DDraceNetwork - developer

d8c9a71 Remove usages of qsort - Learath2 9d39e1d Merge #3046 - bors[bot]

08:45

[ddnet/ddnet] Pull request closed: #3046 Remove usages of qsort

It's useful if you want to separate allocation from initialization (placement new) on objects with const data (avoiding ill-formedness). So it's a low level feature (which should have been a languange feature instead of an STL feature)

[ddnet/ddnet] Issue opened: #3049 Undefined behaviour in snapshots

I found a funny bug again: https://stackoverflow.com/questions/45429780/is-right-shifting-undefined-behavior-for-negative-number-in-cpp-and-in-java I was profiling with MSVC under windows, and for some reason nameplates always called LookupUuid which is quite inefficient (bcs its called for all clients every frame) then i tried under linux.. couldn't reproduce. then i analysed whats the difference between the two versions and found out that in snapshot.h int Type() { return m_Ty...

@Deleted User theorically its implementation-defined not ub monkalaugh

that is ub for me

[ddnet/ddnet] Issue opened: #3050 Go through ubsan and clean up

CC=clang CXX=clang++ CXXFLAGS="-fsanitize=undefined" CFLAGS="-fsanitize=undefined" cmake -GNinja -DWEBSOCKETS=OFF -DMYSQL=ON -DCMAKE_BUILD_TYPE=Debug -DVIDEORECORDER=OFF -DAUTOUPDATE=OFF -DANTIBOT=OFF -DUPNP=OFF -DPREFER_BUNDLED_LIBS=OFF .

This is just on starting client, not even joining game yet: ``` [2020-10-10 11:48:04][host lookup]: host='master1.teeworlds.com' port=0 3 src/base/math.h:50:11: runtime error: left shift of negative value -32 SUMMARY: UndefinedBehaviorSanitizer: ...

09:54

backend_sdl is full of ub

why do we call it find_binary if it does partition_linear

09:55

isnt that linear then xd

[ddnet/ddnet] Pull request opened: #3051 No negative right shift in snaphsot.h (fixes #3049)

Cmd_Texture_Create(static_cast<const CCommandBuffer::SCommand_Texture_Create *>(pBaseCommand));

09:55

looks like doing this is not ok

what does it say

ah no

09:56

downcasting misaligned memory

09:56

address*

09:56

https://github.com/ddnet/ddnet/issues/3050

09:56

its all here

@deen can you quickly look into find_binary, maybe i just dont understand it, but doesnt look like binary search xd

09:56

or what is the sense of find_binary xd

09:56

binary data?

09:59

@Ryozuki mhh yeah, looks like alot of false positives probably

@Deleted User never looked at that code, looks like it calls partition_linear by mistake and should be partition_binary

ok

but weird

https://github.com/ddnet/ddnet/blob/9d39e1da6d6a592d71a3992f4450e37b5cb5616f/src/engine/client/graphics_threaded.h#L586

ddnet/ddnet

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

09:59

i think it comes from here

its just alot of polymorphism casts

10:00

that it detects

10:00

i dunno if it has some option for old cast

ubsan doesn't have false positives btw, everything it finds is an actual problem

&m_CmdBuffer.DataPtr returns a unsigned char

10:00

oh

10:01

no false positives justatest

src/game/client/components/menu_background.cpp:42:40: runtime error: downcast of address 0x5612c7fb0480 which does not point to an object of type 'CMenuMap' 0x5612c7fb0480: note: object is of type 'CBackgroundEngineMap'

10:01

looks wrong imo

10:01

it clearly creates a cmenumap

10:02

even used override to not fall into the signature trap

10:02

or is the vtable not available in constructors

@deen https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81705

10:03

are you sure?

we use the clang version, not gcc. and also "no false positives assuming no bugs in ubsan"

i still dont see a bug in that code

10:05

yeah

10:05

so i think the undefined behaviour is virtual functions

10:05

makes sense and good to know

10:06

about memory alignment

10:06

in x86 it probs doiesnt matter too much

10:06

but we could fix all xd

exactly, we should still fix them for other cpu archs

but thats alot of work actually

10:07

the whole command buffer just doesnt care about alignment

what would you need to do to fix this?

align memory

10:07

alignof(ptr_type)

10:07

is 8 on 64bit

10:08

so always add paddings

10:08

to all commands

10:08

well if we go even further

10:08

fix all floating point alignments

10:08

they make more trouble under arm

alignas(alignof(unsigned)) unsigned m_Cmd; ?

