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 2023-07-28 00:00:00Z and 2023-07-29 00:00:00Z

how's this look for the ui? the text disappears & the list view occupies that space once again when finished

Image attachment

00:37

looks weird how it is but all i did was add that text block & pad for it lol the weird gap above was there before

Ewan

seems there's still a problem. loading the textures all at once creates a tremendous lag spike w/ my 2000 skins, i guess this should be expected but this is exactly the sort of thing i was trying to avoid. but instead of it happening in the loading screen, now it can potentially happen while in the game

How huge is the lag in milliseconds? Xd

wdym

01:02

we are talking like

01:02

15+ seconds on each of the threads

01:02

i have added a 10ms delay on the other thread for testing and it's pretty smooth, as much as i hate this it might have to be

Wtf

yea it just takes a sec for skins to fill in but with 2k that should be expected & with a normal user's rational amount of skins it'll be completely unproblematic

01:03

the average amount of time it takes the average user's pc to load the average amount of skins is still probably less than the time it would take to even join the game

01:03

with this 10ms

Mh i dunno

01:04

I'd have expected multi threaded reading of 2k skins takes around let's say 200ms max

xddd

Ok let's say a second

01:05

But not 15 seconds

you know i just have 1 other thread for all the skins

01:05

not a pool

01:05

it's multithreaded in the sense that it's not on the main thread :D

Oh so there isn't really an advantage for client loading

my internet is shitting the bed

01:06

it takes that time off of the initial loading time

Just for downloaded skins

no

01:06

it's for downloaded skins

So u download 2k skins at once? Xd

01:07

That explains your favorite skins list xd

the other thread reads the file data, decompresses the png, and prepares the CImageInfo for the main thread on a skin-by-skin basis, in the order they are returned by the OS's list directory mechanism

01:07

yea

01:07

oh my favorite skins list is all the worst ones

01:07

they are so funny

Ewan

the other thread reads the file data, decompresses the png, and prepares the CImageInfo for the main thread on a skin-by-skin basis, in the order they are returned by the OS's list directory mechanism

I see sounds good

yea and then whenever the main thread gets a chance it uploads this CImageInfo to the gpu basically

01:08

it's a bit more complex than just that but most of the processing is done on the other thread

01:08

like for example the process of splitting up the image into parts is on the main thread still

01:08

but pretty inconsequential

But the process takes 15 seconds.. but it creates no lag on the main thread, does it?

not 15 seconds total, 15 seconds per turn of hogging the mutex

01:09

which usually happens 3 or 4 times lol

Ok I would sleep lmao. I can't think rn

the other thread's and the main thread's loops don't line up

01:10

i give them a chance to & eliminate lag by adding the delay on the other thread

01:10

or at least a chance for the main thread's loop to try and lock the mutex so it can do its thing often enough

01:11

i don't think i'm ready for review yet but i would appreciate some from you, specifically as it pertains to the graphics code because i'm kinda tapping into it an uncomfortable amount

I'm not quite sure what u even trying with some second loop etc. xd The idea about skin loading sounds correct. But why don't u simply use the job engine for that. It is also a thread pool

01:20

That would speed the loading up

I'm not quite sure what u even trying with some second loop etc. xd

what do you mean? it has to iterate on the files

01:20

and yeah i would have used the job engine if i knew about it

01:20

it's quite late for that

01:21

and having a bunch of threads doing disk work at once might speed things up but it would chew up the disk unnecessarily

01:21

a low and steady load like this is probably better

01:21

especially since it will matter pretty much 0 for most people

Ewan

a low and steady load like this is probably better

Mh on HDDs maybe

01:21

But I'm future human

01:21

I have ssd

Ewan

I'm not quite sure what u even trying with some second loop etc. xd

what do you mean? it has to iterate on the files

Yeah I guess I currently cannot imagine your implementation^^

01:22

So it's not even the main thread preparing the list of skins to load?

01:23

Or does it get the list and is like... Here we go

main thread gives it the criteria and other thread gives it the textures

01:26

p much

01:26

but it's not quite so simple

01:26

you'd have to see it

If we'd properly implement it, we'd need to use futures anyway

01:29

Blocking a while thread while reading the disk is probably not the best solution

01:29

Whole

yeah

[ddnet-maps:master] 1 new commit

3b5f6d0 A run_event - ddnet-maps

[ddnet/ddnet] Pull request opened: #6941 improve offset adaption for envelope zoom

Checklist

[x] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added coverage to integration test
[ ] Considered possible null pointers and out of bounds array indexing
[ ] Changed no physics that affect existing maps
[ ] Tested the change with [ASan+UBSan or valgrind's memcheck](https://github.com/ddnet/ddnet/#using-ad...

