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 2020-10-13 00:00:00Z and 2020-10-14 00:00:00Z

[ddnet-web:master] 1 new commit

1fa64ae server stats - def-

[ddnet-web:master] 1 new commit

2f69f31 template server stats - def-

oh

07:32

nice

got annoyed by all the copy paste at some point

[ddnet/ddnet] Pull request opened: #3082 Set correct value on layer saving (which was recently uninitialized)

Didn't check out the teeworlds source while setting the value the last time. teeworlds sets those values to 0.

[ddnet:master] 3 new commits

06e9050 Use the same value as vanilla - Patiga a252097 Fix comments - Patiga 05fdd02 Merge #3082 - bors[bot]

08:14

[ddnet/ddnet] Pull request closed: #3082 Set correct value on layer saving (which was recently uninitialized)

the bg animation isn't smooth when you switch the tabs Settings -> DDNet

10:37

and please why this terrible sun with wrong sunrays

smooth like the camera?

10:38

or the actual animation

10:39

@Lady Saavik we are happy to take better maps, the current ones are from vanilla

oh it's so random

if u have laggs loading the settings, it might be bcs of loading something from ur harddrive (edited)

it's annoying that the sun sometimes moves so far and sometimes not

the clouds resetting is the most annoying thing

10:43

as said the maps arent really good, except maybe jungle night

10:43

winter night is ok, rest is not really nice (edited)

idk, the bug is gone .-.

10:44

omg it's back

10:44

I know

10:44

now

are you using 15.1 btw?

10:45

or master

I was in wrong tab in settings

10:45

Image attachment

ok

I can record it

its not like i dont believe it

10:47

but go ahead

idk if you understand me

ok

Tee and Controls have almost the same position in the background xd

xd

same with General and Player

10:49

not smooth

yeah

ah

true

same

probably bcs the distance is too short

but it works the other way xd

10:53

when I click on "Run server" and close the black window, why does it still say "Stop server"

Internet and Sound have the same position

the client doesnt track the server process probably

10:54

u can probably also close the client and it wont stop the server

10:56

setting for example cl_rotation_radius 15

10:56

fixes it, so its very probably the distance

@Дядя Женя please don't make a huge PR with everything. Split features into smaller PR's. That has benefits: you get feedback early and will make less work on refactoring; others will get features sooner; subtasks will not stall the development; you will get better motivation seeing something is finished already; ~~we don't risk you quit development at all not making a single PR cause it's "unfinished"~~

12:16

In my point of view although chat/ messages are technically related but are functionally separate from scores. (edited)

k, I'll try to make some margin, fix bugs and clear some useless code I made on failed attempts and make a PR

12:23

I was afraid making any pull requests mainly because I don't think it's ok to build release version of client with unfinished feature, which not everyone would like on that stage

12:24

Maybe I should add client config variable for that to be enabled

12:25

Also half of the padding is hardcoded now, 'cause I had no idea how to calculate it properly (edited)

There is such thing as WIP-PR or drafts. A way to get a feedback from reviewers before the feature gets to the stage. Though chances are if your feature is only lacks some visual polishing only it still could be merged.

12:26

Sadly I don't know anything about client. But I am sure others ddnet devs will suggest better solutions when reviewing your PR. Also you can direct them to lines in code commenting it in PR by yourself and asking questions there.

12:27

Image attachment

12:27

https://github.blog/2019-02-14-introducing-draft-pull-requests/

12:28

Before draft pr's were introduced a title containing "WIP" keyword was indicating that this PR is not yet complete but you need a review to get a fresh look or suggestions.

12:28

:s/but/but still

12:29

(I hate this SED edit syntax enforced in this channel)

[ddnet/ddnet] Pull request opened: #3083 Make menu background camera impl more robust

The vanilla code isnt really robust against short distances, hope this fixes almost all edge cases

[ddnet/ddnet] Issue opened: #3084 #3068 breaks Alloc for data

i guess its just the data allocation that is broken, not the commands itself #3068

@Deleted User what is broken about data allocation? It didn't look like it required anything specific to me, that's why I didn't touch it

yeah i dunno yet either, but OGL 3.3 doesnt work in entities anymore @Learath2

@Deleted User You sure it's alloc? entities works completely fine for me on macOS with 3.3

i bisect and its 100% this commit

15:14

but im not 100% sure what is causing it

I'll try linux

15:17

Well I'll be, for once something works on macOS doesn't on linux

xD

failed to allocate data for command buffer

y<eah

15:20

i think i found the issue

15:20

the current commands rely on the buffer size too much

15:20

i'll write the patch if thats ok for u

15:21

so its not ur impl

15:22

(except that it always adds the alignment)

15:22

which increases the buffer size always

Yeah, I came to the same conclusion, it genuinely doesn't seem to fit

15:22

I'll fix the wasted bytes

15:23

Oh, I wonder if that would be enough actually

yeah thats the question

Yep, it's enough

15:24

But still, is it expected that space is so tight in the data buffer?

no, but it will use the full buffer if the alloc fails once

15:24

it just clears it and uses full space

That was apparently wasting so many bytes that even the full space wasn't enough

yes, its always 16 bytes too much

15:25

so it can never allocate the space it wants

Oh I forgot to ask, what should the alignment for data be?

well, actually doesnt matter probably, alignof(ptr) should be fine for almost all architectures

15:27

dunno if floats have higher alignments somewhere

15:27

maybe maximum of floats and ptrs xd

15:27

but that would be the same issue for commands

AllocData is just byte storage, right?

they sometimes contain floats too

15:28

yes, but it should still be aligned correctly

15:28

arm cant read floats that are missaligned

15:28

(so too driver dependend) to make it byte aligned

(iirc new arm can do unaligned read/write, it's quite costly though)

well for floats it crashes

15:29

i had a bug in arm android some day

15:29

i am speaking about arm32 btw

I guess I'll keep max_align_t for data, just to be safe

64 is fine

ah, arm32 yeah

yeah just keep it, should be safe

Should I close the issue by fixing the waste or do you want to do something about it too?

yeah just do it by waste

15:32

i dont see anything that could break else

[ddnet/ddnet] Pull request opened: #3085 Don't waste bytes when we are already aligned

This fixes #3084, seems to be enough to make all the data fit in the buffer again.

