fe2f73b518
Add new client component `CTouchControls` for playing the game with touch inputs. The touch controls can be enabled/disabled with the config variable `cl_touch_controls`, which defaults to `1` on Android and `0` on other platforms. The touch controls consist of various on-screen touch buttons. The different buttons are only shown when they are usable depending on the context. The touch button configuration is stored in `touch_controls.json`. Default touch button configuration ---------------------------------- Movement buttons for left, right and jump actions are arranged in a `⊥`-pattern similar to WASD controls. For the fire and hook action, two modes are implemented: 1. Direct touch input: the mouse is moved exactly where the player touches on the screen. 2. Virtual joystick: a button is used to emulate a joystick, which moves the mouse relative to the center of the screen. In either mode, a button is used to switch between the active actions (fire and hook). While the virtual joystick is being held down, this button uses the other action directly instead of switching it. The direct touch input mode can be enabled/disabled separately for ingame/spectating, to prevent accidental direct touch input when using a joystick. When spectating, direct touch input is used to allow panning the map directly like in an image/map viewer. Virtual joysticks can also be used to pan the map while spectating. Two separate buttons are shown to switch to the previous and next weapons. A hamburger menu button `☰` is used to toggle the visibility of lesser used touch buttons. This includes buttons for showing the scoreboard, showing the emoticon selector, showing the spectator menu, opening team and team chat, voting yes/no, and zooming. Long pressing the hamburger menu button will open the regular menu. When the dummy is connected, a button for swapping between main and dummy is shown. The emoticon selector and spectator menu are activated with the respective touch buttons and can be deactivated by touching outside of them or by using the back-button, as toggling them while the ingame touch button is pressed down is currently not feasible and also inconvenient at least for using the spectator menu. Touch button configuration format --------------------------------- The default button layout described above is loaded from `data/touch_controls.json`. This layout can be overridden by creating `touch_controls.json` in the user directory. The following button properties are adjustable: - Position and size (attributes `x`, `y`, `w`, `h`): the X/Y position and width/height are integers on a 1,000,000² grid. The unit grid values are converted to screen grid values at runtime in relation to the aspect ratio of the screen. This means buttons may appear stretched if the resolution is changed, but it allows us to provide a reasonable default for slightly different aspect ratios. If we save the absolute position instead, we would have the problem of buttons appearing outside the screen or not being aligned on different aspect ratios. - Shape (attribute `shape`): determines the shape of the button being rendered. - `"rect"`: rectangle shape. - `"circle"`: circle shape. The button size will automatically be adjusted so that width and height are identical. - Visibility (attribute `visibilities`): an array of predefined visibility classes can be selected and the button is only shown if all conditions are satisfied. An empty array means that the button is always shown. The following visibility classes are defined: - `"ingame"`: player is ingame, i.e. not spectating. - `"extra-menu"`, `"extra-menu-2"`, ..., `"extra-menu-5"`: the extra menu with the given number is activated. - `"zoom-allowed"`: zoom is allowed on the this server. - `"vote-active"`: a vote is currently active. - `"dummy-allowed"`: dummy is allowed on this server. - `"dummy-connected"`: dummy is currently connected. - `"rcon-authed"`: player is currently authed in rcon. - All visibility classes can be inverted by prefixing them with `-`, e.g. `"-ingame"` is satisfied when the player is not ingame, i.e. spectating. - Behavior (attribute `behavior`): an object which describes the behavior of this touch button when it is activated/deactivated as well as its label. The attribute `type` is used to differentiate which type of behavior is used. The behavior is either predefined (hard-coded) or a generic console command (bind). Predefined behavior is only used where necessary, all other buttons are represented as generic binds. - Predefined behavior (attribute `type` set to `"predefined"`): The attribute `id` is set to a fixed string value which determines the specific predefined behavior. The following predefined behaviors exist: - `"ingame-menu"`: