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 2022-04-22 00:00:00Z and 2022-04-23 00:00:00Z

Learath2

Ngl it might be a good idea to do ui with sth like react native

I have no idea how react native works.

03:34

The frontend guy messed up his fetch and show pattern, and everything is queried on the spot and can't retain after a window is closed.

03:34

justatest

03:34

So I thought it might be easier for him if there's a straight up web framework to use

03:35

Since we had TypeScript in unity already. I just dragged mithril.js in and wrote some fake DOM that creates, reorder and delete unity objects.

lynn

@⭐♥ST-Chara♥⭐ hi

Yes (edited)

https://seri.tools/blog/announcing-rust9x/

Rust9x: Compile Rust code for Windows 95, NT and above

06:20

lol

06:21

oldschool cool

i hate

08:12

c++ dep management

08:12

cmake sux

08:13

30 mins trying to find libsodium

08:13

with rust i just do cargo add sodium

08:13

feelsbadman

08:13

trying to port xpanic to ddnet-skeleton

08:13

so its modern

08:14

btw current xpanic stores password in clear text in files

08:14

xd

I could swear it used to work to connect to servers using ./DDNet "password foo" but if i understand the code correctly that does not work anymore since 4 years

08:19

super weird

08:19

https://github.com/ddnet/ddnet/blob/5c98d9227fec63a7617704a35858bb51b695b5e1/src/engine/client/client.cpp#L4185

ddnet/client.cpp at 5c98d9227fec63a7617704a35858bb51b695b5e1 · ddne...

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/client.cpp at 5c98d9227fec63a7617704a35858bb51b695b5e1 · ddnet/ddnet

08:20

So headless clients wont be able to join passworded servers. But i have done that already -.-

Ryozuki

btw current xpanic stores password in clear text in files

it warns about that very loudly before you register. ddnet saves are also plaintext codes, just in a sql db

08:39

ChillerDragon: You have to connect a second time, then it gets sent

well ill try to do it correctly

08:44

even if the password is still sent cleartext over udp xd

08:44

i figured the cmake thing finally

08:44

btw the database interface is p nice

08:44

i have sqlite and mysql support now xd

https://github.com/geekan/HowToLiveLonger

GitHub - geekan/HowToLiveLonger: 程序员延寿指南 | A programmer's guide to ...

程序员延寿指南 | A programmer's guide to live longer. Contribute to geekan/HowToLiveLonger development by creating an account on GitHub.

09:15

How to live longer. Step 1: Learn chinese

Damn, I'm dying early then

looks like a ACM meme more than anything

09:17

oh shit

Image attachment

09:17

looks like you have to sleep exactly 7 hours.

I frequently sleep 9 or more hours, fuck

one of the tips are to eat less egg yolk

and eat more nuts

I love egg yolks :/ I seem to be doing everything wrong

I'm actually wondering

09:22

Is there a way to redirect sunlight into my workplace but not onto my monitor during the whole day

09:22

kek

I can imagine an elaborate setup with mirrors

Sounds like a product I'd like to have

09:23

the tricky part is that the sun moves

i think the data may be biased, for example, older ppl tend to sleep less

09:23

but older ppl may die too

09:23

xd

older ppl die faster as well

and the cause of mortality in the charts is "all-cause"

09:24

which means ppl dieying of old age

based

iirc if u sleep 8 hours ur good

that's what i've heard

i dont think 7-9 matters much

09:25

ppl kill themselves

09:25

e.g drinking

09:25

and nobody cares

09:25

xd

dID yOu JuSt SaY dRiNkInG?

Susge

09:26

PES3_Sus

i'm offended

why

and I don't drink

i do sometimes

09:26

xd

Ryozuki

i think the data may be biased, for example, older ppl tend to sleep less

And chronic depression sufferers sleep more and they also die more, I hope they accounted for those, otherwise worthless research

well then I do sometimes as well

09:26

usually not willingly tho

@Learath2 did u know most research ends up being useless

09:26

also

09:26

academic is plagued with filler research

09:26

just to get numbers

09:26

the more cited u are

09:26

the more "good"

09:26

but thats plain wrong

09:27

u need to find a needle in a haystack to find good research

the page is deliberately being funny than accurate i think (edited)

also that u have to pay to read a paper is plain wrong

09:27

humanity deserves to die

since ACM literally means that ICPC thing which a lot of CS students does.

