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 2020-10-18 00:00:00Z and 2020-10-19 00:00:00Z

Meh, no way to really know which version a save originated on, so all solo and race saves from today have been deleted

@Learath2 I think you asked earlier where DDNet stuff is built

00:12

where was it?

00:13

ddnet.tw?

ddnet.tw

okay

00:16

A bug affecting mostly solo and race servers was noticed earlier today. Only records abusing the bug have been deleted. Since there is no way to know which version a save was created on all saves created today on race and solo were deleted.

00:16

@Learath2 contradiction: only records abusing the bug have been deleted <-> all race and solo records were deleted

All race and solo saves were deleted

00:17

Records are there unless they were played on a broken revision

so remove "Only records abusing the bug have been deleted."?

Well I didn't only delete saves, I also deleted ranks

00:18

I deleted all saves, I didn't delete all ranks

ah

00:18

aah

00:18

I see

Did I word that badly?

no, I didn't see that you were talking about saves

00:19

@Learath2 did the teehistorian git revision for the server work?

Any reason it shouldn't work?

00:20

I extracted it from the json header of the teehistorian file

because someone (I thought you) said they were broken

Oooh

00:20

ooooooooh I didn't think of that

00:20

Ah the SHA was fine

but?

only the version was broken

00:21

because of caching

hum

00:21

how does that work?

00:21

we don't generate the version I think?

I think I fixed it with #3123

https://github.com/ddnet/ddnet/pull/3123

Add workaround for CMake 2.8. Fix version caching by Learath2 · Pul...

@heinrich5991 We kinda sorta do now

Tell me about the bug

We parse version.h then we redefine GAME_RELEASE_VERSION

okay, that's backwards

00:22

@ᶰ°Konͧsti a pull request made some commands only available inside map configs e.g. sv_solo_server but when merging we forgot that we use some of these in our resetfiles not in map configs

00:24

@heinrich5991 I wanted to instead generate version.h entirely, but I got sidetracked validating the shader code with glslang

so maps that didnt have sv solo server in settings were able with dummy?

@ᶰ°Konͧsti yep

@Learath2 we should decide on either doing it from cmake or from source

but most servers hadn't updated yet

00:25

@heinrich5991 we need to be able to override it on the command-line, so it should probably be in cmake

I'd prefer source, so we can still compile ddnet with g++ **.cpp **.c (pkg-config libs)

xd

but you can do command line as well

00:26

we also need to generate source

i remember /map Bumba, vote solo server, do /map Bumba and then chose any solomap to finish with dummy

what is configure_file(src/game/version.h vd.h)

It's a workaround for cmake 2.8 which we claim to support

When make server tunes able for zones

00:28

Like disable hammer or collision

have you tried compiling it on cmake 2.8?

it just copies version.h to vd.h and thus creates a dependence

ah

00:29

I see

honestly, I couldn't be assed to download ancient cmake

yea

00:29

what's not tested doesn't work anyway

00:29

i.e. I doubt it worked on cmake 2.8 before your change

It'd be nice if you could clean up the version stuff

00:30

I think we only need a way to suffix our releases

When make server tunes able for zones

@ᶰ°Konͧsti aren't all tunes available in tunezones?

no

00:31

Hammer off isnt available for tune zones

I wonder what that breaks which made us disable some tunes

Only whole map

00:31

its not disabled

it's not a tune

00:31

I think

Just not possible

That would explain it

00:32

@ᶰ°Konͧsti yeah there is no hammer off tune

but a server setting

00:33

and if u want it off for a specific zone only u still have to use spam tiles

00:33

u can set strenght to 0

00:33

But it would still unfreeze

00:34

Collislion zone also doesnt exist i think

I'm off, can't learn vocabulary like that, I'm constantly distracting myself with discord

00:35

good night

@heinrich5991 what vocabulary?

you know it

ah, gl

Gn8 Heinreich

@ᶰ°Konͧsti can't make server settings behave like that sadly

cant just create a tune zone for it? xd

One could maaybe code something like a "config zone" but it'd break a whole lot of things and waste a TON of memory

00:39

Maybe more tunes can be added

[ddnet:master] 3 new commits

ae90bdf Show halloween skins only on Halloween - def- 08045f6 Only assign skin when non-Halloween skin has been chosen - def- 91773cd Merge #3116 - bors[bot]

00:40

[ddnet/ddnet] Pull request closed: #3116 Show halloween skins only on Halloween

Maybe more tunes can be added

@Learath2 thats what i mean

[ddnet/ddnet] Pull request opened: #3131 Get CMake 2.8 working

It only required fixing one weird CMake 2.8 oddity. I guess we should merge this if we are keeping to claim 2.8 support.

Mind boggling

Did you change tee collisions or deep fly somehow?

