Teeworlds - bridge

Teeworlds

IRC / bridge

One-way IRC channel bridge. If you want to be able to send messages to IRC, contact @Dune or @heinrich5991. https://www.teeworlds.com/?page=docs&wiki=rules/irc_rules

Between 2020-03-11 00:00:00Z and 2020-03-12 00:00:00Z

I looked at it ... added a black layer behind it to have a better contrast, also added a grid to see tile boundaries.

zooming in also helped

Would you mind if I reimplement allocators in a more C++11 way, or preferably use std::allocator like allocators?

13:16

not C++14/17 mind you, the new allocator_traits stuff is just insane

that's C++03 compatible?

Eh, we still target C++03 don't we? :/

yes

13:44

should ask Oy when he's around I guess :)

The current way we do allocators is good for arrays and stuff, but when implementing a map, I need to rebind the allocator to another type

13:45

Someone give matricks a call, maybe he'd implement a quick hash table for me :P

we probably should have a discussion on containers in general. what should be reimplemented etc

I think it'd be nice if we'd just use stl containers at this point, the performance isn't too bad for us and we don't even use a custom allocator anymore

agree, as long as it doesn't introduce thick ass template abstraction levels

13:49

you don't want to end up with https://twitter.com/mcclure111/status/1078176148582998017

mcc (@mcclure111)

Sometimes the C++ compiler prints something and I feel like I'm trying to talk to Lassie. What is it, Lassie. Is something wrong, Lassie. Is Timmy okay. What the heck are you trying to communicate to me you fucking dog I don't speak Endless Barking

Retweets

790

Likes

3031

13:51

(this is a missing include)

why are you still targeting C++03 btw?

that's a controversial topic and there are many arguments for and against

guess it's too much work for now

Someone give GOD matricks a call,, you missed the prefix.

13:55

xD

@Assa that's not why, adding a compiler flag takes a second

thick ass template abstraction

13:56

c++11 would be nice, but yeah, c++ error messages are just utter garbage

@jxsl13 he missed the suffix too: GOD matricks, guardian of the holy grail

Thou shalt not take the name of the Lord thy God in vain

kinda Doom and Wolfenstein need to be somehow verwurstet(google translate says: sausaged) in that title..

Dune: do you have a good idea how I could pass the hash function for the key?

@ChillerDragon or @fokkonaut mind adding the logging patch that I did in that one issue to ddrace?

15:41

for joining and leaving players

?

this one: https://github.com/teeworlds/teeworlds/issues/2475

As soon as it’s in vanilla I guess

in 5 years then or never

git build failing?

16:49

g++.exe: error: `pkg-config: No such file or directory

16:49

huh, windows broke

16:49

https://github.com/teeworlds/teeworlds/runs/501145892?check_suite_focus=true

teeworlds/teeworlds

A retro multiplayer shooter. Contribute to teeworlds/teeworlds development by creating an account on GitHub.

16:55

hey Oy

hey

Learath2 and I were chatting about when to use STL containers earlier today

16:55

do you have some guidance on that? the baselib redefines some but we use STL at other places

16:55

I think Learath2 is implementing a hashmap of some sort

don't think we use stl containers

16:59

doubt there's aneed for that

for a hashmap?

17:01

grep -rin "include <" src | grep -v ".h>" | grep "<.*>"

17:01

gives <new> and <algorithm>

@oy can you merge the command usage pr

done when it's done

Side note: Since we're okay with changing the network even in vanilla, can we finally get rid of the nethash? If introduced today, I don't believe you'd merge the nethash, yet we keep it in the code. :( It doesn't offer protection either, it's just unneeded complexity…

17:38

why do we have this mentality that things that are in the code should stay, even when there are arguments in favor of removing them.

17:39

flow being the most recent example. we would never merge a half-finished, compiled-out by default feature into teeworlds

17:39

yet we're not removing it either. it can always be ressurected, even if it's gone from the current codebase

of course nethash offers protection

17:46

there's an issue about making flow work, so removing the current state is nonsense

it gives a warning when you change something, right?

just a warning

that issue is https://github.com/teeworlds/teeworlds/issues/2437 ?

