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 2023-06-08 00:00:00Z and 2023-06-09 00:00:00Z

in testing my code i noticed the game does not handle opening a huge file properly

01:44

Image attachment

01:44

segfault here

01:45

winduhs

the file is ~8GB, but io_length() is returning 0 it allocates len (0) + 1 bytes, the resulting buffer is too small for any reading, disaster ensues

Image attachment

02:01

i think unsigned is too small for this, size_t or ulong should be used if anyone's trying to refactor this

02:02

but afaik there are no existing issues with this, it just came up in testing my code which reports if the file is too big

02:02

unrealistic scenario

02:05

but yeah either you can expand io_length to return a larger size or you can return from io_read_all if the result is 0 or negative, this at least lets me handle a file being too big

02:08

actually io_length() already returns a long, i don't know what's going on

io_tell is returning 0 on this file even though its IOHANDLE is non-null (edited)

02:17

this is probably known behavior and i'm just silly

02:17

it's 8425177088 bytes which is well within the boundaries of even a signed long

rip

02:26

my discord number is robbed

1

02:27

now i'm just tsfreddie

sad

02:34

eyah testing on an 8GiB file is giving me 0 lol

// Handle describes an opened non-null IOHANDLE
io_seek(Handle, 0, IOSEEK_END);
long ExpectedSize = io_tell(Handle);
io_seek(Handle, 0, IOSEEK_START);
// ExpectedSize is 0

02:34

this blows

i'm not finding a fstat io_ euqivalent

02:48

agh

agh

anyone ever heard of mingw misunderstanding windows file permissions

03:31

i set the perms of this file through windows security, readable only to admins

03:31

Image attachment

03:31

msys's stat compiled with mingw doesn't get it

03:31

my checks using std::filesystem::perms also wrongfully report it as readable

03:32

i'd have to compile with msvc to make sure this isn't a different issue but setting up vcpkg or whatever would be required to compile ddnet sounds like a nightmare

unrelated but what's the naming convention for C++ style loop iterator names

03:43

i usually do it (lowercase) but would It be more appropriate?

symlink detection is also screwed

[ddnet-maps:master] 1 new commit

64f7523 A Jomp - ddnet-maps

@Ewan i hope u are aware u compiled for 32bit

05:13

At least it looks so

05:14

Besides that, yes our system CPP uses int very often instead of the correct usize

05:14

So you probably run out of luck with ddnet for such stuff

05:17

You could read it in chunks. Maybe then it works ^^

Jupstar ✪

@Ewan i hope u are aware u compiled for 32bit

what makes you think that

Ewan

what makes you think that

The memory location is 32bit in your image

the address width in the screenshot? it's automatically truncated

05:40

it's a 64 bit binary

05:41

mingw w64 and PE header says 64

Mh ok weird

i'm not sure what's going on with the file loading but i'm convinced that it's not actually ddnet's fault

05:44

yes it should use long wherever needed, but io_tell (which wraps ftell directly) has nothing between the caller and the libc that should narrow the number, and still has this issue

05:44

it may be a mingw issue, but i am using the 64 bit compiler

05:44

the limit is 2GiB exactly

05:44

1 byte less works

What function signature does it have

this is classic 32 bit issue

Jupstar ✪

What function signature does it have

returns long

05:44

ill look up actual sig

Long would be 32b on windows xd (edited)

05:45

But yeah dunno. Windows is always the problem

Jupstar ✪

Long would be 32b on windows xd (edited)

I think this is specific to MSVC

05:49

oh i guess not

05:50

ug

05:51

ftell vs. ftello

05:52

ftell returns long which is 32 bit and that's the issue

05:52

ftello is the correct function to use in this case

05:52

annoying

I c

05:54

But with std FS it should work tho

idk, what u mean

05:55

i could overcome this by doing what the system.cpp does and just casting my IOHANDLE to a FILE * and using ftello myself instead of io_tell

Std filesystem

the std fs issue is different

What issue?

05:55

I mean try using the std

std filesystem doesn't define an API for opening and reading/writing from/to files

?

what's confusing

I mean there must be some CPP way

I think escape button is broken in the server browser

oh i thought u meant std::filesystem

is it supposed to work in the server browser?

yes i could use stl apis for this but i think most ddnet devs would think that using existing ddnet apis is better

05:58

because they can be wrapped and adapted etc

Tater

is it supposed to work in the server browser?

idk, #bugs

I was wondering if its related to #6709

https://github.com/ddnet/ddnet/pull/6709

Fix escape not working in settings menu anymore by Robyt3 · Pull Re...

The color picker was always consuming the escape hotkey even when already disabled. Regression from #6705 due to changed order of hotkey consumption. Checklist Tested the change ingame Provided ...

If u run old code

Im on latest

Ping rob or create issue then

i wonder if adding templates to ddnet system functions would be the move

06:03

type inference should make existing code still work

[ddnet/ddnet] Issue opened: #6718 Escape doesnt work from server browser.

but the system functions can then use ulong/size_t (ull) and the corresponding libc functions (edited)

06:04

i am also happy to live with this limitation lol nobody needs >2GB files

06:04

but future proofing and whatnot

Ewan

i wonder if adding templates to ddnet system functions would be the move

If u expose it additionally Also no harm in using CPP standard directly

06:05

The system stuff is rather useless

06:05

Most of it

06:05

Not all

well, there are a lot of gotchas when using the libc across systems

06:05

platform and compiler differences make things complicated

06:05

having system apis like this lets us account for it

Yes that's why u don't rely on c

afaik the only cpp apis for this sort of thing are wrapped C APIs (std::fopen, std::fread, etc) and file streams

I mean there is ifstream and stuff

those are file streams

I doubt there is no CPP way xd

ifstream = input file stream

Perfect

06:07

Then u have it

xd

06:07

streaming :(

It's fastest

if i use a cpp api then im mixing cpp apis and ddnet system apis

06:08

which i guess there's nothing wrong with at a small scale like this but that kind of irks me

Mh yeah, ddnets API simply sucks

06:09

Why do u need 8gb btw

i don't, it's just that i want the file loader thing to be future proof

06:10

i think for now i will just use the ddnet apis, since if ddnet ever needs >2GB files, the ddnet system apis will just be updated

Yeah^^

this std::filesystem perms/symlink thing is a real bummer

06:11

i expected more of mingw

06:11

if i want to expose this functionality

06:11

i have to use winapi

I didn't read everything u said sry

Ewan

Click to see attachment 🖼️

yea so i have a hunch that mingw is incorrectly reporting filesystem stuff

06:12

firstly std::filesystem::perms and std::filesystem::is_symlink and the like do not work for me on mingw

Oh

perms incorrectly reporting files as readable. and is_symlink thinking that the symbolic link i made with powershell is indeed not a symlink

Sad

yea

06:13

though, fs_* functions could be added to stat permissions and symlink status

06:14

oh shit

06:15

fs_listdir_fileinfo leverages WIN32_FIND_DATAW on windows, contains a symlink flag

06:16

but it's not exposed by CFsFileInfo, the file info type used by FS_LISTDIR_CALLBACK_FILEINFO which is the fn ptr signature that fs_listdir_fileinfo uses

06:16

annoying!