09:28

do*

09:28

I think my english got worse and worse ever since I came back to china

09:28

kek

09:29

luckily heinrich isn't here tho

xd

TsFreddie

luckily heinrich isn't here tho

heinrich is back Celebrate

Celebrate

09:35

I mean he isn't online

celebrate

09:35

its 11:35

09:35

maybe he is working

09:35

without discord

09:35

monkalaugh

I should be working rn

same

09:35

wht timee is it there

5:35pm

TsFreddie

since ACM literally means that ICPC thing which a lot of CS students does.

This competitive programming thing is one of the most senseless attempts at "ranking" programmers. It's just one of those things that should never be used as a metric

I did it for like 5 years in middle school and high school

09:37

my college invited me but I didn't go

09:38

btw https://mithril.js.org/ is pretty lit

I mean it's fun surely, but it's not really useful. Should be done like a leasure sport

Learath2

This competitive programming thing is one of the most senseless attempts at "ranking" programmers. It's just one of those things that should never be used as a metric

competitive programming tests is just algorithm solving

09:39

imagine if the tests where about figuring a data race in a 200k line code base or smth

09:39

xd

Ryozuki

competitive programming tests is just algorithm solving

more like "problem" solving with a algorithm database baked into your head.

09:40

which you spent three years remembering.

ye

09:40

but there is undeniably smart ppl there

~~like me~~

TsFreddie

btw https://mithril.js.org/ is pretty lit

I'd seen this on lichess' stack but hadn't checked it out yet. If you could describe it in one sentence, what would that sentence be?

https://en.wikipedia.org/wiki/Gennady_Korotkevich

Gennady Korotkevich

Gennady Korotkevich (Belarusian: Генадзь Караткевіч, Hienadź Karatkievič, Russian: Геннадий Короткевич; born 25 September 1994) is a Belarusian competitive programmer who has won major international competitions since the age of 11, as well as numerous national competitions. His top accomplishments include six consecutive gold medals in the I...

09:40

this guy is good

09:41

"tourist"

09:41

his setup screams he is a mathematician tho

09:41

he uses windows

09:41

its actually a weird setup

09:42

Image attachment

09:42

xd

09:42

imagine having this blue screaming screen

Ryozuki

https://en.wikipedia.org/wiki/Gennady_Korotkevich

Has he made anything that's not just golfing code?

Learath2

I'd seen this on lichess' stack but hadn't checked it out yet. If you could describe it in one sentence, what would that sentence be?

small, simple and easy to get into without reading five pages of getting started and can't get it started.

Learath2

Has he made anything that's not just golfing code?

Gennady is the highest-rated programmer on Codeforces,[2] CodeChef,[3] Topcoder,[4] AtCoder [jp][5] and HackerRank.[6] In January 2022, he achieved an historic rating of 3979 on Codeforces, becoming the first to break the 3900 barrier.

09:42

xd

09:42

idk

09:43

Known for Programming prodigy; highly ranked sport programmer from an early age

09:43

lmao

mainly small tho

TsFreddie

small, simple and easy to get into without reading five pages of getting started and can't get it started.

Sounds nice, maybe I'll make something with it to see how it feels compared to vue

the true prodigy ppl imho is ppl like the one i sent a article yesterday

09:43

about sorting algos

09:43

those are the real smart ppl

I've used vue twice

09:44

I think mithril might be better. But I did setup my own webpack config

@deen but ./DDNet "connect localhost;connect localhost" does not connect twice. So for a headless client its non trivial

Ryozuki

those are the real smart ppl

Exactly, imo the amount of time it takes someone to produce work of that quality is not very important

I think vue takes care of boilerplates

Guild Wars 2 uses Mithril (yes, inside the game!)

