naclports is collection of open source libraries and applications that have been ported to Native Client, along with set to tools for building and maintaining them.

The ports themselves live in the ports directory. Each one contains of three main file:

• pkg_info: a description of the package.
• build.sh: a bash script for building it
• nacl.patch: an optional patch file.

The tools for building packages live in build_tools. To build a package run naclports.py build <package_dir>. The build script will download, patch, build and install the application or library. By default it will also first build any dependencies that that the package has.

The build scripts require that certain tools are present in the host system. You will need at least these:

• bash
• make
• curl
• sed
• git

To build all ports you will also need these:

• cmake
• texinfo
• gettext
• pkg-config
• autoconf, automake, libtool
• libglib2.0-dev >= 2.26.0 (if you want to build glib)

Before you can build any of the package you must set the NACL_SDK_ROOT environment variable to top directory of a version of the Native Client SDK (the directory containing toolchain/). This path should be absolute.

The top level Makefile can be used as a quick way to build one or more packages. For example, make vorbis will build libvorbis-1.2.3 and libogg-1.1.4. make all will build all the packages.

There are 4 possible architectures that NaCl modules can be compiled for: i686, x86_64, arm, pnacl. The naclports build system will only build just one at at time. You can control which one by setting the NACL_ARCH environment variable. e.g.:

$ NACL_ARCH=arm make openssl

If you want to build a certain package for all architectures you can use the top level make_all.sh script. e.g.:

$ ./make_all.sh openssl

Headers and libraries are installed into the toolchains directly so there is not add extra -I or -L options in order to use the libraries. (Currently, the naclports scripts will generate a gcc "specs" file to add the required paths.)

The source code and build output for each package is placed in:

out/repository/<PACKAGE_NAME>

Note: Each package has its own license. Please read and understand these licenses before using these packages in your projects.

Note to Windows users: These scripts are written in bash and must be launched from a Cygwin shell. While many of the scripts should work under Cygwin naclports is only tested on Linux and Mac so YMMV.

Applications/Examples that build runnable web pages are published to out/publish. To run them in chrome you need to serve them with a web server. The easiest way to do this is to run:

$ make run

This will start a local web server serving the content of out/publish after which you can navigate to http://localhost:5103 to view the content.

To add a package:

1. If you are using svn, make sure you have a writable version of the repository:

   $ gclient config https://naclports.googlecode.com/svn/trunk/src

2. Add a directory to the ports directory using the name your new package. For example: ports/openssl.
3. Add the build.sh script and pkg_info to that directory.

4. Optionally include the upstream tarball and add its sha1 checksum to pkg_info. You can do this using build_tools/sha1sum.py. Redirect the script to append to the pkg_info file. e.g.:

   $ sha1sum.py mypkg.tar.gz >> ports/openssl/pkg_info

5. Optionally include a patch file (nacl.patch). See below for the recommended way to generate this patch.

6. Make sure your package builds for all architectures:

   $ ./make_all.sh <PACKAGE_NAME>

Each port has an optional build script: build.sh. Some ports, such as those that are based on autotools+make don't need a build script at all. The build script is run in a bash shell, it can set variables at the global scope that override the default behaviour of various steps in the build process. The most common steps that implement by package-specific scripts are:

• ConfigureStep()
• BuildStep()
• InstallStep()
• TestStep()

When implementing a given step the default step can be still invoked, e.g. by calling DefaultBuildStep() from within BuildStep()

Each build is is run independently in a subshell, so variables set in one step are not visible in others, and changing the working directory within a step will not effect other steps.

A variety of shared variables and functions are available from with the build scripts. These are defined in build_tools/common.sh.

When a package is first built, its source is downloaded and extracted to out/repository/<PKG_NAME>. A new git repository is then created in this folder with the original archive contents on a branch called upstream. The optional nacl.patch file is then applied on the master branch. This means that at any given time you can see the changes from upstream using git diff upstream.

To make changes to a package's patch file the recommended workflow is:

1. Directly modify the sources in out/repository/PKG_NAME.
2. Build the package and verify the changes.
3. Use git diff upstream.. > ../path/to/nacl.patch to regenerate the patch file.

Whenever the upstream archive or patch file changes and you try to build the package you will be prompted to remove the existing repository and start a new one. This is to avoid deleting a repository that might have unsaved changed.

Happy porting!