10:11

never used this

10:11

gotta research

no

10:13

not unsigned, that wont matter alot

10:13

its the pointer itself

ah

[ddnet/ddnet] Pull request opened: #3052 Fix virtual call in Constructor

10:18

[ddnet/ddnet] Pull request opened: #3053 Optimize uuid lookup

Already wrote that patch when i noticed its not about this, maybe we can use it, else nvm

we have alot of floating point errors @deen

yes

i fixed alot somewhere

[ddnet:15.1-devel] 1 new commit

20976dd Fix virtual call in Constructor - Jupeyy

10:40

58bd8dc Fix virtual call in Constructor - Jupeyy 80ee8ff Merge #3052 - bors[bot]

10:40

[ddnet/ddnet] Pull request closed: #3052 Fix virtual call in Constructor

ah sad, i probs didnt save most fixes, but atleast one: https://github.com/Jupeyy/ddnet/commit/adb5aabe725cc9f4fdce38f444b5f27d7235c2b3

10:44

sadly most of these things are also in gamecore.cpp or .h

10:44

so i didnt want to touch them i think

10:44

this case i saved bcs the result is sent to the GPU

10:45

but opengl has nan and inf, so not UB directly

find_binary is really find_binary

11:10

it confused me for a while too

why

11:11

wait

11:11

u mean find binary data?

11:11

i dont really get how its binary search, i also switched it and still works xd

Nah, you were talking about partition_binary and partition_linear

yes

11:11

its used in find_binary

11:11

partition_linear

11:12

i changed it in https://github.com/ddnet/ddnet/pull/3053 and still works

Optimize uuid lookup by Jupeyy · Pull Request #3053 · ddnet/ddnet

Already wrote that patch when i noticed its not about this, maybe we can use it, else nvm

Oh find_binary was using partition_linear?

yeah

troll function for sure

@redix fixed it on vanilla a year ago, why wouldn't he share? traitor

xDDD

Only noticed that while fixing the snapshots ^^

welcome in the club

Was a bit confused when I noticed that :D

welcome in the club

11:16

xd

11:18

@deen about the skin fix, we have todo it differently

11:18

else skins like blacktee are broken

too bad, can you change it?

inside ur pr?

The downcast thing is because entities remove and insert themselves from the gameworld, which is just odd

but that wouldnt be ub would it?

11:46

it has to be constructed wrongly somehow

Currently when RemoveEntity is called from ~CEntity which happens after ~CCharacter. At which point it is just a CEntity which is UB to cast to a CCharacter and call a method on

ah alright (edited)

k, good morning. I'll continue asking stupid questions here if you don't mind

about text?

chat

12:45

I managed to align some chat messages text yesterday and even draw roundrect background for each message, but it's too dirty

12:46

The first question is more about cpp I guess, 'cause seems like I suck in it.

if u doing it for ddnet make sure to use buffered rects xd

женя угомонись

12:47

шо ты пишешь

I made a function for drawing chat message background in CTextRender class, which is stupud idea. I probably need a reference to CRenderTools here, but I didn't find the way to get it

u can request it form kernel probably, but do it rather in chat.cpp

@Deleted User i guess, or in your own, both are fine

ok then i'll do it xd

(i just pushed again)

oh

13:34

too late i guess xD

13:34

that was my fear xD

yeah, sorry

no way again in editor game crashed

@Deleted User did you do anything specific?

also what version

14:11

we had quite alot changes in the rc

14:12

newest

15.0.5, right?

yes

14:14

i was copying a lot ideas from old map , saved as new map , go load other map put the all copied ideas in other map and while i was deleting parts that i didn't need it freezed and then quit as always happened in old version ...