Huh, someone already took my game ui idea and ran with it

ok

09:46

i'm doing it too

wats Mithril

never tried to get vue or react into unity yet. but mithril only needs a handful of dom implementation to work

ii only know it from final fantasy

09:46

greenthing

I just wrapped some unity code inside a js class with appendChild insertBefore and removeChild and it just works

09:47

to a certain degree

09:48

kek

@Learath2 https://devblogs.microsoft.com/oldnewthing/20220418-00/?p=106489

The x86 architecture is the weirdo, part 2

It does everything differently, because of course it does.

09:49

a blog from the microsoft guy

09:49

Raymond C

09:49

i think someone mentioned him one day

who

heinrich

09:49

https://discord.com/channels/252358080522747904/293493549758939136/927820019432325161

I posted a couple of his blogposts before aswell

@heinrich5991 hehe

Ryozuki

which means ppl dieying of old age

dieying

i don't see a problem

Ryozuki

@Learath2 https://devblogs.microsoft.com/oldnewthing/20220418-00/?p=106489

https://devblogs.microsoft.com/oldnewthing/20220228-00/?p=106296 the next one he links to is also nice

Zero-cost exceptions aren't actually zero cost

Shifting the cost to the exceptional path.

how can I make the client send a chat message as the player ? (edited)

CChat::Say

Learath2

CChat::Say

from menu settings if i want send Test message in global chat ((CChat *)pUserData)->Say(0, ("Test"))

14:00

but pUserData what is

14:00

void CChat::ConSay(IConsole::IResult *pResult, void **pUserData) { ((CChat *)pUserData)->Say(0, pResult->GetString(0)); } (edited)

I smell someone doing pattern match coding

Learath2

I smell someone doing pattern match coding

idk what u mean xd

I have a feeling you don't actually know C++ but you are just imitating the code we have by form

Learath2

I have a feeling you don't actually know C++ but you are just imitating the code we have by form

exactly xd

14:03

and i have to say it's not very easy haha

Anyway, from CMenus you want to go through the gameclient GameClient(), to get to CGameClient::m_Chat, on which you can just use CChat::Say. So something like GameClient()->m_Chat.Say(0, "Test"));

kio

and i have to say it's not very easy haha

You will have a much better time if you just take some time to learn C++

actually I'm already having fun haha but you are right. however I think I have understood many things thanks to your message above

haha

https://bellard.org, about prodigys (edited)

is this a guy? is this a collection of random projects?

14:17

no about so cant know

if that's one guy that's insane

ye

14:18

qemu alone

14:18

is insane

14:18

xd

Fabrice Bellard yes it is

14:18

Known for "FFmpeg, QEMU, and the Tiny C Compiler." Now that is talent

Fabrice Bellard is a French computer programmer known for writing FFmpeg, QEMU, and the Tiny C Compiler. He developed Bellard's formula for calculating single digits of pi. In 2012

14:18

ye

14:18

both 3 projects are insanely famous

14:19

i see that tiny c compiler mentioned everywhere

Insanely famous and massive in scale

i know another prodigy

14:19

https://stallman.org/

14:20

@Learath2

Thank god Bellard's 50yo and not some 17yo with all those accomplishments

14:20

Makes me feel way better

xd

14:21

lets be real

14:21

most here probs wont achieve even a single project like that

14:21

and thats fine

14:21

accept urself

14:21

BASED

14:21

altho i always feel like programming is relatively young field

14:21

there is shitload more devs now

14:21

and in 40 years they will all be seniors

14:22

who knows what will be done

Ryozuki

most here probs wont achieve even a single project like that

We all contributed to ddnet, the pinnacle of open source multiplayer gaming TOOBASED

TOOBASED

TOOBASED

14:22

i cant wait to be 50y old and preach about the good old rust

14:22

and how todays lang suck

14:23

greenthing

14:23

actually that cant wait

14:23

ii dont wanna be 50y old

14:23

monkalaugh

14:23

can*

14:23

fck i cant type

We are also too late to make simpler ideas like these work. Back when this guy started on ffmpeg there was like 2 video codecs on earth

true but im sure there is a lot to be done

Nowadays your best bet at making influential software is to foray into weird niche fields that require so much damn knowledge

14:25

Back in the day people used to make innovations like "mail but over the internet instead of over human carriers"

Learath2

I smell someone doing pattern match coding

Isn't all human creation pattern matching?