Decide what to do with CFlow · Issue #2437 · teeworlds/teeworlds

CFlow is effectively unused since 0.5. Only Add and Update are called. Neither Get nor DbgRender are used, making these calculations effectively useless. The related config var is commented out //M...

yeah

I think #2437 is more like a question and would advocate for removal

17:47

we should actually make use of it and not have it sitting around in an incomplete state. [...] or it should just be removed,

yeah, should be made working

17:48

question is how the effects should be

ah, okay, that was unclear for me

and while it isn't working, it could be removed? we don't even know whether the code in question does something useful

17:49

I mean it's not entirely improbable that it sits there for another ten years

so? has been there for many years

17:55

it's a starting point to work on it

17:55

no problem with keeping it in

I think both viewpoints make a lot of sense

Oy: one could make the argument that logging of whispers and logging of a warning when nethash changes are similar

18:04

it doesn't change anything technical, but it changes something socially

18:04

currently, one accidentally looks at whispers even if you don't intend to

18:04

currently, one accidentally changes some network stuff but didn't intend to

18:05

with the change, one doesn't look at the whispers accidentally anymore

18:05

with the change, it's harder to notice that you accidentally changed the network code

18:05

re >But hiding that log meesage doesn't change anything - just gives you false-privacy.

one can advocate not to log any chat message in vanilla, people that want to log will do anyway

18:08

following KISS mantra, dead code should be removed

but using heinrich5991's words, I guess it does change something "socially" by pushing people to make it work :p

18:09

(though keep dead code is a bit extreme imo)

following client modding, CFlow could be seen as a still not used powerful feature provide by official

well the conversation looks like the issue author will log whispers if the message will be removed from official source or not

18:13

and people should be aware of that

18:13

that would be the false-privacy

but you don't make people aware by keeping it in the source code

18:13

if you want to make people aware, you put a message next to the whisper dialog

(about CFlow, imo, better removing to re-introduce later if useful)

keeping it in the source code with this reasoning is dishonest imo

18:14

(the whisper log)

18:15

(and seems to be one more example of "we keep it as it is". we probably wouldn't have blocked the whisper PR on it not logging the whisper messages in the server log)

if you want to make people aware, you put a message next to the whisper dialog

18:15

yeah, agreee, that is better

18:16

no chat privacy without p2p or central authentification :/

(about whisper, nothing is private from server-side, teeworlds is not a chat application, peolpe should be aware of it)

yeah, such a message would be good

and removing it from the source would still be good with such a message

18:17

so I as an honest server admin don't accidentally keep logs

i personally don't care if the whisper log message is in or not, was just pointing out people could log it anyway

18:17

and people shouldn't get false-safety

I believe it would be immensely better if it's not logged by default

okay. it sounded like you were against removing the log by default

A bit how cheating is possible in teeworlds, but it's better if we don't ship an aimbot as a console option

(and so a server admin do not have to care about chat message and privacy issues (what about ip?))

eh ips :/

why do we have this mentality that things that are in the code should stay, even when there are arguments in favor of removing them.

people try to argue "it was always there", which is actually not an argument but an excuse not to do anything.

ip is personal information according to heinrich5991 iirc

one step after another. it seems we almost agreed to do something to the whisper thing

off, have not read yet, as this is too active

if you try to fix everything at once, nothing happens

oof*

if I whisper someone my credit card # on an AAA game, and some staff uses it for their purpose, I'd be legitimately upset, even though it's not a chat program per se

if the server chat log messages gets removed it shouldn't be as plain text in the network log anymore

teeworlds doesn't have this privacy and it's probably not intuitive for the end user, and should be made clear

if tackling whisper, why not remove every chat messages ? (not too far from whisper)

rip econ mods that connect chat applications to server chat.

why do you use econ to mod @jxsl13 ? sounds overcomplicated

look at my git repo

18:26

that's why I stmbled upon the broken whisper logging

just answer

yes, I do

18:26

vpn detection

18:27

discord moderation tools that use econ

you should be able to mod the server via econ - don't need a tw client, or server is full

Oy: "mod" here was meant as modification not moderation

mod = modify or mod = moderate

at least my "mod"

jxsl13: fair point

when you said "rip econ mods" you meant moderators?