@Deleted User look what i found on youtube, (probably version 14.6.2) xddd

Image attachment

what video is that

https://www.youtube.com/watch?v=OrhHp5lLobU

15:45

im lurking youtube

15:45

the guy plays the game

15:45

like its normal

15:45

xdd

15:45

feelsbadman

WTF XD

XD

funny how the skins are broken not the map

the map is broken

15:46

look at epix start

thats like the opposite of what we have else

thats not normal

15:46

Image attachment

15:46

i mean kobra

15:47

start

15:47

this is bugged

well anyway with the new verion he'd be in the fallback

He must think we are all on LSD all the time

yeah

but why are the skins broken xD i dont get it

man ppl in 2020 still upload videos to youtube in 480p

15:47

hilarious

4:3 with some 15 years old GPU uploads in 480p

15:48

some ppl just live in the past xD

@Deleted User yep, look at #showroom

15:48

i found a funny chilean

15:48

well u probs dont understand him

ryo lives in a bubble

Image attachment

15:49

xdddd

I don't get it how could valgrind say there is no leak and ASan say there is a leak?

i still dont get why valgrind didnt see the invalid index, what ubsan found

maybe valgrind is not perfect

IT IS

justatest

anyway

15:50

WHO BROKE BORS (edited)

isnt ubsan static analysis and valgrind runtime?

both are runtime

ubsan is runtime too

UBSan and ASan are instrumented directly into the program. Valgrind instruments the program by monkey patching it

is ubsan and asan faster?

Yes

valgrind rly makes the game slow af

15:51

xd

With ASan the game is actually quite playable

yes

15:52

it runs pretty smooth

15:52

> 1k fps

yes we know u have a super vega gpu

its not gpu side xD

monkalaugh

15:52

what cpu u got

3800x

thats pretty good

15:53

isnt it

I wonder if something like ASan exists for GPUs

342€

15:53

yes

15:53

xd

instrumenting the shaders

yeah but like 30% faster than some 5 years old intel

15:53

so not the world xd

will u buy the new amd ryzen 9

15:53

poggers

Well ASan wouldn't be much use in a shader, but something like ubsan might

i only buy hardware all 4 years normally, when i sell my old to my one of my sister xD

xd

15:54

yeah same

15:54

i need to change gpu soon tho

15:54

my r9 390 is getting old

I wish the new generation had some backwards compatibility with DDR3

u still in ddr3

15:54

lol

I want to upgrade but it'll cost a fortune to replace 32g of ram with ddr4

do u rly need 32

15:55

im fine with 16

I do yeah. I use chromium boy

1 tab in chrome = 1gb

why u use chromium

so yes 32 gb

15:55

firefox uses 2gb

The remote debugging

15:55

I use firefox for my own browsing

i even use both at same time

15:55

i dont have ram problems

i know was just jk

15:56

bcs browsers

But that's not what kills my ram, it's unreal engine and maya

android studio eats lot of ram too

15:56

nouis

all java stuff eats ur ram

both open at the same time gets the computer swapping every 10 seconds

15:56

Java is really so irresponsible with the memory given to it

15:57

Not only that, it keeps all the pages hot, all the time with it's GC

ur hot

15:57

15:57

lol

Okay, I don't get it, do you see how src/engine/client/serverbrowser.cpp:502:32 could ever leak?

is it one of those I fixed?

[ddnet/ddnet] Issue closed: #3084 #3068 breaks Alloc for data

16:00

[ddnet/ddnet] Issue closed: #3084 #3068 breaks Alloc for data

16:00

[ddnet/ddnet] Pull request closed: #3085 Don't waste bytes when we are already aligned

bors is dying

Not in #3076 atleast

https://github.com/ddnet/ddnet/pull/3076

Add clang-tidy to CI and enable clang-analyzer checks (supersedes #...

470 warnings treated as errors The actual work will be fixing those or at least NOLINT(clang-analyzer-optin.cplusplus.VirtualCall) them so we can enable this. $ cmake -GNinja -DWEBSOCKETS=OFF -DMY...

then where did you see it leak?

master, ASan

ok, haven't looked into it then

It doesn't make sense though, there is the free right there :/

maybe a buggy libc?

I'd have many more things broken if I had a broken libc, no?

where is the free?

16:03

m_ppServerlist is freed, not ppNewlist

L505

or is the leak at destruction?

16:03

its never freed when closing

Oh, maybe it's reporting it leaking when shutting down

yeah, I guess a free(m_ppServerlist) at the end would help

16:04

seems easy enough to fix in the destructor

Heh, yep. I didn't consider it leaking at the end

there is alot leaking, we could also clean the opengl context before closing

16:06

but so much static stuff that allocates textures, would require alot of work

[ddnet/ddnet] Pull request opened: #3086 Fix 2 leaks in CServerBrowser

The only leaks reported by ASan that are allocated by us. There is one other by SDL and one by X11.

