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 2023-01-31 00:00:00Z and 2023-02-01 00:00:00Z

https://github.com/fish-shell/fish-shell/pull/9512

Rewrite it in Rust by ridiculousfish · Pull Request #9512 · fish-sh...

(Sorry for the meme; also this is obligatory.) I think we should transition to Rust and aim to have it done by the next major release: Nobody really likes C++ or CMake, and there's no clear pa...

10:00

gigachad ngl

Lol is the project so small?^^

it didnt port all at once

10:06

This should be thought of as a "port" instead of a "rewrite" because we would not start from scratch; instead we would translate C++ to Rust, incrementally, module by module, in the span of one release. We'll use an FFI so the Rust and C++ bits can talk to each other until C++ is gone, and tests and CI keep passing at every commit.

>

To prove it can work, in this PR I've ported FLOG, topic monitor, wgetopt, builtin_wait, and some others to Rust. The Rust bits live in a crate that lives inside the C++ and links with it. You can just build it the usual way:

10:07

https://github.com/ridiculousfish/fish-shell/blob/riir/doc_internal/fish-riir-plan.md

10:07

the plan xd

Ryozuki

it didnt port all at once

I just don't understand wtf his plan is

Jupstar ✪

I just don't understand wtf his plan is

to port fish to rust

He wants it in one release but goes over cpp bindings

he didnt say he wants it in one release tho?

Why not complete rewrite xd

10:08

The pr says it

where

Ah he means major release

10:10

This should be thought of as a "port" instead of a "rewrite" because we would not start from scratch; instead we would translate C++ to Rust, incrementally, module by module, in the span of one release. We'll use an FFI so the Rust and C++ bits can talk to each other until C++ is gone, and tests and CI keep passing at every commit.

10:10

Was confused by the"span of one release" part

@heinrich5991 did u consider https://github.com/corrosion-rs/corrosion when adding rust to ddnet?

GitHub - corrosion-rs/corrosion: Marrying Rust and CMake - Easy Rus...

Marrying Rust and CMake - Easy Rust and C/C++ Integration! - GitHub - corrosion-rs/corrosion: Marrying Rust and CMake - Easy Rust and C/C++ Integration!

10:12

@Jupstar ✪

Image attachment

10:12

crab wins fish

Or crab killed fish

xd

Why even a crab

10:13

I like that anime girl more

true

10:13

@Jupstar ✪ also the anime girl as details

10:13

the ampersan as a pin

10:14

its hilarious

10:14

&

10:14

Image attachment

10:14

kek

10:14

where is c++ chan

10:15

10:15

found her

Is that the granny from Kiki's delivery service Then she is cool xd (edited)

Jupstar ✪

Is that the granny from Kiki's delivery service Then she is cool xd (edited)

No thats the howlings (?) moving castle (edited)

10:39

kikis delivery service is the girl riding a broom i think (edited)

10:39

And that kid doesnt get old

default

kikis delivery service is the girl riding a broom i think (edited)

But there is a granny in the show xd

10:45

That wants to fly with the broom

10:45

But yeah it's different xd

Yea I know all Ghibli movies

[ddnet/ddnet] Issue opened: #6319 Vulkan unexpected map rendering behavior

Users are reporting issues on our map since Vulkan been made default on DDNet clients. !renderer-bug The problem arises only on enveloped quads with RGBA EnvPoints values greater than 1, and only in game, not in the editor. Clapping all env points to 1 doesn't completely fix the issue, as Quad Color * Env Color with EnvColor > 1 doesn't necessarly produce a capped 255 value, thus every...

GitHub

Click to see attachment 🖼️

@Jupstar ✪ who should i ask for the fix?

f

@Jupstar ✪ who should i ask for the fix?

Puppet#0463

11:29

I dunno if he already did it, but we talked about it in pm (edited)

Jupstar ✪

Puppet#0463

xD

11:32

i see

11:32

he sent me here

11:32

hahahaha

11:32

i guess i'll have to code the tool then

Oh. You can still ask patiga. He is good with his tool and just clamping env points to 1 sounds ez

11:36

If you ask kindly he might do it^^

clamp is ez i could also manage to do it manually

11:36

but quads should be fixed too

11:37

its not enough to reduce the value on the env

11:37

quads are too many..... a script is needed (edited)

Oh it's not enough? I assume it clamps the quad color itself then?

the final result wont be the same

11:38

we tried...