06:16

i think this is a conversation for later on github

06:17

i either add this stupid functionality or i don't add file permission/symlink status to the file loader

06:20

just saw this for the first time in the loading menu. cheeky

        if(m_Menus.IsInit())
        {
            char aBuff[256];
            str_format(aBuff, std::size(aBuff), "%s [%d/%d]", CompCounter == 40 ? Localize("Why are you slowmo replaying to read this?") : Localize("Initializing components"), (CompCounter + 1), (int)ComponentCount());
            m_Menus.RenderLoading(pLoadingDDNetCaption, aBuff, 1 + SkippedComps);
            SkippedComps = 0;
        }

I'm so glad the translators localize that text

06:21

lol

Yea

06:21

localize all

06:21

it's the way to go

I have no idea why they would allow std::regex in the standard without an error checking mechanism besides a single exception (edited)

06:28

which I can't even use because... exceptions are off

06:29

so i have to just pray the expression passed is valid, or do my own stupid regex parsing

06:29

they don't even have an operator bool()

Yep sucks hard

06:31

I wanted to add regex search support a while ago

they don't provide comparison operators either, so i can't even have a known dud std::regex object for comparison

06:32

you're just fucked

!jup emote and pinger completion now appends a : and a space for better emote spam and pingering

@Jupstar ✪

because pings without space at the end dont hit. like this one @Jupstar ✪! so now you can instantly type after complete no need for : or space

<ChillerDragon> poggers2

poggers2

Epic

<Jupstar>

yeah very nice

07:05

<ChillerDragon22> https://zillyhuhn.com/cs/.1686207916.png

Embedded image

07:05

<ChillerDragon22> is that me being color blind or is that ugly af and super hard to read?

07:06

<ChillerDragon22> oh and html moment it swallowed the spaces

I am colorblind and I can say it is ugly af and super hard to read

<ChillerDragon> men are blind af

07:08

<ChillerDragon> where developer girls?

When fix ur client

Image attachment

deentroll

<ChillerDragon> oh yea @fokkonaut that looks much more sexy

im not colorblind but dark blue on dark background is risky xd

<ChillerDragon> since 50% of the webclient users are color blind i gotta do something

yes

07:12

https://bratgpt.com/

A bratty AI system that listens, learns, and challenges

@fokkonaut do you read your user name here in discord in blue color?

u mean dev role?

yes

no thats purple to me

what color blindness do you have then?

07:13

most ppl have green red

me too, but i think also some other degree. Colors I can easily mix up depending on light level, etc: brown - red green - orange yellow - light green pink - grey blue - purple etc.

i c

sometimes its more clear sometimes not, depending on tone, light, and other factors

don't tell others, but my test to check if gamer girl is girl is to ask if dark purple is blue

07:15

worked a few times already xD

first question people have: "how do you see traffic lights" - well, first im not dumb and green is at the bottom, and next up that green is more white already xD

Jupstar ✪

worked a few times already xD

xDDD

<ChillerDragon> because gril eye more op than male?

yeah

<ChillerDragon> indeed

07:16

<ChillerDragon> smort test

statistically like 10% of men are color blind, but only like 1% of women

07:16

smth like that

<ChillerDragon> so u gamer girl?

I am hehe

i am gamer girliest

<ChillerDragon> UwU

Image attachment

07:20

@Jupstar ✪

<ChillerDragon> international foggonut

for me the top 5 bars are very very similar, except in the 2nd and 3rd row the yellow/green spectrum is a little more bright/pale

fokkonaut

for me the top 5 bars are very very similar, except in the 2nd and 3rd row the yellow/green spectrum is a little more bright/pale

uff, for me all are completely different

Then I am definitely more colorblind than you :P Yet I can see the dev role as purple

Image attachment

07:22

how about this?

fuck you

what color do you see?

thats blue

07:23

xD?

ah ok

IS IT

07:23

??

yeah the color often works

07:23

no its purple too

07:23

but a darker version

it looks more blue to me

seems like they especially struggle with dark purple

its on the edfe

07:23

edge

<ChillerDragon> edge-l

but purple

ryo here when its gettin edgy

07:24

xdd

Jupstar ✪

seems like they especially struggle with dark purple

when i was younger, i painted the sky purple until my teacher said "ehh, why you do that"

lol

thats when i first really noticed xdxd and another teacher was mad at me until i said "dude, im colorblind you fucking cunt" and then she shut up and apologized

07:25

xd

fokkonaut

when i was younger, i painted the sky purple until my teacher said "ehh, why you do that"

lel

@Jupstar ✪ ask me yellow/green ish

07:27

its so hard for me

07:27

tee skins

@fokkonaut yellow/green ish?

skins

dad joke

wat

have to make a skin first

oki

my bad joke flew by

07:28

not good morning

did you skip gym today

im at gym rj

ah

07:28

makes sense

xd

u need to stimulate your brain

07:29

then it works better

07:29

xd

07:29

its like those who need coffee

07:29

or me who needs nictoine :x

this one is hard

Image attachment

thats green

Image attachment

07:30

its closer to yellow than green

where the fuck is green and yellow

07:30

Image attachment

yellow is left

07:30

red to yellow to green

put the slider to where green begins

i think it's hard to say even for non colorblind

Jupstar ✪

its closer to yellow than green

That's true, but it looks very similar to the apple skin, named "apple green"

07:31

Kinda canon?

07:31

Well what the hell does canon mean in this context

Image attachment

I gotta think before I type

i'd around here is "clear" green

Yeah that's sensible

@Jupstar ✪ for example, now in comparison, and with the box around it, i can see the differences pretty good

07:32

comparison is always a good thing for colorblind people

But theres everything in between yellow and you'd typically call a shade of green, rather than a shade of yellow

National Insurance Clerk

Yeah that's sensible

sensitive*

the default skin probably makes it a bit darker

No no

but i'd say it's still yellow

Sensible

07:33

As in makes sense

ah

07:33

ye

07:33

sry i thought u r @ChillerDragon english pro

<ChillerDragon> wot

07:34

<ChillerDragon> pinger

https://github.com/ddnet/ddnet/pull/1272

Treat timeout code as sensible data by ChillerDragon · Pull Request...

07:34

xD

<ChillerDragon> axaxax

"Lime is a color that is a shade of yellow-green, so named because it is a representation of the color of the citrus fruit called limes. " https://en.wikipedia.org/wiki/Lime_(color)

Lime is a color that is a shade of yellow-green, so named because it is a representation of the color of the citrus fruit called limes. It is the color that is in between the web color chartreuse and yellow on the color wheel. Alternate names for this color included yellow-green, lemon-lime, lime green, or bitter lime.

its 5 years ago why i still remember this shit

i'd say its on the yellow part of the lime color xD

lime best

left is the color in the ddnet color picker, then comes lime from wikipedia, then the color the tee has

Image attachment

07:37

Image attachment

default skins has always been more green/beige for me

same colors tee, color picker, wiki

its so warm in the office...

07:40

since i have just opened one of these, which color is this?