Opens the ingame menu immediately when released. - `"extra-menu"`: The extra menu button which toggles visibility of buttons with `"extra-menu"`, `"extra-menu-2"`, `"extra-menu-3"` etc. visibilities. Also opens the normal menu on long press. - The attribute `"number"` specifies an integer between 1 and 5 to associate this button with the respective visibilities `"extra-menu"`, `"extra-menu-2"`, ..., `"extra-menu-5"`. - `"emoticon"`: Opens the emoticon selector (this does not work with binds). - `"spectate"`: Opens the spectator menu (this does not work with binds). - `"swap-action"`: Swaps the active action (fire and hook) for direct touch input and virtual joysticks. - `"use-action"`: Uses the active action with the current aiming position. - `"joystick-action"`: Virtual joystick which uses the active action. - `"joystick-fire"`: Virtual joystick which always uses fire. - `"joystick-hook"`: Virtual joystick which always uses hook. - Bind behavior (attribute `type` set to `"bind"`). Buttons with this behavior work like regular key binds. - The attribute `"label"` specifies as a string the label of the button. - The attribute `"label-type"` specifies as a string the type of the label of the button, i.e. how the attribute `"label"` is interpreted: - `"plain"`: Label is used as is. - `"localized"`: Label is localized. Only usable for the default buttons for which there are translations available. - `"icon"`: Icon font is used for the label. Icons must be encoded using `\uXXXX`, e.g. `\uf3ce` for the mobile phone icon. - The attribute `"command"` specifies as a string the command to execute in the console like a bind when this button is used. Ingame menu buttons ------------------- In addition to the separate on-screen touch controls, a second row is added to the main page of the ingame menu when `cl_touch_controls` is enabled for less frequently used functions which are otherwise not usable without a keyboard: - Buttons to open the local and remote consoles. Opening the local console without the touch controls is useful because error messages would be shown there if the touch controls configuration could not be loaded. - Button to close the menu, which is more convenient than using the virtual back button if it's not always shown. - Checkbox for toggling the touch controls editor UI (see below). Ingame touch controls editor ---------------------------- The user interface to adjust the touch controls is rendered over the main screen of the ingame menu when enabled. For now, the implementation of the touch controls editor is limited to basic configuration management functions. - Saving the configuration to the `touch_controls.json` file in config directory. - Discarding the current changes by reloading the `touch_controls.json` file from config directory. - Restoring the default configuration by reloading the `touch_controls.json` file from data directory. - Displaying whether there are unsaved changes. - Importing and exporting the configuration from and to the clipboard. This is the only way to edit the configuration on newer Android versions, as editing files within apps' storage is not possible anymore. Furthermore, the global touch controls settings can be adjusted in this UI: - Direct touch input can be enabled/disabled separately while ingame and while spectating. While the touch controls editor is active, all buttons are shown regardless of their visibility, to better support arranging the buttons. Future tasks ------------ - Implement more usable UI for adjusting the button layout. - Add `color` property to touch buttons to adjust the default background color? - Support zooming gesture. - Optimize touch button rendering using text and quad containers. |
||
---|---|---|
.cargo | ||
.github | ||
.gitlab | ||
cmake | ||
data | ||
datasrc | ||
ddnet-libs@e78f350e78 | ||
man | ||
other | ||
scripts | ||
src | ||
.clang-format | ||
.clang-tidy | ||
.editorconfig | ||
.gitattributes | ||
.gitignore | ||
.gitlab-ci.yml | ||
.gitmodules | ||
.mailmap | ||
.pylintrc | ||
Cargo.lock | ||
Cargo.toml | ||
CMakeLists.txt | ||
codecov.yml | ||
deny.toml | ||
Dockerfile | ||
Doxyfile | ||
formatting-revs.txt | ||
license.txt | ||
lsan.supp | ||
memcheck.supp | ||
README.md | ||
storage.cfg | ||
ubsan.supp | ||
valgrind.supp |
Our own flavor of DDRace, a Teeworlds mod. See the website for more information.
Development discussions happen on #ddnet on Quakenet (Webchat) or on Discord in the developer channel.
You can get binary releases on the DDNet website, find it on Steam or install from repository.
- Code Browser
- Source Code Documentation (very incomplete, only a few items are documented)
If you want to learn about the source code, you can check the Development article on the wiki.