With experience the patterns just get better

gigachad

2

well i think there needs to be some understanding

deen

Isn't all human creation pattern matching?

With experience the patterns just get better

I guess at some degree it is, but what I was trying to point out there is that pattern matching without knowledge about the patterns is really prone to breaking

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4141622/

Superior pattern processing is the essence of the evolved human brain

Humans have long pondered the nature of their mind/brain and, particularly why its capacities for reasoning, communication and abstract thought are far superior to other species, including closely related anthropoids. This article considers superior pattern ...

14:27

pog

14:28

@Learath2 the new sorting algorithms

14:28

from that article

14:28

xd

i bet learath opens a restaurant in his life

i always thought a restaurant is a nice place to problem solve

14:30

specially with technology

14:30

xd

we always did brainstorming in restuarants with friends back in the school days

14:30

we even did a lan party once xD

xd

they were ok wiht it

as long as u bring lot of ppl who pay

14:31

they are fine with it usually (edited)

14:31

xd

Ryozuki

@Learath2 the new sorting algorithms

Yeah look at those, back in Shell's day he could propose things like "sort by swapping BUT with gaps"

14:31

Now you want to improve sorting you need to do regression learning to drop a single instruction and vector dark magic to utilize cache better

xd

14:32

the same thing happens with math physics etc

Yep, the fields are getting so complicated :/

well i think u have to specialize

14:32

u cant know everything

Unless you are into entrepreneurship you won't achieve shit in your lifetime

if we so advanced why not all games run at million fps

(I won't either, the "you" there was general)

i always thought that

14:33

i want to reach that age

14:33

xd

14:34

the point where u dont even need to thing about how much memory u use

Ryozuki

the point where u dont even need to thing about how much memory u use

We are here

lets build a Matrioshka brain

14:34

https://en.wikipedia.org/wiki/Matrioshka_brain

Matrioshka brain

A matrioshka brain is a hypothetical megastructure of immense computational capacity powered by a Dyson sphere. It was proposed in 1997 by Robert J. Bradbury (1956–2011). It is an example of a class-B stellar engine, employing the entire energy output of a star to drive computer systems. This concept derives its name from the nesting Russian ma...

Learath2

We are here

well im talking to levels like where u accidentally make a program that overflows the stack but u have to wait days till it actually does

14:36

BASED

The world if chromium developers cared about memory usage

Image attachment

xD

14:37

the computer needed to open 20 tabs in chrome

Image attachment

Learath2

The world if chromium developers cared about memory usage

does chromium use lot of mem cuz they leak a lot?

14:38

maybe its just the js engine uses shitload of ram

bcs every chrome tab is a full OS xd

idk the inner workins of that

14:38

xd

Ryozuki

idk the inner workins of that

No one does probably, if there is one thing browser devs care about less than memory usage, it's code line count

xd

14:40

ye isnt chromium larger than linux

most likely

with the drivers included

ouch really?

14:40

over 26mio lines

14:40

sounds horribl

14:40

e

14:41

soon we need 1gb L3 cache or just run smoething smooth

35mil in chromium, 28mil in the linux kernel

14:42

openrc has about 15k LoC, systemd has about 1.3mil

Google claims their internal repo is 2 billion lines of code btw

do they use a monorepo?

ez

yes

how comes that my cl_dummy_hammer is so fast after running 8 days straight? MonkaS

MonkaS

deen

Google claims their internal repo is 2 billion lines of code btw

oh god

ᶰ°Konͧsti

how comes that my cl_dummy_hammer is so fast after running 8 days straight? MonkaS

MonkaS

probs some weird floating point stuff

14:46

is the client running 8 days?

monorepo is really nice when you can just update a lib and all usages at once instead of needing backwards compat

Jupstar ✪

probs some weird floating point stuff

but when i turned off and on it worked normal

ᶰ°Konͧsti

but when i turned off and on it worked normal

ah maybe it has some start time or smth

14:46

for the internal interval

14:47

open an issue on github

14:47

so we dont forget

nah i dont think its worth an issue

14:47

nobody uses normal dummy hammer anyway

then we should still remove it or fix it

i were just afk on that map with dummy hammer on for 8 days

lol

Easy to reproduce then

Truly a bug only konsti could find

afk

4

afk

ᶰ°Konͧsti

i were just afk on that map with dummy hammer on for 8 days

could also be a visual bug like I posted in #bugs

every non finite time calculation should be changed to either use double if not other wise fixable, or fix otherwise xd