<Jupstar> @Ryozuki FFR. What would not have happened this week if you would have used Rust instead of whatever you used? xd

i use rust so idk

07:39

i have a theorical car exam in 1h xd

<Jupstar> wtf xd

07:40

<Jupstar> can u only drive with 21 in spain?

no u can with 18

07:41

i just didnt get it before

07:42

xd

<Jupstar> future human you are

07:42

<Jupstar> no need to learn a skill you dont need

07:42

<Jupstar> ;)

i never needed cuz barcelona metro is nice

<Jupstar> in germany cars are the holy grail

does anyone know if it's possible to code a program that searches for specific coordinates of a select player in a demo and points towards it with a time stamp? (edited)

<Jupstar> possible: i guess so. easy: probably not

Ewan

how's this look for the ui? the text disappears & the list view occupies that space once again when finished

Do the skins appear gradually or all at once when the loading is finished? If they appear gradually, the loading indicator should be drawn as a listbox item. For now just render "Loading...", I already got the loading indicator covered.

ws-client1

<Jupstar> possible: i guess so. easy: probably not

what would i need for that?

08:52

to be clear i dont know anything about programming

<Jupstar> well without coding skills it's not really doable. you need to understand how the demo renderer uses snapshots.. how the position is calculated etc

08:59

<Jupstar> it's not trivial enough to explain it in 5 minutes xd

Iza

to be clear i dont know anything about programming

Some minimum knowledge about coding is probably required. You could try to modify the client to just stop demo playback when a specific player reaches a position that you specified (for simplificy with a config variable and not via UI). Then you can play the demo on fastest speed and it will stop at the position you are looking for. (edited)

Robyt3

Some minimum knowledge about coding is probably required. You could try to modify the client to just stop demo playback when a specific player reaches a position that you specified (for simplificy with a config variable and not via UI). Then you can play the demo on fastest speed and it will stop at the position you are looking for. (edited)

that doesn't sound tooo complicated.

09:02

do i have to know just the basics of c++ for that or more advanced stuff?

Well, the basics of C++ should be enough, but you also need to find your way around in the DDNet source code

09:03

Seems like a good beginning project though

yes it does

09:04

do you recommend any specific guide or topic where I can start off?

https://www.learncpp.com/

The C++ Tutorial

alright thank you

Iza

does anyone know if it's possible to code a program that searches for specific coordinates of a select player in a demo and points towards it with a time stamp? (edited)

Might be a tad easier to extract it from the teehistorian files instead since they record positions directly too

Ewan

15+ seconds on each of the threads

I think theres a lot of people that get 10+ second lag from texture loading.

RUST FRIDAY

Learath2

Might be a tad easier to extract it from the teehistorian files instead since they record positions directly too

from my experience the teehistorian program does not playback the inputs very accurately

[ddnet/ddnet] Pull request opened: #6942 ResetServerGameSettings → ResetGameSettings, remove flag argument

The old code looked pretty weird. It checked for CFGFLAG_SERVER | CFGFLAG_GAME being set above, and then used the passed-in parameter to find the command. I think it'd be better to just look for CFGFLAG_GAME and then find the command using CFGFLAG_GAME.

Checklist

[ ] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related configuration options
[ ] Written a unit test (especially base/) or added covera...

Iza

from my experience the teehistorian program does not playback the inputs very accurately

Not the demo generated by teehistorian2demo, the actual recorded teehistorian file. They are 100% accurate as with demos btw

Learath2

Not the demo generated by teehistorian2demo, the actual recorded teehistorian file. They are 100% accurate as with demos btw

where's that located? i'll have to come back to that after I gained some basic knowledge of c++

exam done

10:24

i think i passed

<Jupstar> do you not get immidiate response?

