Getting Started: Ubuntu Debuild/Pdebuild

In order to control how pdebuild/debuild build, you need to edit files in ${nnstreamer-source}/debian/.

Pdebuild, a sandboxed build environment. (Ubuntu 18.04, 20.04)

This guide uses the nnstreamer PPA to resolve additional build-dependencies (e.g., tensorflow/tensorflow-lite 1.13).

Install build tools for pdebuild:

$ sudo apt install pbuilder debootstrap devscripts

The following example configuration is for Ubuntu 18.04:

$ cat ~/.pbuilderrc
DISTRIBUTION=bionic
COMPONENTS="main restricted universe multiverse"
OTHERMIRROR="deb http://archive.ubuntu.com/ubuntu bionic main restricted universe multiverse |\
  deb http://archive.ubuntu.com/ubuntu bionic-security main restricted universe multiverse |\
  deb http://archive.ubuntu.com/ubuntu bionic-updates main restricted universe multiverse |\
  deb [trusted=yes] http://ppa.launchpad.net/nnstreamer/ppa/ubuntu bionic main |\
  deb [trusted=yes] http://ppa.launchpad.net/one-runtime/ppa/ubuntu bionic main"
$ sudo ln -s  ~/.pbuilderrc /root/.pbuilderrc
$ sudo pbuilder create

Run pdebuild to build and get the package.

$ pdebuild
$ ls -al /var/cache/pbuilder/result/*.deb

Then, you may install the resulting .deb files to your system.

Refer to PbuilderHowto for more about pdebuild.

Debuild, build with system libraries. (Ubuntu/Debian)

After installing all the required packages to your system, you may run debuild to get .deb packages.

In Ubuntu, you may prepare your system easily with the following commands, which installs prebuilt binaries from nnstreamer's PPA repository.

$ sudo apt-add-repository ppa:nnstreamer/ppa
$ sudo add-apt-repository ppa:one-runtime/ppa
$ sudo apt install ssat tensorflow-dev tensorflow-lite-dev libprotobuf-dev nnfw-dev # you may add pytorch and other libraries, too

Note that ssat is required for unit testing. You may download it at Github SSAT.

Refer to debuild command for how to use debuild.

1. Clone the needed repositories

You may skip this if you have downloaded and installed PPA packages of SSAT and tensorflow.

$ git clone https://github.com/myungjoo/SSAT ssat
$ git clone https://git.tizen.org/cgit/platform/upstream/tensorflow
$ git clone https://github.com/nnstreamer/nnstreamer

2. Fix tensorflow build errors

You may skip this if you have downloaded and installed PPA packages of SSAT and tensorflow.

A script, tensorflow/contrib/lite/Makefile, may fail, depending your shell. Replace the ARCH macro (HOST_ARCH in some versions):

From:

ARCH := $(shell if [[ $(shell uname -m) =~ i[345678]86 ]]; then echo x86_32; else echo $(shell uname -m); fi)

To:

ARCH := $(shell uname -m | sed -e 's/i[3-8]86/x86_32/')

3. Prepare for debuild (installing required dependencies).

Skip ssat and tensorflow if you have already installed them.

$ for i in ssat tensorflow nnstreamer; do \
  (cd $i && sudo mk-build-deps --install debian/control && sudo dpkg -i *.deb || break); \
  done

4. Run debuild and get .deb packages.

Skip ssat and tensorflow if you have already installed them.

$ export DEB_BUILD_OPTIONS="parallel=$(($(cat /proc/cpuinfo |grep processor|wc -l) + 1))"
$ for i in ssat tensorflow nnstreamer; do \
  (cd $i && time debuild -us -uc || break); \
  done

If there is a missing package, debuild will tell you which package is missing. If you haven't configured debuild properly, yet, you will need to add -uc -us options to debuild.

5. Install the generated .deb files.

Skip ssat and tensorflow if you have already installed them.

The files will be there at the parent dir. E. g. at nnbuilder/.. directory.

In order to install them (should run as root):

$ sudo apt install ./ssat_*.deb ./tensorflow-lite-dev_*.deb ./tensorflow-dev_*.deb
$ sudo apt install ./nnstreamer_0.1.0-1rc1_amd64.deb

If you need nnstreamer development package:

#apt install ./nnstreamer-dev_0.1.0-1rc1_amd64.deb