11:38

its because of the multiplication

11:38

if the final result is less than 255

11:38

its not you can reduce only the env value

11:38

the equation is not balanced anymore

Ah yeah I see

so the tool should foreach every quad

11:39

and see the max value it reaches with the current env points

11:39

and change both ...

11:39

i really hoped someone else did this already xd

I only remember I talked about it with patiga a few years ago already. But it wasn't really ever a problem, since ddnet maps don't rely on it. Mh yeah kinda annoying, but rn I also cannot help You BCS not at a computer

Jupstar ✪

I only remember I talked about it with patiga a few years ago already. But it wasn't really ever a problem, since ddnet maps don't rely on it. Mh yeah kinda annoying, but rn I also cannot help You BCS not at a computer

im kinda surprised you havent experienced this problem on ddnet maps... is so easy to set a color env value > 1... you see something in the editor thinking that's the effect, than in game it's completely different

11:51

we also just realized this now cause users starting to report this only now, as they got the renderer switched without really being aware of this

f

im kinda surprised you havent experienced this problem on ddnet maps... is so easy to set a color env value > 1... you see something in the editor thinking that's the effect, than in game it's completely different

But it's only important while also having an alpha< 1

yes... the probability this problem arises is proportional to your number of quads and envs

11:52

and we use those massively

f

we also just realized this now cause users starting to report this only now, as they got the renderer switched without really being aware of this

I remember it was reported a few years ago. But thought the maps where fixed back then

i guess thats why

Jupstar ✪

I remember it was reported a few years ago. But thought the maps where fixed back then

i guess it was easier for you to manually fix ye

@Patiga interested in writing a small script. I want to know if ddnet maps rely on it.. what is twgpu doing in this scenario?

cool, let me know if i can help

Jupstar ✪

@Patiga interested in writing a small script. I want to know if ddnet maps rely on it.. what is twgpu doing in this scenario?

twgpu so far just allows > 1 envelope values and uses them as such, didn't see a issue with it back then

12:01

the result is pretty well defined and the color is clamped automatically at the end of rendering

12:02

but I can see how it is sort of unexpected that you can make stuff brighter I suppose, is that the illogical part you are talking about?

Patiga

the result is pretty well defined and the color is clamped automatically at the end of rendering

But does it clamp it manually?

I don't think so

The GPU driver might simply use the alpha value and multiply it with the rgb value and only clamps that result

12:03

So if rgb is 2 and alpha 0.5 result is rgb = 1

huh, I didn't know about that

Old renderer would clamp the rgb

would you be against manual clamping?

12:04

at the end of the fragment shader

I generally prefer to leave it to the GPU and so all kinds of clamping in the editor already

12:04

I'll rationalize it again

12:05

But if it's not ub from Vulkan spec. We should probs allow it

I mean negative values are bonkers and are allowed too

True

up until recently I only saw negative/above 1 values as: influences the interpolation differently compared to 0/1

12:07

but I think values above 1 can be used in a sensible way and I think software such as blender also handles values this way

They often use such information for brightness

12:07

E.g. hdr

12:08

But that works affect the whole scene

12:08

They also probably don't use rgb xd

in the shader editor (the one with the nodes), rgba values can also be set above 1, just checked

12:10

you don't even have to use math nodes

12:11

if you use the slider it will default into the range 0-1, but if you insert 2, the range of the slider is 0-2

12:13

I don't understand hsv yet. there you can't raise hue and saturation above 1, but value and alpha can have any positive values

12:14

honestly I'd be fine with both behaviors, allowing values above 1 would be fine by me

12:14

how do you see it?

Patiga

how do you see it?

I looked into the spec and some khronous opengl wiki. They clearly give the calculation of the alpha blending and say it is clamped after the fragment output. So I guess it's good to allow it generally

12:16

It makes the maps more robust against such changes anyway

12:16

So even if we change it back in future the maps will just work

what changes do you mean?

