Teeworlds - bridge

Oy: I'd like to introduce c++11 dependence with some patch, so we can get some thread safety into teeworlds. is that okay?

00:02

it would get us e.g. std::shared_ptr and atomic variables

huuh :(

00:02

it would get us a lot of bad practices too

00:02

that's a controversial one :p

currently the code around threads is bad

I don't deny that, but a lot of cpp11 additions are undesirable in a project with the philosophy of Teeworlds. Once you accept cpp11, it's harder to say no to people for using auto, abusing lambdas etc.

e.g. the m_Shutdown variable in CJobPool is (according to the standard) undefined behavior

is there no way around that in c++03?

c++03 doesn't have a threading model

00:05

it pretends threading does not exist

pls add threadsafty..

Teeworlds benefit(ed) from a simple, easy code base

no, there is no way around it in c++03

so you can't thread properly in C?

rip c

C11 also brings thread support

ah I see :)

but it is less widely supported than C++11 IIRC (e.g. msvc++)

C11 is not supported by msvc?

at least not until very recently. let me check

@heinrich5991 i found another way now, but still thanks!

we don't have extern C code anymore in Teeworlds right

we have some

btw, discords working again

some of our own code, and our dependencies

grep "extern C" returned no result

grep "extern "C""

ah, mi bad :)

https://developercommunity.visualstudio.com/idea/387315/add-c11-support.html

Add C11 Support - Developer Community

Developer Community for Visual Studio Product family

00:10

so no, not supported

ugh :(

(but that's different from C++11, which is supported)

00:11

C++11 can work together with C90/C99 with no problems

yeah I know

access to sockets should be threadsafe, I'd guess

00:11

printing to console maybe as well ._.?

no, printing to console is not

I don't see Teeworlds having performance issue with those things anyway

I'm running db access in separate threads and the main thread, which is tries to aquire mutexes and if it fails, the main thread's not blocked, if it succeeds it simply processes message queues :o, so threads make sense, if one would like to access e.g. current worldinformation, train his bot detection, etc in a different thread

would rather stick to c++03, if it's not really necessary

hm ok

00:20

the problem with the current CJob interface is that you must guarantee that your CJob struct does not move (or is deallocated) until the job finishes

00:21

that means that if you cannot have a CJob in a structure which is freed before program termination

00:21

-if

that means I need to know the exact amount of requests I want to make a the same time at compile time

that's not a problem for teeworlds currently right?

I checked all the usages of CJob, they look safe

00:34

the favorites name lookup could benefit from parallelism, which isn't really possible right now

why favorites in particular?

the three things that use engine jobs right now are: 1) master server name lookup 2) favorite name lookup 3) sound loading

00:35

master server name lookup and sound loading use one CJob per task

00:35

favorite name lookup uses one CJob for all favorites

right now the jobpool is only using one thread anyway

Dune: no, i only own one xbox360 controller

do you consider practices like 'auto' undesirable?

13:52

std::vector<std::pair<int, CCharacter*>>::iterator it vs auto it

in some cases, they're fine imo. iterators especially, because of nested templates like that

exactly, i think iterators are the most common use case

don't need it for iterators I think because you have for loop sugar in c++11?

?

I don't see it being used very often for some reason I don't know

for(auto it : vector)

for(const int &i : v)

13:55

no need for an explicit iterator there

uh, i didn't know that, that's nice

13:56

std::for_each() for the java people i guess <.<

for (String item : someList) { System.out.println(item); }

13:57

java had it before c++

i know, c++ has it since c++11

yes, but we don't use c++11

13:58

dunno what you mean with std::for_each() for the java people

I usually only need C++ features like map (or hashmap), iterators, shared pointers and (unfortunately) vectors

why unfortunately vectors?

https://en.cppreference.com/w/cpp/algorithm/for_each

for each

those sound better than c++'s hash map support

vectors are like a standard container

Maybe lets dump c++ and recode teeworlds on lolcode

you can often optimize by using another container

https://en.wikipedia.org/wiki/LOLCODE

riir!

I'd say vectors are still the most needed container

and the least flawed, saved for the string :p

instead of using an array for all the Entities, you could use a quadtree and optimize all collision loops, but I always wondered if teeworlds would benefit

14:04

this would require some benchmarks, too

14:05

to be more on topic: I think a greater benefit would be better string rendering xD

for the c++17 people for(auto&[key value]: map)

where is the masterserver requests located? retrieving server list etc?

16:40

are

@jxsl13 the requests are located in src/engine/client/serverbrowser.cpp, the response parsing is done in src/engine/client/client.cpp

thx

does someone remember when I complained about the Dynamic Camera, that it feels like laging when I spin around with it? seems like Fullscreen is the reason for it, window mode fixed my problem but still dont know why it's so "laggy" with fullscreen (edited)

@ShootXen disable vsync

it's already

enable then

can't game itself runs like 30fps for me then

a script that fixes style issues would be nice

yeah

it's not that easy, unfortunately

17:33

I tried writing a script that detects style issues, and even that was quite hard

C++ is notoriously hard to parse :/

I did it with libclang

17:45

C++ is undecidable to parse

yeah, lots of cases you have to consider

17:48

could do it step by step

https://gist.github.com/heinrich5991/2ad5bdd9c7013b31ee1759858bec6c8a

Script to extract identifiers from Teeworlds sources

Script to extract identifiers from Teeworlds sources - extract_identifiers.py

17:50

this extracts all identifiers from the teeworlds source code

17:50

I think Learath2 wrote something to parse its output

clang-format does a good job... but it forces spaces after and before most operators, which basically means reformatting the entire codebase =\

I thought it was about identifiers

Just detection would be fine imo

oh didn't read that... that's another thing

cool that would help with BE or AE