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 2023-01-08 00:00:00Z and 2023-01-09 00:00:00Z

Ryozuki

Click to see attachment 🖼️

What is this game? And what is DDNet steam statistics?

@python311 those are the stats from steam as a whole

1

02:03

https://steamdb.info/app/412220/graphs/

DDraceNetwork (App 412220) · Steam Charts

Want to play the hardest cooperative 2D platformer ever? Want to finish no map ever? Want to be in pain for hours and cry, getting nothing in return? Come play DDNet with a large community of other sufferers! DDraceNetwork Steam charts, data, update history.

1

02:03

^ for ddnet

[ddnet-maps:master] 1 new commit

142b9c0 A Death Rays - ddnet-maps

@Robyt3 i think i found a bug:

go on lan tab -> select a server
go on ddnet tab -> select a server
go back to lan tab -> double click on the already selected server -> it connects to the ddnet server instead bcs server ip is not updated

09:47

and it must be the same index

09:47

between both tabs

https://github.com/Airblader/i3

09:54

️ i3-gaps has been merged into i3.

09:54

@Learath2 poggers

09:55

https://news.ycombinator.com/item?id=34293010

harporoeder

The i3-gaps project has been merged with i3

[ddnet/ddnet] Pull request opened: #6250 Fix server browser selection not being updated based an address input

The selected server browser entry was not being updated anymore when the server address input is changed manually by the user or when selecting an entry in the LAN server list and then switching back to the Internet list. Regression from #6240.

Checklist

[X] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integr...

[ddnet:master] 2 new commits

a0d4a55 Fix server browser selection not being updated based an address input - Robyt3 367cced Merge #6250 - bors[bot]

10:51

[ddnet/ddnet] Pull request closed: #6250 Fix server browser selection not being updated based an address input

(connecting) players aren't filtered anymore

I think they weren't filtered in the server scoreboard before either

11:01

somehow the flag is also larger for them

Image attachment

yeah if they dont have a skin they render like before, looks kinda ugly but usually they are filtered. at least i thought so

11:03

but there is a "filter connecting players" filter

11:03

so its weird they weren't filtered, it should even be default on

yeah, makes sense to hide them from the player list as well, since they are not counted for the total player count with "filter connecting players"

Image attachment

11:07

it suddenly happened

11:08

on a kog server

11:08

and it also stopped after that

"Count players only" also doesn't affect the server browser scoreboard currently. Not sure it's a good idea to hide the spectators in that list if this setting is enabled.

mh ok, questions are:

is it useful to know that there is (connecting) player
should we always render as if there could potentially be a skin, which makes it not use most space available but simplifies the code

- is it useful to know that there is (connecting) player

probably not

- should we always render as if there could potentially be a skin, which makes it not use most space available but simplifies the code

would look better/more consistent IMO

so be it

[ddnet/ddnet] Pull request opened: #6251 Always assume there could potentially be a skin in serverbrowser's pl…

…ayer list

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/d...

Jupstar ✪

Click to see attachment 🖼️

this seems to be related to lags

maybe we should use different return values for all returns in CSnapshotDelta::UnpackDelta, so we know which condition caused the unpacking to fail

12:18

right now we group the failures roughly in a few categories, where -3 means that a value in the delta is invalid

Yes please

@Robyt3 in the previous version, when no server was selected e.g. by switching tabs, you could still use arrow keys to navigate (in this case it would start navigation) now its not working anymore

[ddnet:master] 2 new commits

ed1af6e Always assume there could potentially be a skin in serverbrowser's player list - Jupeyy bda61e5 Merge #6251 - bors[bot]

13:27

[ddnet/ddnet] Pull request closed: #6251 Always assume there could potentially be a skin in serverbrowser's pl…

[ddnet/ddnet] Pull request opened: #6252 Use distinct error return codes for `CSnapshotDelta::UnpackDelta`

Use a different error code for every return statement, so it's easier to determine why unpacking a delta failed. The codes are grouped by the first digit of the error code:

-1xx: not enough data to read
-2xx: value is invalid
-3xx: could not build snapshot item

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (esp...

1

[ddnet/ddnet] Pull request opened: #6253 Allow navigating listboxes when no item is currently selected

It was no longer possible to navigate listboxes (e.g. the server browser) by using keyboard hotkeys when no item is currently selected due to a regression from #6240. Additionally, it was not possible even before #6240 to start navigating the server browser with the keyboard afer manually clearing the server address input, which is now possible as well.

