41

feelsbadman

Maybe I can do it after I finish up the teams i can go through and change that

Didnt u say 0.7 is the oportunity for big changes?

09:42

Ill try to help too specially this weekend

09:45

I wish github had a mobile app

@Learath2 your helper commit misses a lot of things

Probably, not all commits are supposed to work on that branch, im still laying things out

10:51

What did you have in mind?

i will do a pr after shower, you missed basic stuff to get the helper working in general

in case any1 is using windows and discord might be interesting: https://www.bleepingcomputer.com/news/security/discord-turned-into-an-info-stealing-backdoor-by-new-malware/

Discord Turned Into an Info-Stealing Backdoor by New Malware

A new malware is targeting Discord users by modifying the Windows Discord client so that it is transformed into a backdoor and an information-stealing Trojan.

You can log in as helper and there is a console level for helper, I dont see what's missing but sure feel free to pr

did you try logging in?

10:55

Well, probably my mod again having more stuff so more required changes

10:55

I can check later

yo bash haxx0rs? How to compare against newline? :/

11:11

Image attachment

11:11

i can easily match the $c newline but not the $nl

11:11

Image attachment

11:12

oh wait that code is fucked xd

11:12

Image attachment

11:12

thats what i ment but it has same result

11:14

nvm got it

11:14

Image attachment

11:14

that works

[ddnet/ddnet7] Pull request opened: #1 Fix md5 build

11:16

[ddnet/ddnet7] Pull request opened: #2 Fix helper

@Learath2

GWpingSock

1

11:18

u chiller

troll

1

Ah, I missed the default key and setting the actual access level

1

11:19

I'll accept the pr, but I hope you know that the code is open source, and all your contributions are also open source

I know that

11:21

Why do you tell me that so explicitly? :)

Well you asked us not to host your version of the code, I'm just stating upfront that I won't be removing this branch under any condition

11:23

We really should add a CONTRIBUTION file

You can do it. I messaged heinrich about that already

11:24

As I told you now too in pm, I thought about the whole situation

I see

And I definitely apologize for all that.

11:25

I wouldnt make any pr's right now if I would still be against all this

It happens

11:26

Sometimes people can just get on your nerves

:/

11:26

true

11:26

about the md5 thing

the function md5_finish_ is actually called once, with the correct parameters. It just didnt get renamed and failed the build

The md5 thing links fine on ddnet 0.6, and links fine on linux for me, but I'm using hash_openssl

there are two different md5_finish's

11:27

thats the point

11:27

hash_openssl is not used ón win (at least for me)

yeah, the bundled one looks wrong to me aswell

no idea, but this change at least fixed the build

Yeah, it is correct, I just missed it while porting the md5 implementation over

I just took the name with the underscore from where it acutally gets called, there it is correct

11:28

Ah, okay

Should commit the md5 thing to vanilla aswell

11:32

I mean hash_bundled and using openssl if available

[ddnet/ddnet] Issue closed: #1854 Add cl_dummy_resetonswitch 2

11:44

[ddnet/ddnet] Issue closed: #1854 Add cl_dummy_resetonswitch 2

11:44

[ddnet/ddnet] Pull request closed: #1949 Add cl_dummy_resetonswitch 2

11:44

[ddnet:master] 2 new commits

ecf48a8 Add cl_dummy_resetonswitch 2 - d3fault dc6a295 Merge #1949 - bors[bot]

@deen I wanted to remove the last commit before merging those 2 prs

12:04

I can rebase, reorder anyway

12:09

also no need to bother with bors on d7 as I dont run it through ci yet anyway

12:09

it's more of a drawing board

[ddnet7:d7] 2 new commits

e9226b2 Fix md5 build - fokkonaut d4e51fa Fix helper - fokkonaut

12:41

[ddnet/ddnet7] Pull request closed: #2 Fix helper

12:42

[ddnet/ddnet7] Pull request closed: #1 Fix md5 build

12:43

[ddnet7:fb_teams] 1 new commit

2c940d6 Add Teams - Learath2

12:43

[ddnet/ddnet7] New branch created: fb_teams

did u mess with the git history? Should i refork? :/

Were you messing around with d7?

13:19

I thought no one else was doing anything with it, soz, I'll keep the history sane after this

13:20

don't refork, you can just git reset --hard ddnet/d7 your own d7 branch, if you were working on some other branch you'll need to rebase it

what is master and what is d7 used for?

13:44

i think i only used master

d7 is a rewrite, master is fokkonauts port

hi

ah i see

13:52

so rewrite ddnet7 from scratch?

13:52

is d7 planned to replace master?

Just make d7 the main one

14:05

default*

If it ever gets completed, yes that'll be the adopted version

14:08

@Ryozuki do you like the CGameWorlds belonging to the gamecontroller?