Well these were the only two reported by ASan, so I fixed them

oh ok

[ddnet:master] 2 new commits

4c87b0c Make menu background camera impl more robust - Jupeyy 6ee362f Merge #3083 - bors[bot]

16:07

[ddnet/ddnet] Pull request closed: #3083 Make menu background camera impl more robust

Github Actions is also throwing a hissy fit, I wonder what's going on

16:08

I have a feeling it's microsoft still not understanding how one should update their apt repositories

mhh, someone once said here the server browser feels to laggy with OGL 3.3, WHO was it

16:10

cant find him in discord search

index 32767 out of bounds for type 'short [64]' this looks odd

yeah

16:11

we have a huge problem with snapshots

16:11

#3049

https://github.com/ddnet/ddnet/issues/3049

Undefined behaviour in snapshots · Issue #3049 · ddnet/ddnet

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 ...

but the discussion about it moved into some pr

#3051

https://github.com/ddnet/ddnet/pull/3051

Fix some undefined behaviour by def- · Pull Request #3051 · ddnet/d...

?

16:13

I removed the shift commits.

16:13

oh

"yes, my claim is, that appearently the same snapshot has different values between linux and windows So its probs caused by some of the other snapshot code" to quote myself

16:13

it wasnt about the shift

16:13

i assumed it bcs linux returned -1, but i wasnt in windows anymore at that poiunt

Only heinrich knows how this stuff works really well and when the world needed him most, he vanished

100 years passed and Learath and me found a new heinrich (edited)

Okay, from my crude understanding of the snapshot extended objects, I think that type has to be translated first

maybe heinrich already thought about it, he also removed crc code

16:23

@deen have u seen #3077 it litterally doubkes my fps with ogl.. just bcs of the text

https://github.com/ddnet/ddnet/pull/3077

[WIP] UI Elements by Jupeyy · Pull Request #3077 · ddnet/ddnet

Build will fail bcs of clang-format, i'll update this comment to explain.

nice

16:23

I noticed that in menus everything is super slow

I have another question, why is the client even creating a snapshot??

16:24

Ah maybe demo recording

I'M generally in favor of it @Deleted User

good bcs u can still use the old implementation

16:25

i'll just implement the stuff i need whenever i feel there is something too slow

16:25

the rest can stay streamed

Ah, I get what's wrong, the extended types have dynamic size, the itemsizes table shouldn't even be checked

i mean create a pr, and ask heinrich for a review

I will in a minute

a fix until the next version would be nice

@deen where do you think would be a good place to add a COUNT_OF macro? we use it in a couple places and we just have ad-hoc implementations

16:33

I was thinking system.h but there I can't use the more thorough C++ version of the macro

what's a count_of?

It's the macro that everyone uses to get the number of "things" in an array

16:34

sizeof(arr)/sizeof(arr[0])

ah, right

16:35

for all I care make system.c -> system.cpp, I had a PR for that

16:35

then put it in system.h

16:35

otherwise src/base/count.h? I don't really see a fiting place

I'll put the unsafe C version in system.h and we'll all be careful not to pass it anything except arrays

yes lets pls make it cpp

@deen why did you close that PR anyway?

then we can use std::chrono and remove some strange code from windows

If only C++17 wasn't too new, we could replace a lot of the fs_ functions too

i also made a local pr where is changed everything

Maybe I misremember and didn't have it. Foudn only https://github.com/ddnet/ddnet/issues/1444

Upgrading system.c · Issue #1444 · ddnet/ddnet

We are compiling system.c with C90 for some reason. Should we maybe start compiling it with C99 or replace it with a system.cpp(C++11)?

https://github.com/Jupeyy/ddnet/tree/pr_cpp_not_c

DDraceNetwork, a mod of Teeworlds. Contribute to Jupeyy/ddnet development by creating an account on GitHub.

16:37

i made it a few weeks ago

you can remove all the (void) in function signatures in cpp

yeah but too lazy xd

let's ask @heinrich5991 what he thinks, I'd be fine

16:39

sed -i 's/(void)$//' src/base/**/*.{cpp,h} (edited)

will do if we ok with removing c

[ddnet:master] 2 new commits

4854809 Fix 2 leaks in CServerBrowser - Learath2 e11511f Merge #3086 - bors[bot]

16:41

[ddnet/ddnet] Pull request closed: #3086 Fix 2 leaks in CServerBrowser

btw @deen #2157

https://github.com/ddnet/ddnet/pull/2157

Switch src/base from C89 to C++11 by def- · Pull Request #2157 · dd...

Disadvantage is of course losing Vanilla Teeworlds compatibility, so I don't care much either way. Haven't cleaned up much since I don't want to put work into it if it&#...

Probably because of Vanilla compat

who cares xD

but now with our reformatting etc, I guess matters less and less

16:42

so let's just do it

we can still copy their functions

[ddnet/ddnet] Pull request opened: #3087 Fix UB in CSnapshotDelta::CreateDelta

I think this is the proper fix, but it'd be nice to get it confirmed by @heinrich5991 before merging. Fixes #3049, I think?(tm)

i only know that deen tried to fix it and couldnt join the server anymore xD

16:44

but if u can its fine xD

Hm, I wonder if we are talking about the same bug

yes

../src/engine/shared/snapshot.cpp:264:9: runtime error: index 32765 out of bounds for type 'short [64]'
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior ../src/engine/shared/snapshot.cpp:264:9 in

its the code in createdelata and unpackdelta

This one turned out to be in the demo recorder, fixing it like this didn't break anything(tm)

#3050 here are the reports

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

Go through ubsan and clean up · Issue #3050 · ddnet/ddnet

CC=clang CXX=clang++ CXXFLAGS="-fsanitize=undefined" CFLAGS="-fsanitize=undefined" cmake -GNinja -DWEBSOCKETS=OFF -DMYSQL=ON -DCMAKE_BUILD_TYPE=Debug -DVIDEORECO...