Checklist

[X] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with p...

@Robyt3 can i cancel #6252 so you can use enums?

https://github.com/ddnet/ddnet/pull/6252

Use distinct error return codes for CSnapshotDelta::UnpackDelta b...

Use a different error code for every return statement, so it's easier to determine why unpacking a delta failed. The codes are grouped by the first digit of the error code: -1xx: not enough da...

Not sure an enum would help much. We wouldn't use those numbers internally, they are just for understanding console log messages

14:12

You wouldn't do if(UnpackDelta(...) == -301) anyway

hmm ok

14:14

we could use enum to string in dbg_msg instead of raw numbers as well

[ddnet:master] 2 new commits

066768c Use distinct error return codes for CSnapshotDelta::UnpackDelta - Robyt3 7d4fabd Merge #6252 - bors[bot]

14:15

[ddnet/ddnet] Pull request closed: #6252 Use distinct error return codes for `CSnapshotDelta::UnpackDelta`

at that point return a string rather than an enum, I guess

nah, we still need integer data

the function normally returns the size of the unpacked delta

it returns size of data

14:16

and negative for error

so any negative value is an error code

[ddnet/ddnet-web] Pull request opened: #235 Add .well-known files for :ddnet.org matrix server

hmhmm

14:17

what would be the advantage of enums over integers in this case?

enums are integers anyway

only thing i can think of is easier documentation directly inside the enum

yup, that's the point

So bad c++ doesnt have rust style enums

you can have the documentation near the return instead, if you want that

14:18

currently, the returns are unique

some comment you mean?

yea

14:19

or wdym by documentation?

the intention is that the existing return values are not changed or reused again, otherwise we couldn't interpret old log files easily

Ryozuki

So bad c++ doesnt have rust style enums

what are rust enums specificities ?

you could have the function return the size OR an error (e.g. a string)

Ryozuki

So bad c++ doesnt have rust style enums

do tupple enums have an automatic mode for integers? or do you need to do enum { A((1, "Whatever error")), B((2, "...")), } for every entry?

Chairn

what are rust enums specificities ?

They are sum types, and can hold any type of data within

14:20

Result for example

any type or any integer type?

Any type

ah that's nice

heinrich5991

you could have the function return the size OR an error (e.g. a string)

or just a boolean pointer as parameter to see if it worked correctly

Jupstar ✪

do tupple enums have an automatic mode for integers? or do you need to do enum { A((1, "Whatever error")), B((2, "...")), } for every entry?

with automatic i mean like start index = -50000 and then rest goes automatic

it all seems like overengineering though

enum IpAddr {
        V4(u8, u8, u8, u8),
        V6(String),
    }

    let home = IpAddr::V4(127, 0, 0, 1);

    let loopback = IpAddr::V6(String::from("::1"));

@Chairn

nobody commented on my pr

https://github.com/ddnet/ddnet/pull/6219

what Robyte did is instantly useful and doesn't change much code

Ryozuki

enum IpAddr {
        V4(u8, u8, u8, u8),
        V6(String),
    }

    let home = IpAddr::V4(127, 0, 0, 1);

    let loopback = IpAddr::V6(String::from("::1"));

@Chairn

but that doesnt help if we talk about integers tho, so answer me rust pro

return true; // NOLINT(clang-analyzer-unix.Malloc)? ^^

14:23

should we just disable clang-analyzer-unix.Malloc? ^^

heinrich5991

return true; // NOLINT(clang-analyzer-unix.Malloc)? ^^

otherwise clang-tidy complains about potential leak

why did that change?

here you got clang-tidy errors https://github.com/ddnet/ddnet/actions/runs/3841720848/jobs/6542261625

14:24

probably due to inlining of mem_zero function allowing it to see more things?

Jupstar ✪

with automatic i mean like start index = -50000 and then rest goes automatic

Ye

14:26

Idk about enums eith values

Ryozuki

Ye

nice is it a macro or smth? i tried to google this before for tupple enums

14:26

but seems like i suck at googling rust specific stuff

But with enums that look like c its just enum { X = -5000 }

i need for enums with values ^^

I have to check it but im outside so its a hassle

14:27

Mobilr sucks

smth like #[auto_add_convert_into_interger_start_at(0)]

14:27

would be really useful for me at some specific places

I don't think that exists(?)

you can define enums to integer conversion

14:28

enums act like classes

too bad, bcs it looks like not too hard to do with a macro

https://doc.rust-lang.org/rust-by-example/custom_types/enum/c_like.html

C-like - Rust By Example

Rust by Example (RBE) is a collection of runnable examples that illustrate various Rust concepts and standard libraries.

