DDraceNetwork - developer

I think a tee is smaller than a gametile

@Learath2 ask me anything :D Especially the if block with the code you mentioned is really not that nice... but all these checks are needed for the transition from local to serverside starting behavior. The problem is that the client does not know if the racetimer is on until it receives the first snapshot with the correspondint flag set. If there was a timer that is always on the code would be a lot cleaner. About the ddrace check: The client recoginzes the race start one tick earlier than the server. The servers just notices it one tick after the actual movement happened. So i need to correct the tick to keep in sync with old ghosts

00:04

for block tiles the tee is actually larger than gametiles >.<

okay let's see, is the one tick difference always consistent?

so the map would have to be changed?

also btw @redix I think we need physics changes to be 100% backwards compatible

this will be kind of impossible since the old position check was kind of non-working at all

Not sure how many maps would need to be fixed and quite possibly all the ranks would be invalidated on those maps aswell

in general this behavior is rather random since it does also rely on the starting position and not only on the speed. so in terms of race fairness i would prefer to completely remove it...

I don't think it would influence ranks on the map I posted

00:11

or is it different behaviour in general, and not only special cases

problem is that fixing this will break two and four direction stoppers

regarding the ghost stuff, if IGhostRecorder and IGhostLoader are interfaces they do not need Interface pointers passed to them and they do not need helper functions in IClient as you can call them from anywhere in the code

yes you are right... still some remains from the demo recorder code. i used it as a starting point

also what do you mean by serverside starting behaviour and ServerControl, doesn't seem the server is actually controlling the playback and the recording either

server controls means that the starting behavior relies on the race timer instead of checking for the startline client side

when do you need to actually check for the startline client side?

when the race timer is off

00:14

problem is that i only know that it is on after the local start already happened

00:14

so i need all this "transition" code

hmm we really do need some kind of NETMSG_SUPPORTS or an actual way of sending race events

for the ghost a reliable timer is even better for the starting. for the end a new netmsg would be great

components/ghost.cpp:L261 what does this really check?

damn... i will try it tomorrow but i'm pretty sure that the map @jao mentioned won't work anymore =\

00:25

@Learath2 i should have kept the previous version of that check. was better readable i guess if((ServerControl && !Predicted) || (!ServerControl && Predicted)) In the first case the "server side control" might have changed the RenderTick, in the second the local control might have changed it

00:26

i'm not even sure if this line is still needed ^^

tbh I don't get what most of this is doing, always thought the ghost code was a mess :/

this code has been in the render method before

00:28

the onsnapshot method has to deal with so many different cases =\

OnNewSnapshot shouldn't the only thing you are checking be whether the race started or ended somehow?

it is only the start

this is only checking for start?

yes... the rest of the code should be kind of okay... but i know that this method is horrible

could maybe be generalized by abstracting changes in state, e.g. having an OnRaceStateChange hook

this could maybe help... but the thing with the unreliable timer will still bloat the code

00:34

besides the onsnapshot method... any other things you found in the code ^^

that was my main concern, but let me go on reading

okay... each time i fixed some issue with the timer the onsnapshot method got worse

ghosts don't save every tick right?

no only when the buffer of the recorder is full... every 2 seconds with default snap rate (edited)

00:44

about the blockers: i can bring back full entering for one direction blockers but fot 2 and 4 direction blockers it might cause a lot of trouble (edited)

@redix CGhostLoader has only one ghost file loaded at any point, is there any reason to have GetTicks take a CGhostHeader *? GetTicks could just use m_Header as a member of CGhostLoader

GetGhostInfo is used in the menu to get the header

00:52

getticks is actually not used there but gettime is

GetTicks and GetTime then actually belong to CGhostHeader as they only depend on the header

yeah. makes sense

think I'm kinda being nitpicky

haha

ReadData and WriteData should take unsigned char * or void * from what I can read

agree

btw... while changing some things in player.cpp i stumbled across the Predict function. It seems to be part of antiping. A lot of commented out code... it says something about missprediction. some code depends on the ping value from scoreboard... i'm not sure what the guy who wrote it tried to measues... but it looks kinda weird (edited)

oh I'm not sure anyone really knows what that stuff does anymore

01:15

it's mostly @deen's code iirc

01:16

I think except for the things I mentioned and OnNewSnapshot I'm okay with this

https://github.com/ddnet/ddnet/commit/dc6d2305d59446eedacd93149a3efc2d1cea5d9f AntiPing from TDTW client by Psycho.God <3. & kaddy.

Add AntiPing · ddnet/ddnet@dc6d230

why did you need to friend CGhost in CPlayers btw?

okay. i'll fix these things... not sure about onnewsnapshot though. right now this function is very fragile =\ maybe adding a new timer / netmsg and cleaning it up then is a better idea

01:20

for calling RenderPlayer. could also have made it public

this is what I hate about OOP... RenderPlayer doesn't even need to be a member for CPlayers...

01:22

meh I guess it's fine like this

the whole file is messed up. grabbing the predicted values inside a rendering function and stuff >.<

well fix the small stuff, I'll merge then we can look at OnNewSnapshot

okay. gonna do it tomorrow. time to sleep now

01:25

about Predict

yeah way too late

(edited)

the function depends on double d = length(PrevPredPos - Position)/s_Ping; Position depends on IntraTick So PrevPredPos - Position seems to depend on the frame time. I'm pretty sure this is wrong ^^

one more thing before I also get some sleep

01:31

you don't seem to be using File after engine/shared/ghost.cpp:L355 you could just close it

01:31

would simplify your ifs

right. thanks for that hint

I'll get some sleep, gn8

01:36

ofc as always thank you for thy contribution

same here. gn8 :) i wanted to merge this code (from teerace) for a long time

was too lazy until now

https://github.com/ddnet/ddnet/pull/801 Anyone want to fix this up?

Reopen: Added emoji support by heinrich5991 · Pull Request #801 · ddnet/ddnet

Includes: png spritesheet python datasrc script changes :alias: to utf replacer utf/:alias: to png image replacer Needs code review, builds and works well, may need some offset fixes as I wasn't ...

def-

[ddnet/ddnet] Pull request closed: #917 Add DPI awareness support for MSVC compiling for Win8.1 and Win10

07:42

def-

[ddnet:master] 7 new commits

152a715 Add DPI awareness support for MSVC compiling fo... - Jupeyy 907622e don't read the manifest of the target - Jupeyy 162b24c respect the CFG dir - Jupeyy 9a4efe7 use proper CMake 3.4 support for manifest files - Jupeyy 205b054 move manifest to other directory and add MinGW ... - Jupeyy

07:43

def-

[ddnet:master] 1 new commit

acb052e Newline - def-