14:51

or no floating point even better xd

I don't think there is any time calculation for dummy hammer

some interval

it just increments on each OnSnapInput call

mh when turning off and on again fixes it, it kinda must be something related to weird math

I think this could be the cause:

                if(NewPredTick > m_PredTick[g_Config.m_ClDummy])
                {
                    m_PredTick[g_Config.m_ClDummy] = NewPredTick;
                    Repredict = true;

                    // send input
                    SendInput();
                }

Maybe the tick flows over or something?

possible too yeah

I could say we could wait 8 more days to test it again with debug mode on, but, 8 days is 8 days.

justatest

1

if the tick overflows it should actually be unrelated to the days

14:56

can even happen after 1 second

it's only an int, so max 2147483647, but 8 days would only be 34560000 ticks, so not nearly enough

isnt it server time tho

my guess would be that m_DummyFire flows over into negativ , the result of the calculation here: https://github.com/C0D3D3V/ddnet/blob/9dc2141ded307f013402e7d84c66bc793b45cc6c/src/game/client/gameclient.cpp#L422 would always be negativ, so every tick a hammer would be made

ddnet/gameclient.cpp at 9dc2141ded307f013402e7d84c66bc793b45cc6c · ...

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/gameclient.cpp at 9dc2141ded307f013402e7d84c66bc793b45cc6c · C0D3D3V/ddnet

Server time would be 497 days to overflow I guess

is there a benefit from using int to int64

c0d3d3v

my guess would be that m_DummyFire flows over into negativ , the result of the calculation here: https://github.com/C0D3D3V/ddnet/blob/9dc2141ded307f013402e7d84c66bc793b45cc6c/src/game/client/gameclient.cpp#L422 would always be negativ, so every tick a hammer would be made

yes

14:57

sounds reasonable

14:57

bcs it uses floats xd

Ryozuki

is there a benefit from using int to int64

saves u 4 bytes, but i'd say a modern tw could also upper some limits

14:58

server time should not overflow not even after 1 year

ye xd

14:59

@Jupstar ✪ actually

14:59

using int i dont think it guarantees its 32

14:59

it may be 64

it doesnt but our target platforms use 32bit

14:59

in the code we also rely on it sometimes

15:00

remember its teeworlds

15:00

it hates standard

xd

xD we could use mpz_int instead of int

these can get as big as you have memory, or just use rings for ticks and time related stuff

so int32 and stuff like that wasnt used

do 128bit floats exist?

yes

15:01

long double

15:01

but only linux i think

c0d3d3v

my guess would be that m_DummyFire flows over into negativ , the result of the calculation here: https://github.com/C0D3D3V/ddnet/blob/9dc2141ded307f013402e7d84c66bc793b45cc6c/src/game/client/gameclient.cpp#L422 would always be negativ, so every tick a hammer would be made

but that would also require 497 days to overflow I think

deen

but that would also require 497 days to overflow I think

but floats

15:01

floats hate u

m_DummyFire is an int

if((m_DummyFire / 12.5f) - (int)(m_DummyFire / 12.5f) > 0.01f)

15:02

but it gets converted

deen

but that would also require 497 days to overflow I think

mh true

Sure, could be this float calculation getting more imprecise (but not an overflow)

yeah ;D

15:03

we have bugs like these quite often

15:03

i also dont understand why float calculation here is even used

15:03

can just use integer math

15:04

int64 to be safe

Jupstar ✪

if((m_DummyFire / 12.5f) - (int)(m_DummyFire / 12.5f) > 0.01f)

I well just run some test code... to see if it overflows after 8 days ... one moment xD

its not about overflowing

15:04

the calculation will evaluate to true too often

15:04

and then the ticks get updated all the time or smth like that

15:05

or to false

I will check it ...