[ddnet7:fb_teams] 2 new commits

7a16964 Entities and the world itself don't need to go through the gameserver - Learath2 b904b77 Add Teams - Learath2

@Learath2 whats the alternative?

thanks bouncer i got all irc configured fast, but im on fedora and i hate it

CGameContext owning the worlds maybe

wait a bit i look d7 code

@Ryozuki I moved the team stuff to a branch of it's own, fb_teams

i compiled d7

14:22

idk if its fedora stuff but i see one of the things in the bg

14:22

in 4 colors

14:22

like the texture is missing

14:25

Embedded image

welcome to 0.7 @Ryozuki

just 0,7 things

14:28

@Ryozuki did you clone the submodules aswell?

oh

14:28

xd

nobo xd

14:28

wonder that it even compiles without submodules

Learath2: itdoesnt make sense for worlds to be managed by controller i think

14:28

to not be

14:28

*

14:28

ups

14:28

i think

14:29

IGameController::OnEntity(int Index, vec2 Pos, int Layer, int Flags, int Number)

14:29

for example

14:29

i think it makes sense for worlds to manage their own entities

14:29

just like character shouldnt manage itself (inserting itself to the world)

14:29

i think the current hierarchy is not good

14:31

@ChillerDragon it compiles cuz apparently the submodules are just game resources not code

14:31

hmm i still see the same texture stuff after git submodule update --init

14:32

ill clean build

14:34

ah my bad

14:34

i was on master

14:34

on d7 looks fine

yea the d7/master thing is a bit confusing :/

@Learath2 hmm maybe its fine with controller having the worlds

OnEntity handles creating the entities added to the map

but maybe we can move some stuff directly to the world class

like?

hmm wait

14:52

ok nvm

14:55

@Learath2 what about the helper methods from gamecontext being in the gameworld instead?

14:55

(createexplosion, etc)

Oh, hmm, those do sound like they belong there indeed

[ddnet/ddnet7] Pull request opened: #3 teeworlds -> ddnet on cmake

This changes the executable name too. We should also change the icons.

@Learath2 maybe you could provide a todo somewhere on what stuff needs to be done, and the priority

You can grep my entire code for TODOs for small things

15:21

I should make a TODO to lay out the outline too with bigger tasks

Entites inserting themselves does make this a tad difficult

15:49

@Ryozuki hmm, when I create a new CGameWorld for a team, I'll need to iterate all the tiles on the map again to create the entities, do you think that's a good idea? :/

hmm

16:00

maybe we can share entities that doesnt change whether ur in a team

16:00

or is it another thing?

Hmm ,I wonder how entites behave if inserted in two worlds

[ddnet7:fb_teams] 2 new commits

84593a2 Implement world functions cleaner - Learath2 8298ba8 Fix IGameController::OnEntity - Learath2

@Learath2 it would be cool if teams could be completly separated, we could even have the crazy idea to have them in different threads

16:09

idk if that would be profitable

16:09

or possible

yikes multithreaded

in this modern world cpus have multiple cores

16:10

:p

yea agree

16:11

sounds cool but also tricky

and complicated

there are some things that are still shared

I have huge concerns with DoS especially when I have to iterate over the entire map to add entities when the world is created by a team

ye we would need synchronization or whathever

spamming /team

cant u just clone the team

16:11

@Learath

16:11

hmm

what happens if some dickhead joins 64 dummies and does /team {1..64} on those?

yea

xD

what do you guys think about removing teams?

that would iterate through the entire map 64 times

maybe we could rethink the idea of a team yes

maybe I could populate all the gameworlds on server launch, but that would be a waste of memory imho

yeah

keep in mind we want to scale to unlimited slots @Learath2 troll

troll

@Learath2 maybe some entity cache

16:13

or smth

@ChillerDragon yeah I already have it set up so I can extend it to infinite teams or teams with names instead of numbers

yea? ❤

yes, its important that we can reach team 69

1337

the only problem is the fact that creating a new team might lag the server

16:14

which is unacceptable

yep

:/

hmm, maybe I can thread the iteration of the map and dont let the players join the team until the world is populated

entities are shields doors and stuff right?

y

is it rly impossible to clone the world?

16:16

if that makes sense

well it isn't impossible, but in C++ this deep copy would be

16:17

hmmm, how do I put it nicely

we could keep static entities separate from those dynamic

cancer

like character

16:17

a shield will never change

cant u switch shields?

well maybe we could reset the entities to their initial state

but yea non switches shields stay... until you collect them troll

troll

Problem isn't the fact that things change, we can clone characters aswell that's not the issue

16:19

the issue is that the entities don't know their initial states

then we can save them

Hmm, maybe a reference gameworld that no one ever joins and is eternally paused

16:19

I could copy everything over from there

sounds like a good idea

16:19

i wouldnt call it gameworld maybe

16:19

it could be misleading

It will be a CGameWorld we could name it whatever we want though

ah

m_ReferenceWorld?