UnpackDelta looks fine to me, at the very least doesn't suffer from the same bug

yeah cant find unpack delta, so just create

16:47

but 2 times

16:47

src/engine/shared/snapshot.cpp:139:21: runtime error: signed integer overflow: 0 - -2147483648 cannot be represented in type 'int'
    #0 0xede8fb in CSnapshotDelta::DiffItem(int*, int*, int*, int) src/engine/shared/snapshot.cpp:139

I couldn't get that, how did you trigger it?

it was deen

16:49

or was it me

16:49

oh yeah

16:49

it was me xD

[ddnet/ddnet] Issue opened: #3088 Allowed C++ features

We should decide on a list of allowed C++ features, instead of having some unspecified list. Questions:

Replace array with std::vector?
Allow std::string?
Allow auto? Only for iterators?
Allow foreach loops?
…

What do you think?

@Learath2 i dunno, i just played some ddrace and fng

what's the benefit of allowing C++ in system.c? @deen

16:52

@Learath2 you can use the fancy macro in #ifdef __cplusplus__

whats the benefit of using c code xdd

less difference to tw

cpp allows std::chrono std::thread

Stability or system.c

16:52

s/or/of/

the problem with the overflow for m_aItemSizes is due to the protocol extension

oh, that sounds like my fault

16:53

what is it?

it's fixed in 0.7

16:54

the whole thing relies on some random memory being 0

have a link to the PR?

16:54

very nice. sorry for that code

nobody noticed for years ^^

damn my vim is stubborn and doesnt want to insert tabs instead of spaces in the python files we have in datasrc

@Ryozuki use vim-sleuth addon

16:55

https://github.com/tpope/vim-sleuth/

tpope/vim-sleuth

sleuth.vim: Heuristically set buffer options. Contribute to tpope/vim-sleuth development by creating an account on GitHub.

ohh

16:55

thanks

16:56

works perfect

I actually never had to care about indentation again with this vim addon

[ddnet/ddnet] Pull request opened: #3089 Fix out of bounds access in snapshot delta handling

Same as https://github.com/teeworlds/teeworlds/commit/293209e7227c81e92b99c89890ebfdc264374a4a + https://github.com/teeworlds/teeworlds/pull/2684/commits/c87a7b438f08f83385f4e21d1c23a498a6fb4c55 in 0.7

the only problem is that you have to install it

[ddnet/ddnet] Pull request opened: #3090 Make server aware of client prefered language

This an initial pr towards having server side localization to help with https://github.com/ddnet/ddnet/issues/269#issuecomment-706080357 I don't know if we should normalize the language (currently whats send is the language file...), use a enum or what. (This pr doesnt add localization yet)

@redix seems to be the exact same thing I did

oops

didn't notice

redix with another fix to 0.7 without even telling us, traitor angerycat

angerycat

i'm so sorry

xDDDDDDDDDD

nouis

anything else we want to know? XD

17:00

vulkan support for vanilla?

there are so many things that are fixed in either 0.7 or ddnet ^^

yeah its funny xD

Vulkan would probably get us even less compatible with older machines

17:03

I was supposed to have tea like 3 hours ago, then I forgot, now it's too late for tea

never too late for tea

well lets say it like this opengl 2.0 is from 2004 and still like 1% cannot use it xD

I drink tea for breakfast, lunch and dinner

in germany we drink Tees

[ddnet/ddnet] Pull request closed: #3089 Fix out of bounds access in snapshot delta handling

@heinrich5991 I could drink tea with my dinner but I was more looking forward to having a cookie, cookie for dinner is no no

webgpu would be cool as a backend once it's final :D metal for macos + vulkan for the rest

yeah

17:05

its a rly nice project

17:05

but calling it webgpu imho is rly bad

I wonder why they didn't do Vulkan on macOS too

xDDD

17:06

u really wonder about it

I read vulkan came too late

https://gpuweb.github.io/gpuweb/

17:06

webgpu api is rly like vulkan btw

17:06

or atleast from my little experience using it

17:06

i made a triangle

Metal is low level enough that translation from vulkan to metal is pretty cheap

u get a instance, a adapter a device a queue

17:07

etc

MoltenVK is already pretty mature

wgsl is the future too

17:07

https://gpuweb.github.io/gpuweb/wgsl.html

17:08

spir-v ftw

I guess if they don't mind spending the development effort on maintaining a separate metal backend, who am I to judge

i dont like the syntax

17:08

c syntax nice

When are we getting a standard shading language?