14:15

i tryed repeat it but it didn't crashed

[ddnet/ddnet] Issue closed: #3050 Go through ubsan and clean up

14:35

[ddnet/ddnet] Issue reopened: #3050 Go through ubsan and clean up

what about generating automated crash reports in rlsd clients?

dbg symbols are quite huge

I guess even a non debug crash report can be more useful than none idk how exactly that should work without affecting performance

14:37

maybe not even slap a debugger on it

14:38

just track a few function calls and on crash store a backtrace

14:38

or are crashes too rare? I cant tell.

well the best would be cutting the debug symbols after building so we can keep a client with symbols

14:38

and then just stack trace with memory addresses

yeah, works fine with splitdwarf

but does that then require running it inside a debugger?

i think it just sets a hardware expection or smth

14:40

the os knows that the client crashes too without any debugger

i run most of my software w symbols and in debug mode and it works fine but i do not think we should ship something like thatr

14:40

but can the os generate a backtrace?

i dont have too much details on that tbh, but i know that many games do this, and i doubt they trade performance for it

14:41

its like with exceptions

might aswell optionally sent them back to ddnet automatically

it already be a huge win if its saved in some error log

yes

14:43

but i dont know if crashes are even a problem i cant remeber the last crash i had w ddnet

well it always depends

14:44

if you just joining the server and play its most likely that u wont get any crash, bcs thats what the devs found most easily

on what

but editor and stuff is harder for sure

i use the editor tho

14:44

and demos

14:44

and the console

14:45

and the remote console

ok xD, well we had crashes tho

so im just lucky for 5 yrs straight?

yes

lol nice

"is:issue is:open crash "

14:46

;D

i actually lost more progress on windows crashing than ddnet crashing

14:47

yikes 12 open 66 closed

14:47

the current reported one might be #1452

14:48

ddnet bot u here?

14:48

wait is it a discord exclusive feature

14:48

omg

14:48

https://github.com/ddnet/ddnet/issues/1452

Crash when copying multiple layers to another map · Issue #1452 · d...

As reported here: https://forum.ddnet.tw/viewtopic.php?t=6747 If u trying copy all game tiles (swither, tele,game itc ) with hold shift, then load map , then if u trying paste switch and speedup,cl...

It's really hard for us to generate any meaningful backtrace in our application after it crashed

hm why?

I'd guess the OSs have support for that

14:58

linux has coredumps, windows probably also has something like that

Well the OS killed your application for a reason, at that point even if you catch the signal, there is no guarantee that anything makes sense

14:59

You could have completely overwritten the stack e.g.

yea, but that's how we usually debug ^^

15:00

the debugger also just looks at the stack (that might have been overwritten already)

Well we get to run it a couple times, we get to try different tools, it's acceptable when the crash is happening on my machine

15:01

On iOS and Android there are many solutions that monitor the call stack by sampling it, it does cost a bit, but you get real nice crash reports

oh, interesting

15:02

got a name/link?

Is the server send plain text as system messages and not events?

yes

When will it start sending events? (edited)

Mh, I can't seem to remember where I saw it, must have been a talk on debugging but the guy had access to the call stack at 150hz for the last 5 seconds

15:05

"many solutions" might have been exaggerating a bit, I'm not really sure how common it is actually

15:08

Oh I have an idea, how about stealing crashpad from chromium?

that might work

15:08

should be opt in though, might contain personal data

We just capture it and ask the user if he wants to send it?

ye

15:09

that's opt-in in my view

Or maybe we capture it and just save it in a dir, so they can provide it with their report?

15:09

I guess the first one would result in more people reporting though

that would also work

Heh, if you sign your application with a fancy expensive certificate, you can gain access to the error reports sent to microsoft

Or you can use Crashpad

15:36

though it will require using google toolchain which is a bit too complex

where can I find code that renders this top right list ?

