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 2020-08-31 00:00:00Z and 2020-09-01 00:00:00Z

Not sure if this is right place to talk about this, but I just found out that 0.7 has 1 frame delay -- this means that on top of needing antiping, it needs anti-frame-delay

00:08

So let's say someone like Cammo, who plays with 8 ping, plays on a 60Hz monitor -- that's at minimum ~300% delay

00:09

So everybody plays worse and have no idea why

00:10

Not to mention significantly lower FPS (though I don't think most people would be affected by this) (edited)

0.7

kek

0.6 > 0.7

@Pathos try gfx_finish 0 and gfx_vsync 0, then 0.7 should be as good as ddnet old renderer

is there a size limit for audio files

@Pathos try gfx_finish 0 and gfx_vsync 0, then 0.7 should be as good as ddnet old renderer

@Deleted User Already have those settings

05:26

Thanks for looking out though

@Pathos what graphics card are you using? 0.7 has the async rendering turned off by default. You could try gfx_asyncrender 1. Might help but it can also make it worse =\ For me both ddnet and 0.7 feel good in terms of latency

[ddnet-web:master] 1 new commit

5347611 Add ZiKkie live stream to #51 - def-

Why are there no odd numbers on ranks?

1000 / 50 = 20

08:24

so 20ms per tw tick

ah

can we have a command /teamranks [name] [times x] ?

08:38

it should list all teamranks from a player on this map

5

[ddnet/ddnet] Pull request opened: #2683 Add platform and arch in console output

[ddnet/ddnet-libs] Pull request closed: #11 Upgrade SDL2 for Linux Steam release to 2.0.12 (fixes ddnet#2679)

10:19

[ddnet-libs:master] 2 new commits

a3958b0 Upgrade SDL2 for Linux Steam release to 2.0.12 (fixes ddnet#2679) - def- d3ea375 Merge pull request #11 from ddnet/pr-sdl-linux-steam - def-

10:24

[ddnet/ddnet] Pull request closed: #2683 Add platform and arch in console output

10:24

[ddnet:master] 2 new commits

2b0e13d Add platform and arch in console output - def- c864677 Merge #2683 - bors[bot]

10:31

[ddnet/ddnet] Pull request opened: #2684 Make steam_api dependency static on non-macOS non-Steam builds

This allows building without rpath modifying on Linux which seems to be needed for openSUSE packaging. Fixes #2669.

10:34

[ddnet/ddnet] Issue closed: #2679 Umlauts stopped working?

10:34

[ddnet:master] 3 new commits

3de4bc5 Upgrade SDL2 for Linux Steam release to 2.0.12 (fixes #2679) - def- 44b1014 Load serverbrowser more fairly - def- 2795082 Merge #2678 #2681 - bors[bot]

10:34

[ddnet/ddnet] Pull request closed: #2681 Upgrade SDL2 for Linux Steam release to 2.0.12 (fixes #2679)

10:34

[ddnet/ddnet] Pull request closed: #2678 Load serverbrowser more fairly

10:39

[ddnet/ddnet] Issue opened: #2685 Mouse cursor still works ingame when console is open?

Reported by becc on Discord, can't reproduce on Linux:

When you open console and click any mouse button, the click will happen in background on the teeworlds-game-cursor, so you could accidentally join a server or something

[ddnet/ddnet] Issue opened: #2686 Autoupdater often fails with permissions problems

Some ideas:

Do we still need an autoupdater or is the Steam one good enough?
Should we reimplement the autoupdater as a separate executable launching DDNet?
Could we auto-create the update json file based on diffing two directories?
Could we show permission problems more clearly in the UI?

