/libuavcan

Portable reference implementation of the UAVCAN protocol stack in C++ for embedded systems and Linux.

Primary LanguageC++MIT LicenseMIT

UAVCAN stack in C++

Coverity Scan Travis CI Gitter

Portable reference implementation of the UAVCAN protocol stack in C++ for embedded systems and Linux.

UAVCAN is a lightweight protocol designed for reliable communication in aerospace and robotic applications via CAN bus.

Documentation

Library usage

Dependencies

  • Python 2.7 or 3.3 or newer

Note that this reporitory includes Pyuavcan as a submodule. Such inclusion enables the library to be built even if pyuavcan is not installed in the system.

Cloning the repository

git clone https://github.com/UAVCAN/libuavcan
cd libuavcan
git submodule update --init

If this repository is used as a git submodule in your project, make sure to use --recursive when updating it.

Using in a Linux application

Libuavcan can be built as a static library and installed on the system globally as shown below.

mkdir build
cd build
cmake .. # Default build type is RelWithDebInfo, which can be overriden if needed.
make -j8
sudo make install

The following components will be installed:

  • Libuavcan headers and the static library
  • Generated DSDL headers
  • Libuavcan DSDL compiler (a Python script named libuavcan_dsdlc)
  • Libuavcan DSDL compiler's support library (a Python package named libuavcan_dsdl_compiler)

Note that Pyuavcan (an implementation of UAVCAN in Python) will not be installed. You will need to install it separately if you intend to use the Libuavcan's DSDL compiler in your applications.

It is also possible to use the library as a submodule rather than installing it system-wide. Please refer to the example applications supplied with the Linux platform driver for more information.

Using with an embedded system

For ARM targets, it is recommended to use GCC ARM Embedded; however, any other standard-compliant C++ compiler should also work.

With Make

Please refer to the documentation at the UAVCAN website.

With CMake

In order to cross-compile the library with CMake, please follow the below instructions. You will need to provide a CMake toolchain file, Toolchain-stm32-cortex-m4.cmake in this example. If you're not sure what a toolchain file is or how to prepare one, these instructions are probably not for your use case; please refer to the section about Make instead.

mkdir build
cd build
cmake .. -DCMAKE_TOOLCHAIN_FILE=../cmake/Toolchain-stm32-cortex-m4.cmake
make -j8

Library development

Despite the fact that the library itself can be used on virtually any platform that has a standard-compliant C++11 compiler, the library development process assumes that the host OS is Linux.

Prerequisites:

  • Google test library for C++ - gtest (dowloaded as part of the build from github)
  • C++11 capable compiler with GCC-like interface (e.g. GCC, Clang)
  • CMake 2.8+
  • Optional: static analysis tool for C++ - cppcheck (on Debian/Ubuntu use package cppcheck)

Building the debug version and running the unit tests:

mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Debug
make -j8
make ARGS=-VV test

Test outputs can be found in the build directory under libuavcan.

Note that unit tests suffixed with "_RealTime" must be executed in real time, otherwise they may produce false warnings; this implies that they will likely fail if ran on a virtual machine or on a highly loaded system.

Contributors, please follow the Zubax C++ Coding Conventions.

Vagrant

Vagrant can be used to setup a compatible Ubuntu virtual image. Follow the instructions on Vagrantup to install virtualbox and vagrant then do:

vagrant up
vagrant ssh
mkdir build
cd build
mkdir build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Debug -DCONTINUOUS_INTEGRATION_BUILD=1

Note that -DCONTINUOUS_INTEGRATION_BUILD=1 is required for this build as the realtime unit tests will not work on a virt.

You can build using commands like:

vagrant ssh -c "cd /vagrant/build && make -j4 && make test"

or to run a single test:

vagrant ssh -c "cd /vagrant/build && make libuavcan_test && ./libuavcan/libuavcan_test --gtest_filter=Node.Basic"

Developing with Eclipse

An Eclipse project can be generated like that:

cmake ../../libuavcan -G"Eclipse CDT4 - Unix Makefiles" \
                      -DCMAKE_ECLIPSE_VERSION=4.3 \
                      -DCMAKE_BUILD_TYPE=Debug \
                      -DCMAKE_CXX_COMPILER_ARG1=-std=c++11

Path ../../libuavcan in the command above points at the directory where the top-level CMakeLists.txt is located; you may need to adjust this per your environment. Note that the directory where Eclipse project is generated must not be a descendant of the source directory.

Submitting a Coverity Scan build

First, get the Coverity build tool. Then build the library with it:

export PATH=$PATH:<coverity-build-tool-directory>/bin/
mkdir build && cd build
cmake <uavcan-source-directory> -DCMAKE_BUILD_TYPE=Debug
cov-build --dir cov-int make -j8
tar czvf uavcan.tgz cov-int

Then upload the resulting archive to Coverity.

Automatic check can be triggered by pushing to the branch coverity_scan.