chromium-notes

Testing

Google Tests (General)

Example: the blink_platform_unittests & blink_unittests (previously webkit_unit_tests) targets:

./out/Debug/blink_unittests --gtest_filter="HTMLPreloadScannerDocumentTest.XHRResponseDocument"
./out/Debug/blink_platform_unittests --gtest_filter="ResourceFetcherTest.UseExistingResource"
./out/Debug/blink_platform_unittests --gtest_filter="ResourceFetcherTest.*"

Useful gtest flags:

• No timeout: --test-timeout=100000000 --test-launcher-timeout=1000000
• No retry: --test-launcher-retry-limit=0
• Multiple inclusions/exclusions stackoverflow: --gtest_filter="POSITIVE_PATTERNS:AND_MORE[-NEGATIVE_PATTERNS:AND_MORE]"
  • Example: --gtest_filter="ResourceFetcherTest.*-ResourceFetcherTest.StartLoadAfterFrameDetach:ResourceFetcherTest.Vary"

Browser tests

SSH & Display

Running a non-unit-test from a remote device over SSH will likely result in the test failing due to display issues. This requires running export DISPLAY=:20 to fix.

Misc tips

Browser test tips: https://chromium.googlesource.com/chromium/src/+/master/content/public/test/README.md.

Unit tests

Integrating RuntimeEnabledFeatures

Runtime enable features have automatically-generated Scope* classes to help enable features for unit tests. See https://source.chromium.org/search?q=Scoped.*ForTest%20file:test.cc%20file:blink&sq= for example.

WPTs & Layout Tests

Misc useful links:

• WPT GitHub
• testharness.js API docs
  • Super useful to learn testharness APIs i.e., that promise_tests run sequentially, how to use t.step_func(), or the difference between add_result_callback() and add_completion_callback(). Also it is not super easy to find for some reason I think
• Chromium Running Web Tests docs
• WPT local setup
• Trusting root CA
  • To easily get https tests working locally

Running layout tests:

third_party/blink/tools/run_web_tests.py -t Debug http/tests/security/referrer-policy-conflicting-policies.html external/wpt/workers

Useful flags:

• --no-retry-failures
• --fully-parallel
  • When running WPTs, your machine spins up an “appropriate” number of content shells for parallelization. As the number of remaining WPTs decrease, your machine decreases the number of content shells running in parallel. Even when running up to ~80 tests, your machine may start many content shells and quickly reduce the number, slowing down the run. This flag forces the machine to use as many content shells as it would for an enormous number of tests, all the way through the end.

WPT results

The last set of test results are always available at:

./out/Debug/layout-test-results/results.html

Server output

Python server output can be found at:

./out/<Dir>/layout-test-results/wptserve_stdout.txt
./out/<Dir>/layout-test-results/wptserve_stderr.txt

Useful run_web_tests.py flags

• --driver-logging
  • Enable Chromium log streaming to the terminal
• --additional-driver-flag=--no-sandbox
  • Needed(?) when attaching a debugger to a renderer process during a WPT
• --timeout-ms=100000000

Cross-origin tests

If you want to use a cross-origin in a test, you have two options:

1. Substitution tests
   • Make the test name of the format test-name.sub.html or test-name.sub.https.html
   • This makes the test run through the server-side substitutor(?) before it is served
   • Therefore you can use things like the following to create origins:
     • http://:/html/
     • http://:/html/
2. Import the get-host-info helper
   • <script src="/common/get-host-info.sub.js"></script>
   • URL(get_host_info().HTTPS_REMOTE_ORIGIN + '/worklets/resources/referrer-checker.py');

Public and private networks

Sometimes you have to test Chrome’s behavior on public and private networks. See the following resources for doing so:

• https://source.chromium.org/chromium/chromium/src/+/main:third_party/wpt_tools/wpt.config.json;l=9-13;drc=053d474806089cfbe11c4f2419b753bfe99d5cec
• https://source.chromium.org/chromium/chromium/src/+/main:third_party/wpt_tools/wpt/tools/wptrunner/wptrunner/environment.py;l=177-181;drc=db5b385f525ee477c733a0170dea6ffe7163a338
• https://source.chromium.org/chromium/chromium/src/+/main:services/network/public/cpp/network_switches.cc;l=113;drc=f8c933c2bd17344ce7ac61be2ac7725ed840b19f

Here’s an example:

./out/Debug/chrome --no-sandbox --enable-experimental-web-platform-features --enable-features=FencedFrames,SharedStorageAPI --ip-address-space-overrides=127.0.0.1:8445=private,127.0.0.1:8446=public https://fenced-frames.glitch.me/

Misc WPT/testharness things

• Without HTML boilerplate: https://web-platform-tests.org/writing-tests/testharness.html#without-html-boilerplate-window-js
• Image progressive rendering. There exists a python resource that responds with a progressively-rendered image, at /element-timing/resources/progressive-image.py. You can see it being used in the following CL: https://chromium-review.googlesource.com/c/chromium/src/+/2138994/3/third_party/blink/web_tests/external/wpt/preload/subresource-integrity-partial-image.html#25.