Hound -- named after the dog with excellent sense of smell -- is a library for Linux intended to easily and performantly read data from sensors on embedded devices. Common sensors handled include CAN and many analog sensors (e.g. accelerometer, gyroscope, and anything else handled by the Linux Kernel Industrial I/O subsystem). The idea is to abstract away the differences between sensor types by binding all the sensor to a common API. In doing so, application code can write to a single API without concern for the underlying device, driver. In addition, application code can worry less about a host of performance issues, like queueing samples, minimizing latency, etc.
-
meson:
pip3 install meson -
ninja:
pip3 install ninja -
libgps:
sudo apt install libgps-dev -
libYAML:
sudo apt install libyaml-dev -
libmosquitto:
sudo apt install libmosquitto-dev. Also a mosquitto broker for the unit tests:sudo apt install mosquitto. -
libmsgpack:
sudo apt install libmsgpack-dev. -
xlib: compile from source: https://github.com/XevoInc/xlib
-
yobd: compile from source: https://github.com/XevoInc/yobd
-
(optional, for developers)
clang-tidy. This is used for static analysis and is thus a build but not runtime requirement. -
(optional, for developers)
valgrind. This is used in unit tests. -
(optional, for developers) Python requirements, as documented in
dev-requirements.txt. These are some build-time tools that are not required for runtime. They are used for sanity-checking schemas. You can install them from your distro or viapip install -r requirements.txt.
Note that, if any of your build prerequisites do not come from standard distro packaging, you will need also need to tweak the following env vars:
PKG_CONFIG_PATHneeds to be set only when you runmesonand doesn't matter after that. It should be set to the directory containing the.pcfiles used by the prerequisite you built.LD_LIBRARY_PATHneeds to be set whenever you actually load the Hound library, such as when you run the unit tests withninja test. It should be set to the directory containing the built prerequisite libraries.
To get your build requirements, you just need to run:
sudo apt-get -y install meson ninja-build
To get your build requirements, you just need to run:
sudo dnf -y install meson ninja-build
Note that on fedora you will substitute the ninja-build command instead of
the ninja command for the various build commands on this page.
mkdir build
cd build
meson ..
ninja
To rebuild at any time, you just need to rerun the last ninja command:
ninja
You can run this command from any directory you like. For instance, to rebuild
from the top level directory, use the ninja -C option to point ninja at the
build directory:
ninja -C build
Also, there is nothing special about the directory name build. If you prefer a
different name than build, this is not a problem, and you can have different
build directories with different configurations; meson and ninja don't care.
If you want to use clang, it's the usual meson methodology:
mkdir build.clang
cd !$
CC=clang meson ..
ninja
To run the unit tests, first use the can-setup script to setup virtual CAN
devices. Note that this script requires sudo:
ninja can-setup
After that, you can use either of these commands to run the tests:
ninja test
meson test
To run the tests under valgrind, you can use:
meson test --setup valgrind
Note that ninja test actually calls meson test.
Before checking in, you should run:
ninja check
Which runs unit tests, does static analysis, and anything else that is deemed
"good hygiene" prior to checking in. This list may change over time, but the
check target should remain valid.
Static analysis uses clang-tidy and can be run with:
ninja clang-tidy
Note that you will need to install the clang-tidy tool, either via distro or
some other method.
Code documentation is handled by doxygen and can be built with:
ninja docs