Image attachment

id say thats green like the leaf on it

07:41

grass green maybe

07:41

bit lighter

green yes, but the green from the leaf is way darker

its a bit darker yes

in real life at least

07:41

but ye its green i guess so too

07:42

@Jupstar ✪ u work in an office?

sometimes xd

u have a table that u can change the height?

yes

do you use it

no

im using it right now for the first time, its pretty nice actually

i used it 5 times and then decided it sucks

i dont think it sucks tbh, been standing for one hour almost

generally its cool to adjust the height but i probably prefer to work while sitting

me too, but its nice and healthy

07:43

ofc not 8h standing

07:43

but being able to stand up when ur back hurts or whatever

office jobs are always bad for your health xd

07:44

i slowly feel that im getting older xD

ye but mostly back problems

07:44

working on the roof for years is also not good

07:44

or on baustel

07:45

baustelle

07:45

construction side worker

07:45

whatever

[ddnet/ddnet] Pull request opened: #6719 Fix escape not working in server browser and demo browser

Closes #6718.

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/ddnet/...

conclusion: work is not healthy

conclusion: life sux

cant argue about that

07:48

xddd

fuckyousnail

but its still worth it

use rust to ease the pain

pepedrive

Ryozuki

use rust to ease the pain

sorry i dont need more pain

greenthing

Sticker

<ChillerDragon> wait i still didnt add :fuckyousnail: id?

07:52

<ChillerDragon> ryo go leak id

Embedded image

<ChillerDragon> ty

07:56

<ChillerDragon2> pog pog fuckyousnail

fuckyousnail

@fokkonaut

hi

you are lurking

i like ur tortilla picture

cyberFighter

you are lurking

no

07:58

i am working laptop + 2 screens so i can spot your pings faster than the light travels

ok

WAIT

?

is your profile picture

07:58

this one thing

<ChillerDragon2> ?xd

wait

what thing

https://www.youtube.com/watch?v=WtO3AHMBePY

Dancing triangle pumped up kicks Tongo #DancingTriangle

07:59

this

no

ok

ws-client

<ChillerDragon> ryo go leak id

u should consider using the discord api

08:00

instead of asking for every emoji xd

Ewan

instead of asking for every emoji xd

troll

<ChillerDragon2> @Ewan i got all

nice

<ChillerDragon> all i need*

[ddnet:master] 2 new commits

a3abc27 Fix escape not working in server browser and demo browser - Robyt3 a7a0ba7 Merge #6719 - bors[bot]

08:26

[ddnet/ddnet] Issue closed: #6718 Escape doesnt work from server browser.

08:26

[ddnet/ddnet] Pull request closed: #6719 Fix escape not working in server browser and demo browser

[ddnet/ddnet] Pull request opened: #6720 Improve scrollbar option UI and settings, improve joystick settings UI especially on 5:4 and 4:3 resolutions

Screenshots (not an exhaustive list):

Controls settings:
- Before:

After:

Controls settings (5:4):

Before:

!controls_5-4 old

After:

![controls_5-4 new](https://github.com/ddnet...

i hate css flip animations

09:10

why is it so dumb

flippin css

09:11

i am still waiting for a browser with proper front end error messages for html and css

09:11

because i refuse to use bloated css and html generators

09:12

we should replace ddnet desktop client with a mobile first electron app

great idea

tauri

write a DDNet launcher in rust, so graphics settings can be changed outside of the client

oh yea ddnet launcher like polyMC would be cool

a settings editor should be ez

09:15

i could make it

09:15

with egui

we just need to be able to change the graphics settings that crash the client for some people

different config directories and different bot versions in one launcher

and buttons to open the config directory, website etc would be nice

true

09:17

ill look into it

are there configs that crash the client on launch even?

09:17

never seen it

it could also detect if the client crashes and show a helpful message

ye!

for some graphics cards and backends it seems to crash

09:18

and we can't detect this in the client, so we need a launcher

how would it work with steam tho?

09:18

does the steam game launch the ddnet launcher?

yeah, but we could at an option to the launcher to skip it maybe

09:18

add*

ye

since when heinrich not admin but this weird role xd

Image attachment

sounds like a lot of work but would be cool

since always

really

@fokkonaut actually idk if its cuz me

09:19

probs not

i never saw this role before

heinrich refuses to in game admin him self

but back then

09:19

i rly was anti heinrich being admin

09:19

xd

heinrich is ssh admin only

so i guess they made it not obvious

ah so he is admin but not officially

09:19

ye

09:20

very unobvious

09:20

xd

he is also technically never authed in game

xd

you can ez kickvote him

09:20

i know from a friend

kek

2

we all know he gets to decide most stuff

decide yes

09:20

but no ingame powa

09:20

so no regular admin

thats not the definition of admin

he has

you cant pinger him for ingame moderation

ingame admin is the least of problems lel

09:21

like i couldnt care less

hot topics

ye other than that he might be the person in tw with the most power

forgot to mention chiller

09:21

docker is cool for auto-restart when crashing, no need to make a bash script for that

bash script is cool for auto restarz when crashing, no need to use docker for that

09:22

while :; do DDNet-Server; done

heinrich for example decided (mostly alone?) the hold hammer on freeze thing

09:22

which has been a before and after for ddnet

that was intense change yes

btw @Robyt3 is this a good place for the Preview button?

Image attachment

feels weird being together with the buttons that change the sorting

@Mr.Gh0s7 was it hard to implement?

09:23

i didnt look at the pr but im curious

Ryozuki

@Mr.Gh0s7 was it hard to implement?

Since I managed to do it nope

nice feature

09:24

xd

Robyt3

feels weird being together with the buttons that change the sorting

Hmm it must be on the Filemanager/explorer though else it'll be frustrating to change it

yeah. what if it's on the right side like you can expand the server browser in vanilla

Mr.Gh0s7

Since I managed to do it nope

well I just hooked it up to the editor code so I didn't do much tbh

09:25

ok will do that ;)

Image attachment

Ryozuki

which has been a before and after for ddnet

ye and either did he not think about it or just went for it

Robyt3

Click to see attachment 🖼️

No multiline texts?

maybe not outside if that causes issues

Cellegen

No multiline texts?

I think they exist but vanilla chose not to do it

Or they didn't need it

hi roby

we didn't use the whole "Playing/Spectating ... on ... " text either

09:28

hi fokko

kokkonaut

ryozoozki

09:29

zooz

fun fact, i dont zooz on fng

banhammer

1

yet

troll

fun fact, i have a proxy which fills up fng server so every part of the map is visible for at least 1 client and then all the info gets sent to my main client so i can zooz in peace

which fng though?

its a joke

fokkonaut

fun fact, i have a proxy which fills up fng server so every part of the map is visible for at least 1 client and then all the info gets sent to my main client so i can zooz in peace

Ah yes, you just invented dos

no?

09:31

i find it funny, 90% of the things you say are random

kek

not the dos operator

09:31

dos attacks

so random xd

sorry, I'll write more specifically in the future

fokkonaut

i find it funny, 90% of the things you say are random

please try /cellegen random to make me say random bs (don't make a bot which does that justatest

justatest

) (edited)

fokkonaut

i find it funny, 90% of the things you say are random

always been like this

09:33

justatest

Ryozuki

always been like this

that's why you guys hate me so much, and I'm living for it

ws-client

<ChillerDragon22> https://zillyhuhn.com/cs/.1686207916.png

What syntax highlighting lib are you using?