@Deleted User u can use glsl

17:08

and compile it to spir-v

then its fine

@Learath2 I'd include the vanilla fix instead of yours, is that ok?

justatest

@Deleted User I see more out of date intel gpu drivers causing troubles, maybe adopt the blacklist? ^^

@heinrich5991 It's the same fix, but sure, go ahead

@heinrich5991 well idc, but what would that help, telling them to upgrade? moving them down to gl 1.x?

telling them to upgrade

17:12

what you'd manually say to them

well the guy in #general didnt, bcs too much work for him xD

I guess it defines an extra enum to avoid arr/arr[0]

but as said, i'd be ok with it @heinrich5991

17:12

still better than nothing

@Deleted User could give them a small warning and fall back to gl1

the problem is, they will never go back to higher gl versions then

Who cares? Not like their toaster will ever get a driver update

nono

17:13

its new intel gpus

17:13

funny enough my 10years old laptop works fine

17:14

while louis with a 2017 gpu has problems

I wonder how Unreal and Unity solve the issues

directx

17:14

very simple

17:14

these problems are mostly on windows

[ddnet/ddnet] Pull request reopened: #3089 Fix out of bounds access in snapshot delta handling

mac always the problem

Image attachment

17:15

compile in clang @Ryozuki

UE4 is coming out with Vulkan as the backend, I wonder how they'll handle it there

well maybe just drop support for older gpus?

compile what in clang

hdiutil: create failed - Resource busy

most tripple A titles have pretty high miniums

hmmm

@Ryozuki ok nvm if its just a bug

17:16

thought some clang warning xD

hdiutil thing happens every once in a while

17:16

just re-run the job

is that a known bug of hdiutil so we should just retry in the script?

17:16

or are we holding hdiutil wrong?

I have no idea whether it's a known bug

https://stackoverflow.com/a/18738960

Error creating disk image using hdutil

I'm trying to build dmg of .app file using following hdutil command: hdiutil create -srcfolder /Users/me/My.app My.dmg It works as expected as it creates My.dmg correctly. Problem started happen...

17:19

wtf

HAHAHAHA classic apple

17:20

Of course creating an archive on the commandline would require a trash folder

hm. can you check whether such a trash folder is actually created?

17:21

if you have a mac handy

17:21

(just to check the plausibility and whether I should really implement that solution…)

[ddnet:master] 2 new commits

d172286 Fix out of bounds access in snapshot delta handling - axblk 0689466 Merge #3089 - bors[bot]

17:22

[ddnet/ddnet] Pull request closed: #3089 Fix out of bounds access in snapshot delta handling

It would be the first time you open the folder in finder, let me check

[ddnet/ddnet] Pull request closed: #3087 Fix UB in CSnapshotDelta::CreateDelta

eh, I mean by the hdiutil

17:23

i.e. can you run a build (like on CI) and check whether the generated archive contains a trash folder

Sure

@heinrich5991 u mean server sending msg ids and the client displaying them?

17:28

how would u work around dynamic conent

17:28

content

it'd need our own printf parser

Or we could do it in a very insecure way

no

Oh cmon, it'd be fun

rce is a feature, not a bug

17:29

user freedom!

Server sending a %n to do arbitrary writes

17:30

Error copying file "/Users/learath2/Desktop/C/ddnet/ddnet-libs/ffmpeg/mac/lib64/libavcodec.58.dylib" to "pack_DDNet-15.1.2-osx_dmg/DDNet.app/Contents/Frameworks/".

[ddnet/ddnet] Pull request closed: #3090 Make server aware of client prefered language

uhm, idk how this is supposed to work

17:32

Why is it even looking there? we don't ship that

I guess you have an up-to-date submodule?

Even if I disable -DVIDEORECORDER it's looking for it

17:33

Yeah, up-to-date

[ddnet/ddnet] Issue opened: #3091 Client Side translation of Server Messages

The server would send a packet instead of a chat message with a id and arguments, then the client would display the translated messages with those values. We would then need our own printf parser according to @heinrich5991 We also need to take care of compatibility with unsuported clients. Any more ideas?

idk tbh @Learath2

Oh, btw how are we supposed to deal with unsupported clients? The version number is an ugly fix

can oyu comment that out?

17:35

I'd send the untranslated string as well

17:35

that'd be some network traffic though

@deen I think you forgot to commit the mac build for ffmpeg to ddnet-libs

17:36

@heinrich5991 and it would require very ugly string matching to remove the untranslated string

hm? no

17:36

I was thinking:

especially since there is nowhere to hide a mark in a chat message

maybe add a hidden unicode characte

17:37

at the start

17:37

and use it to hide on modern clients

17:37

superh acky hack

17:37

nouis

a new message type TRANSLATED_SERVER_MESSAGE, with an ID (UUID), untranslated message (string), format parameters (various, string, int)

but the problem here is old clients

17:37

they dont know about this packet

17:37

they only want chat messages

yea, distinguish them by version

then why we need to send the untrasnlated messages o nnew clients?

The new message type is simple

i dont find the reason

Ah by version

because they might not support all the strings we need

damn i type like a retard

17:38

missing keys

17:38

xd

17:39

ah thats true

17:39

but

17:39

it can fallback to the english

17:39

ah

17:39

we dont have a english

17:39

its directly in the src

yes, that's the untranslated string

well then im fine with that

17:39

btw add any final decicion to the issue

17:39

this will be lost

17:39

here

yes, true

17:39

thanks for the reminder

So we check the version, if it's too old we send preformatted english string. If new enough to know this string, send unformatted message with parameters

and english

no, if new enough to know the TRANSLATED_SERVER_MESSAGE message

17:40

then we send it

Ah, you want to send the entire string

then the client can decide whetheri t knows this particular message

in case its not translated

17:40

we need a fallback

meh, atleast it's not more traffic

well, my solution is more traffic?!

yes its more

17:41

but i dont think it matters

the UUID + the formatting parameters

why a UUID?

