Teeworlds - bridge

Teeworlds

IRC / bridge

One-way IRC channel bridge. If you want to be able to send messages to IRC, contact @Dune or @heinrich5991. https://www.teeworlds.com/?page=docs&wiki=rules/irc_rules

Between 2020-08-21 00:00:00Z and 2020-08-22 00:00:00Z

can we do something about the wrong prediction in 0.7?

00:27

if i hook a tee the hookpos is so false

00:27

even noticable on good ping

00:27

only on 0.7 clients (edited)

there is no prediction for other tees in 0.7 and when you are hooking someone else the client just uses the hook pos from the server does it happen on all servers?

yes

otherwise i guess the server is just sending a wrong hook pos

00:58

https://github.com/teeworlds/teeworlds/blob/751bfb92b313e336819b362429a971db8d6e67df/src/game/client/components/players.cpp#L107

nop

00:58

for 0.6 clients it works

no prediction involved here. it just uses the pos from the server

no

00:58

the character order doesnt match the servers' one

00:58

thats the issue in 0.7

00:59

thats why this is happening

00:59

oy said something like that iirc

00:59

that atleast makes sense to me and explains the weird behaviour

that was just the same in 0.6

nop

01:00

i dont know the cause but there is a bug, i know it

it was. snapshot diffs have no clue about the order

01:00

ddnet has additional code for the order but 0.6 never had

look

01:02

try to join a server with 0.7 vanilla client (can also be vanilla server), preferable high ping, then hook players and drag them through the air

01:02

you will notice the hook pos doesnt match the character pos at all time

01:02

then try it with ddnet client, there its butter smooth in the center of the tee, always.

01:03

can also try it out on bridge servers, as its really just a clientside thing

it's wrong until the server tells you about the hook. same in 0.6. seems like ddnet fixed that

in ddnet it's a side effect of the antiping implementation

01:45

the vanilla code mixes predicted and unpredicted values

01:45

@fokkonaut try this https://github.com/axblk/teeworlds/tree/cleanup_prediction

will test it tomorrow

but i am 100% sure this wasnt the case in 0.6

01:47

like also not in vanilla 0.6

01:47

will test that too tomorrow

just tested it in 0.6. same as 0.7

01:49

that code actually didn't change between versions

lol

01:53

so it actually was always buggy?

Yes. It checks the unpredicted value for whether you are hooking someone. So the client still thinks your hook is in the air. Then it uses the predicted position of the hook, which is already pulling the other tee back.

xd

Actually you cannot use the predicted hook position at all when you are hooking someone, as the position of the other tee is not predicted. You need to use the unpredicted position of the tee as hook position

How does ddnet handle it then

Thanks to antiping the corresponding function in ddnet does not even know the unpredicted values. They rather fixed that accidentally

lol

Exported 46 message(s)