09:41

All of them i tried sucked

09:41

I ended up using monaco editor guts and it works pretty well

nothing is better than tree sitter

https://github.com/ewancg/ewancg.github.io/blob/4bb740396c4e847788073dd240d67c4a6ee3f76c/gist.js#L77

but idk if it has a js lib

09:42

tree sitter is best

09:42

neovim uses it

yea i meant web ones

its so good its on par with a lsp backed highlighter

09:43

yeah tree sitter is standalone too

09:43

https://github.com/edg-l/treelight

GitHub - edg-l/treelight: A syntax highlighter for the web using tr...

A syntax highlighter for the web using tree-sitter. - GitHub - edg-l/treelight: A syntax highlighter for the web using tree-sitter.

09:43

i made this

09:43

but may be a bit broken

09:45

i should retake it xd

yea my website surely needs a bit of love too

09:45

i just don’t do the web stuff often

09:45

the vanilla js approach sucks more and more

09:45

as time goes on

[ddnet:master] 5 new commits

8bdffee Improve scrollbar option UI - Robyt3 651cfd9 Improve joystick settings UI especially on 5:4 and 4:3 resolutions - Robyt3 e0bdad8 Remove DoScrollbarOptionLabeled - Robyt3 4c92309 Use DoScrollbarOption instead of DoScrollbarH for settings - Robyt3 6bffc85 Merge #6720 - bors[bot]

09:46

[ddnet/ddnet] Pull request closed: #6720 Improve scrollbar option UI and settings, improve joystick settings UI especially on 5:4 and 4:3 resolutions

hey @Robyt3 while you're doing all these dope UI improvements, should u add a horizontal mode to the scroll viewer?

do we need horizontal scrolling anywhere though?

i did for some stuff, my implementation is outdated now, and it blows anyway

it's somewhat inconvenient I'd say, because most users don't have horizontal scrolling mouse wheels either

as long as it's not bi-directional scroll main mouse wheel works too

09:50

also, shift or alt + scroll wheel is the convention for that

I'll consider adding it eventually

all my implementation did was map x/y to y/x and w/h to h/w, etc.

09:52

swap dimensions

09:53

it was fine

09:53

it just would have been better if there was more time put into it and less fear of breaking other shit

09:57

oh yeah, it would be nice if when the game is maximized on close it'll be maximized when opened again

Image attachment

11:13

sweet

11:17

i added fs_is_readable and fs_is_symlink (edited)

11:17

the former is easily expandible to account for write and exec access too, i just have it that way atm

how does fs_is_readable work? as far as I know, you can never know if you can truly read/write a file unless you try

11:19

at least on Windows

i was originally using std::filesystem for checking readable/symlink status, but it turns out that on MinGW the platform differences of how permissions and softlinks work are so different, they just decided screw it & to make the API report wrong stuff

11:19

on Windows i'm using security APIs for is readable

well, as long as nobody tries using the game on a network drive

2.46 KB

11:21

generic_access_rights on line 10 is saying to only check for read access, we can easily bind this enum & add another parameter if anyone cares (edited)

what do you need this for?

i'm creating an object that will read files in batches & since it's going to be relatively abstract i want it to be versatile and the errors to be descriptive

11:23

i plan to make an async version and use that for skin and asset loading. but it's not needed for those atm, though it would be nice to say why a read failed

couldn't you call GetLastError on Windows or check errno on UNIX after the operation already failed?

PANIC

Image attachment

Jupstar ✪

PANIC

@RYOzu

lel

shit

nice ping

11:24

Hi RYO

11:24

welcome to dev chat

Jupstar ✪

PANIC

ddnet_engine_shared.lib(std-391022a4250a8b9a.std.feb3b897-cgu.0.rcgu.o) : error LNK2019: unresolved external symbol __imp_NtCreateFile referenced in function _ZN3std3sys7windows2fs20open_link_no_reparse17h9b3d5753800c4b4fE [D:\a\ddnet\ddnet\debug\game-server.vcxproj]
ddnet_engine_shared.lib(std-391022a4250a8b9a.std.feb3b897-cgu.0.rcgu.o) : error LNK2019: unresolved external symbol __imp_RtlNtStatusToDosError referenced in function _ZN3std3sys7windows2fs20open_link_no_reparse17h9b3d5753800c4b4fE [D:\a\ddnet\ddnet\debug\game-server.vcxproj]
ddnet_engine_shared.lib(std-391022a4250a8b9a.std.feb3b897-cgu.0.rcgu.o) : error LNK2019: unresolved external symbol __imp_NtReadFile referenced in function _ZN3std3sys7windows6handle6Handle16synchronous_read17h609a485d92162e8dE [D:\a\ddnet\ddnet\debug\game-server.vcxproj]
ddnet_engine_shared.lib(std-391022a4250a8b9a.std.feb3b897-cgu.0.rcgu.o) : error LNK2019: unresolved external symbol __imp_NtWriteFile referenced in function _ZN3std3sys7windows6handle6Handle17synchronous_write17h5e143db420a86fa8E [D:\a\ddnet\ddnet\debug\game-server.vcxproj]
D:\a\ddnet\ddnet\debug\DDNet-Server.exe : fatal error LNK1120: 4 unresolved externals [D:\a\ddnet\ddnet\debug\game-server.vcxproj]

This is the third time a CI build failed on Windows like this in the last two days

Robyt3

couldn't you call GetLastError on Windows or check errno on UNIX after the operation already failed?

not afaik, seems unintended with io_* because it's all wrapped

11:25

and it wouldn't be as descriptive anyway

Even better, github is not loading for me at all

wow this blows

oh man

I'm wondering how we can break linking non-deterministically. My assumption is that is a VS problem?

damn my new laptop is great, but not graphic card rip

VS forget the mutex

ReiTW

damn my new laptop is great, but not graphic card rip

hardware info or did not happen

I have a intel iris xe

Robyt3

do we need horizontal scrolling anywhere though?

Maybe the sliders in settings? Some games do that

i5-1240p, 16gb ddr5 4800MHz, and a beautiful oled screen 2.8k

that is some fast ram

murpi

Maybe the sliders in settings? Some games do that

yeah, we have the value sliders, but we don't have any scroll regions where the content scrolls horizontally

ReiTW

i5-1240p, 16gb ddr5 4800MHz, and a beautiful oled screen 2.8k

nice, but how can it not have a GPU then

it's a pro laptop, not for gaming

11:34

zenbook 14 ux3402za

Robyt3

yeah, we have the value sliders, but we don't have any scroll regions where the content scrolls horizontally

fwiw i added it in my status bar settings gui, it was sort of unique https://cdn.discordapp.com/attachments/806786136474910730/1093164149846061177/2023-04-05_07-23-04.mp4

