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 2021-07-04 00:00:00Z and 2021-07-05 00:00:00Z

https://youtu.be/ITogH7lJTyE

Async Await try-catch hell

03:33

justatest

[ddnet-maps:master] 1 new commit

b8143e7 M Skill Rises, M Ton - ddnet-maps

@Deleted User slow death eh?

stencil

@Deleted User slow death eh?

what

does someone know where this variable gets set? https://github.com/ddnet/ddnet/blob/b7adc5a690216d56fa8636c5efbfd91a5e4fe07d/src/game/editor/editor.h#L338

DDraceNetwork, a cooperative racing mod of Teeworlds - ddnet/ddnet

11:33

I just did a CEditorMap m_Map and m_Map.m_pEditor was set magically to a editor object o.O ?and the construcor only calls clean

void CEditorMap::Clean()
{
	for(int i = 0; i < m_lGroups.size(); i++)
	{
		m_lGroups[i]->m_lLayers.delete_all();
	}
	m_lGroups.delete_all();
	m_lEnvelopes.delete_all();
	m_lImages.delete_all();
	m_lSounds.delete_all();

	m_MapInfo.Reset();

	m_lSettings.clear();

	m_pGameLayer = 0x0;
	m_pGameGroup = 0x0;

	m_Modified = false;

	m_pTeleLayer = 0x0;
	m_pSpeedupLayer = 0x0;
	m_pFrontLayer = 0x0;
	m_pSwitchLayer = 0x0;
	m_pTuneLayer = 0x0;
}

magic

chillerdragon

I just did a CEditorMap m_Map and m_Map.m_pEditor was set magically to a editor object o.O ?and the construcor only calls clean

void CEditorMap::Clean()
{
	for(int i = 0; i < m_lGroups.size(); i++)
	{
		m_lGroups[i]->m_lLayers.delete_all();
	}
	m_lGroups.delete_all();
	m_lEnvelopes.delete_all();
	m_lImages.delete_all();
	m_lSounds.delete_all();

	m_MapInfo.Reset();

	m_lSettings.clear();

	m_pGameLayer = 0x0;
	m_pGameGroup = 0x0;

	m_Modified = false;

	m_pTeleLayer = 0x0;
	m_pSpeedupLayer = 0x0;
	m_pFrontLayer = 0x0;
	m_pSwitchLayer = 0x0;
	m_pTuneLayer = 0x0;
}

I'd guess UB, since m_pEditor is set in CEditor::Init()

Defaulting to UB upon first glance seems like a weird idea to me. but im not a wizard.

Deleted User

between those

yes so sorry im late, i tried the first one and i got this. i just dont know why i have this, everything look ok

Image attachment

12:43

remove the mentioned functions

12:43

u probably tried the other solution

12:43

and forgot to remove them now

12:43

ResetDragVelocity and the other

12:44

? thats not good

Image attachment

12:45

well thats the other solution isnt it?

12:45

the other way to remove weak

12:46

u havent implemented it, and u also dont need it

12:46

i deleted all this things i added with the second way

12:47

i mean i tried this tsFreddie link not your

12:48

yeah but u need to remove the things from tsfreddie

12:48

if u go the other way

12:48

ok, maybe i forget some things

12:49

i retry a last time with ur

i have many errors

Image attachment

13:05

+20

13:05

i think i do smth wrong

13:05

probably

yes

13:19

you probably forgot to implement the pretick

13:19

just do it as in the commit

13:19

with virtual functions etc

13:25

smth is really wrong, why is CCharacter in gamecore

13:27

omfg

@Deleted User do u accept private message?

13:46

doesn't understand what that mean here (edited)

Image attachment

TsFreddie

Defaulting to UB upon first glance seems like a weird idea to me. but im not a wizard.

I mean if he literally just created a CEditorMap and hasn't even initialized it there is no other conclusion to come to

ye i guess

- The face, the arms and the hands of the candidate as well as the paper where the candidate writes must be in the view field of the camera during the whole test.

Does a camera with such a FOV even exist?

i guess its called being one meter away

I checked even a meter away it won't fit all requirements

14:00

To get my face in the same frame I need to sit with my head to the desk

[ddnet-web:master] 1 new commit

a056092 Removed Arii - Learath2

f Arii

[ddnet:master] 6 new commits

9dfea55 Use cpp - Jupeyy 3e1d4ad Use chrono - Jupeyy a663799 uint64 -> uint64_t, int64 -> int64_t - Jupeyy 69aa97f Remove (void) parameter of functions - Jupeyy 2cb3c57 Merge master into pr_use_cpp - def- 94cfb33 Merge #3913 - bors[bot]

14:18

[ddnet/ddnet] Pull request closed: #3913 Use C++

Deleted User

doesn't understand what that mean here (edited)

i think it just calculates it based on the current entity list

is the dummy occasionally having half the fps of the main tee a known ddnet issue or some buggo in my crack client?

dunno, someone asked this before, was that u too?