10:24

<Jupstar> my exam took like 3 minutes and i directly got the result xd

10:24

<Jupstar> speedrun

Ewan

can we get cpp20 so i can do atomic<shared_ptr<T>> troll

troll

(edited)

you can do that before c++20 I believe

10:25

I think I do it in the logger

idk they said at 6 today

10:25

i did it on a tactile computer xd

Iza

where's that located? i'll have to come back to that after I gained some basic knowledge of c++

Currently not available but we can provide them for you to test, and/or your tool can be added to the teehistorian tools available to mods online

Iza

do i have to know just the basics of c++ for that or more advanced stuff?

this looks like a pretty complicated problem. as a start, try compiling ddnet. then maybe add a chat command to the server that just prints something

Ryozuki

i did it on a tactile computer xd

what's a tactile computer?

heinrich5991

what's a tactile computer?

touchpad

10:30

my bad

10:30

xd

10:30

it made sense in my spanish mind

10:30

catalan*

10:30

xd

Ryozuki

exam done

What sort of exam?

in spanish u say tactil

10:31

@Learath2 driver license

10:31

but the theorical one

10:31

not practical

Bizarre that you don’t get immediate feedback on a computerized exam. I think I got mine within like 5 minutes

heinrich5991

this looks like a pretty complicated problem. as a start, try compiling ddnet. then maybe add a chat command to the server that just prints something

looks like i have to start at the very beginning then, that'll be fun

@Robyt3 Shouldnt Shift + Right click be Shift + right click or even Shift + right-click?

marmare_314

@Robyt3 Shouldnt Shift + Right click be Shift + right click or even Shift + right-click?

right-click would be the most correct one, it should also be used for the tangent handles then

github-merge-queue[bot]

[ddnet:master] 2 new commits

c695fd7 Remove inverted round rectangle corner drawing - Robyt3 564a595 Merge pull request #6940 from Robyt3/Graphics-Rect-Remove-Inverted-Corners - heinrich5991

10:44

[ddnet/ddnet] Pull request closed: #6940 Remove inverted round rectangle corner drawing

github-merge-queue[bot]

[ddnet:master] 2 new commits

62ee646 allow multiple selection in envelope editor - Marmare314 8b209f0 Merge pull request #6872 from Marmare314/envelope_contextmenu - Robyt3

11:37

[ddnet/ddnet] Pull request closed: #6872 Multiple selection in envelope editor

Learath2

Bizarre that you don’t get immediate feedback on a computerized exam. I think I got mine within like 5 minutes

In those situations, I like to think that someone has to take your exam, print it on a piece of paper, evaluate it, digitize it back again, then it gets shown to you

Cipy29

In those situations, I like to think that someone has to take your exam, print it on a piece of paper, evaluate it, digitize it back again, then it gets shown to you

pepeW

github-merge-queue[bot]

[ddnet:master] 2 new commits

7069d74 Check for errors when reading and decompressing datafile data - Robyt3 8b0f317 Merge pull request #6939 from Robyt3/Datafile-Data-Decompress-Check - heinrich5991

11:58

[ddnet/ddnet] Pull request closed: #6939 Check for errors when reading and decompressing datafile data

Learath2

Bizarre that you don’t get immediate feedback on a computerized exam. I think I got mine within like 5 minutes

i wonder too, maybe its in case someone is detected cheating, to have a grace time

[ddnet-libs:master] 1 new commit

5e607ed Add libssp-0.dll - def-

12:20

[ddnet-libs:master] 1 new commit

59d64db Add libssp-0.dll - def-

github-merge-queue[bot]

[ddnet:master] 2 new commits

f3353e8 improve offset adaption for envelope zoom - Marmare314 c237959 Merge pull request #6941 from Marmare314/zoom-adaption - Robyt3

12:36

[ddnet/ddnet] Pull request closed: #6941 improve offset adaption for envelope zoom

kinda sad that rust decided to let Drop take self as reference partial moving is really cool, and not depending on Default or Option can make sense too if i implement destroy (self) it would work, but ofc requires explicit call :/

What would you have drop take instead? It makes sense to me that drop is moved to as the final "drain"

i mean the Drop trait

13:33

i'd like it to take self as value

13:33

moved value

13:33

and it order to not drop again it has to do some compiler specific stuff xd

Ah. It can't tho can it? I guess it could be moved to then moved back, but how is that better than a mutable reference?

