gstreamer/sdk-multiplatform-deployment-using-cerbero.md

# Multiplatform deployment using Cerbero

Cerbero is the build and packaging system used to construct the
GStreamer SDK. It uses “recipe” files that indicate how to build
particular projects, and on what other projects they depend.
Moreover, the built projects can be combined into packages for
distribution. These packages are, depending on the target platform,
Windows or OS X installers or Linux packages.

To use Cerbero to build and package your application, you just need to
add a recipe explaining how to build you application and make it depend
on the `gstreamer-sdk` project. Then Cerbero can take care of building
your application and its dependencies and package them all together.

Read [Building from source using
Cerbero](Building%2Bfrom%2Bsource%2Busing%2BCerbero.html) to learn how
to install and use Cerbero.

At this point, after reading the Build from source section in [Building
from source using
Cerbero](Building%2Bfrom%2Bsource%2Busing%2BCerbero.html), you should be
able to build the GStreamer SDK from source and are ready to create
recipe and package files for your application.

In the Cerbero installation directory you will find the
`cerbero-uninstalled` script. Execute it without parameters to see the
list of commands it accepts:

``` bash
./cerbero-uninstalled
```

# Adding a recipe for your application

The first step is to create an empty recipe that you can then tailor to
your needs:

``` bash
./cerbero-uninstalled add-recipe my-app 1.0
```

This will create an initial recipe file in `recipes/my-app.recipe`,
which contains the smallest necessary recipe. This file is a Python
script; set the following attributes to describe your application:

Attribute Name

Description

Required

Example

`name`

The recipe name.

Yes

*name = 'my-app'*

`version`

The software version.

Yes

*version = '1.0'*

`licenses`

A list of licenses of the software (see `cerbero/enums.py:License` for
allowed licenses).

Yes

*licenses = \[License.LGPLv2Plus\]*

`deps`

A list of build dependencies of the software as recipe names.

No

*deps = \['other', 'recipe', 'names'\]*

`platform_deps`

Platform specific build dependencies (see `cerbero/enums.py:Platform`
for allowed platforms).

No

*platform\_deps = {Platform.LINUX: \['some-recipe'\], Platform.WINDOWS:
\['another-recipe'\]}*

`remotes`

A dictionary specifying the git remote urls where sources are pulled
from.

No

*remotes = {'origin': '<git://somewhere>'}*

`commit`

The git commit, tag or branch to use, defaulting to "sdk-*`version`*"*.*

No

*commit = 'my-app-branch'*

`config_sh`

Used to select the configuration script.

No

*config\_sh = 'autoreconf -fiv && sh ./configure'*

`configure_options`

Additional options that should be passed to the `configure` script.

No

*configure\_options = '--enable-something'*

`use_system_libs`

Whether to use system provided libs.

No

*use\_system\_libs = True*

`btype`

The build type (see `cerbero/build/build.py:BuildType` for allowed build
types).

No

*btype = BuildType.CUSTOM*

`stype`

The source type (see `cerbero/build/source.py:SourceType` for allowed
source types).

No

*stype = SourceType.CUSTOM*

`files_category`

A list of files that should be shipped with packages including this
recipe *category*. See below for more details.

Cerbero comes with some predefined categories that should be used if the
files being installed match a category criteria.

The predefined categories are:

`libs` (for libraries), `bins` (for binaries), `devel` (for development
files - header, pkgconfig files, etc), `python` (for python files) and
`lang` (for language files).

Note that for the `bins` and `libs` categories there is no need to
specify the files extensions as Cerbero will do it for you.

Yes\*

*files\_bins = \['some-binary'\]*

*files\_libs = \['libsomelib'\]*

*files\_devel = \['include/something'\] files\_python =
\['site-packages/some/pythonfile%(pext)s'\]*

*files\_lang = \['foo'\]*

`platform_files_category`

Same as *`files_category`* but for platform specific files.

No

*platform\_files\_some\_category = {Platform.LINUX: \['/some/file'\]}*

\* At least one “files” category should be set.

Apart from the attributes listed above, it is also possible to override
some Recipe methods. For example the `prepare` method can be overridden
to do anything before the software is built, or the `install` and
`post_install` methods for overriding what should be done during or
after installation. Take a look at the existing recipes in
`cerbero/recipes` for example.

Alternatively, you can pass some options to cerbero-uninstalled so some
of these attributes are already set for you. For
example:

```
./cerbero-uninstalled add-recipe --licenses "LGPL" --deps "glib,gtk+" --origin "git://git.my-app.com" --commit "git-commit-to-use" my-app 1.0
```

See `./cerbero-uninstalled add-recipe -h` for help.

As an example, this is the recipe used to build the Snappy media player:

```
class Recipe(recipe.Recipe):
    name = 'snappy'
    version = '0.2+git'
    licenses = [License.GPLv2Plus]
    config_sh = 'autoreconf -fiv && sh ./configure'
    deps = ['glib', 'gstreamer', 'gst-plugins-base', 'clutter', 'clutter-gst']
    platform_deps = { Platform.LINUX: ['libXtst'] }
    use_system_libs = True
    remotes = {'upstream': 'git://git.gnome.org/snappy'}

    files_bins = ['snappy']
    files_data = ['share/snappy']

    def prepare(self):
        if self.config.target_platform == Platform.LINUX:
            self.configure_options += ' --enable-dbus' 
```

Cerbero gets the software sources to build from a GIT repository, which
is specified via the `git_root` configuration variable from the Cerbero
configuration file (see the "Build from software" section in [Installing
on Linux](Installing%2Bon%2BLinux.html)) and can be overridden by the
`remotes` attribute inside the recipes (if setting the `origin` remote).
In this case where no “commit” attribute is specified, Cerbero will use
the commit named “sdk-0.2+git” from the GIT repository when building
Snappy.