01:57

or hammer logic

About those weird crashes on Windows, i made this thing: https://github.com/ccomrade/c1-launcher/blob/master/Code/Launcher/CrashLogger.cpp It catches all unhandled fatal SEH exceptions within the process and logs stacktrace and few other things. No external tools or libraries are needed. It uses just basic WinAPI. The code is not so nice because it works with WinAPI and it also needs to be compatible with VS2005 compiler that has no C++11 support. To obtain nice stacktraces with source file names and line numbers, you need to provide symbol files (.pdb) next to your EXEs/DLLs. In CMake you can use -DCMAKE_BUILD_TYPE=RelWithDebInfo to generate them. Also, it seems MinGW doesn't use the Microsoft's proprietary debug info format, so it can't generate symbol files that VS or Windows itself can parse probably. If it can help, feel free to use the code. Both 32-bit and 64-bit code is supported. VS debugger is still much more powerful than this thing, but at least something.

ccomrade/c1-launcher

Open-source Crysis executables. Contribute to ccomrade/c1-launcher development by creating an account on GitHub.

omg new ddnet servers messes with my skin how can i opt out @heinrich5991 ?

08:15

wait is it forced?

08:15

should be events 1 imo

08:16

guess its time to downgrade the server :/

08:16

oh sorry heinrich looks like @deen is to blame

08:16

sorry for the spam im lil pissed xd

oh btw @deen the demo that recorded the kill feed bug does not exist. Probably another bug that kills autorecording on timeout protection.

08:25

i rly dont understand the halloween code and how to turn it off neither events 1 nor events 0 helps

[ddnet/ddnet] Pull request opened: #3132 Fix halloween logic

It showed halloween skins all the time :/

[ddnet:master] 2 new commits

c81eb50 Fix halloween logic - ChillerDragon b3769a0 Merge #3132 - bors[bot]

08:44

[ddnet/ddnet] Pull request closed: #3132 Fix halloween logic

ChillerDragon: halloween only gets set to true when events 1 is set

i fixed it in a pr @deen

08:49

u were only showing halloween skins when halloween is not set

08:49

i guess uv seen it xd

08:50

maybe i can get a thsirt after all this yr :D

[ddnet:master] 2 new commits

4f57396 Get CMake 2.8 working - Learath2 e88a7de Merge #3131 - bors[bot]

08:52

[ddnet/ddnet] Pull request closed: #3131 Get CMake 2.8 working

meh can u guys even tell a difference? https://zillyhuhn.com/cs/.1603012286.png

Embedded image

Hey, i'm back

wb

@ChillerDragon delete IRC when

Yes, left is vector, right is old

correct deen but still

[ddnet/ddnet] Pull request opened: #3133 add a checklist for PRs (fixes #3130)

[ddnet/ddnet] Issue opened: #3134 Various memory leaks in MySQL on real server

@Zwelf Could you take a look? Recorded on real server. (GER2, 8305) ``` ================================================================= ==4360==ERROR: LeakSanitizer: detected memory leaks Direct leak of 360712 byte(s) in 11 object(s) allocated from: #0 0x51fb22 in operator new(unsigned long) (/home/teeworlds/servers/DDNet-Server-asan+0x51fb22) #1 0x561d77 in CServer::GetAnnouncementLine(char const*) /home/teeworlds/src/master/src/engine/server/server.cpp:3563:20 #2 0x...

10:58

[ddnet/ddnet] Issue opened: #3135 Various Undefined Behaviour on real server run

GER2:8309: ``` /home/teeworlds/src/master/src/base/vmath.h:106:17: runtime error: division by zero #0 0x545572 in vector2_base normalize(vector2_base const&) /home/teeworlds/src/master/src/base/vmath.h:106:17 #1 0x7091b2 in CCharacterCore::Tick(bool) /home/teeworlds/src/master/src/game/gamecore.cpp:130:25 #2 0x510354 in CCharacter::Tick() /home/teeworlds/src/master/src/game/server/entities/character.cpp:772:9 #3 0x5e828e in CGameWorld::Tick() /home/teeworlds/src/master/s...

10:59

[ddnet/ddnet] Pull request opened: #3136 Fix some asan issues

[ddnet/ddnet] Pull request opened: #3137 Add gfx debug

If someone wants to try, you have to execute the client with a config that contains dbg_gfx 1 If opengl debug is supported it will show "Enabled OpenGL debug mode" in the console

a seemingly empty name on ger1 :8298

14:25

ㅤ:

@redix what's your github again? You made the F-client for 0,7, right?