11:35

mute video

Jupstar ✪

nice, but how can it not have a GPU then

some CPUs don't have graphic drivers compatibility

what

Cellegen

some CPUs don't have graphic drivers compatibility

wat xd

Ewan

fwiw i added it in my status bar settings gui, it was sort of unique https://cdn.discordapp.com/attachments/806786136474910730/1093164149846061177/2023-04-05_07-23-04.mp4

fresh music

sorry

11:36

this is old video

11:36

i forgot to mute

Jupstar ✪

wat xd

As I wrote, some intel CPUs cannot use intel graphic drivers like UHD, because it is not compatible

by gpu I meant a nvidia chip or smth like that, It is using the iGPU obv

Cellegen

As I wrote, some intel CPUs cannot use intel graphic drivers like UHD, because it is not compatible

wat xd

Cellegen

As I wrote, some intel CPUs cannot use intel graphic drivers like UHD, because it is not compatible

that makes no more sense

ReiTW

by gpu I meant a nvidia chip or smth like that, It is using the iGPU obv

yeah i read it like that

Cellegen

As I wrote, some intel CPUs cannot use intel graphic drivers like UHD, because it is not compatible

the only case that i know where the iGPU could not be used was when the monitor was plugged into the gpu (even for laptops this exists)

11:37

so it was forced to use the high perf GPU instead

11:38

even then it might™️ still be usable offscreen

11:38

depending if the driver allows this

Ok, in technical terms, what I mean is, some CPUs' Processor Graphics don't have a driver for you to install and instead uses the system based Graphics driver to display

Cellegen

Ok, in technical terms, what I mean is, some CPUs' Processor Graphics don't have a driver for you to install and instead uses the system based Graphics driver to display

https://www.intel.com/content/www/us/en/support/articles/000058958/graphics.html

Warning: Installing This Graphics Driver From Intel May Overwrite...

Describes why a computer manufacturer installation warning appears before updating a graphics driver via the Intel® Driver & Support Assistant (IDSA).

So yes, you cannot use drivers on some CPUs like mobile CPUs cuz it only uses the system graphics

Only if the igpu is really really old

11:39

and that's moot since a driver did exist for it at some point

its called OEM and u can disable that/ignore that

Ewan

Only if the igpu is really really old

Some 2010 computers / laptops are still around (edited)

It would have to be even older

even they simply got the drivers from intel and simply customized them a bit

11:40

bcs what do i know.. they maybe added a wifi controler or stuff like that

11:41

that would break else

11:41

back in the old days the GPUs also often were not part of the CPU

11:41

but of the motherboard

I think who knows how an old 2000's computer ran would know this info

Why are you sassing

anyway anything before 2015 is dead for me

11:42

xd

11:42

oh we in 2023 already

11:42

then lets up it to 2018

11:43

5 years support

11:43

chairn trigger warning ^

Jupstar ✪

anyway anything before 2015 is dead for me

I think for today's standards, thats still a pretty good date

computer industry evolves so fast, everything gets bloated, nobody understands anything anymore

11:44

AMD releases integrated gpus and drop support after only like 2 years lmao

Jupstar ✪

anyway anything before 2015 is dead for me

I'll keep reporting problems on my 2009 laptop though

2000 what?

11:52

the funny thing about computer industry is that it's completely different than any other industry it evolves insanely fast almost no backward compability open source (free premium quality luxus stuff, wtf :D) relatively progressive community (complete opposite of most "normal" ppl)

11:52

i mean in just 20-30 years we got so far xd

Robyt3

maybe not outside if that causes issues

I didn't managed to do it outside the file{manager,explorer} so I did it inside

. Also I made it so the View resizes whether map preview is active

that's why u have to stay on the train

11:53

friss oder stirb

cuz we care of this tech, people make insane money off of it and we found ways to exploit many things in the world with computers alone

Mr.Gh0s7

I didn't managed to do it outside the file{manager,explorer} so I did it inside

. Also I made it so the View resizes whether map preview is active

I should also mention I still haven't managed to successfully clip the contents of some maps. idk what breaks

Mr.Gh0s7

I didn't managed to do it outside the file{manager,explorer} so I did it inside

. Also I made it so the View resizes whether map preview is active

Can you snap the camera to the checkpoint 1 entity, so that wherever that cp1 entity is, it'll default the camera's position to that?

Cellegen

Can you snap the camera to the checkpoint 1 entity, so that wherever that cp1 entity is, it'll default the camera's position to that?

I haven't reached that part but yeah I will do that ;)

Mr.Gh0s7

I didn't managed to do it outside the file{manager,explorer} so I did it inside

. Also I made it so the View resizes whether map preview is active

nice desktop envirement ⭐

justatest

1

Teero

nice desktop envirement ⭐

it's dwm with gaps and picom with round corners and stuff

ah nice

11:56

what distro?

artix

Mr.Gh0s7

I didn't managed to do it outside the file{manager,explorer} so I did it inside

. Also I made it so the View resizes whether map preview is active

does it now use a viewport, or did u solve it differently?

Mr.Gh0s7

artix

what do you use for the bar?

Teero

what do you use for the bar?

dwm has a bar and I use that

hmm okay

Jupstar ✪

that's why u have to stay on the train

I'm extremely skeptical of new developments in the software industry. It's layers upon layers of abstractions which are hiding all the interesting interactions and make debugging and end-to-end testing nigh impossible. It works for megacorps, but for smaller companies I think it can actually be a competitive advantage not to buy into all the hype.

Jupstar ✪

does it now use a viewport, or did u solve it differently?

It still as it is I didn't manage to use the background menu's inner workings pepeH

pepeH

deen

I'm extremely skeptical of new developments in the software industry. It's layers upon layers of abstractions which are hiding all the interesting interactions and make debugging and end-to-end testing nigh impossible. It works for megacorps, but for smaller companies I think it can actually be a competitive advantage not to buy into all the hype.

you are not wrong. but the mega corps give the direction, that's why i wrote "friss oder stirb" that's their spirit. I also agree that it can get very annoying that abstractions destroy everything, but not every abstraction comes because of making stuff easier, but are also build for backward compability. If you'd redesign an OS nowadays u probably have to add many layers to it, just to emulate existing stuff

12:00

i'd generally say modern APIs are often bloated bcs they want to support to many varrying hardware controllers, joysticks, what do i know what exists and users having one of the unsupported things get directly mad xD

Ewan

Click to see attachment 🖼️

you probably want to open the file and report the error of that instead

12:04

checking access rights is incredibly error-prone and not actually guaranteed to work

that’s not the only thing I’m checking lol (edited)

heinrich ur name is so hard to make a funny nickname of

12:05

but when python huffman?

hmmmm :/

this summer?

theoretically I'd love to have a release of uniffi

12:05

but I guess I can't wait for that

any ETA?

so maybe(!) I can look into it this evening

how fast is their dev cycle?

no ETA

release sounds good

so let me try without

12:06

sorry for the wait already

np

12:06

i told u im ok w wait

(in the evening, maybe(!))

i rather wait and have it clean than getting deprecated

12:07

you can also code libtw2 snapshot support while waiting for release :p

you mean wireshark dissector snapshot support? (edited)

12:17

is that still interesting to you?