The change that the old renderer clamped the value in the cpu

ah

Instead of letting the GPU do it. Only problem remains it. If this means the old renderer has to upload floating points now

12:18

Instead of normalized color valley

12:18

Values

12:19

I changed that one day to match opengl 3.3 behavior. But i think vanilla uses fp

eh, yeah that is unfortunate, wasn't aware of which renderer introduces which behavior

12:19

and vanilla also doesn't clamp?

Jupstar ✪

Instead of normalized color valley

couldn't you still upload the original values, only let the gpu convert them to floats and divide by 1024?

Without shaders it's hard xd

ow haven't touched opengl yet

12:21

btw that is also an issue, many mappers think that 1000 = 1, and not 1024

12:22

I even found that in some mapper tutorial somewhere :/

Yeah xd

12:22

If we want to talk about map format. I can give u many issues xdd

heh ^^

12:22

I'd call that one an editor issue tho

12:23

the users shouldn't have to work with fixed point numbers

Ah I see xd

@Jupstar ✪ @Patiga if this might help you: noticed that the map is buggy also with opengl 3.3.0, but not with the other versions of opengl (edited)

1

Ryozuki

@heinrich5991 did u consider https://github.com/corrosion-rs/corrosion when adding rust to ddnet?

I used it to understand how to embed rust into cmake. however, it didn't really do much and wasn't worth the extra dependency

i see

I am trying to add flag to dfng (which is based on ddnet). I have problems when trying to compile.

dfng-16.5flag/src/game/server/gamecontext.cpp: In member function ‘virtual void CGameContext::OnInit()’:
dfng-16.5flag/src/game/server/gamecontext.cpp:3339:37: error: expected type-specifier before ‘CGameControllerCTF’
 3339 |                 m_pController = new CGameControllerCTF(this);
      |                                     ^~~~~~~~~~~~~~~~~~

15:51

I used the 0.6 source for the flag entity and the ctf mod with some changes.

do you define CGameControllerCTF somewhere?

c++
class CGameControllerCTF : public IGameController
{
public:
    CGameControllerCTF(class CGameContext *pGameServer);

(edited)

15:56

on ctf.h under game/server/gamemodes

[ddnet/ddnet] Pull request closed: #6292 rewrite int64_t to CMask

i love c++ cmake

17:09

1 hour to add a dep

17:10

finally i get to code

Mr.Gh0s7

on ctf.h under game/server/gamemodes

Did you #include ".../ctf.h" in gamecontext.cpp?

deen

Did you #include ".../ctf.h" in gamecontext.cpp?

yeah that was it thanks although I get segfault now xD. I'll put everything under DDRace.cpp.

run with debugger to figure out the crash

it segfaults at this line 78: ((CGameControllerDDRace *)(pGameServer->m_pController))->m_pInitResult = InitResult; on score.cpp under game/server

17:27

So i'll just add the flag-ing to the ddrace controller ;p

[ddnet:master] 4 new commits

1cd9eac rewrite int64_t to CClientMask - Valentin Bashkirov 79f72a5 move typedef to protocol.h - 0xfaulty 056ae40 removed bitmask.h - 0xfaulty 9bd4379 Merge #6293 - bors[bot]

17:34

[ddnet/ddnet] Pull request closed: #6293 rewrite int64_t to bitset for clients mask

Again my curiosity: since ddnet has controller support: would it be possible (just technically) to release ddnet for xbox/ps5? (edited)

I guess, as long as the consoles can reach regular servers

Haha, cool, that would be so funny.

xbox is windows anyway

17:40

thats also why u can play xbox games on win10+

Is it all intel 64 bit?

amd64

17:41

but yeah

What Is Error ?

Image attachment

'拏 MAhdiyar

What Is Error ?

i dunno what exactly its used for maybe for the checksum or for the git commit displayed at the version string but i guess smth like that

17:42

but no idea how to fix xd

omg

'拏 MAhdiyar

omg

did you not use git clone

17:43

but just downloaded the source?