SPOILER

SPOILER_message.txt

2.32 KB

15:50

All launch crash from start

SPOILER

SPOILER_message.txt

28.23 KB

15:55

killmessages

what Graphics()->MapScreen() does?

translate map coordinates into screen coordinates

16:11

(or the other way around, not sure)

what's the reason for ddnet work a lot worse with significant fps drop and freezes on connecting to server when playing right from the compiled version from source code? Is that something about preprocessors?

17:06

I guess something heavy is not compiled for the public version?

debug version?

17:07

or what do u mean

yes

bcs in debug versions, there are no optimizations and sometimes the compiler's libs have additional things that helps for debugging but slow the execution down

I just worried if that's my code causes long freeze before joining the server

well switch to releasewithdebuginfo/release and you'll know

@Soreu I finished replacing all textures with SVGs on https://zillywoods.github.io/ZillyWoods-web/ it scales so well on my 4k screen <3 The grass looks a little bit different maybe ill change it. Could you send a screen how it looks like on ur broken end now? :/

ZillyWoods Client

Custom client for Teeworlds version 0.7+ coded by ChillerDragon

I'd like to remind you that it is broken for me because of the code I wrote

ChillerDragon 144hz > 4k screen

@Soreu fix

will do

did u upload to ur database?

...someday justatest

i want to have svg :D:D

He's replacing single tiles from grass_main

atleast smth

because I did cut the mapres in pieces to use only the needed ones

keep it up chillerdragon, looks really good

https://www.github.com/TeeworldsDB/images/tree/master/realistic%2Ftransparent

TeeworldsDB/images

A collection of free to use and edit images for other art work. No actual game skins. - TeeworldsDB/images

yeah the single tiles, but i'll wait for the full mapres xd

there they are somehow

17:39

I don't even understand how he pulled part of different repo into the website

17:40

but I barely understand GitHub anyway xd

submodule magic i am a git wizard

17:47

yes Jupstar ofc everythin in the repo and doodads are almost finished but keep in mind i wont produce tilesets from last decade but 0.7 standards :D

17:47

Ryozuki: nah 60hz the screen is older than time it self and actually affordable xd

17:48

oh wait u did a compare

17:48

ok fair guess it depends i have more hz but less pixels on my laptop and i prefer my 60hz big 4k screen

chillerdragon what u mean with tilesets from last decade?

17:49

can 0.7 client set single tiles or what xd

teeworlds uses a new incompatible mapres format

17:50

to the one that ddnet uses

wtf, how that

im all for backwards compability but sometimes u have to move forwards :)

but pls explain

how did u not know?

since when?

0.7 rls

but i used 0.7 tilesets

17:51

they work fine

well "work"

17:51

if u embedd them

i honestly dunno what u mean

sec

17:52

https://raw.githubusercontent.com/teeworlds/teeworlds/master/datasrc/mapres/grass_main.png https://raw.githubusercontent.com/ddnet/ddnet/master/data/mapres/grass_main.png

its a normal png

if u place shadow in ddnet client and open it in vanilla client its gone

17:53

do u know what i mean?

well that not a real problem, if we have svg with shadows, ddnet has shadows again

thats the thing

its just missing in the png

17:53

or opposite

i wont make shadows :p

17:53

i will make 0.7 tiles

yeah but we dont need to name it grass_main

17:54

we could call it grass_4k

well ok fine

;D

yes then sure no problemo

17:54

i thought more like replacing the standard tiles

well we can build uppon your tileset

ill pay a lawayer to craft me a 0.7+ license

17:55

xd

BSD license.. yes ;D

suggestion: add more emotes like the once from twitch.tv to the discord group? like widepeepoHappy, Pogu , monkaW etc etc...? - https://betterttv.com/emotes/top (edited)

BetterTTV - Top Emotes

The top shared emotes across all of BetterTTV.

add emotes to chat again too

add emotes to chat

not this again xd

someone allready suggested it

? (edited)

even implemented

18:41

not these emotes tho

spam more pls Aoe xd

18:42

wait are we talking about tw chat or discord emotes?