@Jupstar ✪ anything before 2020

12:22

justatest

Ryozuki

@Jupstar ✪ anything before 2020

rust stable cammostripes

cammostripes

gigachad

12:24

my case is pre 2020, old hardware

12:24

gotta replace

xDDDDD

robsterbyte, u have to change server config and restart

12:37

friendly reminder

@heinrich5991 probably always will be even after i finished all my protocol implementations and 0.7 support

12:54

it would be a nice to have for traffic analsis in general but yea also would help a lot to create test cases for my implentations but i can also manage without

12:55

i implemented snap items in ruby already so yeah .. but would be nice anyways

when are fast emotes back

[ddnet/ddnet] Pull request opened: #6721 Improve selection popups and dropdown menus, add dropdown menus to select screen and active controller

Screenshots:

Controls settings:
- Before:

After:

Graphics settings (collapsed):

Before:

!graphics_collapsed old

After:

![graphics_collapsed new](https://github.com/ddnet/dd...

just edit your client to render emote event instead of sending emote msg

bencie

when are fast emotes back

yes pls

13:31

i die so hard

isnt the change already in

then it feels fast eventho u dont send anything xd

yes

not all servers restarted yet

still need to be recompiled

13:31

ah, did you already recompile?

fast fast

13:31

pepeFASTJAM

u accidentially use evil emote, cant even correct for 1 seond all ppl on server completely mad on you

<ChillerDragon2> yea ik i miss emote 50% correction so important xd

13:33

<ChillerDragon2> nice emote ryo xd

13:34

<Jupstar> oh, why can u see it

13:34

<Jupstar> why are messages grouped so weirdly

13:35

<Jupstar> confusing af xD

13:35

<ChillerDragon2> because i did hotreload while deving

13:35

<ChillerDragon2> to see ryo emote xd

13:35

<Jupstar> lel

13:35

<ChillerDragon2> i have a 6 letter alias to deploy to prod

13:36

<Jupstar> does hotreload work now?

13:37

<Jupstar> did u document that command?

@ChillerDragon why dont you send the webhook message properly btw, by using the username field for the actual username?...

It's not a discord webhook. It's an irc bot actually

ah yea, true

<ChillerDragon> ssh chiller@zillyhuhn.com -t "cd /var/www/discord-irc/ && source ~/.nvm/nvm.sh && git pull && npm run build"

leak

<ChillerDragon> @Jupstar no hot reload just effortless deploy to production for me

13:55

<ChillerDragon> fakin newlines are so hard ._.

what u doing

14:00

wdym by fake

14:00

new lines

14:00

p { white-space: pre; }

14:00

or pre-line

ye ye ik pre

14:01

but somehow my fake irc multi line causes newlines to be duped

14:01

its not really a css issue its more a chiler brain issue

[ddnet/ddnet] Pull request opened: #6722 Add highlight color when hovering DDNet/KoG type/country filters

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/ddnet/#using-addres...

Robyte full time ddnet employee

14:36

Image attachment

he's separating his work more cleanly, makes for more commits

14:49

but he's also been active quiiiite a lot

better delete 1.3 million whatever

Image attachment

14:52

@Ryozuki did your contribution mapping finally pay out

nope

14:54

github is dumb

14:54

so im missing 100 commits

[ddnet:master] 3 new commits

60f56f4 Support arbitrary number of type/country filter buttons - Robyt3 bc70726 Add highlight color when hovering DDNet/KoG type/country filters - Robyt3 7b93e41 Merge #6722 - bors[bot]

15:16

[ddnet/ddnet] Pull request closed: #6722 Add highlight color when hovering DDNet/KoG type/country filters

skill issue

15:16

github gaming

Jupstar ✪

better delete 1.3 million whatever

I wonder how that happened

let's guess

15:31

libraries

prolly

Jupstar ✪

better delete 1.3 million whatever

i believe that's when someone added boost to the repo

wat, why did they remove it ;~;

boost has bad compile times, I think

17:14

factorio removed it due to that

if thats an argument you should never use rust xD

1

pepeH

2

<ChillerDragon> was about to say that

Jupstar ✪

wat, why did they remove it ;~;

because you don't add boost to a git repo, you just use it and add it to the dependency list

where is the dependency then :^)

https://discord.com/channels/252358080522747904/293493549758939136/989718018005299260

how do you create these good looking message links?

17:36

https://discord.com/channels/252358080522747904/293493549758939136/1116419017549283430

17:36

ah

17:36

discord added a parser for them

rust programs compile faster than boost

17:59

source: gentoo user

what, did mozilla make gentoo as well?

sup

Ryozuki

rust programs compile faster than boost

the correct answer would be: Buy hardware from 2020+

Jupstar ✪

the correct answer would be: Buy hardware from 2020+

i disagree

boomer

18:23

welcome to stone age

Jupstar ✪

if thats an argument you should never use rust xD

I heard rust build times are faster than C++ with heavy template usage

18:44

due to SFINAE

mh i can't say. Compile times are obviously already higher bcs of the static linking. Generally coding with rust tools (the analyzer etc.) feels slower than using the same with clangd so either the tool is badly written or it has a harder time to reparse certain code snips

Jupstar ✪

mh i can't say. Compile times are obviously already higher bcs of the static linking. Generally coding with rust tools (the analyzer etc.) feels slower than using the same with clangd so either the tool is badly written or it has a harder time to reparse certain code snips

yeah clangd is surprisingly fast

18:55

i wonder how they differ

18:55

i know rust analyzer is like another compiler frontend

18:55

idk

anyway

18:56

just buy better hardware is good solution

ye

i got 16 cores for 250€

m1 is surprisingly fast

nobody can tell me thats too much money xd

nearly on par to 5800x

18:57

or on par

18:57

idk

i wish we'd have m2 with proper software

idk the cost ratio to a 5800x tho

sad af

18:57

lmao

yeah m2 is fail right

18:58

i want m1 on a linux laptop

18:58

with non propietary bloat

18:58

a completly open source processor

i am interested so much in m2 bcs it has almost desktop class GPU and shares the memory with CPU

but i heard it gets too hot

so u can zero copy images etc into GPU memory

Ryozuki

but i heard it gets too hot

i'd give it water cooling

18:59

if it wouldnt be apple

xd

18:59

but its meant for laptops

18:59

no?

soon they release an imac

18:59

with m2

Jupstar ✪

nobody can tell me thats too much money xd

all the people who have to eat for the rest of the month with 50€ can

Chairn

all the people who have to eat for the rest of the month with 50€ can

yayayaya

Ryozuki

i know rust analyzer is like another compiler frontend

llvm no?

no

19:01

rust analyzer emulates rustc

19:01

to find errors

19:01

A Rust compiler front-end for IDEs

19:02

https://rust-analyzer.github.io/

rust-analyzer: building a better Rust IDE

Chairn

llvm no?

llvm is the backend

19:02

clang is frontend

19:02

clangd is probs rly interwined with clang

19:02

so it works fast

19:02

idk rly the ins and outs

19:04

@Jupstar ✪ making a rly good LSP is haaaard

19:04

rust analyzer has to identify broken rust code

19:04

and show what it can

19:04

autocomplete

19:04

u have to autocomplete even when u have bad code

19:04

they do some blog posts

it also often doesnt work for me inside code inside macros :/

https://rust-analyzer.github.io/blog/2020/09/16/challeging-LR-parsing.html

Challenging LR Parsing

This post is a direct response to Which Parsing Approach?.If you haven’t read that article, do it now — it is the best short survey o...

19:04

https://rust-analyzer.github.io/blog/2020/10/24/introducing-ungrammar.html

Introducing Ungrammar

rust-analyzer is a new

19:05

https://rust-analyzer.github.io/blog/2021/11/21/ides-and-macros.html

IDEs and Macros

In this article, we’ll discuss challenges that language servers face when supporting macros.This is interesting, because for rust-analyzer, macros are ...