Cloning
To clone this repository with full history and external libraries (~350 MB):
git clone --recursive https://github.com/ddnet/ddnet
To clone this repository with full history when you have the necessary libraries on your system already (~220 MB):
git clone https://github.com/ddnet/ddnet
To clone this repository with history since we moved the libraries to https://github.com/ddnet/ddnet-libs (~40 MB):
git clone --shallow-exclude=included-libs https://github.com/ddnet/ddnet
To clone the libraries if you have previously cloned DDNet without them, or if you require the ddnet-libs history instead of a shallow clone:
git submodule update --init --recursive
Dependencies on Linux / macOS
You can install the required libraries on your system, touch CMakeLists.txt
and CMake will use the system-wide libraries by default. You can install all required dependencies and CMake on Debian or Ubuntu like this:
sudo apt install build-essential cargo cmake git glslang-tools google-mock libavcodec-extra libavdevice-dev libavfilter-dev libavformat-dev libavutil-dev libcurl4-openssl-dev libfreetype6-dev libglew-dev libnotify-dev libogg-dev libopus-dev libopusfile-dev libpng-dev libsdl2-dev libsqlite3-dev libssl-dev libvulkan-dev libwavpack-dev libx264-dev python3 rustc spirv-tools
On older distributions like Ubuntu 18.04 don't install google-mock
, but instead set -DDOWNLOAD_GTEST=ON
when building to get a more recent gtest/gmock version.
On older distributions rustc
version might be too old, to get an up-to-date Rust compiler you can use rustup with stable channel instead or try the rustc-mozilla
package.
Or on CentOS, RedHat and AlmaLinux like this:
sudo yum install cargo cmake ffmpeg-devel freetype-devel gcc gcc-c++ git glew-devel glslang gmock-devel gtest-devel libcurl-devel libnotify-devel libogg-devel libpng-devel libx264-devel make openssl-devel opus-devel opusfile-devel python2 rust SDL2-devel spirv-tools sqlite-devel vulkan-devel wavpack-devel
Or on Fedora like this:
sudo dnf install cargo cmake ffmpeg-devel freetype-devel gcc gcc-c++ git glew-devel glslang gmock-devel gtest-devel libcurl-devel libnotify-devel libogg-devel libpng-devel make openssl-devel opus-devel opusfile-devel python2 SDL2-devel spirv-tools sqlite-devel vulkan-devel wavpack-devel x264-devel
Or on Arch Linux like this:
sudo pacman -S --needed base-devel cmake curl ffmpeg freetype2 git glew glslang gmock libnotify libpng opusfile python rust sdl2 spirv-tools sqlite vulkan-headers vulkan-icd-loader wavpack x264
Or on Gentoo like this:
emerge --ask dev-db/sqlite dev-lang/rust-bin dev-libs/glib dev-libs/openssl dev-util/glslang dev-util/spirv-headers dev-util/spirv-tools media-libs/freetype media-libs/glew media-libs/libglvnd media-libs/libogg media-libs/libpng media-libs/libsdl2 media-libs/libsdl2[vulkan] media-libs/opus media-libs/opusfile media-libs/pnglite media-libs/vulkan-loader[layers] media-sound/wavpack media-video/ffmpeg net-misc/curl x11-libs/gdk-pixbuf x11-libs/libnotify
On macOS you can use homebrew to install build dependencies like this:
brew install cmake ffmpeg freetype glew glslang googletest libpng molten-vk opusfile rust SDL2 spirv-tools vulkan-headers wavpack x264
If you don't want to use the system libraries, you can pass the -DPREFER_BUNDLED_LIBS=ON
parameter to cmake.
Building on Linux and macOS
To compile DDNet yourself, execute the following commands in the source root:
mkdir build
cd build
cmake ..
make -j$(nproc)
Pass the number of threads for compilation to make -j
. $(nproc)
in this case returns the number of processing units.
DDNet requires additional libraries, some of which are bundled for the most common platforms (Windows, Mac, Linux, all x86 and x86_64) for convenience and the official builds. The bundled libraries for official builds are now in the ddnet-libs submodule. Note that when you build and develop locally, you should ideally use your system's package manager to install the dependencies, instead of relying on ddnet-libs submodule, which does not contain all dependencies anyway (e.g. openssl, vulkan). See the previous section for how to get the dependencies. Alternatively see the following build arguments for how to disable some features and their dependencies (-DVULKAN=OFF
won't require Vulkan for example).