[ddnet:master] 2 new commits

8845a15 Allow navigating listboxes when no item is currently selected - Robyt3 a2737de Merge #6253 - bors[bot]

14:36

[ddnet/ddnet] Pull request closed: #6253 Allow navigating listboxes when no item is currently selected

Is #6249 ready, @Jupstar ✪ ? Looks like the run failure is just a macOS fluke

https://github.com/ddnet/ddnet/pull/6249

Fix DoLabelStreamed by using offsets, instead of rebuilding when x,...

Checklist Tested the change ingame Provided screenshots if it is a visual change Tested in combination with possibly related configuration options Written a unit test (especially base/) or add...

[ddnet/ddnet] Pull request opened: #6254 Update Korean translations by CHaBek

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-ad...

Robyt3

Is #6249 ready, @Jupstar ✪ ? Looks like the run failure is just a macOS fluke

i cancelled it bcs there was a bug, but its ready

Chairn

nobody commented on my pr

https://github.com/ddnet/ddnet/pull/6219

I'm looking into it

Chairn

nobody commented on my pr

https://github.com/ddnet/ddnet/pull/6219

i didnt even saw you saying this ^ before 5 seconds ago xd

im a very discrete man

what I find a bit weird is that the new mem_zero doesn't actually zero the mem in some cases

because it shouldn't, it has to respect default constructor

14:52

we can rename it to mem_init

yes, it shouldn't because it's UB

but that's less clear in most cases

perhaps these calls could instead be replaced by calls to the constructor?

ehm https://github.com/ddnet/ddnet/pull/5690#pullrequestreview-1056912261

https://github.com/ddnet/ddnet/blob/a2737de1c5a8f5d06b69222892dbee6757415c12/src/engine/shared/network_console.cpp#L10-L11

ddnet/network_console.cpp at a2737de1c5a8f5d06b69222892dbee6757415c...

DDraceNetwork, a free cooperative platformer game. Contribute to ddnet/ddnet development by creating an account on GitHub.

14:53

for example this ^

14:53

this seems to be very misleading after the PR

it does zero it because there's no constructor and no complex type

man this hurts my eyes

btw, mem_zero in constructor would probably endless loop

[ddnet:master] 2 new commits

7815be7 Fix DoLabelStreamed by using offsets, instead of rebuilding when x,y changes - Jupeyy 1347101 Merge #6249 - bors[bot]

15:01

[ddnet/ddnet] Pull request closed: #6249 Fix DoLabelStreamed by using offsets, instead of rebuilding when x,y changes

[ddnet/ddnet] Pull request opened: #6255 Add new french translations for 16.7

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-addresssan...

15:35

[ddnet/ddnet] Pull request opened: #6256 Typesafe `mem_zero` function

My preferred alternative to #6219. Thanks to @Chairn for laying the groundwork for making this safer. This PR simply doesn't allow to call mem_zero on nontrivial types. I like that better than calling the destructors/constructors when the function name suggests it only zeros memory.

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especia...

GitHub

Click to see attachment 🖼️

I thiink I like this better too

running out of time right now, ill review it tonight. Gotta go see Avatar 2

have fun

Tho am on phone, so I'll think about it when im hom

TIL Avatar 2 exists

ye avatar is surprisingly niche

I do get some clang errors did i miss something?

erm... #off-topic ?

../src/base/vmath.h:141:38: error: constexpr function never produces a constant expression [clang-diagnostic-invalid-constexpr]

ChillerDragon

../src/base/vmath.h:141:38: error: constexpr function never produces a constant expression [clang-diagnostic-invalid-constexpr]

Which version of clang? I added constexpr to more functions in #6244. It works for me locally and in the CI.

https://github.com/ddnet/ddnet/pull/6244

Minor base refactoring by Robyt3 · Pull Request #6244 · ddnet/ddnet

Checklist Tested the change ingame Provided screenshots if it is a visual change Tested in combination with possibly related configuration options Written a unit test (especially base/) or add...

Voxel

erm... #off-topic ?

Uhm actually #developer is offtopic for @Devs

https://fanpu.io/blog/2023/latex-tips/

The Art of LaTeX: Common Mistakes, and Advice for Typesetting Beaut...

