DDraceNetwork

Tater

you could actually end up hitting a freeze tile you didn't hit if you are moving fast enough (edited)

This needs to be featured in a map

@Skystrife you can do a map with it!

my "x" and "d" dont work somehow

06:22

my keyboard works fine, so its not the keyboard problem

06:24

with the keyboard thing

idk try binds in f1

https://discord.com/channels/252358080522747904/1299737820683833354/1299739182691913749 cant vote map (edited)

where are we lol

Yoko_3.1_239.480_Atsuko.demo

1.78 MB

Devinci

bwoah that's so cool tho. Maybe it's just me but I think that those "features" might create some interesting ideas. (if it's not fixed I'm probably stealing this x)).

just fyi if it wont get fixed there will be a map where players explore that bug soon (not planned but on accident)

Ah gg. Well, i would only use this for a trick or two at most anyways. I do enjoy these kind of bugs :p

Patiga

@AssassinTee do you think this is a thing that could be fixed? :)

what?

AssassinTee

what?

https://discord.com/channels/252358080522747904/757720336274948198/1299482522941522052

I can list related issues I already pinged. This is not the only case this can happen

18:23

It's because freeze is calculated in ticks instead of subticks

18:24

and because ddnet decided to take effects between two ticks in a straight line which is wrong. This is reproducable even without bouncyness

18:24

@Patiga

18:26

related issue @Robyt3 https://github.com/ddnet/ddnet/issues/4523

wrong ddnet-tile collision in subtick · Issue #4523 · ddnet/ddnet

Hi, I was checking some DDNet code and stumbled across this function: https://github.com/ddnet/ddnet/blob/master/src/game/collision.cpp#L920 This is used to determine for example when start and fin...

18:26

notice how it's from 2021. (edited)

18:28

The pictures from @Tater are great, this can also be done by simple hitting another tile with speed

18:31

black is wall

AssassinTee

Click to see attachment 🖼️

the bounce tile interaction appears to be much more apparent, as it interferes with the normal function of the tile ^^

if you want this fixed, add a "freezed" flag to the move box function and check on the bounce if you hit a freeze, this is quick and dirty and doesn't fix the underlying issue, it's just an ugly workaround. I highly advocate for fixing this for all ddnet interactions (edited)

18:40

this is also the explanation why the jump flag is added

AssassinTee

if you want this fixed, add a "freezed" flag to the move box function and check on the bounce if you hit a freeze, this is quick and dirty and doesn't fix the underlying issue, it's just an ugly workaround. I highly advocate for fixing this for all ddnet interactions (edited)

changing deep physics like that is typically not an option

also this doesn't fix the issue of hitting a freeze you never actually hit

server settings can be moved and layers selected with the same keybind, if you open the server settings menu, select a setting, then click on a layer

Patiga

changing deep physics like that is typically not an option

The implications are a lot of other undesirable side effects that are not so apparent, only other known effect are wrong finish/start times, because the subtick counter is in the mud (edited)

18:49

next Issue i can think of: Freeze now fixed, but can skip start/finish/switches/weapon grap(?)/whatever entitiy you'd like

the movebox function would need to handle the ddnet tiles

18:51

if implemented "properly"

1

18:51

because the information cannot be obtained outside of it

18:51

but as Patiga said, changing that sounds difficult

could the function just be called twice on a bounce?

you mean like adding a third position like m_bouncePos and then calling that straight line function twice? (edited)

19:02

that actually sounds neat

yea, like towards the reflection point or smth

this could also work for the other case, with the wall

you could have an arbitrary number of bounces, so you'd have to return a list, but that would still work

true

I think the only good solution is having move box handle the unskippable tiles but that also seems like a massive breaking change so I don't see how this gets fixed

22:48

returning a list is really slow

we can trade one bug for a different one if we set m_LastPos to the most recent bounce position, this wouldn't affect other physics.