Clone this repo:

Branches

  1. 14c2bd7 Video: Collect dropped frame statistics using Javascript by Ethan Cheng · 17 hours ago main
  2. 5b22778 heap dumps in open-close by Charles William Dick · 26 hours ago
  3. ef3a154 Roll third_party/crossbench/ 1584e1021..6874c8656 (33 commits) by Charles William Dick · 26 hours ago
  4. 05af571 Add multi-process-cycling variant to tab-stress test by Shintaro Kawamura · 2 days ago
  5. 17a31d8 Roll third_party/webpagereplay/ 2c5049abf..9057e5d94 (22 commits) by Sharana K U · 8 days ago

web-tests

web-tests contains:

  • Definitions for CUJs implemented using crossbench's loading benchmark
  • Configuration files for running benchmarks that are built in to crossbench (such as speedometer)
  • Metric definitions and queries for CUJs and benchmarks.

Setup

Do not git clone web-tests. Use the fetch command included with depot_tools.

mkdir src cd src gcloud auth login fetch web-tests cd web-tests 

Don't forget to run gclient sync every time you pull new changes from origin.

Poetry

web-tests uses poetry to manage python dependencies.

sudo apt install python3.11 python3-poetry 

Alternatively, install poetry in a python venv:

python3 -m venv web-tests-venv source web-tests-venv/bin/activate pip install poetry 

Check that you have poetry on your path and make sure you have the right $PATH settings.

poetry --help || echo "Please update your \$PATH to include poetry bin location"; # Depending on your setup, add one of the following to your $PATH: echo "`python3 -m site --user-base`/bin"; python3 -c "import sysconfig; print(sysconfig.get_path('scripts'))"; 

Install the necessary dependencies from the lock file using poetry:

cd cuj/crossbench/runner export PYTHON_KEYRING_BACKEND=keyring.backends.null.Keyring poetry env use 3.11 poetry install 

Setting PYTHON_KEYRING_BACKEND to keyring.backends.null.Keyring disables keyring and prevents poetry install from getting stuck waiting for user input in the GUI.

Running Tests

Android

Before running a test against an android target, make sure your device is connected through adb:

adb devices > List of devices attached > 192.168.20.194:5555 device 

Replace <DEVICE ID> below with the actual device ID from adb devices:

cd cuj/crossbench/runner poetry run python run.py --platform adb --device <DEVICE ID> 

ChromeOS

Before running a test against a ChromeOS target, make sure passwordless SSH is available to the device. Either an IP address or a SSH host is supported as the device id.

Replace <DEVICE> below with the IP address or hostname of your device:

cd cuj/crossbench/runner poetry run python run.py --platform cros --device <DEVICE> 

Local

Running against a local browser on linux is minimally supported, but may require manual changes to test configuration.

cd cuj/crossbench/runner poetry run python run.py --platform local 

Specifying Tests and Variants

The minimal invocation of the runner will attempt to run all benchmarks, CUJs, and corresponding variants in series.

To run a subset of tests, use the --tests flag. --tests supports Python regex format for matching the test names.

poetry run python run.py --platform adb --device <DEVICE ID> --tests speedometer.* 

To specify only certain variants of a test, you can use the --variants flag. --variants also supports Python regex format for matching variants.

poetry run python run.py --platform adb --device <DEVICE ID> --tests local-conference --variants 16p 

Specifying Browsers

By default the runner will use ‘Chrome’ if the --browser flag is not specified. The format accepted by the --browser flag depends on the platform. For Android, use the package name of the installed browser. For ChromeOS or local, use the path to the browser executable.

Secrets

Some CUJs require secrets to perform privileged actions (such as a test account username/password, or auth tokens for the Google Meet Bond API). Place your secrets in secrets.hjson and pass the file to the runner:

poetry run python run.py --platform adb --device <DEVICE ID> --secrets /home/me/secrets.hjson --tests docs 

Looping Tests

Tests can be repeated for a number of iterations or for a specified amount of time using the --playback flag. This flag is supported by crossbench and will iterate the post-setup sections of a CUJ and collect metrics for the entire invocation (instead of splitting metrics by iteration).

poetry run python run.py --platform adb --device <DEVICE ID> --tests tab-stress --variants blank-tab --playback 50x 
poetry run python run.py --platform adb --device <DEVICE ID> --tests tab-stress --variants blank-tab --playback 2h 

Contributing

To get started contributing to web-tests, refer to the docs.