When was the first time you had to use LaTeX? If you are like most people, it was probably suddenly forced upon you during your first math or CS class where you had to start writing proofs, with minimal guidance on how to get started. Unfortunately, this meant that while many people have good operational knowledge of LaTeX, there are still many ...

Debian clang version 11.0.1-2

17:31

deb11

I guess you can just remove as many constexpr as necessary until it compiles for you and then make a PR

or maybe its fixed in a later clang version eventho dropping deb11 support sounds bad i think if the github CI passes its okaay

17:33

deb12 has to release soon anyways

Ryozuki

https://fanpu.io/blog/2023/latex-tips/

I like everything in this except that they said to always use \dots. Really, you should be using \dotsc, \dotsb, \dotsm, etc

[ddnet/ddnet-web] Pull request closed: #235 Add .well-known files for :ddnet.org matrix server

19:56

[ddnet-web:master] 2 new commits

abc59c5 Add .well-known files for :ddnet.org matrix server - Zwelf 13e77b6 Merge pull request #235 from Zwelf/pr-ddnet-matrix - def-

@Zwelf will we be able to get matrix accounts based on ddnet domain?

(edited)

Anyone here got experience with a Docker container not having working DNS lookup? @Miguilim asked me for the Brazil server, but I couldn't find anything weird, and tried removing my iptables rules, but to no effect

Mh, I had issues with podman and cni, but migrating to netavark fixed all my issues

[ddnet:master] 2 new commits

7437c5c Add new french translations for 16.7 - Chairn fb2ec3f Merge #6255 - bors[bot]

21:58

[ddnet/ddnet] Pull request closed: #6255 Add new french translations for 16.7

heinrich5991

have fun

sorry, i'll do it this week then

@Jupstar ✪ do u know what could be the cause of a GPU not used much on windows ?

22:22

have it used 20% only on some big games

i mean its the games code fault

22:23

or maybe ur cpu is bottlenecking

ReiTW

@Jupstar ✪ do u know what could be the cause of a GPU not used much on windows ?

teeworlds or generally

well have a pretty good config, but this game is damn not optimized at all lmao

22:24

ig only Escape From Tarkov, they released a map and the average FPS for everyone is around 40-50 (edited)

22:24

Image attachment

22:24

justatest

xd

ReiTW

Click to see attachment 🖼️

change first graph to logical cores

22:26

see if one core is at 100%

Jupstar ✪

change first graph to logical cores

u mean for cpu? nothing is at 100

Jupstar ✪

@Zwelf will we be able to get matrix accounts based on ddnet domain?

(edited)

No for now

. And it will probably be never be completely open to registration. For now only as a bridge + moderation tool.

Jupstar ✪

change first graph to logical cores

mayb it bottlenecks cuz of the gpu mem being full

ReiTW

mayb it bottlenecks cuz of the gpu mem being full

possible

a 3060 how much has

trying mip streaming from unity (yes they did a big game using a game engine not good for that, game would run way better using UE) 12G for mine (edited)

my 3080 has 10 xd

22:30

i wonder how it using ur full vram tho

22:30

kinda insane kek

definetly no clue

Image attachment

ReiTW

definetly no clue

but still lot of usage

yeah well

deen

Anyone here got experience with a Docker container not having working DNS lookup? @Miguilim asked me for the Brazil server, but I couldn't find anything weird, and tried removing my iptables rules, but to no effect

does it have a reasonable /etc/resolv.conf?

22:50

does dig work? (edited)

dig doesn't work, I tried fixing the /etc/resolv.conf to use 8.8.8.8, but still didn't work, pinging 8.8.8.8 worked though

22:52

so a bit weird that ICMP goes through but UDP not

heinrich5991

does it have a reasonable /etc/resolv.conf?

it seems something related to dnsmasq

22:55

/etc/resolv.conf only has: nameserver 127.0.0.1 (edited)

That is our local server which we used for ddnet, but it forwards everything. I didn't see that used in docker though

deen

That is our local server which we used for ddnet, but it forwards everything. I didn't see that used in docker though

I found this in the official docker documentation https://docs.docker.com/engine/install/troubleshoot/#dns-resolver-issues

23:00

Image attachment

deen

Anyone here got experience with a Docker container not having working DNS lookup? @Miguilim asked me for the Brazil server, but I couldn't find anything weird, and tried removing my iptables rules, but to no effect

We are running our Server on K3s but with a OCI standard image - so docker compliant

23:08

docker uses its own resolver

23:08

internally

why use kubernetes at all

23:11

or docker

23:11

just use nixos

23:11

1 config, replicate everywhere

23:11

the future is here

23:11

its reproducible too!