discord emotes, what do you mean spam ?

ohh

18:42

discord xd

why was the in-game chat emotes removed

thought ingame

18:42

bcs they are distracting and annoying xd

it was funny

we have tee emotes

18:43

tee emotes best

i miss

in chat :(

they were so annoying

2

1

18:43

no they werent

yes

why werent they just made opt-in

18:44

after all the code was added to support them

18:44

just as easy to add a setting

because then u see the emote names instead of the emote kek

ye so

thats even more spam

18:45

even more annoying

ye the solution is to just add back emotes

:antijao:

18:45

which is more annoying

the solution is to remove them

but people can still use the version with emotes

but they wont have new features

and still can type :smile: into the chat with any vr

18:45

idc

F4 emotes in chat sux

i still use ddnet 10

why would they use it

to see emotes

no one else sees emotes anyways

18:46

so ur like talking with urself

same as using a custom skin that none else has

18:46

some ppl do it

KENZO LIE

(tbh i agree its totally useless feature lol)

but using a custom u see all the time urself

18:47

custom tee*

ye and u can see ur own emotes

but no one talks with himself

18:47

no one that’s mentally stable

im never used emoji version but im writing emote names ingame anyway

18:48

xd

its like when people who dont have nitro say :kek: on discord

they cant see the emote but who cares xd

u see people saying kek

in ddnet?

yes

lol

18:49

Much

and :feelsamazingman:

lol no

lol yes

18:49

xd

ive literally seen it

18:50

wym

18:50

also the :danAmazing:

yes ppl say kek

no they never showed up in tw chat

18:50

but ppl still say it with :: anyway

dead

me and Suxep even had danAmazing binds xd

norway u

1

example: *** 'Konsti' called vote to kick 'jao' (lol) *** 'jao' has left the game (You have been banned for 5 minutes (Banned by vote)) Konsti: :danAmazing: nameless tee: :danAmazing:

2

xddd

18:52

Fucked up

:danAmazing:

to minimize spam on ddnet just add chat filter on server side

18:53

its about emotes

F4

client side obivously

They would ban me from saying n word

people who dont want to see spam can choose to filter it out

18:53

the default should b to allow all

18:53

same could apply for 18+ language i guess

nword= noby ofc

u

im about emoji spamming

noone really spams emojis

18:53

just says them someitmes xd

trolls can

no

18:54

ppl spam emojis to me

trolls cna spam non emojis

trolls can

everyone has a different defn of 'spam'

18:55

trying to filter it serverside is

18:55

i think its literally impossible beyond simple ratelimiting

oh also ddnet should be less restrictive on the automute ratelimit

they should just let people setup their own chat filter in client

18:55

and +1

i always get automuted just cus i type fast

press return less often helps but

stop typing short messages

lmfao

i cant not type short messages

18:55

typing long messages is boring

i have the same problem because i type rly fast

18:55

even when i type long messages

@louis <:nouis:745612528834445313>

or when i correct messages that i mistype

18:56

it sometimes automutes me and i have to a: buse

18:56

and unmute myself

a: buse

i swear im not spamming tho

18:56

if u send a short sentence every couple seconds like im doing rn

18:56

it can mute u

just do like on noby server, u cant spam the same messages

yes

18:59

i would make a PR for ratelimit

18:59

if they wanted it

18:59

19:00

i use this instead of auto mute and i know u can just spam and change the last character but it still helps alot xd

19:00

itts just set to only let u say the same thing 3x/min

19:00

but the limit and time could be anything

then you can add emotes feelsamazingman

i mean

19:02

a 64/64 ddnet ddrace server might have higher standards of chat moderation than my smaller fng server .. xd

ig spam protect should be or serverside only or clientside only

19:03

and last one would be better

i think the only serverside spam protection should be simple rate limit

do you know clientside

19:04

eh, @Deleted User can you help with that?