makes more sense

Jupstar ✪

i also dont understand why float calculation here is even used

I think because 12.5 felt good to me for flying, and that's not an integer

xd

15:07

many games in the past used floats to store current game time(added every time the game loop runs).. after like 20 minutes they are around 50ms off

jupstar is right xD

https://seri.tools/blog/announcing-rust9x/

Rust9x: Compile Rust code for Windows 95, NT and above

the condition is not true way before the int overfloats

15:18

it gets already imprecise at m_DummyFire = 13107213, then it hammers 3 times in a row, at some value it hammers constantly

15:20

10 times in a row at 67108869

15:20

50 at 104857646

15:21

100 at 104857696 1000000 at 105857596 (edited)

15:21

I used this code:

    int m_DummyFire = 0;
    int hammerd = 0;
    int debug = 0;
    while(m_DummyFire != -4000)
    {
        if((m_DummyFire / 12.5f) - (int)(m_DummyFire / 12.5f) > 0.01f)
        {            
            m_DummyFire++;
            hammerd = 0;
            continue;
        }
        if(hammerd == 1)
        {
            debug = 0;
        }
        if(hammerd == 10)
        {
            debug = 0;
        }
        if(hammerd == 50)
        {
            debug = 0;
        }
        if(hammerd == 100)
        {
            debug = 0;
        }
        m_DummyFire++;
        hammerd++;
    }

(edited)

15:25

so after 2.14 days it is imprecise

15:25

after 11 days it hammers 10 times in a row

15:27

by changing 12.5f to 12.5 and 0.01f to 0.01, it does not get inpercise before the int oveflows to -2147483647

just change it to ints if you want to fix the "problem"

you mean unsinged int

15:29

with int we will always have the wrong calculation after 497days

15:29

xD

15:29

who ever plays that long

after 497 days, something else probably had a bug first

we could just use m_DummyFire % 25 == 0 and reset m_DummyFire to 0 every hammer

15:33

I make the patch

nice

Why not just make m_DummyFire unsigned and do % 25?

15:37

Don't ever need to reset to 0 that way

But it will be inaccurate for one time after 497 days

1

Learath2

Why not just make m_DummyFire unsigned and do % 25?

Yes you are right... I will do that... I also wanted to do that first, but thought, why not reset it every time, we do not need the value anyway. We could also just use a byte to save memory xD lol

deen

But it will be inaccurate for one time after 497 days

If we use an exact power of two it won't be, let's check if % 32 feels good for dummy fly

heathen

c0d3d3v

Yes you are right... I will do that... I also wanted to do that first, but thought, why not reset it every time, we do not need the value anyway. We could also just use a byte to save memory xD lol

but a byte would be inaccurate more often

15:42

instead of once every 497day teeworlds session

(they were saying "by resetting it to 0 each time, you don't even need 32 bit, you can fit it into 8 bit")

Hm, I may have misinterpreted, but anyway in all reality both are performant enough that this discussion is just for fun

yes

I guess no reset to 0 every 25 snaps is very technically better assuming the cpu is completely inept and can't pipeline it

1

15:46

s/technically/slightly/

15:46

If we ever hope to run ddnet on an 8086 we should be careful of these

or use just floats std::stack<float> push 25 times and then clear them again

16:08

also nice solution use some unsorted array use a sorting algotihm that sorts this array in 25 steps do one step per tick and check if sorted

16:08

then reset

16:09

open 25 chrome tabs

16:09

xd

I like the chrome tabs idea. A friend of me usually has ~1k tabs in firefox open. The tabs get to sleep at some point and are no longer in memory, thats why it works (edited)

The standalone 16.0 rc1 build doesn't have the update button to the newest version in the main menu thonk

thonk

16:42

do we not update the client on beta builds?

probs not

beta and rc builds have no update functionality

16:58

most convenient way to use them is through Steam

16:58

If we wanted to have updates for them, we'd need two additional update streams through the website

17:04

I didn't think that's worth the additional complexity

Learath2

I have a feeling you don't actually know C++ but you are just imitating the code we have by form

isn't that how you're supposed to do with a code base you don't know?