10:58

[ddnet/ddnet] Issue opened: #2687 Official maps show the red texture warning

@Jupeyy is it supposed to show for graphics on official maps? Should we go through them and fix all? !screenshot-20200831@125808

[ddnet/ddnet] Issue opened: #2688 New graphics warnings aren't localized

11:08

[ddnet/ddnet] Pull request opened: #2689 Add localization for texture warning, improve wording (fixes #2688)

@deen how do you add the translation texts, have a script for it? we should add an option to disable graphic warnings in the graphic settings tab and maybe https://github.com/ddnet/ddnet/issues/2676 remove these instead

Remove texture compression and high texture quality from graphic se...

seems like ppl are really clicking it: Also they probs not worth it too much. We can still leave them as config. And smth else: maybe for rescaling we should now use the resize function its cubic a...

[ddnet/ddnet] Issue closed: #2688 New graphics warnings aren't localized

11:20

[ddnet/ddnet] Pull request closed: #2689 Add localization for texture warning, improve wording (fixes #2688)

11:20

[ddnet:master] 2 new commits

ffa9550 Add localization for texture warning, improve wording (fixes #2688) - def- 3bbf657 Merge #2689 - bors[bot]

11:25

[ddnet/ddnet] Pull request opened: #2690 Remove Quality Textures & Texture Compression from graphics menu (fixes #2676)

Since people just randomly click it and then complain about bad looking graphics

11:31

[ddnet/ddnet] Issue closed: #2669 libsteam_api.so: cannot open shared object file: No such file or directory

11:31

[ddnet:master] 2 new commits

32734ac Make steam_api dependency static on Linux non-Steam builds - heinrich5991 7a4284c Merge #2684 - bors[bot]

11:31

[ddnet/ddnet] Pull request closed: #2684 Make steam_api dependency static on Linux non-Steam builds

@deen yes, build with -DSTEAM=ON for the steam releases (for windows/macOS it doesn't matter right now, but it doesn't hurt either)

[ddnet/ddnet] Issue closed: #2676 Remove texture compression and high texture quality from graphic settings menu

11:47

[ddnet:master] 2 new commits

9dc6a1c Remove Quality Textures & Texture Compression from graphics menu (fixes #2676) - def- c05e22d Merge #2690 - bors[bot]

11:47

[ddnet/ddnet] Pull request closed: #2690 Remove Quality Textures & Texture Compression from graphics menu (fixes #2676)

[ddnet/ddnet] Pull request opened: #2691 Also check for divisibility by 16 of height (fixes #2687)

what this sign mean?

Image attachment

13:58

seems like background texture was not initialized or something

13:59

reproduced in 100% with old OpenGL

yeah i think i looked that up some day it was blob.png or some debug texture

14:00

cant remember rn

correct (its blob.png) (edited)

i think blob without blending

14:01

if i remember correctly, i just enabled glblend default in 3.3

14:01

that's maybe why its not there

should I open an issue for this?

14:02

this thing here for a looong time

i dunno if its fixable, dont have a profiler that works with < gl 3.2

14:02

but i can try later

14:05

m_State.m_BlendMode = CCommandBuffer::BLEND_ALPHA;

14:05

can you try that @BannZay in graphics_threaded.cpp

14:05

sadly i barely see the loading screen

your compucter is too fast

or i have to take screenshots automtically to test

yes I will try

14:06

I used obsstudio to catch it

ah ok

14:07

in the constructor btw

you are genius my friend

14:08

it helped

ok nice, you can create the pr if u want

[ddnet/ddnet] Pull request opened: #2692 Set blend mode on background render

Old OpenGL, 100% reproducible: !image

[ddnet/ddnet] Issue opened: #2693 Feature: Time Limit Tile

I propose a feature that is similar to a countdown timer, where after you hit the initial 'start entity', you have X amount of seconds to get to the 'end entity'. If you do not get to the end entity within X seconds, you either die, or you are teleported to the beginning (as specified by the editor.)

@Pathos what graphics card are you using? 0.7 has the async rendering turned off by default. You could try gfx_asyncrender 1. Might help but it can also make it worse =\ For me both ddnet and 0.7 feel good in terms of latency

@redix I will try this, thanks

@Learath2 do u know why would u use a single element tuple (x,) instead of just x (in langs that support them)

15:11

hm, maybe to keep an API consistent?

tee_thinking

15:12

@Learath2 apparently haskell doesnt have them

15:12

they consider them useless

single element tuples?