13:34

(As in it can't take self)

13:35

Or hm, maybe it could work

if i want to reuse resources i might want to move a member out

fn destroy(self) {
  self.a.destroy();
  self.b.destroy();
}

where a and b also implement destroy(self)

Why can't you do it with &mut self?

if a and b want an explicit call i'd need a and b be inside Options

13:37

bcs i cannot move them

When self drops, won't self.a and self.b also get dropped if a and b are owned by self?

yes, but what if a and b carry some resource i want to manually free at some point it would require a Option

Jupstar ✪

kinda sad that rust decided to let Drop take self as reference partial moving is really cool, and not depending on Default or Option can make sense too if i implement destroy (self) it would work, but ofc requires explicit call :/

legacy decision. nowadays you'd probably give an option to take ManuallyDrop instead

13:38

perhaps there's an RFC for it

Jupstar ✪

yes, but what if a and b carry some resource i want to manually free at some point it would require a Option

And how would a moved value help with that? You'd still need the option when you free the resource in a and b, no?

my problem is that i cannot move a and b out of a reference

13:40

&mut self

13:40

is Drop trait

I'm just curious why one might need to move them

bcs then the type does not require to implement Default

because you e.g. want to call a hypothetical File::close which takes ownership of the File

13:43

you can wrap the thing into a ManuallyDrop if you want to @Jupstar ✪

13:43

requires some unsafe in handling though

heinrich5991

you can wrap the thing into a ManuallyDrop if you want to @Jupstar ✪

yeah, sadly i use it quite regulary, so i'm not sure about that

heinrich5991

because you e.g. want to call a hypothetical File::close which takes ownership of the File

I see. Is this a common thing to have cleanup functions take final ownership?

Jupstar ✪

yeah, sadly i use it quite regulary, so i'm not sure about that

no problem, ManuallyDrop derefs to its inner object

Learath2

I see. Is this a common thing to have cleanup functions take final ownership?

no, because of the problem @Jupstar ✪ mentions

heinrich5991

no problem, ManuallyDrop derefs to its inner object

but i'd have to store the wrapper in an object

Jupstar ✪

but i'd have to store the wrapper in an object

can you give a more complete example

13:44

?

Fwiw, I'd have thought drop would indeed take final ownership. Curious decision that it doesnt

the naive way to implement it doesn't work

github-merge-queue[bot]

[ddnet:master] 2 new commits

d964a3d handle big value ranges on envelope zoom reset - Marmare314 b33c9ca Merge pull request #6934 from Marmare314/zoom-reset-big-values - Robyt3

13:47

[ddnet/ddnet] Issue closed: #6926 Envelope editor default zoom can be larger than manual zoom

13:47

[ddnet/ddnet] Pull request closed: #6934 handle big value ranges on envelope zoom reset

Learath2

Fwiw, I'd have thought drop would indeed take final ownership. Curious decision that it doesnt

because then, you'd have a destructor call at the end of the destructor

heinrich5991

can you give a more complete example

e.g. for resource tracking i use a Rc inside an index struct to make sure indices are cleanup properly it takes the index by moved value (so its defs not used anymore) and the Rc will 100% decrease in use count

13:47

stuff like that mostly

fn drop(x: Struct) {
    // oops, implicit call to drop inserted
}

Jupstar ✪

e.g. for resource tracking i use a Rc inside an index struct to make sure indices are cleanup properly it takes the index by moved value (so its defs not used anymore) and the Rc will 100% decrease in use count

I guess wrapping these in an option would sound the best

heinrich5991

because then, you'd have a destructor call at the end of the destructor

Hm, and I guess it's rather ugly to have no implicit drops in drop

i dislike it

13:49

i like to think about programming as "it exists or you didnt want to create it in first place"

do you only want to drop it in the destructor?

13:49

wait

13:49

the dropping of the Rcs already works

13:50

do you have some sort of manual cleanup function?

yes

Learath2

Hm, and I guess it's rather ugly to have no implicit drops in drop

that's probably halting-problem complex. what do you do if self is moved somewhere else. what happens if another object of the same type is constructed in the destructor, conditionally? etc.

13:51

can you explain why there's a need for a manual cleanup function?

13:51

what happens if the destructor is called before the manual cleanup? resource leakage?

rn it assumes resource leakage and panics

heinrich5991

that's probably halting-problem complex. what do you do if self is moved somewhere else. what happens if another object of the same type is constructed in the destructor, conditionally? etc.

Not allowed to move anymore drop is the final destination, no moving self out of drop

that's bad; destructors can be called implicitly, e.g. when panicking

Jupstar ✪

rn it assumes resource leakage and panics

re this

13:52

e.g. when unwinding

its a fatal error

panic while panicking kills the process IIRC

so abort

ah

Learath2

Not allowed to move anymore drop is the final destination, no moving self out of drop

hmm. and struct fields only get dropped if they're not moved out. I guess this could work?