The following is a non-exhaustive list of build arguments that can be passed to the cmake
command-line tool in order to enable or disable options in build time:
-
-DCMAKE_BUILD_TYPE=[Release|Debug|RelWithDebInfo|MinSizeRel]
An optional CMake variable for setting the build type. If not set, defaults to "Release" if-DDEV=ON
is not used, and "Debug" if-DDEV=ON
is used. SeeCMAKE_BUILD_TYPE
in CMake Documentation for more information. -
-DPREFER_BUNDLED_LIBS=[ON|OFF]
Whether to prefer bundled libraries over system libraries. Setting to ON will make DDNet use third party libraries available in theddnet-libs
folder, which is the git-submodule target of the ddnet-libs repository mentioned above -- Useful if you do not have those libraries installed and want to avoid building them. If set to OFF, will only use bundled libraries when system libraries are not found. Default value is OFF. -
-DWEBSOCKETS=[ON|OFF]
Whether to enable WebSocket support for server. Setting to ON requires thelibwebsockets-dev
library installed. Default value is OFF. -
-DMYSQL=[ON|OFF]
Whether to enable MySQL/MariaDB support for server. Requires at least MySQL 8.0 or MariaDB 10.2. Setting to ON requires thelibmariadbclient-dev
library installed, which are also provided as bundled libraries for the common platforms. Default value is OFF.Note that the bundled MySQL libraries might not work properly on your system. If you run into connection problems with the MySQL server, for example that it connects as root while you chose another user, make sure to install your system libraries for the MySQL client. Make sure that the CMake configuration summary says that it found MySQL libs that were not bundled (no "using bundled libs").
-
-DTEST_MYSQL=[ON|OFF]
Whether to test MySQL/MariaDB support in GTest based tests. Default value is OFF.Note that this requires a running MySQL/MariaDB database on localhost with this setup:
CREATE DATABASE ddnet;
CREATE USER 'ddnet'@'localhost' IDENTIFIED BY 'thebestpassword';
GRANT ALL PRIVILEGES ON ddnet.* TO 'ddnet'@'localhost';
FLUSH PRIVILEGES;
-
-DAUTOUPDATE=[ON|OFF]
Whether to enable the autoupdater. Packagers may want to disable this for their packages. Default value is ON for Windows and Linux. -
-DCLIENT=[ON|OFF]
Whether to enable client compilation. If set to OFF, DDNet will not depend on Curl, Freetype, Ogg, Opus, Opusfile, and SDL2. Default value is ON. -
-DVIDEORECORDER=[ON|OFF]
Whether to add video recording support using FFmpeg to the client. Default value is ON. -
-DDOWNLOAD_GTEST=[ON|OFF]
Whether to download and compile GTest. Useful if GTest is not installed and, for Linux users, there is no suitable package providing it. Default value is OFF. -
-DDEV=[ON|OFF]
Whether to optimize for development, speeding up the compilation process a little. If enabled, don't generate stuff necessary for packaging. Setting to ON will set CMAKE_BUILD_TYPE to Debug by default. Default value is OFF. -
-DUPNP=[ON|OFF]
Whether to enable UPnP support for the server. You need to installlibminiupnpc-dev
on Debian,miniupnpc
on Arch Linux. Default value is OFF. -
-DVULKAN=[ON|OFF]
Whether to enable the vulkan backend. On Windows you need to install the Vulkan SDK and set theVULKAN_SDK
environment flag accordingly. Default value is ON for Windows x86_64 and Linux, and OFF for Windows x86 and macOS. -
-GNinja
Use the Ninja build system instead of Make. This automatically parallelizes the build and is generally faster. Compile withninja
instead ofmake
. Install Ninja withsudo apt install ninja-build
on Debian,sudo pacman -S --needed ninja
on Arch Linux. -
-DCMAKE_CXX_LINK_FLAGS=[FLAGS]
Custom flags to set for compiler when linking. -
-DEXCEPTION_HANDLING=[ON|OFF]
Enable exception handling (only works with Windows as of now, uses DrMingw there). Default value is OFF. -
-DIPO=[ON|OFF]
Enable interprocedural optimizations, also known as Link Time Optimization (LTO). Default value is OFF. -
-DFUSE_LD=[OFF|LINKER]
Linker to use. Default value is OFF to try mold, lld, gold. -
-DSECURITY_COMPILER_FLAGS=[ON|OFF]
Whether to set security-relevant compiler flags like-D_FORTIFY_SOURCE=2
and-fstack-protector-all
. Default Value is ON.
