This is the platform support for my Raspberry Pi-based flight controller.

Hardware:

• Raspberry Pi Model B
• Pololu Micro Maestro serial servo controller
• CH Robotics UM-6 and Pololu GPS GP-9 via Pololu USB-to-serial adapter

• Get CMake to export a statically linkable library
• Delete main.c
• Get sending commands to the IMU to work properly
• Get IMU health report to be more meaningful (ensure has GPS lock at least)
• Write some kind of tests

• Add Rpi tools repository as a submodule rather than cloning it into toolchain Docker image
• Connect to and read from IMU
• Create C API for driving servos
• Create C API for reading IMU
• Get Rpi to wiggle a servo using Maestro example bash script
• Ship source code from laptop to Rpi
• Set up cross compiler
• Cross-compile, deploy binary, and run on Rpi in one command
• Get Rpi to wiggle a servo using Maestro example C code
• Bring platform/src to root
• Move src/maestro to src/platform/servo_controller

• Get Docker on Rpi dumb idea
• Build minimal development container with compiler needed a cross-compiler instead
• Compile and run source code within development container run a binary on the Rpi natively

Install Arch Linux on an SD card. Make sure to resize the partition to take up the available space. (Use fdisk to rewrite the partition table and resize2fs to do the actual resizing. Google it.)

Make sure you've got Docker on your laptop and your docker command is set up to talk to a Docker daemon. On OS X, I use boot2docker. Ensure boot2docker is up and running;

$ boot2docker up

The core build script builds a Docker container with a compiler toolchain that targets the Rpi and then invokes compiler.

$ ./build.sh

The first time you run this it will take a while to build the docker image. Successive runs will be faster.

We scp the resulting binary onto the Rpi (make sure your public key is on the Rpi so you don't have to type your password) and then run it.

./deploy.sh [email protected]

Congratulations, you're hacking!

For some reason the program doesn’t always die on the Pi when you expect it to. You might need to:

$ ssh [email protected] killall fcs-platform

I wanted to be able to work on the Pi without having to plug it into a router, rather, I wanted to plug it into my Macbook Air directly via ethernet and the Thunderbolt ethernet adapter. OS X Internet Sharing was being a contrarian as I attempted to do this. After much wailing and gnashing of teeth I arrived at this process:

1. Disable Internet Sharing
2. Delete the following files: /Library/Preferences/SystemConfiguration/com.apple.nat.plist /Library/Preferences/SystemConfiguration/NetworkInterfaces.plist /Library/Preferences/SystemConfiguration/com.apple.network.identification.plist /Library/Preferences/SystemConfiguration/com.apple.airport.preferences.plist
3. Reboot.
4. Re-enable Internet Sharing.
5. It worked!

Create /etc/bootptab to cause the Pi to always get the same IP address:

%%
raspberry-pi 1 b8:27:eb:c:45:4f 192.168.2.200 

Do not go about trying to hand-edit /etc/bootpd.plist, because the Internet Sharing dialog in System Preferences is meant to stomp on that file. Tailing /var/log/system.log was instructive.

I also modified /System/Library/LaunchDaemons/bootps.plist to add the -dv flag so bootpd would emit more interesting log messages:

...
<key>ProgramArguments</key>
<array>
    <string>/usr/libexec/bootpd</string>
    <string>-dv</string>
</array>
....

TIL that this is useful for cat-ing the raw output from the GPS:

$ stty -F /dev/ttyUSB0 57600 raw