Atleast that's what I'd explore if I were designing the language, but I do understand why the mutable borrow is what they went for, much saner

I remember the discussions around this

Learath2

Atleast that's what I'd explore if I were designing the language, but I do understand why the mutable borrow is what they went for, much saner

I think this might have been something that wasn't proposed

13:55

can you give an example of such a cleanup function?

Learath2

Atleast that's what I'd explore if I were designing the language, but I do understand why the mutable borrow is what they went for, much saner

well sane vs convenience IMHO you also do fn new() -> Self

13:55

so it kinda would make sense to do the opposite of this

13:55

fn destroy(self)

Not being able to have uninitialized stuff is pretty hard for me too. I wish I could have uninit stuff without maybeuninit and unsafe everywhere

I thought we had a solution for you the other day?

13:56

you can have locally uninitialized variables without unsafe, as long as you only use them in paths after you initialize them

Those resources will only ever be an Option while the struct is being constructed/deconstructed, so it's rather sadge to have unwrap everywhere

@Learath2 you want them for stack allocations i assume?

heinrich5991

I thought we had a solution for you the other day?

Yeah there usually is a solution, it's just a bit hard for me to find them until I get more experience

Learath2

Those resources will only ever be an Option while the struct is being constructed/deconstructed, so it's rather sadge to have unwrap everywhere

if you want to have it for members, I think there are also solutions for that

Jupstar ✪

@Learath2 you want them for stack allocations i assume?

For any allocation, sometimes initializing a struct in one go is rather hard

can you give an example?

but if its not on the stack u can also use Vec

13:59

or other heap stuff

It's usually that your program structure includes loops though which when/if you break them makes it simple again

heinrich5991

can you give an example?

E.g. the context I had a week or so ago, it had a loop so to initialize one struct I had to have already initialized the other and viceversa

14:00

Adding an extra layer of indirection and an init function fixed it but I still had to think more

maybe u can solve the problem without this kind of "self"-depenency

Self referential stuff just have to be eradicated in rust, atleast at my experience level

Jupstar ✪

maybe u can solve the problem without this kind of "self"-depenency

Yeah, it's just not how I'm used to thinking about my programs is all

well for performance reasons it can make sense.. but then i'd maybe use unsafe anyway and point to whatever i reference

14:02

so maybe it does make sense xd

It's just habit. Having convenient context pointers around just makes it easier for me, but it does lead to worse hierarchy

yeah... i guess rust simply isnt the best option for prototyping xD

I haven't written large applications from ground up in rust yet

14:04

the masterserver was very smooth though

i have to say the ddnet stuff i wrote was mostly smooth too

Like CComponent having a pointer to it's parent and the parent having a pointer to it. Technically you can structure that in another way with a message bus and stuff, but again it's just not what I'm used to. Hard to teach an old dog new tricks

Learath2

Like CComponent having a pointer to it's parent and the parent having a pointer to it. Technically you can structure that in another way with a message bus and stuff, but again it's just not what I'm used to. Hard to teach an old dog new tricks

that for example i gave up on

14:05

our Components are way too mighty

14:05

it makes it basically impossible with borrow checking, Rc, RefCell etc to stay sane

Jupstar ✪

kinda sad that rust decided to let Drop take self as reference partial moving is really cool, and not depending on Default or Option can make sense too if i implement destroy (self) it would work, but ofc requires explicit call :/

beware with drop and unsafe, panicking is often a cause of double frees in unsafe code

Ryozuki

beware with drop and unsafe, panicking is often a cause of double frees in unsafe code

im not using any unsafe xD

ah xd

14:06

i didnt read convo

14:06

tldr?

it doesnt exist

14:06

xd

14:07

i have to live with a wrapper or manual calls for now

14:07

but probably manual calls for now, since they at least take additional params

14:07

so i can save some Rcs

14:07

but sadly not forced ;~;

14:07

one can forget to call it

14:08

(i like non runtime checks)

what you want are linear types, I think

14:08

rust doesn't have those (unfortunately?)

Jupstar ✪

it makes it basically impossible with borrow checking, Rc, RefCell etc to stay sane

Hm, I think I would just have the client have the only strong pointer and make the rest weak

heinrich5991

what you want are linear types, I think

yeah maybe i what i want isn't easy to have but since i cant have what i want in Drop, i also don't want to introduce Rc and RefCell (to allow automatical drops) so i'm kinda in a dilemma

It's not exactly pleasant using weaks but I can't think of any other sane hierarchy that uses direct references, else you can just bite the bullet and have message channels everywhere

heinrich5991

can you give an example of such a cleanup function?