Once the recipe is ready, instruct Cerbero to build it:

``` bash
./cerbero-uninstalled build my-app
```

# Adding a package for you software

To distribute your software with the SDK it is necessary to put it into
a package or installer, depending on the target platform. This is done
by selecting the files that should be included. To add a package you
have to create a package file in `cerbero/packages`. The package files
are Python scripts too and there are already many examples of package
files in `cerbero/packages`.

Now, to create an empty package, do:

``` bash
./cerbero-uninstalled add-package my-app 1.0
```

This will create an initial package file in `packages/my-app.package`.

The following Package attributes are used to describe your package:

**Attribute Name**

**Description**

**Required**

**Example**

`name`

The package name.

Yes

*name = 'my-app'*

`shortdesc`

A short description of the package.

No

*shortdesc = 'some-short-desc'*

`longdesc`

A long description of the package.

No

*longdesc = 'Some Longer Description'*

`codename`

The release codename.

No

*codename = 'MyAppReleaseName'*

`vendor`

Vendor for this package.

No

*vendor = 'MyCompany'*

`url`

The package url

No

*url = 'http://www.my-app.com'*

`version`

The package version.

Yes

*version = '1.0'*

`license`

The package license (see `cerbero/enums.py:License` for allowed
licenses).

Yes

*license = License.LGPLv2Plus*

`uuid`

The package unique id

Yes

*uuid = '6cd161c2-4535-411f-8287-e8f6a892f853'*

`deps`

A list of package dependencies as package names.

No

*deps = \['other', 'package', 'names'\]*

`sys_deps`

The system dependencies for this package.

No

*sys\_deps= {Distro.DEBIAN: \['python'\]}*

`files`

A list of files included in the **runtime** package in the form
*“recipe\_name:category1:category2:...”*

If the recipe category is omitted, all categories are included.

Yes\*

*files = \['my-app'\]*

*files = \['my-app:category1'\]*

`files_devel`

A list of files included in the **devel** package in the form
*“recipe\_name:category1:category2:...”*

Yes\*

*files\_devel = \['my-app:category\_devel'\]*

`platform_files`

Same as *files* but allowing to specify different files for different
platforms.

Yes\*

*platform\_files = {Platform.WINDOWS:
\['my-app:windows\_only\_category'\]}*

`platform_files_devel`

Same as *files\_devel* but allowing to specify different files for
different platforms.

Yes\*

*platform\_files\_devel = {Platform.WINDOWS:
\['my-app:windows\_only\_category\_devel'\]}*

\* At least one of the “files” attributes should be set.

Alternatively you can also pass some options to `cerbero-uninstalled`,
for
example:

``` bash
./cerbero-uninstalled add-package my-app 1.0 --license "LGPL" --codename MyApp --vendor MyAppVendor --url "http://www.my-app.com" --files=my-app:bins:libs --files-devel=my-app:devel --platform-files=linux:my-app:linux_specific --platform-files-devel=linux:my-app:linux_specific_devel,windows:my-app:windows_specific_devel --deps base-system --includes gstreamer-core
```

See `./cerbero-uninstalled add-package -h` for help.

As an example, this is the package file that is used for packaging the
`gstreamer-core` package:

```
class Package(package.Package):
    name = 'gstreamer-codecs'
    shortdesc = 'GStreamer codecs'
    version = '2012.5'
    codename = 'Amazon'
    url = "http://www.gstreamer.com"
    license = License.LGPL
    vendor = 'GStreamer Project'
    uuid = '6cd161c2-4535-411f-8287-e8f6a892f853'
    deps = ['gstreamer-core']

    files = ['flac:libs',
            'jasper:libs', 'libkate:libs',
            'libogg:libs', 'schroedinger:libs', 'speex:libs',
            'libtheora:libs', 'libvorbis:libs', 'wavpack:libs', 'libvpx:libs',
            'taglib:libs',
            'gst-plugins-base:codecs', 'gst-plugins-good:codecs',
            'gst-plugins-bad:codecs', 'gst-plugins-ugly:codecs']
    files_devel = ['gst-plugins-base-static:codecs_devel',
            'gst-plugins-good-static:codecs_devel',
            'gst-plugins-bad-static:codecs_devel',
            'gst-plugins-ugly-static:codecs_devel']
    platform_files = {
            Platform.LINUX: ['libdv:libs'],
            Platform.DARWIN: ['libdv:libs']
    } 
```

At this point you have two main options: you could either have a single
package that contains everything your software needs, or depend on a
shared version of the SDK.

### Having a private version of the SDK

To have a private version of the SDK included in a single package you
don't have to add the `deps` variable to the package file but instead
list all files you need in the `files` variables. If you decide to go
this road you must make sure that you use a different prefix than the
GStreamer SDK in the Cerbero configuration file, otherwise your package
will have file conflicts with the GStreamer SDK.

### Having a shared version of the SDK

If you decide to use a shared version of the SDK you can create a
package file like the other package files in the GStreamer SDK. Just
list all packages you need in the `deps` variable and put the files your
software needs inside the `files` variables. When building a package
this way you must make sure that you use the same prefix and
packages\_prefix as the ones in your Cerbero configuration file.

Finally, build your package by using:

``` bash
./cerbero-uninstalled package your-package 
```

Where `your-package` is the name of the `.package` file that you created
in the `packages` directory. This command will build your software and
all its dependencies, and then make individual packages for them (both
the dependencies and your software). The resulting files will be in the
current working directory.