Building Gateway or Adapter on CentOS

abhijit.nathwani · August 28, 2019, 7:42am

Hi Team,

I was exploring building the gateway/adapter on the CentOS. The build fails while trying to build sqlite. It fails with the toolchain arm-linux-gnueabihf-gcc not found. I tried installing the toolchain externally but none of them work.

Do you have an exact package name for the required toolchain or Do we support building on CentOS or not?

Thanks,
Abhijit Nathwani

mrstegeman · August 28, 2019, 3:26pm

The gateway’s README includes instructions for Fedora, which should be very close (if not identical) to what’s needed for CentOS.

For adapters, there is typically just a package.sh script that needs to be run to build.

I’m not sure why you’re hitting an error for an ARM toolchain. What are you running CentOS on?

abhijit.nathwani · August 29, 2019, 5:54am

Hi @mrstegeman,

When running the ./build-gateway.sh script on CentOS to build the gateway image, it tries to download the node module gateway-addon-node which in turn downloads sqlite3. The pre-build binaries for sqlite3 are not found on the url it tries to fetch, so it fall backs to compile it from source. That’s where the ARM toolchain comes into play. and it fails on CentOS. I tried various other relevant toolchains, but it is failing at exactly this step everytime.

Please find the attached screenshot of the error.

Thanks,
Abhijit

abhijit.nathwani · August 29, 2019, 5:58am

The ARM CC specified here in the script:

github.com

mozilla-iot/gateway/blob/0fa650e6c4ddbe18293f847426d3eb087b7af4bf/image/build-gateway.sh#L44


# it to the same path under /rpxc/sysroot
#
# My guess is that this would be fine for chrooted apps, but I don't
# think that the cross compilers run chrooted.
LIBUDEV_SO=/lib/arm-linux-gnueabihf/libudev.so.1.5.0
sudo mkdir -p $(dirname ${LIBUDEV_SO})
sudo ln -s ${SYSROOT}/${LIBUDEV_SO} ${LIBUDEV_SO}


# Setup Cross compiler vars which are used by node-gyp when building native node modules.
OPTS="--sysroot=${SYSROOT}"
export CC="arm-linux-gnueabihf-gcc ${OPTS}"
export CXX="arm-linux-gnueabihf-g++ ${OPTS}"


# Install and configure nvm
curl -o- https://raw.githubusercontent.com/creationix/nvm/${NVM_VERSION}/install.sh | bash
# The following 2 lines are installed into ~/.bashrc by the above,
# but on the RPi, sourcing ~/.bashrc winds up being a no-op (when sourced
# from a script), so we just run it here.
export NVM_DIR="$HOME/.nvm"
[ -s "$NVM_DIR/nvm.sh" ] && source "$NVM_DIR/nvm.sh"  # This loads nvm
nvm install ${NODE_VERSION}

mrstegeman · August 29, 2019, 3:10pm

That script is specifically part of our Raspberry Pi image creation process (for Raspbian).

If you’d like to build and run the gateway yourself, you’ll need to follow the instructions in the README.