15:13

does indeed sound like an oxymoron to me

@Learath2 https://github.com/microsoft/vcpkg

microsoft/vcpkg

C++ Library Manager for Windows, Linux, and MacOS. Contribute to microsoft/vcpkg development by creating an account on GitHub.

15:19

what do u think about this?

15:19

it has cmake support

hadn't seen it before

15:21

very skeptical, there have been attempts at a c/c++ package manager before, there is a reason none stuck on

10k stars

15:22

xd

is it a popularity contest?

justatest

15:25

@Learath2 well, i think it is more good for windows than linux

15:25

since distros have their own package managers

15:25

but windows not

one issue I see is that it uses the cmake toolchain file, many projects use that for setting up cross compilation instead (including ddnet)

15:27

Even in linux c/c++ library management is quite awful

Image attachment

2

No good way to pin versions being the major issue imho

i got it

18:45

my dummy couldn't connect because dummy didn't have name

18:45

it just was empty (edited)

omg

also, i could enter to other mods (edited)

18:46

with dummy

is this can be releated with steam name update?

yes

1

can we have a visual clue in the leaderboard when a team is in practice mode?

@Im 'corneum I was going to add this, but it requires a rather tough patch to replace netobjects in snapshots

i dont know what that means but i hope you get it working

It requires a motivation boost and some extra time

@Learath2 i believe in you!!!❤️

19:19

... now you only need some extra time

send time

️

19:39

[ddnet/ddnet] Pull request opened: #2694 Fix mouse interaction when game console is open

fixes #2685

Image attachment

How is it performing with that many people around?

i had 128 players at once already :D

19:44

for some minutes, 2 days ago

19:44

works quite good

19:44

also regarding the 0.7 side of the server, seems to work good after i updated the algorithm

19:48

@Learath2 when i tried to send more than 64 players in the client count, like 1-2 years ago it wasnt possible, only the max clients, now it also works for the num clients

[ddnet:master] 2 new commits

9375b29 Set blend mode on background render - BannZay 4afec9e Merge #2692 - bors[bot]

19:55

[ddnet/ddnet] Pull request closed: #2692 Set blend mode on background render

@fokkonaut btw, why not add 128 players to official DDNet? Did we have that discussion already?

8192 characters when

in theory its not hard @deen, you have the algorithm already basically, for me it was only hard because i have to handle 0.7, which uses net msgs, not snapshots

well, we also support 0.7

ah, true, lol

19:58

also, i have no client-support

19:58

its all based on the id-maps

19:58

just like ddnets 64 player version in 0.6

20:00

If you are interested, I could add it, my algorithm isnt perfect tho and I do think that it can cause micro laggs, but I think your servers are better anyways

20:01

I wrote my own UpdatePlayerMaps algorithm, to handle some things slightly easier and to have it a bit more improved (i hope, never actually checked, but by feeling its better :D)

Personally I think it's interesting, not sure if @Learath2 and @heinrich5991 agree

[ddnet/ddnet] Issue closed: #2685 Mouse cursor still works ingame when console is open?

20:03

[ddnet/ddnet] Issue closed: #2685 Mouse cursor still works ingame when console is open?

20:03

[ddnet/ddnet] Pull request closed: #2694 Fix mouse interaction when game console is open

@deen i needed to improve stuff, such as all players from one ip having the same faked ids among them, to make sure hammerfly and stuff with dummy works fine

20:14

The initial commit bascially has everything already but ofc i fixed some smaller bugs with time

@Learath2 any tips/ideas on finding protocol bugs which leads to 0.7 client freeze? First one I've found was CNetObjPickup which had different size on 0.6 and 0.7

20:54

Image attachment

20:54

I've fixed it like this and it worked like a charm

20:55

But it was easy to fix because I knew the pickup will appear and what is causing it to appear.