Ryozuki

also that u have to pay to read a paper is plain wrong

use sci-hub

ik

ᶰ°Konͧsti

nobody uses normal dummy hammer anyway

i do, but i suck at dummy

i have to do that when someone join in your server and check if he is in your friend list

18:07

someone know?^

Chairn

isn't that how you're supposed to do with a code base you don't know?

You are talking about a higher level of pattern matching, if you can break down the patterns you see into fundamental language constructs there is no problem

kio

someone know?^

I think the client has no way of knowing who joined the server besides those " *** x joined the server" chat messages Someone pls tell me if I'm wrong xd

can someone reproduce this too? https://discord.com/channels/252358080522747904/757720336274948198/967069481279045633 with cl_predict on

So just use that string to extract the name of whoever joined

xD you could make a diff of the user list xD

18:18

but the join message should be easier

Cipy29

I think the client has no way of knowing who joined the server besides those " *** x joined the server" chat messages Someone pls tell me if I'm wrong xd

You can keep track of the clients in the snap

that is also what I mean by making a diff*

Ah yes ofc

someone has an Idea what settings causes this, because I'm mostly sure I have never seen this, but I tried it before to hold left click

18:28

and I always played with cl_predict and antiping

18:29

it is the same in latest master, the video is on the version from the commit I tracked down (https://github.com/ddnet/ddnet/commit/eb1bacb969b3edc121934991380d1d03718cd718) (edited)

Update prediction with latest server physics · ddnet/ddnet@eb1bacb

it is this added code:

    if(m_FrozenLastTick)
        FullAuto = true;

https://github.com/C0D3D3V/ddnet/blob/8c6a7ef38a89d123cf4b9c5e495e2c25971efe92/src/game/client/prediction/entities/character.cpp#L263-L264 serverside it was added here: https://github.com/ddnet/ddnet/commit/f19220f1bbbb647954d6544865e3c2a2e9953037 but why does noone else see this bug

Allow to hammer directly on unfreeze · ddnet/ddnet@f19220f

Previously, this was only doable with frame-perfect input if you were going to get frozen again the next tick. Now, you only have to hold your hammer to immediately hammer once you get unfrozen. T...

ddnet/character.cpp at 8c6a7ef38a89d123cf4b9c5e495e2c25971efe92 · C...

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/character.cpp at 8c6a7ef38a89d123cf4b9c5e495e2c25971efe92 · C0D3D3V/ddnet

18:46

ah it is broken if cl_m_Predict_DDRace is off

thats why I have never saw this, because then m_FrozenLastTick is not set to false (edited)

18:47

Why do we even need so much settings for predict and antiping

can we not get rid of them

Cipy29

I think the client has no way of knowing who joined the server besides those " *** x joined the server" chat messages Someone pls tell me if I'm wrong xd

I think there is a render in the chat to see when a line appears, and if you match it you look among the online ones put in an array and you see who is new, but I was hoping there was a way more ez

Cipy29

I think the client has no way of knowing who joined the server besides those " *** x joined the server" chat messages Someone pls tell me if I'm wrong xd

0.7 client does

19:20

not 0.6 one

gerdoe

0.7 client does

explain

kio

explain

explain what

how know who jin

19:23

join

You already know who's new from the *** x has entered and joined the game

19:27

Right?

19:28

Then you see if they're your friend

Cipy29

You already know who's new from the *** x has entered and joined the game

not all servers have exactly same join message

kio

join

afair 0.7 clients accept info about joined player as clientid

19:32

and then localize join message

I assumed he would be using it on ddnet servers

c0d3d3v

Why do we even need so much settings for predict and antiping

can we not get rid of them

@c0d3d3v yes, let's decrease the number of settings!