Nope. It's from @fokkonaut

f-client by fallenKN :troll:

14:41

@onbgy

14:42

@trml here?

no u

[ddnet/ddnet] Pull request opened: #3138 Fix out of bounds access in GetSolo

As discovered in #3135

14:55

[ddnet/ddnet] Pull request opened: #3139 Use NUM_CLIENTS for dummy array length

Should make code more readable by explaining that the array is used for dummy. Also makes modders life easier if there is a need to scale up dummy count. I left out two places that do not include client.h https://github.com/ddnet/ddnet/blob/3be8a592e52a03e555b6aa29a8dea529797bb645/src/game/client/prediction/gameworld.h#L92 https://github.com/ddnet/ddnet/blob/3be8a592e52a03e555b6aa29a8dea529797bb645/src/game/gamecore.h#L212 Moving the enum to protocol.h for example is probably not ...

@deen i think server commands that return a time should also use hours

15:17

in accordance with the client

15:18

Image attachment

[ddnet/ddnet] Issue opened: #3140 /mapinfo doesnt show time with hour format

i think only mapinfo is the one left out

Image attachment

15:41

demos don't do it as well

We have to be careful with the format for "finished in: ". It's parsed by the client

15:51

I'll fix the "Finish time:"

We should probably add a warning comment for the "finished in" string

will do

[ddnet/ddnet] Pull request opened: #3141 More hours

Will also look into #3140 and fix here.

why does the client need to parse that?

for demos and ghost names

oh

16:06

then we can hack the client

16:06

and display it with hours

16:06

after parsing

nono

monkalaugh

i think there's a finish time packet already, use that in the client

16:07

and then wait a few months and you can change the format in server

we will have to remember

you can add a static_assert that checks the system time and once some date passed fails the compilation with an error message that reminds you

16:07

16:08

(not a serious suggestion)

justatest

Amazing idea

a.k.a. how to make package managers hate you

[ddnet/ddnet] Pull request opened: #3142 Fix two more serverbrowser filter out of bounds

what do you guys think of some way to make it easier to maintain a custom ddnet client? Like a api. Functions to hook in a plugin system or even some lua thingy like ath has. Something where one can modify the ddnet client without touching the source so hard that there is a merge conflict every month.

16:45

i played some minecraft these days and its awesome how users can combine client mods as easy as drag and dropping them into a mods/ folder. Would be fun for ddnet imo.

not really fun how bad minecraft performs tho

haha yes its the worst

16:46

i never looked into ath but maybe their system is not too bad

a clean code base would be nicer, than spamming stuff with more and more plugin systems

do we have a system already?

no

or is more and more a hyperbel of 1?

the plugin system would be required to maintained just as the code, so it would break stuff here and there too i guess.. the code isnt ready for this in a bigger style

hm

16:48

there is a way to provide a antibot lib already hm?

16:48

that is something similar isnt it

well i dunno what u need but ofc its doable, as said it just needs cleanup

what is dirty atm?

adding hooks everywhere is dirty

hm

16:50

so any ideas how to do it nice?

allow more polymorphism with better structured code (edited)

yikes

16:50

arent there nice hook loops already for all the components

mh? you cannot overload them can you?

https://github.com/ddnet/ddnet/blob/master/src/game/client/gameclient.cpp#L320

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

16:52

just add an api component

16:52

or something like that

u just want to add a component?

16:52

how does lead to merge conflicts

im nub im just spitting out ideas

16:52

what do you mean?

add your component somewhere.. finished

oh no i did never add a component everytime i changed something i did it directly in place

then u have all access to the component interface

Polymorphism is costly with how shitty compilers are at devirtualization

https://steamcommunity.com/app/412220/discussions/0/5118863332777631303/

ddnet-info.json failed :: DDraceNetwork General Discussions

что делать если при заходе в игру пишет Dowload ddnet-info.json failed и больше не находит карты и сервера?? ːsteamsadː

17:06

all 3 people have 1 in common, they are all russian

I actually had an interesting project for a while. I stole the idea from unreal engine. You tag the function with a macro and the preprocessor generates pre and post hooks for it

I've reported a similar issue few days ago

@Learath2 sounds interesting xd

https://discordapp.com/channels/252358080522747904/293493549758939136/766319349232435200

@murpi best is you ask @deen to silence that warning again, it might be too aggressive

PeepoShrug

@murpi would be great to see a log when it happens next. Also there is #3080

https://github.com/ddnet/ddnet/issues/3080

Make the ddnet-info.json warning less fatal looking · Issue #3080 ·...

It does look very worrying for a quite recoverable error :D

how do get log's though? (edited)

maybe they just blocked by ddnet.tw?

17:10

or vice versa

@murpi you can either have your console on, or you can add logfile client.log to your config

where do you find the said log?

I think deen said it appears in the config directory, but I have absolutely no idea tbh with you

Polymorphism is costly with how shitty compilers are at devirtualization

@Learath2 is llvm shit at that?

for me it says it's in C: but it's not

17:11

hm okey

@Ryozuki llvm is currently the best, though idk if the improvements landed in stable yet

i kinda like a lot the idea of llvm

but polymorphism is clean and still fast, its just an additional pointer loaded

Gcc has made progress too but theirs is just too agressive for now to be enabled

from vtable

@Deleted User yeah but it is a cost to be kept in mind and imho avoided if possible

17:13

RTTI isn't too pretty either

@Learath2 is it wise to compile cmake stuff with clang instead of gcc?