Jupstar ✪

it also often doesnt work for me inside code inside macros :/

ye macros are hard

19:05

Image attachment

19:05

@Jupstar ✪ last post is about macros

19:05

and why they hard

is it really useful to keep cargo.lock on a project or you add it to gitignore ?

if its a binary keep it

19:07

if its a library

why that

keep it if u want to have consistent CI

reproducable builds

otherwise not

but it also sucks when u have merge conflicts

19:07

at least for npm package locks i had it few times, then u delete and recreate it xd

ReiTW

why that

this is for reproducible builds indeed, rust follows semver

19:08

so

19:08

if when u build the binary u want to use the frozen lockfile u use --frozen

19:08

otherwise rust may find a minor newer version on a dep

19:08

and use it

19:08

if a package follows semver this is fine

ah

ReiTW

is it really useful to keep cargo.lock on a project or you add it to gitignore ?

my opinion: always add a lockfile

19:08

it gives you reproducible builds which is very nice

i recommend https://github.com/killercup/cargo-edit too

GitHub - killercup/cargo-edit: A utility for managing cargo depende...

A utility for managing cargo dependencies from the command line. - GitHub - killercup/cargo-edit: A utility for managing cargo dependencies from the command line.

19:09

cargo upgrade

19:09

will find new deps

you can cargo update if you want to update your dependencies. other people using your library will ignore your lockfile

ok ok thank you both for explaining (+ jupstar too !) (edited)

i also have cargo install-update to manage deps

19:10

@heinrich5991 update ur deps within the semver

19:10

upgrade modifies the toml (the extension i sent)

ye

it also allows managing ur crate version

19:10

xd

19:10

cargo set-version --bump major

19:11

i also use cargo install-update to manage binaries installed with rust

is there any good article/book to learn macro-magic?

and cargo cache

gerdoe

is there any good article/book to learn macro-magic?

this is a thing u need to try urself too to learn

19:12

apart from reading

19:12

syn and quote ur friends

Ryozuki

this is a thing u need to try urself too to learn

yeah i know that i need practice

you use syn to receive the ast tokens

19:12

and quote to easily create the generated code

i also had a task that i wanted to do with macro derives but stuck on first steps

gerdoe

i also had a task that i wanted to do with macro derives but stuck on first steps

copy some existing crate doing it I guess

https://doc.rust-lang.org/book/ch19-06-macros.html

Macros - The Rust Programming Language

19:14

@gerdoe i made this useless thing back in time https://github.com/edg-l/formy

GitHub - edg-l/formy: A rust derive macro to generate HTML forms fr...

A rust derive macro to generate HTML forms from structs. - GitHub - edg-l/formy: A rust derive macro to generate HTML forms from structs.

I think I've never had to write a non-declarative macro before

heinrich5991

copy some existing crate doing it I guess

oh i tried to learn something from reading alkahest macro derives but its too complicated

19:14

gave up instantly xd

heinrich5991

I think I've never had to write a non-declarative macro before

u mean proc macros?

19:15

@gerdoe check this out https://github.com/edg-l/formy/blob/master/formy_derive/src/lib.rs

yeah im lurking around rn

19:16

thanks, cute lib xd

its easier if you have dealt with AST before

19:16

abstract syntax trees

like never

fn get_meta_list(nested_meta: &syn::MetaList) -> Result<Vec<(&syn::Path, &syn::Lit)>, TokenStream> {

19:16

for example

messed around with cpp templates and c macroproc

syn is a crate that has most of rust ast representations

19:16

here you get a meta list

19:17

if i remember correctly its #[stuffhere]

Ryozuki

u mean proc macros?

ye

this functio receives the meta list

19:17

and returns a list of paths and literals

proc macros are cool xd u can write a completly new language in it

paths are type symbols for example

19:17

proc macros are epic

19:17

@Jupstar ✪ proc macros are even better when the diagnostic RFC lands

19:18

it allows macros to report errors cleanly

so i get metadata about struct/struct field and use it to do stuff right

in that function yeah

19:18

fn get_meta_list(nested_meta: &syn::MetaList) -> Result<Vec<(&syn::Path, &syn::Lit)>, TokenStream> {
    let mut list = vec![];
    for v in &nested_meta.nested {
        match v {
            syn::NestedMeta::Meta(m) => {
                if let syn::Meta::NameValue(value) = &m {
                    list.push((&value.path, &value.lit));
                } else {
                    return Err(
                        quote_spanned! {m.span()=> compile_error!("Must be a named value.")}.into(),
                    );
                }
            }
            x => {
                return Err(
                    quote_spanned! {x.span()=> compile_error!("Invalid meta type.")}.into(),
                );
            }
        }
    }

    Ok(list)
}

19:18

this is my old rust code tho

19:18

not perfect

19:18

quote! allows you to construct TokenStreams

19:18

using rust syntax directly

19:19

its epic

19:19

otherwise u would need to do ugly stuff

19:19

build it manually u know

19:19

quote_spanned is like quote! but adds spans where the macro is expanded

19:19

so u see i put a compile_error!

19:19

it will show a compiler error there

19:19

if u use a wrong meta attribute

19:20

(thanks to the spans)

huhuhu

here u can do anything

19:20

u can do a http request

19:21

and generate a struct from a json u received

okey i'll try to do some kind of serialization to learn more

19:21

thanks a lot ryozoozki

use dbg!

19:21

everywhere

kekw

did u know about dbg?

nope actually

https://edgarluque.com/blog/rust-dbg-macro/

The Rust dbg! macro

A short article about a quite unknown but useful macro in Rust.

santatrollet

1

19:21

kek

19:24

@gerdoe its hilarious cuz im doing heavy proc macro stuff at work rn

19:24

and tbh idk how it will look

19:24

we literally call libclang to parse c++ stuff and gen stuff

19:24

cxx didnt look good enough

cxx -> c**

19:25

cжж

justatest

sounds complicated

19:25

does it mean you parse c++ to make it rusty or something

19:26

don't get it

you're doing c++ rust interop at work, @Ryozuki?

y

heh, interesting

im sure c++ purists would die

19:27

anyway

19:29

there is a reason why people make bindings to c and not cpp

19:29

xd

19:29

but well, using libclang, we analyze the classes etc

19:30

we store them as [x;u8]

19:30

and get the align and size

19:30

with libclang

19:30

and more weird stuff

19:30

my coworker is more c++ savy than me

u still didnt rewrite llvm in rust?

no

mhh

i would

19:30

llvm has millions of lines

19:30

if im paid i would

19:30

it for sure would make me have a 2y job

19:30

xdd

probably longer xd

xd

well depends on how useful most of the code is

19:31

maybe u make a clean version

Ryozuki

there is a reason why people make bindings to c and not cpp

cpp parsing is very complicated isn't it tho

19:31

cpp has so much in it

yeah but that doesnt matter

19:31

cuz we use libclang

name me one github project (binary) that is older than 5 years and isnt bloated

@gerdoe u dont make bindings to c instead of cpp cuz its hard to parse

19:32

its cuz c is simpler, has less edge cases, and idk seems more "stable" or smth

c has stable abi yup

idk if thats right

19:32

C does not have a standard ABI.Dec 20, 2010

19:33

> >

C defines no ABI. In fact, it bends over backwards to avoid defining an ABI. Those people, who like me, who have spent most of their programming lives programming in C on 16/32/64 bit architectures with 8 bit bytes, 2's complement arithmetic and flat address spaces, will usually be quite surprised on reading the convoluted language of the current C standard.