I meant modifications

18:28

and implicitly the moderation would be impossible as well

18:28

without joining the server

18:28

with a tw client

18:29

I got like votes shown in discord and chaz + team chat

fair enough

chat

Dune: ah ok

if people write "@mods"

18:29

the moderators in the discord server get pinged

18:30

might take a look at the chat (got logs of last 24h)

18:30

and if there is no chat that indicates anything, they can join the game and help there if needed

18:30

could also chat via server messages if needed, if it's only some question asked

18:30

and no actual funvoting or other offenses

It's not exactly trivial to implement a good hashmap and it's even more difficult with the way we use allocators in the tl right now. I don't really see any benefit to maintaining our own tl now that we don't even use a fancy allocator anymore

18:51

I highly doubt base/tl/array.h is better then std::vector, at the very least it has a much worse way of resizing itself (+1 element each time)

wasn't array resizing by 50% each time Oo.

@jxsl13 yeah it is https://github.com/teeworlds/teeworlds/blob/a50f2750c4a8fa5fd59e0a7d6821be34fb7e9856/src/base/tl/array.h#L320

lol, in add it says size +1

oh wow...

well, both used or one implemented but not added, but the other one documented

nonono, redix is right, it is 50%

size vs capacity

std vector grows by 100%?

that's an implementation detail and yeah it is most commonly 100%

19:00

vector is a really simple one, but I'm working on a hashmap, which is most commonly implemented with a rb tree for stl and usually uses open addressing, both are much more efficient then what we have in the codebase and what I would implement

has anyone ever done some benchmarks of stl vector vs tw vector? i guess the main reason for reimplementing it was that the stl vector pulls quite a lot of code into the binary... i doubt that it has a serious performance benefit

I bet for the tw use case the tw one is better

I highly doubt both drawbacks, not like we can avoid linking to libstdc++ nor is the naive implementation of an array in tw anything too well optimized for the usecase

I wonder why the tw array is used rather rarely, I think most uses are in the client, nothing or nearly nothing on the server side.

because in tw world dynamic memory allocation is a thing from hell

ah, k

19:08

satan begone. well perrormance-wise it is.

I don't even remember what I was coding the hashmap anymore...

but I think the server code is currently not the part that gets inefficient ober time.

19:08

ober

19:08

over

19:08

you wanted some privacy for whisper messages?

19:09

maybe the eeitor preview for images ;D

19:09

editor

Dynamic memory surely isn't very efficient but it's not the end of the world...

well you should not use dynamic allocations in performance critical code, yeah... but avoiding it like tw does is a bit too much i think

matricks rule was: "no allocation in the main loop" IIRC

well, if there is only one loop, yeah :D

we only have one main loop in client and server

Okay since we have to stay in 2003, can you review my hashmap heinrich5991?

Wolfenstein, same priciples :D..

gimme

the server has only one loop?

only one main loop?

I mean are there multiple threads anywhere

19:12

well, main loop is kinda implicitly only one?

ah. they're not used except for busy waiting and doing a host lookup every hour

The server does spawn threads to lookup for the maasters iirc

ah, k.

19:13

but full server logic seems to be in like thread.

no, only one thread

19:13

the main thread

one*

19:13

thread x)

@redix @jxsl13 if I did benchmarks, would that change your mind based on the result?

heinrich5991: sadly redix and jxsl13 aren't the ones we need to convince :P

I'm fully behind using c++17

19:19

behind you*

no thanks

And I'm highly in favour of refactoring Cvarint

19:20

c++11 would be enough actually :D

i honestly don't care about the vector/array thing too much. the tw implementation does not really limit us right now it was more of a general thing. i would be interested in some benchmarks though

C++14 is as far as I'm willing to go, after 17 things start to get weird, very weird

(Y)

19:21

the only thing about c++17 I like are structured bindings, that's about it

c++ brings a load of good things and also a load of bullshit. in the absence of strict coding guidelines, c++03 seems like a fair compromise to me

19:21

with great power comes great responsibility :p

19:21

c++11*

c++ should slowly be approached imo.

19:22

11

sadly I agree with dune

you can't write thread-safe code in pre-c++11

but using some stl (e.g. hashmap) and c++11 might be good

we have code that "works"

atomics, shared pointers, ...

well, every step forward are 10 shots of bullets in each of your legs :D

19:25

next step would be when the whole headers crap gets removed and replaced by some modular system.

+

in another 10-15 years.

Dune: we can regulate the amount of template mess we accept, we've been up to C++14 on ddnet for a year now, not that much bs made it in

c++ modules seems like a far future better move to rust then

Learath2, I'm sure it can be done, but it's very delicate, brings controversy, and I guess rejecting PRs for code style is not really in the spirit of the repo

well oy didn't like my goto :P

goto bad, use for loops with breaks >:D

19:30

to emulate goto

I prefer not having goto in the code

or while loops

I also removed all the remaining switch-case fallthroughs

fallthrough bad :D

gotos are the best way to handle errors and escape from nested loops

is there no feature in c++ to break outer loops?