19:04

or you are render boy (edited)

19:04

maybe ryo knows

͍

tldr?

add back emotes

client side spamprotection is useless

19:22

when the client is open source xd

chat spam is rare a problem tho

client side spam protection = add_foe

or do it like the browser console and show the amount of occurences using a number

19:25

@Deleted User why? its just :-: filter (edited)

so u want a chat filter, mh, that might be unwated by some servers

19:29

or do u want chat filter like removing bad words

unwated by some servers

uhm why?

mods*

19:32

well its a forced limit

19:33

so do u want a rate limit, or a chat filter?

19:33

im not sure rn

uh oh

19:34

maybe add new flag/netobj that will be sended to new ddnet users with rate limit count

19:34

or dk

19:35

add to existing (edited)

well u can always create an issue or pr, maybe someone will help you

okey (edited)

heinrich5991

[ddnet/ddnet] Pull request opened: #3054 Remove CRC in all possible places

Some places remain due to net/file format compatibility.

[ddnet:15.1-devel] 1 new commit

e17a2b1 Update 15.1 release date - def-

@deen release on tournament day?

22:00

Yeah

[ddnet/ddnet] Pull request closed: #2975 Version 15.1

22:01

507f7f6 Version 15.1 - def- ef9c85c Update 15.1 release date - def-

22:03

[ddnet-web:pr-15.1] 1 new commit

c790d78 Version 15.1 - def-

what could go wrong?

3

@deen u coulkd also just wait after the tournament

22:18

but idc

@heinrich5991 @Learath2 i recorded a demo with 10.6.2 client and im trying to send it to people to view with new client

22:18

and noone can open it..

I feel like nothing more was reported since last RC, a few days passed

they keep getting map errors like map differes from server

22:18

and deleting from downloadedmaps changes the error to a different error but doesnt fix

22:18

and afaik downloading the version i used doesnt work either

22:18

what can i do to show them?

use new client

...

22:19

it doesnt open

i mean u the recorder xd

yes but i already recoded it

22:19

i cant re-record

@deen ok, then lets hope it works xd

idk if theyll ever come back

mhh ok

so

22:20

how can i show them the demo

22:20

short of just using custom screen recorder

stop using old veresion

fix the demo I guess

how

22:21

acqwerty no u

22:21

it doesnt matter what version i use, demo is already created

then you should have switched before recording

but i didnt

22:21

so thats why i aske for help

too bad

u

fix the demo binary xd

how

hex editor

ye but what do i edit

hm

22:21

check what crc/sha256 is included

22:22

try to find the correct map

demo i want

22:23

random demo recorded with new vrs

22:23

er

22:23

dk why its different actually hmm

22:23

22:26

if they put the version of fng.map that was on server 2 weeks ago in their downloadedmaps then it will work? xd

I think it'll work if they put that map in their downloadedmaps

@Skeith

[ddnet/ddnet-web] Pull request closed: #103 Version 15.1

22:28

[ddnet-web:master] 2 new commits

c790d78 Version 15.1 - def- b43bd3a Merge pull request #103 from ddnet/pr-15.1 - def-

@noby you'll have to provide them the map

ok ty ill ask if someone has it

22:30

is there no way to extract map from demo?

22:31

ive seen it done before afaik

well, I think your demo has the wrong map

22:31

causing this behavior

22:31

are you currently using an up-to-date client?

1

no

22:33

i recorded with 10.6 client

22:33

im trying to show it to ppl

yes, I understand, but maybe you should start using an up-to-date client ^^

22:34

(for the future)

xd

f3

but

22:34

the demo is of someone that rarely joins server

I realize

ok

I thought I helped you fix it

i will look in to finding that exact map file

22:35

ty

The "Start server" button in main menu is supposed to start server from the directory of client (Steam ver) and look for myServerconfig.bat in same directory, right?