14:39

i have same fps

might be me i think this issue is old

14:39

lemme check irclogs

14:40

ddnet/2021-03-25.log 63:17:10 <+ChillerDragon> anyone having expirience with the dummy having only 1/4 of the fps when running the client with gdb after some time?

14:40

ye

14:40

was me

14:40

now its only half and no gdb

14:41

but seems like back then i did not use a custom client or did i hmm _D

14:41

why is it always that when i write code i trade features against performance and stability :/

i dunno, but should be ez to profile i guess

how would u profile it?

callgrind i guess

yikers

gperf

valgrind?

one of those

14:43

yeah

hmm that sounds better

14:43

but then i have some callgraph and idk whats going on :D

14:43

but sure ill give it a try

callgrind has kcachegrind, which is good to look at the gcc instructions only generate a text file

14:44

but also ok to look at

i can not run a client with valgrind

14:44

in fact idk what even runs with valrgind

gprof is it, not gperf*

yea

gotta compile with -pg

all the same isnt it

14:45

ye ik

ok

14:45

i can run almost anything with valgrind

thats fast

u should use fair shed for multithreaded apps

can you run a client with valgrind and use it?

yes

wtf

but only with 15 fps ofc

14:46

not like playable

14:47

ah callgrind btw is much faster

14:47

also it has a toggle

14:47

valgrind --tool=callgrind --instr-atstart=no

ah

14:47

i only do valrgind

callgrind_control -i on/off

14:48

valgrind is just something like a vm iirc

ah true i have 15 fps too

14:48

but 15 fps is crazy

memtool is for memory and callgrind for performance stuff

i have 4 fps in scoreboard how would i tell if dummy is worse :D

see on the generated output what functions are slow

ye valgrind is a vm

it should look similar if you run the same stuff

definitely lost: 21,422,992 bytes in 81 blocks

at closing?

btw i always wonder if that means the code has bugs and something has to be done

it probs means u didnt deallocate memory when closing

14:49

most textures arent deallocated for example

i closed the client with X in the UI

14:49

vanilla ddnet btw

14:50

i think no tw client or server has 0 bytes lost

its nice to have, but also not insanly important it should still be a goal for better code

14:51

if u deallocate correctly without crashes thats a sign of good code imo

yea sounds like a nice goal

14:52

also if upstream has 0 i (the cracky fork maintainer) can see quickly when i messed something up :D

14:53

but some memory leak isnt neccesary related to performance issues is it?

if its inside the prgramm executation could defs be

14:57

but when closing the app.. doesnt matter then anyway

callgrind is very heavy

17:24

perf and gprof are much more usable for client performance profiling

17:24

valgrind isn't a vm, it's a runtime instrumenter, it injects code

does the one exclude the other?

chillerdragon: are you actually playing in fokkonaut server or its a fake?

do you really care xd

yes cause i tried multiple time to ping him about teeworlds code and he talking about shit or idk

18:38

tried to scam the rcon pass from my srv (edited)

isn't it enough to understand that it is a trolldragon (edited)

that not funny

i mean

18:42

fake chillerdragon

fake with +level 40

18:44

where's the staff

his ping 40+

18:45

afaik chillerdragon lives in germany

18:46

okey idk

Deleted User

chillerdragon: are you actually playing in fokkonaut server or its a fake?

he does

Deleted User

where's the staff

Here

@fokkonaut is possible to implement function /give? you type /give ZimE and me then type /accept and we switch position? when i fail other player i can give my position to other player pls?

chillerbot: you are god of comedy

19:57

send rcon pls

CGameContext is a non trivial type, it has virtual functions, normally using virtual functions also implies using a virtual destructor. Calling the destructor of CGameContext and then new on the pointer with the base class CGameContext is just bad code

20:09

@Deleted User is it not possible to reinitialize an object using its actual class? it feels like this should be possible

heinrich5991

@Deleted User is it not possible to reinitialize an object using its actual class? it feels like this should be possible

thats my point, its a base class, so if you inherited from that and use delete new on the base class it looks really bad to me

ah, so you mean that because someone in the future could decide to inherit from CGameContext, it's bad design to use this reinitialization?

yes i think its just a bad design choice, i dunno if its actually UB c++ wise, the stackoverflow question was closed xD

20:41

https://en.cppreference.com/w/cpp/language/lifetime#Storage_reuse

cpp/language/lifetime

20:41

this is the requirement apparently, i don't see it breaking the definition tho

20:42

the mem_zero is possibly UB bcs the constructor doesnt initialize the values, and the mem_zero is allowed to be optimized

20:42

away (edited)

mem_zero is definitely not undefined behavior

20:49

zeroing bytes is fine

20:50

you could ask whether it's allowed for the constructor to mangle bytes of the underlying storage

yes

Exported 152 message(s)