The prejudice against it has to end, that one stupid blogpost about goto just made everyone into militants

that's code racism :D

escape from nested loops is ok, because the language can't handle it

19:33

errors can already be handled with return and RAII semantics, no need for it in C++ (edited)

nesting things into more functions is not a good solution

can rust break outer loops?

yes iirc

hm, newschool languages doing stuff right.

That switch case there is beautiful and clear, now I'll need to replace it with an ugly if-else chain or an ugly failure flag

golang has a great keyword called "defer"

19:35

why is there no sich thing in c++ :(

19:35

such

I'd love a break n; in C++ to break out of n blocks in general

go handles is with those assem ly style labels.

19:35

it

19:35

break label.

same for rust and java

hm, nice.

19:38

defer is like encapsulating a function call in some object and calling it within the destructor of that object.

@heinrich5991 do you recommend any rust books other than the one on the website?

no, I learned it using a much older version of the book

19:50

and writing code in a language you want to learn is absolutely viable

19:50

*vital

19:54

@jxsl13 what do you want to achieve in rust?

currently nothing, I'd love to take a look at the concepts used there, but have no actual project ideas yet

you should have something to program though, otherwise language learning can be hard

I know, especially wirh the learning curve in rust

19:56

steep

@heinrich5991 isn't c++03 thread-safe with proper libraries, eg boost?

@Dune no

20:02

there's even a paper about this

20:03

"Threads cannot be implemented in a library – Hans-J. Boehm"

20:04

the problem is that the language needs a memory model

20:05

and c++03 doesn't have that. you can't have thread-safety in c++11

heinrich5991: when does c++ become thread-safe? c++14?

c++11

but you said "you can't have thread-safety in c++11"

the last sentence had a typo

20:09

meant c++03

I see, is there any good reading on this? I'm interested in why a memory-model is required

Learath2: you could read the paper I quoted :P

@heinrich5991 does rust have somethinf like go channels?

rust has async-await, you'd usually use this

21:02

but there are also channels in the standard library if you prefer:

21:02

https://doc.rust-lang.org/std/sync/mpsc/fn.channel.html

std::sync::mpsc::channel - Rust

API documentation for the Rust channel fn in crate std.

I recently learned golang and its quite enjoyanle, im using it for web backends. I might learn rust next

21:03

Enjoyable*

21:04

golang is plenty fun

Yea it is very C like but super convenient

21:04

U can bind functions to structs

I got a makro for if err != nil { return nil, err }

21:04

makro key

And the interface{} and reflection are super powerul

21:05

I also like the annotation system

I like the forced conventions

Same

exporting stuff

it's only like C in its features

I also love u import modules by their git repo

you have a garbage collector

Yeah

[10:05 PM] Ryozuki: I also love u import modules by their git repo

google themselves just got bit by that btw

21:06

because they changed hosting providers

Well bad for them

21:06

the gc gets improved quite well

21:10

dependency management is great

21:10

defer = ❤️

21:10

python-like workflow

I dislike both python's and go's dependency workflow

21:11

they're "better than C", but that's the best that can be said about them

21:11

there are languages with working dependency management

go get

21:11

Bam

21:11

Done

not centralized and the forced package naming conventions are a drawback

U are using modern go modules right?

that's about it?

21:12

since 1.11 you are not forced to work in the gopath ._.

apparently something changed. can't comment about it then

21:12

according to @Ryozuki

Yeah in 1.11 they introduced go mod tooo

21:12

Tool

I have learnt it recently, so I can't comment on how the behaviour was pre 1.11

21:13

pythons dependency management is a paint

21:13

pain

https://blog.golang.org/using-go-modules

all I do is execute go mod init, iirc

21:14

with a name

21:14

in some project folder

21:15

and start programming in that folder

Yeah

https://github.com/jxsl13/twapi ;D

21:16

if anyone wanted to use that, the do import "github.com/jxsl13/twapi", then execute go get -d and can use that

21:16

they

The more I implement the more I find in the stl implementations, gcc and clang both have caching for the hash function :/

rabbit hole :0?

Exported 324 message(s)