the same so link xd

19:34

then question remains open kekw

<ChillerDragon> @Jupstar ✪ totally worth it to spend the whole day polish code snippets xd

19:51

<ChillerDragon> https://zillyhuhn.com/cs/.1686253851.png

Embedded image

19:51

<ChillerDragon> kinda pog that this is single line irc under the hood

19:52

<ChillerDragon> oh no robi wide snippet added the sideway scroll again xd

19:52

<ChillerDragon> https://zillyhuhn.com/cs/.1686253932.png

Embedded image

Nice

chiller

20:22

not to say anything but

20:22

the snippet is wrong

20:22

looks like u cant handle <>&

Html moment

Oh no

Use customized input tag

20:30

There u don't need these html hacks

Sounds good but I sleepy now

chillerdragon: I tried doing the python module

Ryozuki

we literally call libclang to parse c++ stuff and gen stuff

Huh, cxx really wasn't enough?

but I wasn't able to

Learath2

Huh, cxx really wasn't enough?

idk its a bit more hassle

20:33

unique ptrs everywhere u cant pass by value

20:33

with libclang we can cuz we know about the data layouts etc

20:33

also we can find if a method is inline, constexpr, etc and wrap it or etc

20:33

this is all inside a proc macro xd

I thought the primitive types were mapped so you can pass by value

i meant classes

20:37

i think u kinda can do in cxx if u mix cbindgen or smth

20:37

or impl the trait

20:38

i wonder how cxx works around inline methods

hm, can you even really pass classes by value to another language? Classes are too large to really pass by value down at the ABI level

it seems to work

20:38

on rust we just store the bytes as a array of u8

20:38

we get the align from clang

and apply that align to the struct using some black magic?

repr

20:39

repr(C, align(x)))

but when it comes down to calling a c++ function on it, don't you end up having to use a pointer?

yeah

20:40

this as the first param

20:40

tbh i just kinda went with my coworker wanted to do xD

20:41

walking landmines

anyway, becareful, it's really easy to invoke nasal demons when working at C++ in this level

ikr

20:41

he knows too

20:41

if this works

20:41

kinda deserves a blog post

chillerdragon: couldn't figure out how to run custom code (build the rust project) using the python build stuff

20:42

so I didn't manage to do it this evening

Learath2

anyway, becareful, it's really easy to invoke nasal demons when working at C++ in this level

i just wish llvm was made in rust

20:43

but we cant use c++

20:43

because we need to use rust libraries

20:43

and doing this the other way around

20:43

is more hell

20:43

kek

Ryozuki

i just wish llvm was made in rust

make it in rust

So wait, you said passing classes by value, for member functions that's not the case, they use a pointer at the abi level. For normal functions I thought that entailed a copy or move to the stack

I think larger structs are passed by pointer

Learath2

So wait, you said passing classes by value, for member functions that's not the case, they use a pointer at the abi level. For normal functions I thought that entailed a copy or move to the stack

i mean

20:45

imagine u have a class and constructor

20:45

we call it

20:45

the instance of the class

20:45

we store it by value on rust side

20:45

for method calls it seems to accept a pointer on *this yeah

20:45

atleast on linux and macos

20:45

xd

20:46

we didnt go to windows hell

Ah, I think I understand what you mean, and libclang darkmagic gives you the alignment and size you need to allocate on the rust side for the new object

also macos seems to put __ before symbols or smth

20:46

weird stuff

Learath2

Ah, I think I understand what you mean, and libclang darkmagic gives you the alignment and size you need to allocate on the rust side for the new object

ye

20:47

let me copy some stuff

I guess cxx instead let c++ do the allocations

ye

heinrich5991

chillerdragon: couldn't figure out how to run custom code (build the rust project) using the python build stuff

anyone experience with python packaging? ^^

https://pastebin.com/FHXqxsP2

pub mod ffi { #[repr(C, align(8))] pub struct DialectRegistry...

Pastebin.com is the number one paste tool since 2002. Pastebin is a website where you can store text online for a set period of time.

20:48

@Learath2

20:48

cargo expand

20:48

of the proc macro

20:48

well this is a bit older version

20:49

we now wrap automatically inline functions

20:49

and something with placement new

20:49

i forgot

20:50

extern "C" void wrap___ZN4mlir19registerAllDialectsERNS_11MLIRContextE(mlir::MLIRContext & context) {
    return registerAllDialects(context);
}

extern "C" void wrap___ZNK4mlir8Location4dumpEv(mlir::Location self) {
    return mlir::Location::dump(self);
}

extern "C" void wrap___ZN4mlir7BuilderC1EPNS_11MLIRContextE(mlir::MLIRContext * context) {
    new(context) mlir::Builder(context);
    return;
}

the last method

20:50

is a inline constructor

20:50

builder

Excellent mangling btw

xDD

20:51

anyway its interesting for sure

A decade looking at this language and it still takes me so long to decode them

which is why i love this current job

20:51

even if i could have more paying

20:51

but i doubt i can do this crazy stuff elsewhere easily

Learath2

A decade looking at this language and it still takes me so long to decode them

xd

20:52

@Learath2 did u look at rust mangling

20:52

u get 100+ names

Haven't yet, sounds excellent

Image attachment

20:53

from the blog post from yesterday

20:53

its not all the name

20:53

Image attachment

Do you know which language has amazing mangling?

no

20:53

C

20:53

none

C ofc ❤️

xdd

Learath2

Do you know which language has amazing mangling?

Does c literally guarantee no mangling? I mean if a compiler would generate it, would it leave c standard xd

no, it wouldn't

21:03

msvc mangles the names a litle

Msvc is probably not the best example

21:04

But good to know

C doesn't care

theorically rust either

21:26

but the only impl does it

21:26

kek

[ddnet/ddnet-scripts] Pull request opened: #45 Add better translation support for "Settings and Commands" page on the wiki

https://wiki.ddnet.org/wiki/Settings_and_Commands Assigns each string a permanent translation id

21:38

[ddnet/ddnet-scripts] Pull request closed: #45 Add better translation support for "Settings and Commands" page on the wiki

21:38

[ddnet-scripts:master] 2 new commits

9c52ab0 Add better translation support for "Settings and Commands" page on the wiki - Zwelf cce80db Merge pull request #45 from Zwelf/pr-wiki-settings-translation - def-

huh russian cpp community shared deen's article about haskell

21:47

unexpected

[ddnet/ddnet] Issue opened: #6723 Vote-kick cooldown can be undefined

This can happen with the following server configuration:

sv_solo_server 1
sv_vote_kick_delay 10

Steps to reproduce 1. Vote a map 2. Let the vote fail 3. Attempt to kick a player !image Expected behavior It should instead say "You can kick only your team member" because you can't kick others on a solo server.

Exported 1,049 message(s)