17:42

and not a simple id

It's sliightly more traffic

@Ryozuki because people are going to add their own messages and we don't want to conflict

ah

e.g. noby adds a message for fng

Why bother with an id at all actually?

you are thinking outside ddnet too

17:43

good i forgot

@Learath2 because the untranslated string might change without the translations needing change

We can send the untranslated string unformatted

(was my thought)

i prefer a uuid

17:44

this get stuff translation is rly funny

17:44

cuz there are repeated translations

17:44

atleast i found that when translating

How can I know what causes permanent freeze on my build?

I sort of dislike having redundant information, but I guess I can't come up with a better solution

if u cant then its not that redudant i guess

well it is redundant in the sense that we control client and server

@Дядя Женя run it in a debugger, when it freezes, send SIGTERM, the debugger will tell you where it stopped. Repeat a couple times to see whereabouts the client is looping

but not redundant if we don't assume that

It's redundant because we actually know on the server whether a client knows a string or not via the version number

@heinrich5991 why did u decide not to add "Map: " on the steam status?

Image attachment

but I thinkn that's kind a ugly

i find it really funny tho

@Ryozuki redundant, people will learn I display the map

17:47

wasted screen space otherwise

17:47

(my thinking)

btw there may be ppl who dont play the game

17:47

and can see ur status

nice

17:47

I'm In Da House

xd

17:48

btw remember to add info to the issue in github

17:48

xd

17:48

we will rly forget

@heinrich5991 hm, well I'm sure we could come up with some scripts to make it pretty, generate a ClientKnowsString(const CUuid &Uuid, const int Version)

17:49

but your solution is trivial really

what about allowing auto too when u have a cast

17:52

auto x = (sometype)a;

Can I take someones time to answer my question about rendering optimization

well i should probs write it in the issue

So, there are "text containers" where text is being drawn and then, if nothing new is happening, game just draws entire container at once

17:53

Skipping drawing characters for text individually (edited)

17:54

I want to draw not only text, but also background for it, RenderTee, maybe some quads and text as well

17:54

In a container ofc, so it doesn't need to be redrawn every time

17:54

Do you have that kind of buffering already implemented

CQuadContainer is the generic version, the text container is also implemented on top of it

Or should I make my own containers for that

17:55

I'll explain. I want to draw chat messages with some extra graphics, margin, background and so on (edited)

17:56

So the best way to do it is just render the message when it was received from server and store it somewhere

17:57

I'm just not sure how to do it and if you have any logic for that already

@Deleted User would know best, but the best way would probably be to render it on the first OnRender rendering when you receive it is a meh idea

[ddnet/ddnet] Pull request opened: #3092 Use pointer directly for tile layer building

Huge maps like back in time 2, will always "only" upload 2MB of data at once, and then wait for the graphics thread to finish building/copying the VRAM, this might increase performance a bit

Doesn't matter. The idea is to render only once and then buffer

Currently chat messages are rendered onto text containers if they haven't been rendered yet. You can probably replace the text container with a quad container

I need some help with "buffer" part

it builds VRAM on the GPU

What is quad container

just like u build HEAP on RAM

17:58

dont think of GPUs to different from CPUs

isnt text already optimized?

GPUs are CPUs with more cores xD

I'm remaking and redesigning it @Ryozuki

he wants to add a ui rect behind it

17:59

CreateRoundRectQuadContainer

and probably a tee image to the left of it from what I gather

i already added a function for that

ah u want to display a tee

I've already done all that

will u add the tee on the chat?

18:00

its already bloated

Just need to know the best way to buffer it once and not redraw every frame

heart + id + repeated counter

18:00

and now a tee

18:00

monkalaugh

you want to have framebuffers?

18:01

sounds like u want to render to a texture

18:01

we dont have this and if we would have it, only for gl 3.3+

if u add that, use it on video recording too

yes

u said smth like that time ago

Embedded image

i know, but no motivation for that feature xD

18:02

OBS is too good xD

@Deleted User couldn't you render a tee into a quad container? Not that it'd help much

Embedded image

yeah

18:02

obs best

that's an example of what I have right now

@Learath2 well u can draw a tee with quad containers, i already do it, but not all tee parts at once

did u test with chinese

18:02

and stuff

it be still good enough tho

Imagine second screen as one message container. I want to draw it, store and then redraw whole container for optimization purposes

aslong as the VRAM is built, its pretty fast

the tee there looks so out of place

18:03

xD

yes, because margin in source code sucks

@Deleted User why can't you do the entire tee?

that was already hard enough to margin text like that

@Learath2 bcs multi texture support is cancer, when the drivers are buggy xd

@Дядя Женя don't bother, all of those functions you use buffer wherever they can anyway

18:04

Ah, I keep forgetting the shitty drivers

Image attachment

18:05

this happens every OnRender

18:05

I can leave it like that if you'll say that's ok

Completely fine, we render rectangles everywhere

what if I'll move DrawRoundRect in a loop?

