DDraceNetwork - developer

is it intended that pressing run server wont open up a console window with the password?

02:50

steam gnome fedora

I guess the goal is passwordless automatic authentication on servers launched from the client

ChillerDragon

[ddnet/ddnet] Pull request opened: #9399 Replace magic numbers with compile time string length

The sizeof() macro is evaluated at compile time. The string passed to it does not end up in the final binary and there is also no counting happening at runtime. The sizeof() operator counts the null terminator too. So sizeof("foo") is 4 and not 3. But this comes in handy because we want to skip over the length of the command and the slash character in front of it.

Checklist

[X] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Test...

<ChillerDragon> @Evelyn before we do any 128 player debugging could you update your server? I pushed an update to the feature_128 branch

05:17

<ChillerDragon> @deen btw while at it i also sent you the solved conflicts for the 128 pr in case you want to save 5 minutes https://github.com/def-/ddnet/pull/3

128: rebase 1 by ChillerDragon · Pull Request #3 · def-/ddnet

Solve conflicts and rebase the 128 branch onto the latest master

05:18

<ChillerDragon> somehow gh forced me to create a merge commit to be able to pr a force pushed branch ._.

05:19

<ChillerDragon> but you can also git reset --hard HEAD^ to pop off the merge commit and then force push that your self that should be a clean rebase

better antiping smoothing algorithm I'm working on, the first 20 seconds is the old version for comparison.

ddnet-maps

[ddnet-maps:master] 1 new commit

4e60bcb A Edge Jump Pro, A One Block - ddnet-maps

ChillerDragon: i remember you stating ddnet or 0.6 clamps NetIntRange instead of dropping like in 0.7, do you remember how sure were you when you found that out?

Tater

better antiping smoothing algorithm I'm working on, the first 20 seconds is the old version for comparison.

YES

05:35

please we need a better antiping algo

Tater

better antiping smoothing algorithm I'm working on, the first 20 seconds is the old version for comparison.

oh, i remember doing something similar when i was working on that ddnet clone for a chinese company. i found it extremely helpful to interp or smooth separate axises instead of a single vector on 2d characters given that jump is often quite violent. not sure if it is worth doing on ddnet given how often the character is in the air, but might be worth a try. we also did stuff on 3d characters where we smooth xz and y separately in our VR game, similar principle i guess. basically if there is a clamping or a entropy system, we run it one dimensionally on two axis separately.

that's basically what I'm doing but I decompose the prediction vector based on a certainty vector so it's not biased by any axis (edited)

05:42

it generalizes to infinite dimensions in theory

jump and gravity usually just work on y only which where that system comes from since it behave differently than x.

05:43

might only work how i was doing it tho, just sharing.

I actually haven't added the "smoothing" part yet, this is just designed to give a more accurate position a higher % of the time. It still has instantaneous movements so I'll need to add another layer to smooth those.

05:44

the smoothing reduces accuracy for smoothness

05:46

I was thinking about doing a special case for jumping but there's too many types of actions that behave the same way as jumping so I don't feel it's worth it.

TsFreddie

oh, i remember doing something similar when i was working on that ddnet clone for a chinese company. i found it extremely helpful to interp or smooth separate axises instead of a single vector on 2d characters given that jump is often quite violent. not sure if it is worth doing on ddnet given how often the character is in the air, but might be worth a try. we also did stuff on 3d characters where we smooth xz and y separately in our VR game, similar principle i guess. basically if there is a clamping or a entropy system, we run it one dimensionally on two axis separately.

wait you said it's a ddnet clone but you have 3d charecters?

Tater

wait you said it's a ddnet clone but you have 3d charecters?

that's a different game

05:50

i worked on like 3 different projects that has similar gameplays in 3 years

so how did your 2d smoothing algorithm work?

05:53

gather some input changes and state changes from past few ticks and get a "entropy" value, it's just how often it changes, and slowly decrease the delay as the movement gets more predictable (as in decreasing that magic entropy value). and that looks too floaty and we just run it twice on x and y separately and it looks better (edited)

yeah that's very similar to what I did https://github.com/sjrc6/TaterClient-ddnet/blob/6382bfc13e2f9ebb5b9652cd4f67fa4e9193b945/src/game/client/gameclient.cpp#L2507-L2697

TaterClient-ddnet/src/game/client/gameclient.cpp at 6382bfc13e2f9eb...

Extended DDNet Teeworlds client. . Contribute to sjrc6/TaterClient-ddnet development by creating an account on GitHub.

i'm just saying doing whatever that we were doing and just run it twice 1-d might look better

idk what you mean by twice 1-d

05:59

this is once 2d

if you are detecting changes in vel, detect changes in velx and vely separately and do the posx and posy prediction separately using the two changes info you gathered instead of treating the velocity as a 2d vector.

06:02

like whatever you are doing, break it in to x and y and do it separately, maintain two states, and if you are doing inputs, map direction to x's changes and jump to y's changes and hooks to both.

I used angle between prediction vectors to detect changes instead of velocity because it doesn't need any tuning (edited)

that is quite different from how i imagined it, but you could just run the angle as a 2d vector and detect changes in x and y separately. but again, just sharing. do whatever you feel like is best.

06:06

like replace the dot with just distance delta of x and y or something, but value is definitely different.

but if you have 2 axis then you can't measure instability along a 45 degree

not sure if it is worth doing on ddnet given how often the character is in the air, but might be worth a try.

06:07

i don't have any more info to share

06:07

i'm just saying i observed that

06:07

i don't have mathamatical proof

like in this case the 1d*2 system will see this as instability in all axis and won't find the stable axis

06:09

we can safely put the tee at the end of the red line, but if you do 1D twice it will have to place it at the server position

our entropy is more input driven, again, do whatever you believe is the best

hmm

i really don't have any proof

well

06:09

neither do I

06:09

Idk what you mean by proof

like i can't proof the thing i did is mathematically closer to ground truth