Running tests (Debian/Ubuntu)
In order to run the tests, you need to install the following library libgtest-dev
.
This library isn't compiled, so you have to do it:
sudo apt install libgtest-dev
cd /usr/src/gtest
sudo cmake CMakeLists.txt
sudo make -j$(nproc)
# copy or symlink libgtest.a and libgtest_main.a to your /usr/lib folder
sudo cp lib/*.a /usr/lib
To run the tests you must target run_tests
with make:
make run_tests
Code formatting
We use clang-format 10 to format the C++ code of this project. Execute scripts/fix_style.py
after changing the code to ensure code is formatted properly, a GitHub central style checker will do the same and prevent your change from being submitted.
On Arch Linux you can install clang-format 10 using the clang-format-static-bin AUR package. On macOS you can install clang-format 10 using a homebrew tap:
brew install r-lib/taps/clang-format@10
sudo ln -s /opt/homebrew/Cellar/clang-format@10/10.0.1/bin/clang-format /opt/homebrew/bin/clang-format-10
Using AddressSanitizer + UndefinedBehaviourSanitizer or Valgrind's Memcheck
ASan+UBSan and Memcheck are useful to find code problems more easily. Please use them to test your changes if you can.
For ASan+UBSan compile with:
CC=clang CXX=clang++ CXXFLAGS="-fsanitize=address,undefined -fsanitize-recover=address,undefined -fno-omit-frame-pointer" CFLAGS="-fsanitize=address,undefined -fsanitize-recover=address,undefined -fno-omit-frame-pointer" cmake -DCMAKE_BUILD_TYPE=Debug .
make
and run with:
UBSAN_OPTIONS=suppressions=./ubsan.supp:log_path=./SAN:print_stacktrace=1:halt_on_errors=0 ASAN_OPTIONS=log_path=./SAN:print_stacktrace=1:check_initialization_order=1:detect_leaks=1:halt_on_errors=0 LSAN_OPTIONS=suppressions=./lsan.supp ./DDNet
Check the SAN.* files afterwards. This finds more problems than memcheck, runs faster, but requires a modern GCC/Clang compiler.
For valgrind's memcheck compile a normal Debug build and run with: valgrind --tool=memcheck ./DDNet
Expect a large slow down.
Building on Windows with the Visual Studio IDE
Download and install some version of Microsoft Visual Studio (At the time of writing, MSVS Community 2022) with C++ support.
You'll have to install both Python 3 and Rust as well.
Make sure the MSVC build tools, C++ CMake-Tools and the latest Windows SDK version appropriate to your windows version are selected in the installer.
Now open up your Project folder, Visual Studio should automatically detect and configure your project using CMake.
On your tools hotbar next to the triangular "Run" Button, you can now select what you want to start (e.g game-client or game-server) and build it.
Building on Windows with standalone MSVC build tools
First off you will need to install the MSVC Build Tools, Python 3 as well as Rust.
To compile and build DDNet on Windows, use your IDE of choice either with a CMake integration (e.g Visual Studio Code), or by deprecated using the CMake GUI.
Configure CMake to use the MSVC Build Tools appropriate to your System by your IDE's instructions.
If you're using Visual Studio Code, you can use the CMake Tools extension to configure and build the project.
You can then open the project folder in VSC and press Ctrl+Shift+P
to open the command palette, then search for CMake: Configure
This will open up a prompt for you to select a kit, select your Visual Studio
version and save it. You can now use the GUI (bottom left) to compile and build your project.
Cross-compiling on Linux to Windows x86/x86_64
Install MinGW cross-compilers of the form i686-w64-mingw32-gcc
(32 bit) or
x86_64-w64-mingw32-gcc
(64 bit). This is probably the hard part. ;)
Then add -DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/mingw64.toolchain
to the
initial CMake command line.
Cross-compiling on Linux to WebAssembly via Emscripten
Install Emscripten cross-compilers (e.g. sudo apt install emscripten
) on a modern linux distro.
