DDraceNetwork - developer

DDraceNetwork

Development / developer

Development discussion. Logged to https://ddnet.tw/irclogs/ Connected with DDNet's IRC channel, Matrix room and GitHub repositories — IRC: #ddnet on Quakenet | Matrix: #ddnet-developer:matrix.org GitHub: https://github.com/ddnet

Between 2020-04-15 00:00:00Z and 2020-04-16 00:00:00Z

@Learath2 i think you or @Ryozuki broke vanilla whisper. on master you cant use /w and /whisper anymore. it just sends the msg as a normal chat message leading to No such command: w

well do you delete w as a command?

of course i dont

09:08

it even shows up

09:09

I am guessing this is because of @Ryozuki's disable whisper thing

09:09

he never tests stuff

tell me exactly how to reproduce it, including the versions

master, just try to /w

on what server

no matter

09:09

try /w target

09:10

or /w target msg

09:10

it just gets sent as chat msg, on ddrace server leading to no such command, on ctf leading to just a chat msg

"I am guessing this is because of @Ryozuki's disable whisper thing he never tests stuff" fuck you

no u

kid

well, did you test it?

it's my fault not his

yes i tested

ok, sorry then @Ryozuki

idk how to fix it tbh

whats the problem :D

not without unfixing the other bug

which commit caused it?

https://github.com/teeworlds/teeworlds/pull/2532

Fix small bug with commands by Learath2 · Pull Request #2532 · teew...

mh

When you put w then press space, you are typing a command, but the selected command becomes -1

but for custom commands like /stats <name> it works

09:20

ah, probably not

I'm pretty sure that only works because it goes through the old interface

probably it just gets executed as a chat command

09:20

yes

09:21

idk how to fix it tbh

Now I understand what you meant xd

Hmm I can change the other fix

@fokkonaut that should take care of it, it'd be nice if you could give it a test

@Ryozuki you never test stuff, so that means maybe you would release my map feelsamazingman

feelsamazingman

https://www.taiwannews.com.tw/en/news/3916690

China to ban online gaming, chatting with for... | Taiwan News

Say goodbye to your guildmates in China, as CCP extends its political censorship to online gaming world

12:02

China to ban online gaming, chatting with foreigners outside Great Firewall

Not a big surprise considering that non-Chinese social networks etc are also banned

haha rip this minecrap server with all the press for china

12:20

i wondered how long this loophole will stay open

12:21

if one is bored and wants to write a method that converts decimal to base93 for me that would be awesome :D

12:21

i hate numbers

12:22

https://paste.zillyhuhn.com/xA

12:22

this is what i have so far

12:24

ok that code was incomplete https://paste.zillyhuhn.com/P8

ubuntu > win

debian > ubuntu

discord > ChillerDragon

https://paste.zillyhuhn.com/1Y

12:41

everything works until 92 and then at 93 it breaks im sure there is just a -1 or a +1 somewhere in the pack and/or unpacker

12:41

cmn brains i know u know it :D

just like you would do with any base

13:43

first get rid of the sign

13:43

then % 93 to extract the last digit

13:43

then / 93 to get rid of the last digit

13:43

repeat

base93? who even uses base93 for anything?

some random blog post cited by wikipedia, apparently. also the only google result: https://kiwigis.blogspot.com/2013/09/base-93-integer-shortening-in-c.html

Base93 - Integer Shortening in C#

This post will describe a technique of shortening an integer by approximately 50%. For example, the approximate coordinate extent of cont...

13:55

@ChillerDragon why do you implement base93?

yay i did it! it was some +1 and some index instead of 0

14:05

i hate numbers

14:05

@heinrich5991 i want to compress integers for my network protocol and i use only printable ascii range

14:05

my network packets are simple ascii strings

have you considered using base 10, base 16 or base 64?

yea

all of these are standard

yea but base93 > base64

only on efficiency

14:07

not really on human-readability, which you seem to be after with your ascii-only protocol?

my initial issue suggested me to implement 64 but then i did a test base64 encode and got a string that is longer than 1 character

14:07

so i decided to build my own and max out on 93 printable chars

@fokkonaut /kill doesn't seem to be working on your server

yea i want it to be easier to debug also in wireshark

why is it easy to debug with base93?