@Jupstar ✪

Learath2

Hm, I think I would just have the client have the only strong pointer and make the rest weak

yeah dunno, but u also need RefCell if u want mutibility

Learath2

It's not exactly pleasant using weaks but I can't think of any other sane hierarchy that uses direct references, else you can just bite the bullet and have message channels everywhere

channels are bad though, stuff would not happen immediately, but "in the next tick" or so

and that's at runtime again xD

Yeah, I'm not sure if you can avoid interior mutability here

heinrich5991

@Jupstar ✪

well i use the index, send it to the vulkan backend, where it frees some memory the fact that i have to use indices/ids/whatever is already ofc not nice.. but probably inevitable. So at least make sure a resource must be explicitly free'd Also yes i could rely on Option or Default etc. my point is really not that there aren't other solutions. But that it would be this tiny bit more elegant imo

Anyway, it’s fairly hard for me to think about the hierarchy in any larger program, especially since I’m not a very frequent user of C++ smart pointers either

Jupstar ✪

well i use the index, send it to the vulkan backend, where it frees some memory the fact that i have to use indices/ids/whatever is already ofc not nice.. but probably inevitable. So at least make sure a resource must be explicitly free'd Also yes i could rely on Option or Default etc. my point is really not that there aren't other solutions. But that it would be this tiny bit more elegant imo

can you have more than one vulkan backend?

heinrich5991

can you have more than one vulkan backend?

i knew that would be comming xD

14:15

perhaps you could impl drop in there and just throw it at the global object

well i kinda like to free what i allocate and i dont want to put the whole backend in a global or Rc

14:16

well tbf i dont have a very strong opinion on this

i think top down one way is usuallyt prefered

I just fear extensive Rc + RefCell usage

doesnt it allow better paralelization

@Jupstar ✪ do you have the code public somewhere? I’m kinda curious now

Jupstar ✪

I just fear extensive Rc + RefCell usage

global variable would work there, no?

children being in a "black box" and not knowing much

heinrich5991

global variable would work there, no?

He also fears globals

heinrich5991

global variable would work there, no?

but that would introduce quite a bit of unsafe, wouldnt it

I would think you could contain it to a couple of functions (edited)

Learath2

He also fears globals

hate is smth else than fear

14:18

see i really like to solve problems by passing stuff around when i see Demo()->isPaused() in our c++ code while i simply stop the timer of the demo renderer (bcs the timer is always passed and can simply be swapped between client timer and demo timer) it enforces some style

14:18

it has a clear style

can you give an API link to such a cleanup function?

14:18

of vulkan

Ryozuki

i think top down one way is usuallyt prefered

For hierarchy? Yes but I never really heard much in the way of arguments for it. It’s just much much easier to think about 1 way relationships all over, but then you’ll need something akin to a message bus, good for easy parallelism, but not the best for performance iirc

heinrich5991

can you give an API link to such a cleanup function?

well it takes a vulkan instance... that is the one that was to be active while u allocate memory from it, but let me find it

heinrich5991

can you give an API link to such a cleanup function?

https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/vkFreeMemory.html well it takes a device which u choose after the instance is created

14:20

device = GPU

Learath2

For hierarchy? Yes but I never really heard much in the way of arguments for it. It’s just much much easier to think about 1 way relationships all over, but then you’ll need something akin to a message bus, good for easy parallelism, but not the best for performance iirc

Or callbacks everywhere, but callbacks are pretty hard combined with async, so your easy parallelism benefits can be lost

[ddnet/ddnet] Pull request closed: #5398 implement tuning values for elasticity

14:22

[ddnet/ddnet] Issue closed: #5124 feature request: bouncy blocks

Btw @heinrich5991 be very careful with #6942, the code looks much cleaner which scares me

https://github.com/ddnet/ddnet/pull/6942

ResetServerGameSettings → ResetGameSettings, remove flag argument b...

The old code looked pretty weird. It checked for CFGFLAG_SERVER | CFGFLAG_GAME being set above, and then used the passed-in parameter to find the command. I think it'd be better to just look fo...

github-merge-queue[bot]

[ddnet:master] 2 new commits

e1ba77d implement tuning values for elasticity - AssassinTee 03fd1b4 Merge pull request #5398 from AssassinTee/feature/elasticity-tuning - heinrich5991

Learath2

Btw @heinrich5991 be very careful with #6942, the code looks much cleaner which scares me

much cleaner than what?

