Merge pull request #3 from bakercp/feature-0.8.4
|6 years ago|
|README.md||6 years ago|
|Vagrantfile||6 years ago|
|fixQualifiedLibraryPaths.sh||6 years ago|
|openframeworks-pi.yml||6 years ago|
|playbook.yml||6 years ago|
This package uses Vagrant and Ansible to spin up a virtual machine that runs on Mac, Linux, or Windows(?). In no time at all, you’ll have an environment that:
You won’t have to do any of the things on the Raspberry Pi Cross-Compiling Guide page. I have contributed to that guide to make the process more smooth, but it’s still a mess.
Here, Vagrant automates the process of setting up a Virtualbox virtual machine and Ansible automates the process of setting up the cross-compiler, the NFS server, and OpenFrameworks.
The Raspberry Pi is slow. This environment will let you compile OpenFrameworks applications on your fast desktop.
Though I built this virtual machine with OpenFrameworks in mind, it’ll work just fine for any cross-compiling task.
If you’re writing code that runs on a single Raspberry Pi, NFS-booting lets you edit code locally on your desktop. There’s no need to SSH into a Pi to edit and compile.
Because the root partition is loop-mounted from a .img file, you can later create a standalone SD card by merely dd’ing it SD cards.
The magic comes when you’re building a cluster of Raspberry Pis. There’s no need to rsync or to reflash a stack of SD cards. The latest code is accessible on every Pi at all times. At Two Bit Circus, a common design pattern is to have a cluster of NFS-booted Raspberry Pis all choreographed by a single Linux server. This virtual machine serves as the starting point for that design pattern.
ln -s 2014-06-20-wheezy-raspbian.img image.img
tools-master.zip. Leave it compressed.
You only have to do this if you’re not using 2014-06-20-wheezy-raspbian.
The NFS root with which you’re booting the Pi lives in the image.img file. You need to calculate the offsets to the boot and root partitions on that device. On OSX you can just type
file image.img. The relevant information here is the start sector for the boot (1st) and root (2nd) partitions. Multiply the start sector by the block size (512) to get the byte offset. Put these numbers in
offset_root in playbook.yml.
10.0.0.2/255.0.0.0. The IP address of the virtual machine is hard-coded to
vagrant up. It will probably ask you to select the network interface to bridge to. Select your wired connection.
vagrant provisionto retry the provisioning process.
vagrant sshto connect to and begin using your new environment.
The provisioning process in the preceding section has already modified your SD card image to enable NFS booting. We need to write the boot partition and not the root partition to an SD card. Insert an SD card into your computer. It can be tiny. We’re only writing about 60MB to it.
dd if=image.img of=/dev/rdiskX ibs=512 obs=1m count=<root_offset_sectors>. The root offset is the offset of the root partiion from before, but divided by 512. On my SD card, that number is 122880. This particular incantation only copies the first partition (the boot partition) to the SD card. We don’t want a root partition on this card, because it’ll be using the NFS share. Remember to unmount the partition but not eject the disk.
cmdline.txtfile in the newly minted SD card. It assigns the static IP address 10.0.0.101 (which you can change for subsequent cards) and designates 10.0.0.1 as the NFS server
It’s worth noting that the ansible provisioning process has already altered the /etc/fstab on the root partition on the SD card image to inhibit mounting the root partition from the SD card.
The virtual machine is set up with the tools to do this automatically.
vagrant sshto drop into a shell on your virtual machine.
ssh firstname.lastname@example.org log in (the password is
raspberry). Immediately log back out by typing ‘exit’. We did this to add the host to the known_hosts file.
ansible-playbook --ask-pass -i "10.0.0.101," /vagrant/openframeworks-pi.yml(the password is ‘raspberry’). Note: I may have gone overboard here. This does the same as running install_dependencies.sh on the Raspberry Pi from /opt/openframeworks/scripts/linux/linux_armv6l. If you go this route, there’s an http_proxy running at http://10.0.0.1:8888
When you’re ssh’ed into your virtual machine, you can access the root partition in /opt/raspberrypi/root. Dig deeper, and you’ll find /opt/raspberrypi/root/opt/openframeworks. This is the armv6 OpenFrameworks directory, uncompressed and ready to go. It’s symlinked to /opt/openframeworks for simplicity.
The virtual machine has an alias
rmake which calls the cross-compiler with the appropriate options.
From your vagrant shell:
cd /opt/openframeworks/apps/myApps/emptyExample rmake
From your Raspberry Pi:
cd /opt/openframeworks/apps/myApps/emptyExample bin/emptyExample
Note: If you find yourself running out of disk space you might consider removing several large packages that are installed by default on the 2014-06-20-wheezy-raspbian image. To do this, ssh into the pi (
vagrant ssh and then
ssh email@example.com) and then issue the command
sudo apt-get remove wolfram-engine squeak-vm scratch. This should free up about 500MB.
Big deal, right? Well, consider what you can do with this development environment.
Want to run a network of Raspberry Pis all with the same codebase, or with access to the same shared media? Flash some more SD cards. Change
cmdline.txt to set a different IP address for each. They’ll all boot up and have access to that same root partition.
Are you finished developing and want to flash a stand-alone SD card that doesn’t require NFS booting? Simply
dd the entire image.img file to an SD card. You’ve been editing that image all along!
There’s some things you’ll have to do first:
If you put an OpenFrameworks project in the rpi-build-and-boot directory, and change config.make to point the OpenFrameworks root at /opt/openframeworks, you can compile in XCode on the Mac side AND compile from the /vagrant directory.