and @heinrich5991 do u wanna go thru the process of bisecting again sometime in the nearish future to see why old client still gives me better/more stable fps

my myServerconfig.bat points to exec "server/config.bat" and it doesn't get executed no matter where I put the first filter

we did find the most significant cause and u fixed it

the first file*

but still it is lower

we can try again

22:48

make sure to disable hidpi on all versions while testing

ya highdpi was the most signifiant problem from last time

yea, but only disabled on latest

22:49

so you need to manually disable it everywhere

22:49

to get comparable results

i will try to do this testing tomorrow, been awake since yday morning so abit tired now

22:50

ill try to compare based on cpu usage or sth

Didn't u say with refresh rate 0 it's when faster

22:57

Snowby

but thats bad bcause

22:57

it can get up to 2500+ fps with uncapped rate on both

22:57

but it runs the computer very hard and fans spin at max the whle time

22:58

i remembr u suggested switching the rates

But I already told u old client fps is different

and i think it did help

22:58

yea

22:58

the way it estimates fps is different

22:58

but the way it makes fans spin seems more intensive on new client

Also with cl refresh rate u won't get more fps anymore

22:59

In old client u could get 600fps with 480 cl refresh

stable fps above 400-500 is fine, what i remember seeing is that it was less stable and appeared to use more cpu but i havent done a controlled test to back it up

Hey guys :D Can anyone help me?

Just ask xd

I patched the latest ddnet client and my antiping bar disappeared

23:01

intentional removal

[ddnet/ddnet] Pull request opened: #3055 Add back Australia...

23:01

[ddnet:15.1-devel] 1 new commit

1a9abee Add back Australia... - def-

you can access it via the console @Protokol T'nP

why would you need the antiping bar

well fixing gfx_refresh_rate helps the fan spin issue a bit but it still does spin

@plsplsplslol Its easier for me there

@Protokol T'nP rather, what do you use it for?

@noby I'd say it was https://github.com/ddnet/ddnet/pull/1055

Keep track of lost frames and update time by Jupeyy · Pull Request ...

The problem a friend was occuring was, that on a fresh Windows install settings like cl_refresh_rate are set to 480. Since the render calls take longer time than no render calls, it happened that c...

23:07

That made the refresh rates more useful

23:07

During the time I fixed some stuff tho

does it make the fps counter more accurate in realtime

23:08

idk if it works like this but it might be interesting to have an option to display "old" fps count at least for debugging

wouldn't help with bisecting now

the fixes for wait time make sense and i dont see why anything in that commit by itself would cause higher cpu load but maybe im missing smoething

We had a different fps counter once but deen didn't like it bcs it litteraly updated every frame

the one on newer ddnet seems to fluctuate alot between every frame

23:11

i remember mentioning it and having it explained as the new one being more accurate

Not higher CPU load, but probs less frames bcs CPU time is more accurate

i mean i dont notice the difference between 400 and 600 fps tbh but i do notice when the fans start spinning when they otherwise wouldnt have

23:12

before they fixed highdpi it was this same problem but 5x worse

23:12

extreme fan spinning always

an fps or frametime graph would probably be nice to have

now its just moderate

We have

@Learath2 I think that already exists xd

Ctrl shift d and then g

old ddnet the fps looks more stable in the counter

23:14

15.0.4

23:14

different servers but same map

To me new looks more stable

but lower fps and it spisn the fans

If the second one is the newer one it does look muuuch more stable

Then lower your stupid fans xdd

23:17

But pls use different fps measure anyway

23:17

Steam fps or smth

on mac you don't get to control much about your fans

The vanilla one was bit wrong

if i lower the fans it will just kill the cpu so no ty

23:18

and its not rly about fps i think

Under 70c is safe xd

I don't get why you'd get more cpu usage though

thats what the bisect would be for i guess

The newer renderer does more on the gpu, the cpu should be more relaxed

i have a laptop with better cpu and barely any gpu

I first want to see the real fps from outside fps program

how do i use an external program to monitor fps

Steam e.g. had an overlay

23:19

Dunno under macos tbh

23:20

There will be some program xd

7dc33ab Add back Australia... - def- 1356fb3 Merge #3055 - bors[bot]

23:21