20:55

Now I have some different network protocol packet which only being sent after a warmup end and it hangs 0.7 client only.

20:56

And I don't know what it could be.

Hm lemme think

Client freezes or hangs due to invalid snapshot?

Seems like that

can you open menus like settings, f1?

20:57

If yes, its invalid snaos

20:57

snaps*

yes, that's not client itself freezes

ok

20:57

then make sure to use correct Sizes

20:57

for all snap objects

it just seems fail to parse correctly snapshots

I forgot to adjust one size, then the snap was invalu

20:58

invalid*

but client doesdn't disconnects so it at least thinks it doing something useful and sends packets to a server

Yea, its exactly that

20:58

Fix all sizes

Any thoughts on how to get info from client itself?

20:59

maybe add some debug line somewhere

Just check all your snaps in the server

20:59

https://github.com/fokkonaut/F-DDrace/commit/116af771d0366da7ff069748b386bcd0ebe3d2cc

Fix heartgun for 0.6 · fokkonaut/F-DDrace@116af77

20:59

I forgot this one, after I fixed it it worked.

20:59

Check all snaps you send

first one was HEARTH pickup yeah

21:00

but now I have only 2 clients in snapshots and idk what could be wrong

21:00

tbh I don't know what is in snapshots ofc

:D

it just game don't have anything besides 2 characters

21:00

and a map ofc

Okay, do you have any own custom snap Objects?

plenty of them

Then fix all of them

21:01

The sizes need to be correct

but they're not spawn at the start of a game

21:01

no custom object is in game

Do you have a github?

yes but the project is closed source for some period

21:02

at least while there is a threat of clone servers (edited)

21:02

I don't want players to split on servers

21:02

so no

21:02

it is something related to characters I think

Its exactly what I said :)

The netobj translation is just not there :/

any tips on what could be added to client itself to raise exception on invalid snapshot state?

turn on debug mode and check console

21:05

it should say failed delta unpacking

21:05

with information on what object

hm... I could build a client with debug on... right

No

21:05

just f1: debug 1

21:05

then check console window

21:05

it doesnt get printed to f1, just to the console window

wow thanks

21:07

so much info

21:07

and hud is really nice

21:08

it prints some net packets errors which is awesome

21:08

[2020-09-01 00:07:11][game]: invalidated index=7 type=-1 ((out of range)) size=8 id=0
[2020-09-01 00:07:11][game]: invalidated index=8 type=-1 ((out of range)) size=8 id=0
[2020-09-01 00:07:11][game]: invalidated index=9 type=-1 ((out of range)) size=20 id=0

21:08

console is full of this

21:08

and a

[2020-09-01 00:08:31][client]: prediction time reset!
[2020-09-01 00:08:32][client]: prediction time reset!
[2020-09-01 00:08:33][client]: prediction time reset!
[2020-09-01 00:08:34][client]: prediction time reset!

21:09

I think this is something I can work with

Why dont you listen? :/

21:12

It is the snap sizes, because they differ in 0.6 and 0.7

21:12

You have to send the correct objects to the specific clients

[ddnet/ddnet] Issue opened: #2695 Websockets stopped working

Might be related to websockets update, I'm on Arch Linux using libwebsockets 4.0.20-3 Using https://teewebs.net/, compiled server with -DWEBSOCKETS=ON, getting this on the server: ``` [2020-08-31 23:12:02][websockets]: connection established with 0.0.0.0:55272 , assigned fake port 2 [2020-08-31 23:12:02][websockets]: connection with fake port 2 closed [2020-08-31 23:12:03][websockets]: connection established with 0.0.0.0:55274 , assigned fake port 6 [2020-08-31 23:12:03][websockets]...

I add a dbg_break to delta unpacking failed while testing the bridge

21:14

I also add a dbg_msg printing out every object in the snap in delta unpack and one in snap new item

Why dont you listen? :/ It is the snap sizes, because they differ in 0.6 and 0.7 You have to send the correct objects to the specific clients

I've listened, still want to know which exact packets are wrong. And this debug info helps me to find.