ye

Well I'll still need to write copy constructors for all of the entities but it's a good starting point

we must definitly make them not self insert imho

Let me quickly try to rework that

16:22

I wonder however why matricks decided to make them self insert in the first place

16:22

I have a feeling I'll find his reason in a minute

:o

16:23

@Learath2 probs so he can do new CProjectile() and forget

Well found one problem already

well idk

16:24

whats it

I wonder if it's an actual issue, but the character is only marked alive after it's inserted into the world

well

eh, I'll try

its a feature

16:25

when the character is dead

16:25

it isnt even allocated

16:25

iirc

16:25

look into CPlayer

16:25

it allocates it on every respawn

16:25

iirc

16:26

https://github.com/ddnet/ddnet/blob/master/src/game/server/player.cpp#L494

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/ddnet

16:26

https://github.com/ddnet/ddnet/blob/master/src/game/server/player.cpp#L565

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/ddnet

16:26

see

16:26

i guess its same on vanilla

I can't seem to find where CFlag is ever constructed

16:31

oh it was probably in ctf.cpp

[ddnet7:fb_entities] 1 new commit

25e4e28 Don't self-insert entities - Learath2

16:59

[ddnet/ddnet7] New branch created: fb_entities

@Ryozuki I don't like it

xd

17:18

Let me see

@Learath2 u still make the entity insert itself kinda

17:23

hmm

17:24

@Learath2 InsertInto is a misleading name

17:24

if its what i think it is, it should be OnInsert

Probably, I just quickly whipped it up to see what changes would entail

i think its good to have a OnInsert

17:26

so you can clearly see what happens when its inserted

17:27

@Learath2 while you are at it maybe u can add a getter World() or something

There is a getter for it

17:27

GameWorld()

you dont use it

Where?

oh wait

Problem is that some entities need the gameworld to construct themselves so you need to wait until insertion to get the gameworld

the gamecontroller doesnt have it too?

17:28

which ones?

The ones that have Server() or a function call in their InsertInto's

17:29

Gamecontroller doesn't have a getter for world because there is a problem there

17:30

I didn't feel like breaking the vanilla gamecontroller, but the fact that now we have multiple gameworlds the vanilla controller started falling apart

??

17:30

ye

17:30

u can always pass for example starttick and team directly

17:30

on the constructor

17:30

so u dont have to wait for insert

17:30

on cproj

Maybe I should just cannibalize the gamecontroller

you can also mkae a helper function on the world

17:31

World()->CreateProjectile()

17:31

with sensible defaults

17:32

@Learath2 u will chuckle but there is a thing called factories

17:32

xd

17:33

idk if this concept is used in c++ a lot tho

You mean public static final void __AbstractWorldEntityFactoryManagerEx?

well u make the worst case but ok xd

17:35

@Learath2 one thing it doesnt make sense to me is that cprojectile gets a team on insert

17:35

why is that

17:35

shouldnt the world abstract that away somehow

@Ryozuki we don't have access to the server until then either

thats why a factory would solve that maybe

17:37

if u create it through it

17:37

the factory has access to the world

17:37

by factory i mean a way to generalize World()->CreateProjc()

17:37

idk

17:38

maybe im just talking bullshit

17:38

idk anymore

17:38

xd

@Ryozuki you mean the world itself creates and inserts the entity?

y

I could probably abuse C++ lambda functions to achieve that

try it

17:42

:o

I don't really like the fact that I have to call CEntity::InsertInto in all my overriden InsertInto's tho

17:42

Any idea how to get that fixed?

we could also provide always the tick when the entity was created

17:42

hmm

17:42

ye

17:42

creating the from the world

Oh, I should add the gameworld back to the constructor if i'm letting the world construct the entity

is a way tofix that

17:43

why u dont just pass the parameters needed

17:43

instead of the world

I'm going out for quik food, then I'll try that

xd

@Ryozuki because some entities do actually require the world

like which

laser e.g. needs intersectcharacter

17:47

hmm, actually I can't add the gameworld back to the constructor of centity

17:47

what if someone constructs an entity with one world then inserts it to another

why you dont simply pass a reference to the charcore

17:48

if it needs it

17:48

why would u need to do that

the laser intersects with others in a future time

17:48

not intentionally, but that is known at compile time, it should be preventable

well if this solution just complicates things then its not good

17:49

xD

eh,. theere should be no harm in setting things OnInsert

17:49

As the entity won't even start ticking before insert

you could actually run one ddnet server with valgrind to see what problems getinfo causes

20:44

as long as it had no players

20:44

or use gdb and press ctrl-c a few times, the poor man's sampling

20:45

an easy way might be to cache the getinfo response for each second

@deen I ran a ddnet server under gprof

[ddnet7:fb_teams] 1 new commit

63d51df Don't self-insert entities. Add testing command for team. - Learath2

@fokkonaut or @ChillerDragon either of you here and have a moment?

Exported 338 message(s)