If you need to compile the ddnet-libs for WebAssembly, simply call
# <directory to build in> should be a directory outside of the project's source directory
scripts/compile_libs/gen_libs.sh <directory to build in> webasm
from the project's source directory. It will automatically create a directory called ddnet-libs
.
You can then manually merge this directory with the one in the ddnet source directory.
Then run emcmake cmake .. -DVIDEORECORDER=OFF -DVULKAN=OFF -DSERVER=OFF -DTOOLS=OFF -DPREFER_BUNDLED_LIBS=ON
in your build directory.
To test the compiled code locally, just use emrun --browser firefox DDNet.html
To host the compiled .html file copy all .data
, .html
, .js
, .wasm
files to the web server. (see /other/emscripten/minimal.html for a minimal html example)
Then enable cross origin policies. Example for apache2 on debian based distros:
sudo a2enmod header
# edit the apache2 config to allow .htaccess files
sudo nano /etc/apache2/apache2.conf
# set AllowOverride to All for your directory
# then create a .htaccess file on the web server (where the .html is)
# and add these lines
Header add Cross-Origin-Embedder-Policy "require-corp"
Header add Cross-Origin-Opener-Policy "same-origin"
# now restart apache2
sudo service apache2 restart
Cross-compiling on Linux to macOS
Install osxcross, then add
-DCMAKE_TOOLCHAIN_FILE=../cmake/toolchains/darwin.toolchain
and
-DCMAKE_OSX_SYSROOT=/path/to/osxcross/target/SDK/MacOSX10.11.sdk/
to the
initial CMake command line.
Install dmg
and hfsplus
from
libdmg-hfsplus and newfs_hfs
from
diskdev_cmds
to unlock the package_dmg
target that outputs a macOS disk image.
Importing the official DDNet Database
$ wget https://ddnet.org/stats/ddnet-sql.zip
$ unzip ddnet-sql.zip
$ yaourt -S mariadb mysql-connector-c++
$ mysql_install_db --user=mysql --basedir=/usr --datadir=/var/lib/mysql
$ systemctl start mariadb
$ mysqladmin -u root password 'PW'
$ mysql -u root -p'PW'
MariaDB [(none)]> create database teeworlds; create user 'teeworlds'@'localhost' identified by 'PW2'; grant all privileges on teeworlds.* to 'teeworlds'@'localhost'; flush privileges;
# this takes a while, you can remove the KEYs in record_race.sql to trade performance in queries
$ mysql -u teeworlds -p'PW2' teeworlds < ddnet-sql/record_*.sql
$ cat mine.cfg
sv_use_sql 1
add_sqlserver r teeworlds record teeworlds "PW2" "localhost" "3306"
add_sqlserver w teeworlds record teeworlds "PW2" "localhost" "3306"
$ mkdir build
$ cd build
$ cmake -DMYSQL=ON ..
$ make -j$(nproc)
$ ./DDNet-Server -f mine.cfg
Installation from Repository
Debian/Ubuntu
$ apt-get install ddnet
MacOS
$ brew install --cask ddnet
Fedora
$ dnf install ddnet
Arch Linux
$ yay -S ddnet
FreeBSD
$ pkg install DDNet
Windows (Scoop)
scoop bucket add games
scoop install games/ddnet
Benchmarking
DDNet is available in the Phoronix Test Suite. If you have PTS installed you can easily benchmark DDNet on your own system like this:
$ phoronix-test-suite benchmark ddnet
Better Git Blame
First, use a better tool than git blame
itself, e.g. tig
. There's probably a good UI for Windows, too. Alternatively, use the GitHub UI, click "Blame" in any file view.
For tig
, use tig blame path/to/file.cpp
to open the blame view, you can navigate with arrow keys or kj, press comma to go to the previous revision of the current line, q to quit.
Only then you could also set up git to ignore specific formatting revisions:
git config blame.ignoreRevsFile formatting-revs.txt
(Neo)Vim Syntax Highlighting for config files
Copy the file detection and syntax files to your vim config folder:
# vim
cp -R other/vim/* ~/.vim/
# neovim
cp -R other/vim/* ~/.config/nvim/