@Learath2 may I change the type of MACRO_CONFIG_INT config variables to std::atomic_int so we can actually access it from multiple threads? ^^

21:43

or do you have an idea on how to access stuff like g_Config.m_ConsoleLogLevel from multiple threads?

heinrich5991

@Learath2 may I change the type of MACRO_CONFIG_INT config variables to std::atomic_int so we can actually access it from multiple threads? ^^

An atomic really sounds like the least impacting of all the available options

22:30

The other options I can think of are far worse like locking

https://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html it compiles down to the same instructions even, just a source complexity

22:31

at least on x86 or if we used relaxed memory order

Don't think relaxed would be what we want here

it's basically the current behavior, just without the UB

22:33

for console_log_level, it'd probably be okay, at least I can't think of any weird stuff

Ah yep

heinrich5991

@Learath2 may I change the type of MACRO_CONFIG_INT config variables to std::atomic_int so we can actually access it from multiple threads? ^^

is there any risk of conflict? I don't think people change config that often, otherwise, just declare it volatile?

volatile is wrong there

22:36

volatile is not a thread synchronization primitive

oh this discussion again haha

22:37

@Learath2 i remember ur rant

Volatile is definitely wrong here :D

22:37

It won't even emit a fence

relaxed won't emit a fence either

https://wiki.sei.cmu.edu/confluence/display/c/CON02-C.+Do+not+use+volatile+as+a+synchronization+primitive

heinrich5991

relaxed won't emit a fence either

Hence I said I don't think relaxed is the best idea here

22:38

But it will definitely fix the UB

i do believe volatile is enough for this use case

22:40

it forces to reload the value from memory

22:40

let hardware coherency do its job (edited)

bad bad bad

It doesn't actually fix the UB as far as the standard is concerned

22:42

Volatile doesnt provide a synchronization point, so the two threads will be accessing memory out of sync

config variable change once every 36 of the month, not a big deal

@Chairn you get the same behavior with relaxed atomics, except it's without the UB

[ddnet/ddnet] Pull request opened: #5012 Rename `Act` → `Cur`, some "actual" → "current"

Seems like it got introduced by means of the false friend "actual" ≠ "aktuell" = "current".

Why would you want to invoke nasal demons when you can avoid them?

22:47

Even compiler people can't agree on what the hell volatile can guarantee, I doubt the standards people are sure either. Relaxed atomics on the other hand have very well defined semantics

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

Deprecating volatile - JF Bastien - CppCon 2019

22:48

if you have one hour of free time (edited)

[ddnet/ddnet] Pull request opened: #5013 Unify logging infrastructure between `IConsole` and `dbg_msg`

This makes the "black console window" less important on Windows (or anywhere else, for that matter), lets you see logs from other threads in the f1 console, and removes the distinction between IConsole::Print and dbg_msg.

anyone think it would be fixable to patch "Please balance teams!" for more than just race modes?

23:08

heinrich suggested MACRO_CONFIG_INT for it

no MACRO_CONFIG_INT pls

wait what should we use

I suggested making it a flag the server can send

i guess i assumed you meant sv_

23:08

but maybe you mean just in snap

just in snap, I meant

ye makes sense

23:09

i guess so

23:09

what netobj

[ddnet:master] 2 new commits

103b663 Rename Act → Cur, some "actual" → "current" - heinrich5991 b14c916 Merge #5012 - bors[bot]

23:09

[ddnet/ddnet] Pull request closed: #5012 Rename `Act` → `Cur`, some "actual" → "current"

Hey what do you guys think of this pr? I think deen wants some more opinions before merging https://github.com/ddnet/ddnet/pull/4997

we have some game info object

wow so nice but margins in a more square ratio look kinda small

Image attachment

23:19

like with feet and above

OT: maybe some of u can help saved a game and cant remember save name... where i can find it?

#questions and it's in ddnet-saves.txt

Yeah that could be bigger

ty

hard to work with existing space already. i suggest making them a lot smaller

23:21

and adding a label "Default eyes"

23:21

unclear maybe at first what it does

Hm I'm actually a little intrigued whether a better answer exists to this kind of synchronization problem where writes are much less frequent than reads

I mean on x86, it works perfectly fine with AcqRel semantics of atomics

23:23

@Learath2 https://github.com/ddnet/ddnet/pull/5013#pullrequestreview-950687001 asked some questions if you're interested

I can answer some tomorrow, I really need to sleep asap but I have atomics on my mind

https://www.ctan.org/pkg/minted best package to include some code in your latex

ngl it was so weird to come here and see "emit a fence" without context

23:35

i was like wtf is that

this a fence

Image attachment

1

23:38

now imagine your computer throwing this at your face because you didn't synchronize your threads (edited)

Exported 511 message(s)