Chromebook Dev. Setup

Jan 17, 2020

Had to blow up my Chromebook setup due to some bad lib linking in R. Here's a blow-by-blow of getting back to where I was. Useful if you what a Chromebook that can be used for coding in Python or R in Jupyter Notebooks

Step 1. Get into dev mode

  • Hold ESC and Refresh at the same time, then hit Power.
  • Ctrl-D when screen with yellow "!" appears

Step 2. Update .bashrc file

While the aliases aren't needed, they're useful and once you get used to them, they're hard to live without. The exports are needed later on and you might as well add them now.

alias cls='clear'
alias gh='history | grep'
alias ll='ls -laF'
alias h='history | tail -n 10'

# Below are needed for R to work correctly
export EDITOR='vim'
export TMPDIR=/usr/local/tmp
export PAGER=/usr/local/bin/less

Step 3. Set developer passwd

This is recommended to help offset the vulnerability you create by running the machine in Developer mode.

$ sudo chromeos-setdevpasswd

Step 4. Install Chromebrew

Chromebrew is a package manager which allows you to install programs onto your Chromebook. It's similar to the Mac Homebrew package manager.

  • Download and install Chromebrew installer: 
$ cd ~/Downloads # This is the only dir that is writable to you at the moment
$ curl https://raw.githubusercontent.com/skycocker/chromebrew/master/install.sh -o install.sh
$ bash ./install.sh

This will download and install a compiler toolchain which includes gcc and other useful items that are needed to compile source code on your machine.

NOTE: The list of installed packages is contained in /usr/local/etc/crew/device.json. It is often better to remove a package entry from here and then reinstall it, rather than 'crew remove <pkg", which can cause problems when done on low level pkgs like gcc8 and libiconv.</p>

Step 5. Install some pre-requisites

NOTE: When installing from Chromebrew, I always use the "-s" flag to the install command. This builds the package locally from source, which I find to be much more reliable than installing from precompiled packages.

  • Install hdf5 library (needed for NetCDF)
 $ crew install -s hdf5
  • Reinstall libxml2 (used by ALOT of things. Best to reinstall it from source)
$ crew remove libxml2
$ crew install -s libxml2
  • Download and install NetCDF
    Go here, http://github.com/Unidata/netcdf-c/releases, and download the .tar.gz of the version you want. I'm using the latest (4.7.3) here. We need this before we build and install GDAL and R so that they have support for working with .nc files.
$ ll ~/Downloads/netcdf*
-rw-r--r--. 1 chronos chronos 18410112 Jan 18 11:06 netcdf-c-4.7.3.tar.gz

$ cd /usr/local/share
$ cp ~/Downloads/netcdf-c-4.7.3.tar.gz ./
$ tar xzvf netcdf-c-4.7.3.tar.gz 
$ cd netcdf-c-4.7.3
$ ./configure --prefix=/usr/local --libdir=/usr/local/lib64 --disable-dap
$ make check install
  • More prereqs that I'd prefer to install from source
$ crew install -s python27 openjpeg geos proj4 curl
# After removing from devices.json
$ crew install -s libiconv
$ crew install -s gettext
$ crew install -s cairo
  • Make sure numpy is installed before installing GDAL
$ pip install numpy
  • Install GDAL
$ crew install -s gdal

Step 7. Install R

$ crew install -s r

NOTE: This will add a ton of other packages as part of the dependency tree.

Step 8. Install Jupyter

This one is easy. And by all that is holy, do not infect your machine with Anaconda. It might work fine on a normal machine, but here it will create havoc.

$ pip install --upgrade pip
$ pip install jupyter

Step 9. Install IR Kernel

This you do from within R, but first we need to install the system libs for ZeroMQ, and they were incompatible with the latest version of gcc 8 that I had.

Had to download Version 4.3.2 from https://github.com/zeromq/libzmq/releases/download/v4.3.2/zeromq-4.3.2.tar.gz

$ crew install -s zeromq 
$ R
> install.packages("Cairo")
> install.packages('IRkernel') 
> IRkernel::installspec()
  • Set cairo as default device in IRKernel
$ vi /usr/local/lib64/R/library/IRkernel/kernelspec/kernel.json
{"argv": ["R", "--slave", "-e", "options(bitmapType='cairo') ; "IRkernel::main()", "--args", "{connection_file}"],
 "display_name":"R",
 "language":"R"
}

Step 10. Install Python dependencies

$ pip install matplotlib

Step 11. Debug

  • R notebook won't open because it lacks a PNG device
$ crew remove cairo
$ crew install -s cairo
  • Fix Python ascii decode error
$ vi /usr/local/lib/python2.7/site-packages/nbformat/sign.py

# Add to the beginning of the script
reload(sys)
sys.setdefaultencoding('utf8')
Report abuse