k3s

23:12

much simpler than k8s

whather that is

its for edge services

sry im not up to date with the hip words xD

I always thought k3s is people writing k8s wrong

https://k3s.io/

r u running ddnet on arm?

no

23:13

but our servers are not powerful enough to run k8s

23:13

k8s takes a lot of ressources

23:14

k3s is smaller and simpler

anyway this looks like uneeded complexity and overhead to me

23:14

specially for a game server

23:14

nixos ftw

nah its not

23:14

you can run minecraft next to teeworlds without dependency problems

23:14

you can run different teeworlds versions without dependency problems at all

so can u with nix

23:14

and u dont have a container overhead

but learning k8s is great for some jobs

23:15

cloud native and so on

i know there is lot of fuzz about this and jobs

23:15

but i dont like it idk

23:15

something in it

23:15

i prefer bare metal

deen

dig doesn't work, I tried fixing the /etc/resolv.conf to use 8.8.8.8, but still didn't work, pinging 8.8.8.8 worked though

how did you call dig? did you include the resolver IP?

solutions

23:15

@Jupstar ✪ do u agree

dig ddnet.org @8.8.8.8

Ryozuki

@Jupstar ✪ do u agree

tl;dr

@Jupstar ✪ using nixos over e.g docker containers since they run on the native kernel and they are reproducible, thus, u can deploy like docker

docker also runs on the native kernel btw

but it has overhead

is docker open source'd?

what kind of overhead do you envision?

23:17

it's just using linux namespaces

"The cost of security"

Jupstar ✪

is docker open source'd?

Yes

>

One of these security mechanisms is seccomp, which Docker uses to constrain what system calls containers can run. Older versions of seccomp have a performance problem that can slow down operations. Docker still hasn’t enabled this performance fix.

docker is not using an open source license though

that too

But talking about docker is different

https://pythonspeed.com/articles/docker-performance-overhead/

Docker can slow down your code and distort your benchmarks

In theory, Docker containers have no performance overhead. In practice, they can actually slow down your code and distort performance measurements.

Its built using multiple technologies

23:18

Embedded image

23:18

the alternative to runc is crun

23:18

a docker OCI runtime

there is another too

23:18

made in rust

there are multiple

23:19

thats why its soo cool

23:19

you can switch them without a hassle

well its not cool to me but i get it

23:19

https://github.com/containers/youki

GitHub - containers/youki: A container runtime written in Rust

A container runtime written in Rust. Contribute to containers/youki development by creating an account on GitHub.

heinrich5991

how did you call dig? did you include the resolver IP?

/var/lib/pterodactyl # nslookup ddnet.org 8.8.8.8
;; connection timed out; no servers could be reached

@Ryozuki the article says you can disable that via --privileged

deen

/var/lib/pterodactyl # nslookup ddnet.org 8.8.8.8
;; connection timed out; no servers could be reached

can I access that server somewhere to poke it a bit?

@heinrich5991 do u know if volumes are slower than normal partitions?

Avolicious

but learning k8s is great for some jobs

Yeah, it's a bit crazy to me, the huge costs everyone seems to accept nowadays to be on cloud

heinrich5991

can I access that server somewhere to poke it a bit?

sure, ssh root@bra.ddnet.org

deen

Yeah, it's a bit crazy to me, the huge costs everyone seems to accept nowadays to be on cloud

this, big time

23:20

aws costs a fortune

23:20

but companies glad to pay

23:21

i think its "google fever" they all think they are on google-scale so they need to use that

I guess I shouldn't complain since it's what's bringing in my salary though

but in reality not

Ryozuki

@heinrich5991 do u know if volumes are slower than normal partitions?

docker uses overlayfs with many layers AFAIK, that might slow things down; but you wouldn't need to do many layers in theory

heinrich5991

docker uses overlayfs with many layers AFAIK, that might slow things down; but you wouldn't need to do many layers in theory

You can fix that

23:23

Its a bit work ngl, but its worth it

it looks like UDP is blocked but TCP isn't

23:35

sorry, no new information from me

23:41

the FORWARD table has a default policy of DROP, this might be one source

23:42

additionally, the DOCKER-ISOLATION-STAGE-2 DROPs everything as well

23:42

only changing the FORWARD table's default policy does not fix the problem

23:45

prepending an ACCEPT rule to the FORWARD table didn't help either

Exported 274 message(s)