Image attachment

18:06

now github is dead xD

$build_status

omg

18:06

whtie theme

18:06

frozen

18:07

https://github.com/StylishThemes/GitHub-Dark

StylishThemes/GitHub-Dark

:octocat: Dark GitHub style. Contribute to StylishThemes/GitHub-Dark development by creating an account on GitHub.

yeah is there black theme?

use the stylus plugin

18:07

with this

is that a browser plugin?=

ye

@Дядя Женя fine, doesn't really matter how much you render

<.< i hate browser plugins for single websites

Image attachment

18:07

its not for

18:07

single websites

look nice

its a plugin for styles

18:07

and u install a style for github

18:07

xd

18:08

it even styles code

18:08

well im using monokai

Image attachment

After recompiling 10 times, there is no .Trash folder @heinrich5991

@Learath2 thanks for trying. should I try anyway? ^^

ah maybe mac can't see the trash folder

18:36

mind sending the dmg?

my client crashes when I'm trying to send really long message (just spamming letters, preferably without spaces)

18:37

I mean it freezes permanently

18:37

But that does not happen on release

18:37

debug only

mhh sounds bad, does that also happen with current ddnet source?

idk, Im here

Image attachment

ok, if you want, create an issue with the text u typed

Ok, now

18:43

1. How do I build game executable

18:43

2. How do I create a pull request for you to review and test (I mean it's clearly not ready to be added in public version) (edited)

puhh, just google git tutorial

Do you have some privacy setting in your repository so I can't break it if I try to push or smth?

u cannot push to ddnet github directly

I can't break it ofc, but would not be cool if I make something u dont want to

18:46

ok nice

you can fork it and push to your forked repository

then I'll click on push button

18:47

xD

18:50

@Deleted User what about building?

what os are u on?

win

visual studio?

yes 2019

just change to release mode

18:50

and build it

18:50

then the executable is accessable in the directory somewhere

I did

/build/Release or smth

Yes, the question is mainly about that "somewhere" thing (edited)

where is your ddnet.sln?

18:51

there should be the Release directory

there is only DDNet.exe

whats what u want isnt it?

I mean the whole game folder. Do I need just replace DDNet.exe in the game directory so it applies all the changes?

yes, that should do the job

ok then, cool

18:53

will try

18:53

Yeah, 4real

18:59

Any advanced git user here?)

thats a bad question

19:01

https://dontasktoask.com/

Don't ask to ask, just ask

19:01

read this website thanks

I've cloned ddnet git repository, build it with CMAKE (I have no idea what it did, really) as written on repository page, made some changes and committed it (edited)

19:01

And now I want to make a pull request on that, which I can't, because I didn't fork the repository before

19:02

So what can I do now

19:02

thats a bad question

@Ryozuki yes, I know, I was typing in the good one at the time, sorry. I'm just chatting that way, that I split one topic on several messages, so the first question mentions on who can help and then goes the problem explaination (edited)

@Дядя Женя ok, let's walk through this

19:29

you cloned the repo and made some changes

Yes. Then locally commit it

hm, now I don't know msvc's git interface

19:32

let me google a bit

After that I realized that I need to fork ddnet repo and clone it instead

no, should still be fine

19:32

but do fork the repo on github please

I did now

nice

19:33

let's google changing the upstream, sec

I use "Fork" program on windows, if it helps you somehow

in msvc?

https://git-fork.com/

Fork - a fast and friendly git client for Mac and Windows

Fork - a fast and friendly git client for Mac and Windows

@Дядя Женя what does the "repository" menu say?

19:35

can you do something like "add remote" there?

Image attachment

hm, show me the push dialog please

Image attachment

hmhm

19:39

@Дядя Женя sorry, distracted. what does the To menu show?

its faster to clone my fork repo and replace all changed files I guess

okay if you want to do that ^^

19:39

(btw, there are also free git clients for windows, I think github has one, for example)

(or use the command line client and you can refer to the hundreds of online tutorials and documentation)

Or may be it's not

19:42

I can use command line, if you have any solution

On the git commandline, it would be just git remote add <name for repo> <url for repo>

so

19:50

git remote add mine https://github.com/yournamehere/ddnet

@Learath2 why do you think so?

Oh, it is there and my ddnet-libs is up to date

20:01

I wonder why it won't download for me

I'm getting stuck somewhere here

Image attachment

20:07

Happens randomly, but with high chance when pressing tab (so chat should change width/height)

[ddnet/ddnet] Pull request opened: #3093 Chat visual update

Includes some critical bugs in it

Player's skin near message author's name
Transparent black background for chat
Text margin when message is multiline
Messages margin

For visual changes you should try to include a screenshot

yeah, sry, forgot about that

20:17

can I edit?

yep

done

20:25

As I mentioned before there is a critical bug that freezes your client completely. seems like infinite loop? See screenshot above

since ur changes or on ddnet master? xd

I have no idea what causes it and need some help

but you've only had the infinite loop happen with your changes?

20:26

so far

Some simple visual bugs occure with my chat's background sometimes (when message got cut instantly without being blended) and there are two problems with text margin if message is 4 lines long or when holding tab, but it's not a big deal, I'll fix it (edited)

20:30

@Deleted User I guess I should choose "staging" branch next time?

20:30

Told ya I suck hard in git

well for the first pr, its acceptable

20:32

i cant rerun the jobs in #3092 i dunno why its so broken, the other pr is much newer and had no problems

https://github.com/ddnet/ddnet/pull/3092

Use pointer directly for tile layer building by Jupeyy · Pull Reque...

Huge maps like back in time 2, will always "only" upload 2MB of data at once, and then wait for the graphics thread to finish building/copying the VRAM, this might increase performance a bit

@Deleted User I reran the jobs

ok lets hope they dont cancel again

20:50

tried already 5 times or smth

just click on merge manually

20:50

github is just having some network problems I guess

it's just for the green checkmark

20:50

bors wouldn't stop the merge for codeql yet I think

20:51

@Pipou looks like they already planned an egyptian tileset in vanilla btw

Ah well.. xd

https://github.com/teeworlds/teeworlds/issues/2778

RFC: New tileset (egyptian?) · Issue #2778 · teeworlds/teeworlds

I just played a round of DDNet and found a real nice tileset that would also fit very well for Vanilla: What do you think about this tileset? The author Pipou seemed willing to release it under a f...