The previous one, it looked weird before now it doesn’t. When I see that I’m suspicious of some unintended change in behaviour

14:25

I thiiiink it’s fine though, still just a heads up

Did Heinrich write the previous code? (edited)

He may have, I don’t remember who implemented it

I did

I assume it was from the original cfgflag game pr

14:26

I was afraid to touch it too much

14:26

Macros scare me

Yeah, cleaning old code is spooky in ddnet, simplest of changes moving around if branches broke stuff in the past

I think his change makes sense though it looks good to me

Yeah, Robyt’s comment does look like a nice imorovement on it too btw

14:28

improvement*

True but it's already working

I'll respond to it

Learath2

@Jupstar ✪ do you have the code public somewhere? I’m kinda curious now

i can create a current snapshot but the code is still very messy and very prototype

Jupstar ✪

i can create a current snapshot but the code is still very messy and very prototype

If you don’t mind sharing I’d love to take a look

14:29

I’ll go back to op amps, cya’ll

bye

Tater

True but it's already working

Is it also working with the new changes? I confirmed that it doesn't crash

I did not test Heinrich pr

14:33

Logically it seems ok to me

14:34

Someone should test it though

@Tater how do I test that the config variables get applied correctly?

@Learath2 https://github.com/Jupeyy/dd-pg/tree/test2 please dont read the code too much xdd

GitHub - Jupeyy/dd-pg at test2

don't use or read this code. Contribute to Jupeyy/dd-pg development by creating an account on GitHub.

I set the client console to initialize with | CFGGAME_FLAG then you can just check them in F1. But you need a map to set them.

14:46

If you use sv_deepfly you can set pred margin to something high like 200 then just test if it's predicted correctly with dummy

14:46

I could test it in like 10 hours if you care

that'd be amazing

Ok

I'll add you as a reviewer

Also idk what this function is doing on the server side. If you're being extra careful you might want to make sure it doesn't have a behavior change there.

[ddnet/ddnet] Pull request closed: #6923 Add grid to envelope editor

@heinrich5991 please do the review for #6931. I dunno if it's important if we fail cmake if ssp is not found etc. i'm static linking elitist now xd

https://github.com/ddnet/ddnet/pull/6931

Fix build which broke in 3 different ways by def- · Pull Request #6...

src/engine/client/backend/opengl/backend_opengl.cpp:1599: undefined reference to `__stack_chk_fail' Still testing if this fixes the build. Edit: works

github-merge-queue[bot]

[ddnet:master] 2 new commits

babdbc4 add grid to envelope editor - Marmare314 88cf4c4 Merge pull request #6923 from Marmare314/envelope_grid - Robyt3

[ddnet/ddnet] Pull request closed: #6916 Minor improvements to Multi-View

15:24

github-merge-queue[bot]

[ddnet:master] 2 new commits

1c231a0 Minor improvements to Multi-View - Vy0x2 5b10547 Merge pull request #6916 from Vy0x2/pr-poc-mv-clean - Robyt3

15:29

[ddnet/ddnet] Pull request closed: #6931 Fix build which broke in 3 different ways

15:30

github-merge-queue[bot]

[ddnet:master] 5 new commits

d37b100 Fix linking wiht -fstack-protector-strong on MinGW - def- a7ae13d Fix glslangValidator.exe executable bit - def- 3497bfd glslangValidator -> glslang - def- 8149529 Add libssp-0.dll - def- b57cd2e Merge pull request #6931 from def-/pr-mingw-stack-protector - heinrich5991

@Robyt3 i rebased #6867 if youre in the mood for a review

https://github.com/ddnet/ddnet/pull/6867

Make point selection easier by Marmare314 · Pull Request #6867 · dd...

Draft to make quad point selection easier. To work properly this will need #6844. Since this change was inspired by blender the hitbox is currently quite big. If two points are within range the nea...

@Jupstar ✪ i passed the exam xd

epic

Robyt3

Do the skins appear gradually or all at once when the loading is finished? If they appear gradually, the loading indicator should be drawn as a listbox item. For now just render "Loading...", I already got the loading indicator covered.

Gradually

Ryozuki

@Jupstar ✪ i passed the exam xd

IDK what exam but congrats

[ddnet/ddnet-scripts] Pull request opened: #47 Update logo font details

Tater

I think theres a lot of people that get 10+ second lag from texture loading.