Jupstar ✪

did you not use git clone

ohhh yes

Jupstar ✪

but just downloaded the source?

yes

@heinrich5991 can we detect it ^

The build should still work with the source download though, if not then that's an issue of the build system

[ddnet:master] 2 new commits

f3f592c Update Persian translations (by Quick) - def- d89434c Merge #6314 - bors[bot]

18:03

[ddnet/ddnet] Pull request closed: #6314 Update Persian translations (by Quick)

i love

18:45

Image attachment

18:45

c++ error messages

18:45

i wanted to do some ddnet coding but im reminded why i stay with rust

18:46

it prints this t o tell me there is a duplicate symbol xd

that's no c++ error messages

18:51

that's build system telling which command failed

same shit

18:52

the build is part of the language

18:52

tooling is essential

it's not

for me it is

here you use Ninja i believe

18:52

ninja is at least 30 years more recent than c++

18:53

it's like blaming the patient when the error is due to the doctor

xd

tooling is essential

rust tooling is better, but in this case u probs only need to read the last sentence anyway

so far 75% of my time has been spent adding a library xd

rest is maybe nice to have actually

18:55

in case u arent sure why the error happens

0x00007ffff582dbf0 in vtable for __cxxabiv1::__si_class_type_info () from /usr/lib/gcc/x86_64-pc-linux-gnu/12/libstdc++.so.6

19:03

i get a segfault here now xd

19:08

https://github.com/ddnet/ddnet/compare/master...edg-l:ddnet:encrypted_messages

19:08

zzz

xd

oh

19:10

maybe cuz we dont use namespaces

19:10

the class names collide

19:10

i expected too much

19:16

fixed xd

I have switched to NixOS recently and the Steam release stopped working for me

2023-01-31 21:04:51 I http: fetching https://info.ddnet.org/info?name=Scrumplex
2023-01-31 21:04:51 I http: https://master2.ddnet.org/ddnet/15/servers.json failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
2023-01-31 21:04:51 I http: https://info.ddnet.org/info?name=Scrumplex failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
2023-01-31 21:04:51 I http: fetching https://info.ddnet.org/info?name=Scrumplex
2023-01-31 21:04:51 I http: https://master1.ddnet.org/ddnet/15/servers.json failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
...

Scrumplex

I have switched to NixOS recently and the Steam release stopped working for me

2023-01-31 21:04:51 I http: fetching https://info.ddnet.org/info?name=Scrumplex
2023-01-31 21:04:51 I http: https://master2.ddnet.org/ddnet/15/servers.json failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
2023-01-31 21:04:51 I http: https://info.ddnet.org/info?name=Scrumplex failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
2023-01-31 21:04:51 I http: fetching https://info.ddnet.org/info?name=Scrumplex
2023-01-31 21:04:51 I http: https://master1.ddnet.org/ddnet/15/servers.json failed. libcurl error (60): SSL certificate problem, verify that the CA cert is OK. Details:
error:14090086:SSL routines:SSL3_GET_SERVER_CERTIFICATE:certificate verify failed
...

Can you open https://master2.ddnet.org/ddnet/15/servers.json in a browser successfully?

yes

20:10

curl also works fine

20:11

I expect there are some quirks with Steam on NixOS here

hmm

20:19

Looks like it works fine when i use the Steam Linux Runtime as a compat tool

see #3909

https://github.com/ddnet/ddnet/issues/3909

DDNet's HTTPS requests fail on nixOS on steam · Issue #3909 · ddnet...

[2021-06-14 09:18:48][http]: libcurl version 7.22.0 (compiled = 7.55.0-DEV) […] [2021-06-14 09:18:49][http]: http https://master4.ddnet.tw/ddnet/15/servers.json [2021-06-14 09:18:49][http]: http ht...

Hello, i can ask questions about coding here?

dont ask to ask, just ask?

How to "substring"?

newlesstee

How to "substring"?

Do you mean in general or in tw/ddnet code?

depends heavily on the language if it's not tw/ddnet

Robyt3

Do you mean in general or in tw/ddnet code?

ddnet code

would it be helpful if there was an additional option like "Has people not playing", so it's more accessible to find empty maps?

Image attachment

5

Exported 217 message(s)