@Deleted User @Ravie @mind I have a question about image rendering and ddnet's new version (edited)

go ahead

If my client is 15.1 and I play old maps with dilated and borderfixed images, will it look funky to me ?

borderfix might make the map very bit less good looking, dilate is still required anyway

@deen so dilate is still needed

sadly teeworlds built upon the borderfix for too long time

20:56

well it auto dilates in the editor now @Pipou

oh

20:56

so I don't need to drag images in the .exe now?

@Deleted User when did that happen? ^^

20:57

how can one check that images are dilated @Deleted User

to counter borderfix, wasn't there borderadd?

20:57

@heinrich5991 you clearly see it ingame

@Pipou what do I need to look out for?

20:58

borders?

without dilate there is an odd outline around the mapres

@heinrich5991 open in gimp and disable alpha layer

k

20:58

teeworlds doesn't require dilate anymore?

Image attachment

20:58

yes

20:59

thye use pre calculated alphas

so I need to re-render all of my images

@Pipou wdym? due to tileset_borderfix no longer being needed?

well all of them are already dilated and borderfixed

21:00

so I need to render them without anything

21:00

fun

yeah especially the border fix

21:01

the dilate doesnt matter really

is dilate idempotent?

21:01

i.e. does it not do anything on the second application?

i dont understand xd

if I apply it to an image A and get another image B

21:02

and then apply it to the image B, do I get the same image B?

if AB were the same before yes

e.g. if pipou puts it into dilate.xe and then it imports it in the editor

21:02

which also dilates

yes

21:02

if he uses our dilate tool

does the second dilating change the image?

it will be the same

kk

and we have a config to disable it

nice

21:03

but why do we need the config then?

21:03

if it's the same anyway

if someone doesnt want to dilate for some reason

ah, at all you mean

dilated images will work fine in 0.7 too, its really just transparent pixels that are getting colored

I see

21:08

yea, thanks for the explanation the other day

21:09

I never knew what dilate does

@Deleted User So if I dilate it myself it's fine ?

yeah, doesn't hurt

oh good great

@Deleted User maybe we should remove the dilate tool if it's not needed anymore?

21:15

so that people don't need to worry about applying it

It's needed for other graphics, like skins

21:16

just not for mapres

okay

21:16

is there anything that doesn't need it?

21:17

so I can add a CI check for dilation?

graphics without transparency

21:17

@heinrich5991 I added one once, my PR must still be around somewhere

but they don't get affected, no?

https://github.com/ddnet/ddnet/pull/2798

Add check_dilate (WIP) by def- · Pull Request #2798 · ddnet/ddnet

Not sure what files to check actually, many are apparently not dilated.

21:17

I just dilated and differed the resulting pictures

21:18

(using identify from imagemagick)

to sum it up, we could go with pre calculated alphas, that would require runtime calculations, for 1024*1024 around 3million float operations per texture, thats not alot, but for 4k images its a bit already on the other side i'd not require dilate at all anymore idc what we do GTA 5 for example does something similar to dilate other games prefer precalculated alphas

3million float operations per texture, thats not alot

computers

poggers

basename "$file" .png

21:19

uh nice trick, didn't know it

with zsh it's just ${i:r}, but apparently no one likes zsh except me

whats the trick

I didn't know you could pass a second parameter to basename

Image attachment

21:21

you should read the manual monkalaugh

monkalaugh

can't read the manual of each command to discover new options, way too many commands

21:22

I can read a manual if I know which tool I need

21:22

I'd probably have googled how to strip the .png in bash

Image attachment

21:22

well

21:22

but yeah

[ddnet/ddnet] Pull request reopened: #2798 Add check_dilate (WIP)

[ddnet/ddnet] Pull request opened: #3094 check_style CI no longer needs git history

[ddnet:master] 2 new commits

bf576de check_style CI no longer needs git history - heinrich5991 fd90740 Merge #3094 - bors[bot]

21:46

[ddnet/ddnet] Pull request closed: #3094 check_style CI no longer needs git history

What should I do with my pull request when I know there is a bug in it and I've already fixed it in my repo?

you can push your new commits

What's gonna happen? Will it be another request or it will be updated somehow magically?

22:03

Oh, I see now

@Deleted User about the PR. in what thread is stuff like CCommandProcessorFragment_OpenGL2::Cmd_RecreateBufferObject executed? main or graphics

22:18

graphics probably?

22:19

https://github.com/ddnet/ddnet/pull/3092/files?w=1#diff-442d7485882d4f383630c81a7061d34995210d7f694acf5eae9366d85a9f66e9

Graphics

okay, so the change is to hand ownership of some buffers over to the graphics thread?

Yeah. In this case only one tho I just implemented it for all for consistency

okay, so as I see it, this can only make thread-safety better

22:21

do I want to ask what ensures that the other pointers are still alive once the graphic thread receives them?

22:21

ah

22:22

the sending function takes care of that?

Yes. If that would fail tw wouldn't render correctly

ye, ok, I see

So it can't fail™️

22:23

Xd

[ddnet:master] 2 new commits

8e9baed Use pointer directly for tile layer building - Jupeyy 1fb4f90 Merge #3092 - bors[bot]

22:31

[ddnet/ddnet] Pull request closed: #3092 Use pointer directly for tile layer building

[ddnet/ddnet] Pull request opened: #3095 Disallow player name changes when the player is muted

Exported 884 message(s)