Chrome for Chromium OS can be built in a couple different ways. After following the initial setup, you'll need to choose one of the following build configurations:
- If you're interested in testing Chrome OS code in Chrome, but not interactions with Chrome OS services, you can build for linux-chromeos using just a Linux workstation.
- Otherwise, Chrome's full integration can be covered by building for a real Chrome OS device or VM using Simple Chrome.
- Use
is_chromeos_device
in GN andBUILDFLAG(IS_CHROMEOS_DEVICE)
in C++ code to differentiate between these two modes.
[TOC]
First, follow the normal Linux build instructions as usual to get a Chromium checkout.
You'll also need to add 'chromeos'
to the target_os
list in your .gclient
configuration, which will fetch the additional build dependencies required for
CrOS. This file is located one level up from your Chromium checkout's src
.
If you don't already have a target_os
line present, simply add this to the
end of the .gclient
file:
target_os = ['chromeos']
If you already have a target_os
line present in your .gclient file
, you can
simply append 'chromeos'
to the existing list there. For example:
target_os = ['android', 'chromeos']
Once your .gclient
file is updated, you will need to run gclient sync
once
before proceeding with the rest of these instructions.
Chromium on Chromium OS uses Linux Chromium as a base, but adds a large number of Chrome OS-specific features to the code. For example, the login UI, window manager and system UI are part of the Chromium code base and built into the chrome binary.
Fortunately, most Chromium changes that affect Chromium OS can be built and tested on a Linux workstation. This build is called "linux-chromeos". In this configuration most system services (like the power manager, bluetooth daemon, etc.) are stubbed out. The entire system UI runs in a single X11 window on your desktop.
You can test sign-in/sync in this mode by adding the --login-manager flag, see the Login notes section.
Run the following in your chromium checkout:
$ gn gen out/Default --args='target_os="chromeos"'
$ autoninja -C out/Default chrome
$ out/Default/chrome --use-system-clipboard
(autoninja
is a wrapper that automatically provides optimal values for the
arguments passed to ninja
).
Some additional options you may wish to set by passing in --args
to gn gen
or running gn args out/Default
:
use_goma = true # Googlers: Use build farm, compiles faster.
is_component_build = true # Links faster.
is_debug = false # Release build, runs faster.
dcheck_always_on = true # Enables DCHECK despite release build.
enable_nacl = false # Skips native client build, compiles faster.
# Set the following true to create a Chrome (instead of Chromium) build.
# This requires a src-internal checkout.
is_chrome_branded = false # Adds internal features and branded art assets.
is_official_build = false # Turns on many optimizations, slower build.
NOTE: You may wish to replace 'Default' with something like 'Cros' if you switch back and forth between Linux and Chromium OS builds, or 'Debug' if you want to differentiate between Debug and Release builds (see below).
See GN Build Configuration for more information about configuring your build.
You can also build and run test targets like unit_tests
, browser_tests
, etc.
Some useful flags:
--ash-debug-shortcuts
: Enable shortcuts such as Ctl+Alt+Shift+T to toggle tablet mode.--ash-host-window-bounds="0+0-800x600,800+0-800x600"
: Specify one or more virtual screens, by display position and size.--enable-features=Feature1,OtherFeature2
: Enable specified features. Features are often listed in chrome://flags, or in source files such as chrome_features.cc or ash_features.cc. Note that changing values in chrome://flags does not work for linux-chromeos, and this flag must be used.--enable-ui-devtools[=9223]
: Allow debugging of the system UI through devtools either within linux-chromeos at chrome://inspect, or from a remote browser at devtools://devtools/bundled/devtools_app.html?uiDevTools=true&ws=127.0.0.1:9223/0--remote-debugging-port=9222
: Allow debugging through devtools at http://localhost:9222--use-system-clipboard
: Integrate clipboard with the host X11 system.
By default this build signs in with a stub user. To specify a real user:
- For first run, add the following options to chrome's command line:
--user-data-dir=/tmp/chrome --login-manager
- Go through the out-of-the-box UX and sign in with a real Gmail account.
- For subsequent runs, if you want to skip the login manager page, add:
--user-data-dir=/tmp/chrome --login-user=username@gmail.com --login-profile=username@gmail.com-hash
. It's also fine to just keep --login-manager instead. - To run in guest mode instantly, add:
--user-data-dir=/tmp/chrome --bwsi --incognito --login-user='$guest' --login-profile=user
Signing in as a specific user is useful for debugging features like sync that require a logged in user.
The Chromium OS build requires a functioning GL so if you plan on testing it through Chromium Remote Desktop you might face drawing problems (e.g. Aura window not painting anything). Possible remedies:
--ui-enable-software-compositing --ui-disable-threaded-compositing
--use-gl=swiftshader
, but it's slow.
To more closely match the UI used on devices, you can install fonts used by Chrome OS, such as Roboto, on your Linux distro.
This configuration allows you to build a fully functional Chrome for a real
Chrome OS device or VM. Since Chrome OS uses a different toolchain for each
device model, you'll first need to know the name of the model (or "board") you
want to build for. For most boards, amd64-generic
and arm-generic
will
produce a functional binary, though it won't be optimized and may be missing
functionality.
Each board has its own toolchain and misc. build dependencies. To fetch these,
list the board under the "cros_boards"
gclient custom var. If you were using
the amd64-generic
board, your .gclient
file would look like:
solutions = [
{
"url": "https://chromium.googlesource.com/chromium/src.git",
"name": "src",
"custom_deps": {},
"custom_vars" : {
"cros_boards": "amd64-generic",
},
},
]
target_os = ["chromeos"]
Once your .gclient file is updated, you will need to run gclient sync
again
to fetch the toolchain.
NOTE:
- If you'd like a VM image additionally downloaded for the board, add it to the
"cros_boards_with_qemu_images"
gclient custom var. That var downloads the SDK along with a VM image.cros_boards
downloads only the SDK. - If you'd like to fetch multiple boards, add a
:
between each board in the gclient var. For example:"cros_boards": "amd64-generic:arm-generic"
.
After the needed toolchain has been downloaded for your ${BOARD}, a build dir
will have been conveniently created for you at out_$BOARD/Release
, which can
then be used to build Chrome. For the amd64-generic
board, this would
look like:
$ gn gen out_amd64-generic/Release
$ autoninja -C out_$BOARD/Release chrome
Or if you prefer to use your own build dir, simply add the following line to the
top of your GN args: import("//build/args/chromeos/amd64-generic.gni")
. eg:
$ gn gen out/Default --args='import("//build/args/chromeos/amd64-generic.gni")'
$ autoninja -C out/Default chrome
That will produce a Chrome OS build of Chrome very similar to what is shipped
for that device. You can also supply additional args or even overwrite ones
supplied in the imported .gni file after the import()
line.
For more information (like copying the locally-built Chrome to a device, or running Tast tests), consult Simple Chrome's full documentation.