WebRTC iOS development

Development Environment

In case you need to build the framework manually or you want to try out the demo application AppRTCMobile, follow the instructions illustrated bellow.

A macOS machine is required for iOS development. While it‘s possible to develop purely from the command line with text editors, it’s easiest to use Xcode. Both methods will be illustrated here.

NOTICE: You will need to install Chromium depot_tools.

Getting the Code

Create a working directory, enter it, and run:

$ fetch --nohooks webrtc_ios
$ gclient sync

This will fetch a regular WebRTC checkout with the iOS-specific parts added. Notice the size is quite large: about 6GB. The same checkout can be used for both Mac and iOS development, since GN allows you to generate your Ninja project files in different directories for each build config.

You may want to disable Spotlight indexing for the checkout to speed up file operations.

Note that the git repository root is in src.

From here you can check out a new local branch with:

$ git new-branch <branch name>

See Development for generic instructions on how to update the code in your checkout.

Generating project files

GN is used to generate Ninja project files. In order to configure GN to generate build files for iOS certain variables need to be set. Those variables can be edited for the various build configurations as needed.

The variables you should care about are the following:

  • target_os:
    • To build for iOS this should be set as target_os="ios" in your gn args. The default is whatever OS you are running the script on, so this can be omitted when generating build files for macOS.
  • target_cpu:
    • For builds targeting iOS devices, this should be set to either "arm" or "arm64", depending on the architecture of the device. For builds to run in the simulator, this should be set to "x64".
  • is_debug:
    • Debug builds are the default. When building for release, specify false.

The component build is the default for Debug builds, which are also enabled by default unless is_debug=false is specified.

The GN command for generating build files is gn gen <output folder>.

After you‘ve generated your build files once, subsequent invocations of gn gen with the same output folder will use the same arguments as first supplied. To edit these at any time use gn args <output folder>. This will open up a file in $EDITOR where you can edit the arguments. When you’ve made changes and save the file, gn will regenerate your project files for you with the new arguments.

Examples

$ # debug build for 64-bit iOS
$ gn gen out/ios_64 --args='target_os="ios" target_cpu="arm64"'

$ # debug build for simulator
$ gn gen out/ios_sim --args='target_os="ios" target_cpu="x64"'

Compiling with ninja

To compile, just run ninja on the appropriate target. For example:

$ ninja -C out/ios_64 AppRTCMobile

Replace AppRTCMobile in the command above with the target you are interested in.

To see a list of available targets, run gn ls out/<output folder>.

Using Xcode

Xcode is the default and preferred IDE to develop for the iOS platform.

Generating an Xcode project

To have GN generate Xcode project files, pass the argument --ide=xcode when running gn gen. This will result in a file named all.xcodeproj placed in your specified output directory.

Example:

$ gn gen out/ios --args='target_os="ios" target_cpu="arm64"' --ide=xcode
$ open -a Xcode.app out/ios/all.xcodeproj

Compile and run with Xcode

Compiling with Xcode is not supported! What we do instead is compile using a script that runs ninja from Xcode. This is done with a custom run script action in the build phases of the generated project. This script will simply call ninja as you would when building from the command line.

This gives us access to the usual deployment/debugging workflow iOS developers are used to in Xcode, without sacrificing the build speed of Ninja.

Running the tests

There are several test targets in WebRTC. To run the tests, you must deploy the .app bundle to a device (see next section) and run them from there. To run a specific test or collection of tests, normally with gtest one would pass the --gtest_filter argument to the test binary when running. To do this when running the tests from Xcode, from the targets menu, select the test bundle and press edit scheme... at the bottom of the target dropdown menu. From there click Run in the sidebar and add --gtest_filter to the Arguments passed on Launch list.

If deploying to a device via the command line using ios-deploy, use the -a flag to pass arguments to the executable on launch.

Deploying to Device

It's easiest to deploy to a device using Xcode. Other command line tools exist as well, e.g. ios-deploy.

NOTICE: To deploy to an iOS device you must have a valid signing identity set up. You can verify this by running:

$ xcrun security find-identity -v -p codesigning

If you don‘t have a valid signing identity, you can still build for ARM, but you won’t be able to deploy your code to an iOS device. To do this, add the flag ios_enable_code_signing=false to the gn gen args when you generate the build files.

Using WebRTC in your app

To build WebRTC for use in a native iOS app, it's easiest to build WebRTC.framework. This can be done with ninja as follows, replacing ios with the actual location of your generated build files.

ninja -C out/ios framework_objc

This should result in a .framework bundle being generated in out/ios. This bundle can now be directly included in another app.

If you need a FAT .framework, that is, a binary that contains code for multiple architectures, and will work both on device and in the simulator, a script is available here

The resulting framework can be found in out_ios_libs/.

Please note that you can not ship the FAT framework binary with your app if you intend to distribute it through the app store. To solve this either remove “x86-64” from the list of architectures in the build script or split the binary and recreate it without x86-64. For instructions on how to do this see here.