17:14

export CXX=/usr/bin/clang++ monkalaugh

monkalaugh

@Ryozuki I do both before pushing nowadays.

i rly want to use clang only

17:14

better error reporting

gcc is still a bit faster probs, but they got closer

17:14

(the executable)

I'm much more excited about lld

17:15

If lld finally becomes stable we can finally ditch the old binutils

LLD is a linker from the LLVM project that is a drop-in replacement for system linkers and runs much faster than them.

17:15

yeah

no

use rust tools + llvm utils

xD

ez

LLD is quite slow yet, maybe in theory its faster xD

thats the first sentence at their web tho

17:15

https://lld.llvm.org/

17:15

monkalaugh

17:16

LLD is very fast. When you link a large program on a multicore machine, you can expect that LLD runs more than twice as fast as the GNU gold linker. Your mileage may vary, though.

17:16

@Deleted User maybe at multicore is better

@Deleted User you sure? I remember the chromium team reporting it was much better

i dunno, i used it a few times, and linking was slower

btw it says gold linker

17:16

apparently a thing used for rly big programs

17:17

instead of ld

Gnu gold is also nice

You can embed LLD in your program to eliminate dependencies on external linkers

:o

17:17

Link-time optimization (LTO) is supported by default.

17:17

@Deleted User ur lto

17:18

monkalaugh

Also lld, clang, llvm have such pretty codebases

Image attachment

17:18

this is their test

If you ever read gcc's source, its actually painful

17:18

Like you want to die amounts of painful

@Deleted User r u sure ld is faster?

17:19

Image attachment

17:19

monkalaugh

when i used it was defs slower

when did u use it

It might have been back when you tried it. It's got a lot of improvements over the last 3-4 years

2010?

but i dont compile huge programs like a compiler

17:20

no like 2-3 months ago

tee_thinking

and im at clang 12 xd

You sure you didnt misconfigure it or sth? :D

don't think so

17:20

only thing i could imagine is as said, the size of the project

i only see claims being lld faster

I never saw any benchmarks with smaller projects tho, maybe there is some overhead thats much more obvious when its slower

17:21

Smaller*

@Ryozuki just tell me a project where u tested both and i test too

can i test it with ddnet

you can

how to change the linker xd

-fuse-ld=lld

isnt there a envar

-DLLVM_ENABLE_PROJECTS=lld directly into cmake?

17:26

dunno

17:26

not really aware of the llvm cmake stuff

17:27

just add to cpp flags

ddnet e88a7dee0

17:28

hm

17:28

but how do i time link time specifically

17:28

lmao

Just do it in CFLAGS and CXXFLAGS

CXX=clang++ CC=clang CXXFLAGS='-fuse-ld=gold' CFLAGS='-fuse-ld=gold' cmake .. -GNinja
CXX=clang++ CC=clang CXXFLAGS='-fuse-ld=gold' CFLAGS='-fuse-ld=gold' time ninja

Well also LDFLAGS

oh shit i used gold

17:28

lmao

weird i can not reproduce the ping crash on debian+gnome

oh its gold

17:29

i mconfused

anyone using arch+gnome here? or any distro w a uptodate gnome version?

didnt have lld installed xd

It would be nice if we could get a -DDEBUG_ASAN to enable Asan easily

real 0m17.006s user 2m32.348s sys 0m8.089s

17:31

clang-10: warning: argument unused during compilation: '-fuse-ld=gold' [-Wunused-command-line-argument]

17:31

why is it using this

17:31

omg

17:32

argument unused during compilation: '-fuse-ld=lld' [-Wunused-command-line-argument]

17:32

monkalaugh

You probably need to clean up your build dir

i do xd

17:32

oh

17:33

it doesnt recognice lld

17:33

as an argument

Image attachment

18:34

monkalaugh

Ryozoozki

[ddnet:master] 2 new commits

bdb53e3 Fix two more filter out of bounds - Jupeyy c04621e Merge #3142 - bors[bot]

20:40

[ddnet/ddnet] Pull request closed: #3142 Fix two more serverbrowser filter out of bounds

@Ryozuki was there anything itneresting in logs?

oh idk

21:39

i dont save the logs :/

[ddnet/ddnet] Pull request opened: #3143 Ignore ddnet-info.json dl failure again

too aggressive, might just be bad internet at the moment

[ddnet-web:master] 1 new commit

d9203d1 Remove Umbra from iCTF and Vanilla mods - def-

@Ryozuki tested now, lld is indeed faster

Exported 328 message(s)