also my whole protocol uses strings so i would get utf-8 errors if i go out of printable range

there are no standard tools to decode base 93 integers

well thats why i built mine

14:08

it was a mess but now it works

I don't get it, at all

apparently it is % 94 and / 93

it's not easy to debug in wireshark if you use base93

for base93

no, that's wrong

14:09

% 93, / 93

ok well then my implementation is wrong

14:09

or my base is actually 94

14:09

well my tests pass

then your tests are wrong

u

I can proof to you that it's % 93, / 93 if you want…

or you haven't created a test that's smart enough

*prove

oh boi i trust you

14:11

here is my test but no need from my side to check it im happy just if ur interested https://paste.zillyhuhn.com/kS

why 93 though? there are 95 printable characters in ascii

because the blog author didn't want to have , and -

well i went from ! to ~

for some reason

14:11

to better compress his json

his code is so complicated 0.0

I feel like one should remove base93 from that wikipedia article for only having a primary source

How does base93 help readability though?

not at all?

that's where I'm lost

u rude

you can copy-paste it?

14:12

perhaps

yes

but you can also copy-paste the hexdump from wireshark

type the numbers on ascii keyboard

14:13

and avoid utf8 errors

14:13

i always send strings over the wire

14:13

i didnt bother building binary based packets

you encode it in base93, you add 0x20 to get ascii characters, so you can copy and paste from wireshark?

@Learath2 it is working, /kill only works after the kil protection time

yea my encoder adds the offset

base64 has the exact same property of being composed of only printable characters AND it's standard almost everywhere

14:14

unless you need the extra compression, I still don't get the idea, but rock on

but is not single character

but a single base93 character isn't something humans can decode either

14:16

unless you have the entire ascii table burned into your head

pepeH

2

sendhelp

2

14:21

not only that, because its not a power of 2, it doesn't align well to byte boundaries, 1 base93 character is worth a nice round 0.81739485138850392378248281789042136653860459116110771616538142207754611495660215200381509901984180335141230364243796154607589618657202135276760265444230954990000188262500247343305991594219461838099269667158834316911854070919771478914536052795385853660643661410874034378300099482440799374831668149376145321093428744538878679558678814981539832844853800974679398724188050886971714870723159757110489346900184725073899821217477661574340213761402609994749634429... bytes

14:22

or 6.53915881110803139025986254312337093230883672928886172932305137662036891965281721603052079215873442681129842913950... bits

14:23

(so it doesn't even align with bit boundries)

14:24

which is only 0.54~ bits better then base64 and that gets you a nice 9%~ increase in compression

standard base64 might not be the way to go, thinking about it

14:25

IIRC it can only encode whole bytes

I'd definitely just go for good old base16

@ChillerDragon you said you still have a bug in base93. can you link the current code again?

@heinrich5991 thanks for asking but i solved it and im very happy all glitches are gone and tests are passing

my packets are fixed size strings so i wanted to be one character long to represent NET_MAX_INT

14:51

yea but for sure its good to go with base64 or standards. Hex might be a good idea but i would have much lower max values

Is there a way for compiling this game for mac in windows?

no, I don't think so

15:18

we only support building from OS to the same OS and from linux to each other OS

So is it possible to compile it for mac from linux?

yes, but it's a little bit annoying

yes, using osxcross

@heinrich5991 do you know of a way to override the python interpreter? on macOS it keeps finding the system python which is py2

in cmake?

hmm actually I better make that script py2 compatible

15:21

FWIW, getting osxcross set up wasn't that difficuly

15:21

difficult*

hm, ok

15:21

I found it hard

well for me it was just copying the sdk over and doing build.sh then build_compiler_rt.sh

whys there no ninja_lifetime tune

because no one added it

imo ninja_lifetime, ninja_length, ninja_force would be very cool

1

16:11

but

16:11

idk how to code so rip

@Deleted User github actions can build teeworlds for you also for macOS

16:16

@louis just go and play around with the source im sure if u invest some time you can get to those goals

Any tutorial?

@Deleted User well github has probably some generic tutorials on github actions you can look into github.com/teeworlds/teeworlds in the .github directory how they did it

17:16

what source base are you trying to build? If its a recent fork of teeworlds and on github you might have a mac build already

17:16

Yo bois?

17:17

Does one know the abbreviation of public domain license?

17:17

Like MIT has MIT what has public domain? Is it UNLICENSED?