21:14

I don't have the whole project and type sizes in my brain @fokkonaut far from that

me neither, but you can use sizeof

21:15

because ddnet has both protocols

I probably still should add a script to check netobj differences between the protocols and only auto translate if they are the exact same

[ddnet/ddnet] Pull request opened: #2696 Fix syntax for primary key tables

To use sizeof I should know what to examine, don't you think?

@Deleted User what's the issue here?

21:16

You are sending an invalid object thats breaking the client?

I am sure "I" do.

Invalidated index is normal, its the extended protocol objects which get dropped by vanilla

ok

21:18

but network errors in hud is a useful sign?

The better way to diagnose this is to first determine how it hangs

they're not many but they are there

[ddnet/ddnet] Issue opened: #2697 Need a way for client to connect using websockets

Maybe something like ws:127.0.0.1:8303 Maybe try websockets when udp connection doesn't work.

hm....

21:18

The better way to diagnose this is to first determine how it hangs

sure, just I don't know how it is happening (edited)

What kind of hang is it, can you just disconnect and reconnect

21:20

?

My steps to reproduce are: 1. Start a server 2. Connect with F-Client to the game 3. Connect a dummy 4. game entering a warmup stage 5. vote to change map 6. map changed 7. I am on new map w/o dummy 8. I click on connect dummy 9. All 2 chararacters are stopped reacting 10. If I connect with DDNet client to this game it looks fine for 0.6 client (edited)

[ddnet/ddnet] Issue opened: #2698 Switch to our own config directory?

At the moment we collide with Teeworlds. If Teeworlds has a default in the config set and you ran it before then the first time running DDNet will not show you the welcome screen and also keep the Teeworlds defaults instead of ours. I guess this should be changed by now. Something like ~/.ddnet instead of ~/.teeworlds

Oooh f client, i wouldnt know, it could be an fclient issue too

21:21

Should ask fokkonaut

Lol

21:23

vanilla 0.7.5 is handing before drawing a map

21:23

even w/o other characters

21:23

so F-Client does it better

21:23

though the problem is there

@Learath2 its the snap sizes...

But that's nice. At least a bug that fails very early

its not a big bug, really

Snap sizes doesnt make sense, do you mean netobj sizes?

Yea

21:25

Just check all your Snap items and make sure to send the Protocol7 version to IsSixup clients and the normal version to non sixup

ok

21:26

Maybe this will be easier.

21:26

Do inspect evevry invocation of NetObjAnything in code

21:26

Ok, thanks. I've got more than I've expected to unstuck myself in investigating this.

https://github.com/ddnet/ddnet/blob/f00cab1b5585858117e2d815e3be0ed738475356/src/game/server/entities/pickup.cpp#L165

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

21:27

@Learath2 you can use sizeof(protocol7::CNetObj_Pickup) here

21:27

and at all the other places where its hardcoded by timakro

Ah I knew I forgot some

[ddnet/ddnet] Pull request opened: #2699 Update norwegian translation

basically also the hardcoded Types @Learath2

21:29

2, 3, etc

21:29

protocol7::PICKUP_x

So it's 12 bytes for 0.7 protocol and variable sized for 0.6?

21:29

for pickups I mean

depends on what object

21:30

Use sizeof

1

21:30

it automatically does that

I will try that Thanks

21:30

Good night

[ddnet:master] 2 new commits

4bef000 Update norwegian translation - trml d87add5 Merge #2699 - bors[bot]

21:39

[ddnet/ddnet] Pull request closed: #2699 Update norwegian translation

What is the @DDNet bot programmed in

python: https://github.com/12pm/ddnet-discordbot

12pm/ddnet-discordbot

Contribute to 12pm/ddnet-discordbot development by creating an account on GitHub.

oh epic

22:08

python???? [superiority complex engaged] umm why not use C???

(edited)

22:10

JK i dont care what programming languages poeple use

Exported 273 message(s)