Hmm

github-merge-queue[bot]

[ddnet:master] 2 new commits

862c0ca make point hitbox bigger - Marmare314 1f1cc80 Merge pull request #6867 from Marmare314/point_selection - Robyt3

18:25

[ddnet/ddnet] Pull request closed: #6867 Make point selection easier

18:29

[ddnet/ddnet-scripts] Pull request closed: #47 Update logo font details

18:29

[ddnet-scripts:master] 2 new commits

04a024c Update logo font details - murpii 0794ab7 Merge pull request #47 from murpii/pr_update_logo_font_details - def-

guys what about do like this design for #map-releases

Image attachment

yea the bot could do that

18:47

that would be pretty cool

18:47

you'll put poor snail out of work tho

Ewan

you'll put poor snail out of work tho

im can do it if they send me list of map release

18:47

im just send them source they put the webhook and send message

18:47

done

they would probably have @DDNet do it

18:48

not a webhook

Ewan

they would probably have @DDNet do it

thats gonna be harder i think

18:48

not for them rly

18:48

the msg data would be the same

18:49

are u doing it in json or what

yeah

18:49

9.36 KB

18:50

even its harder lmfao

18:52

easier send me list

18:52

and im do all

18:52

and they send just with webhook

nah i don't think they'll go for that

sad

18:53

event its a good design

18:53

:I

do you know python

not good fr

k

Jurai!

easier send me list

There is no list, it's more like you might have a short conversation about the map releases or snail picks out suitable maps that just fits atm (edited)

Cøke

There is no list, it's more like you might have a short conversation about the map releases or snail picks out suitable maps that just fits atm (edited)

yeah too possible

i was just imagining a bot command with name, author, difficulty, release time and thumbnail as parameters. would be pretty much the same process, but then you can benefit from the cool markdown and link stuff in the embed

Jurai!

guys what about do like this design for #map-releases

What u think about it @snail ?

[ddnet/ddnet] Pull request opened: #6943 Use std::shared_ptr for editor resources, fix crashes with brush

Simplify memory management of groups, layers, images, sounds and envelopes by using std::shared_ptr for all of them. Clear brush when creating a new map to prevent crash due to unloaded textures being used in tile layer. Closes #6935. Also fix crash when restoring a saved brush after removing an image that is being used by it.

Checklist

[X] Tested the change ingame
[ ] Provided screenshots if it is a visual change
[ ] Tested in combination with possibly related config...

hi so basically last time when i compiled ddnet everything went right, today im a bit sleepy etc, and resetted my windows also and wanted to compile ddnet once again in vsc, and when im building the cmake, i get this: (return code 1) i don't know if its the right channel, sorry for that

cmake_compile.txt

2.89 KB

20:07

the dependencies was not found but i have it in the ddnet-libs, and git downloaded recursived (edited)

What is the actual error?

20:08

I think the last lines are missing

20:08

I have a feeling this could be my ssp change from today having broken it

tried to download it without recursive and after downloading the ddnet libs, and trying cmake but i get this, and getting this also if i try with --recursive

20:09

rust,cargo and python ofc downloaded

What is the error message?

@Ryozuki could you help him pls, he has trouble compiling on windows xd

this is when im trying to build it, crazy

264.97 KB

20:13

i dont really know whats going on never happened before, lol

ok, this is a Rust linking error

20:25

Maybe you have changed the Rust version?

20:26

It's probably still a bug and @heinrich5991 might know something about the Rust build process that could be going wrong

20:29

Something like this: https://github.com/rust-lang/rust/issues/101134

__CxxFrameHandler3 still referenced even though panic=abort · I...

Code I tried this code: fn main() {} With this section in Cargo.toml: [profile.release] panic = 'abort' lto = true And also disabled msvcrt.lib. Unfortunately rust does not have a good way ...

20:29

@lean can you check if a week old DDNet version builds fine?

my friend autojoined the brasil tutorial server even though he's from USA

deen

@lean can you check if a week old DDNet version builds fine?

version 17.1?

20:33

also, if i try to compile it in visual studio (not code) it compiles, and starts running the exe, but then i get that these dll's are missing:

libpng16-16.dll
avformat-6.dll
SDL2.dll
libfreetype.dll

lean

version 17.1?

17.1 (https://github.com/ddnet/ddnet/releases/tag/17.1) builded successfully

via vs. (edited)

Release 17.1 · ddnet/ddnet

Fix online clan friends not being removable via UI

[ddnet/ddnet] Issue closed: #6787 Layers should keep being selected after dragging and dropping

wv

Exported 404 message(s)