17:23

omg i thought there is one pub domain which is same as unlicensed but this list says there are multiple :/ https://spdx.org/licenses/

17:24

oke looks like i found it => "Unlicense"

Can someone explain me how antiping works?

it predicts the position of objects in advance at the simplest level

22:21

it predicts where they (might) be in your ping milliseconds

22:21

so if ur ping is 300, it predicts where they will be in 300 milliseconds

22:21

helps mostly with drag parts where the tee is stationary and prediction is easy

where the tee is frozen*

trying to phrase it differently: your client takes over the physics that the server normally would, giving you feedback on your own input instantaneously. this works super well for your own tee, since it knows your input, but not quite as well for other tees, since your client does not know their input. for other tees, it simply simulates the physics as if they keep pressing the same buttons they just did. if another tee suddenly changes direction, your client predicted them to still walk in the same direction and corrects that as soon as it hears that from the server. that way antiping tries to simulate 0 ping, it depends on a stable ping, not as much on small ping

I see

22:34

So when I use antiping on a server with high ping and play solo and fall in the freeze, it does not predict that. So the freeze is delayed. Is there any change in playing with or without antiping when playing solo?

this works super well for your own tee, since it knows your input

it should be able to predict you perfectly

22:35

since nothing interfers, which your client wouldnt be able to predict

So you think playing with antiping is always better when playing solo?

yes

22:36

i would always suggest playing with antiping

what is your ingame name?

also Patiga :)

22:37

antiping also helps for example whenever you have to catch a frozen tee. since its just falling down, your client can predict that perfectly

Ok. You can try to go on chile gores server and fall in the freeze with antiping on and off. there is no difference. its both delayed. but if its the way you say, shouldnt antiping predict it and freeze you instantly? I mean make it visible

meaning you dont have to care about ping

22:38

in theory it should exactly do that, if the implementation was perfect

22:38

my guess is that the antiping correctly predicted your position, but didnt make your tee look frozen?

I always thought antiping is only "used on" the other players. and does not effect solo gameplay at all

22:38

@Patiga yes

:/

22:39

ye then it sounds like it does well with the position, but doesnt predict you being frozen, but waits for the server response for that

22:39

that sounds like it could be fixed at some point

but wouldnt that mean I could not play with 250 Ping gores at all? Or actually, I still dont understand how exactly the position prediction works.

your client takes over the physics that the server normally would, giving you feedback on your own input instantaneously. this works super well for your own tee, since it knows your input, but not quite as well for other tees, since your client does not know their input. for other tees, it simply simulates the physics as if they keep pressing the same buttons they just did. if another tee suddenly changes direction, your client predicted them to still walk in the same direction and corrects that as soon as it hears that from the server.

i thought this would cover it quite well

22:42

your client knows the map and it also knows your input, meaning it should be able to predict well

22:42

(maybe not moving freeze tho)

130 ping and below antiping works almost perfectly

22:43

anything above that and it is really annoying

Oh do you play league of legends? when i have 250 ping there, it actually is impossible to play. thats right. because its delayed. but in teeworlds idk, i can play on chile server 250ping witout antiping just as good

i still dont get how smurfer can play with 200 ping tho

no delay at all

@Shyzo do u have antiping turned on

22:43

also make sure antiping limit is 0 or whatever

22:44

the slider

usually I play with antiping on, yes

22:44

yea its on 0

and make sure u have all the boxes checked

22:44

below

i do :D

i just tried out chile gores, and it worked well, apart from when my ping suddenly changed a lot

22:45

(at ~250 ping)

yes. but with and without antiping, its the same

22:45

how?

aah

22:45

1 sec

22:46

you are right, it works just as well

22:46

my guess is that antiping is in vanilla

22:46

so ddnet antiping is just extending it

ah

22:46

that could be it

22:47

lol,its funny. hook and a/d is fine

22:47

but gun fire

22:47

is deelayed xd

interesting, then normal antiping is probably movement only

22:48

and ddnet antiping adds weapons and support for their blocks

cl_predict affects client side calculations about your own tee

22:49

turn it off and u have server side

22:49

antiping for pure gores solo is useless, but if u use rockets etc its nice to have

oof I thought cl_predict and antiping is the same. oopsi daisy. i just tried it. now thats how it should be xD

22:54

and can someone explain me the difference between cl_refresh_rate and gfx_refresh_rate ? All I know is that, when I cap with cl, my entities blocks are a bit laggy

yeah cl is affects all calculations, even the ones that aren't drawed frames(invisible updates basically) while gfx is really just your fps

22:59

so with cl on 1000 u will never get 1000fps bcs ur GPU isnt inifinitly fast

22:59

i also dont like this implementation, but best is probs to let cl be 0 and gfx and anything u want

23:02

to say it better, it may skip frames makes the updates less smooth

Ok. I am wondering why it is the case that cl_refresh_rate 480 is a default setting

23:17

Or used to be default. I am not sure. But some people have it like this

old client xd

Exported 234 message(s)