Chromium‎ > ‎DOMUI Testing‎ > ‎

WebUI browser_tests

Note: This page is out of date. If you are using modern practices, you are using a BrowserProxy for your JS/C++ interactions. The BrowserProxy document covers some testing practices for that case.


See Also domui-testing.

WebUI contains Javascript, which runs in the renderer and a C++ handler, which runs in the UI thread of the browser process. While this is a necessary part of the design, testing across these boundaries of both language and process/thread is cumbersome at best.


Make it possible to test the Javascript portion of WebUI in Javascript:
  • Write WebUI tests in Javascript.
  • Run in browser_tests so that the page is loaded in a real chrome browser.
  • Allow WebUI handlers to be mocked in Javascript.


The solution comes in the following parts:
  • Support libraries: chrome/test/data/webui/test_api.js, chrome/third_party/mock4js/mock4js.js.
  • gyp rules for js2webui generator feeding results into browser_tests.
  • WebUIBrowserTest C++ class in chrome/browser/ui/webui/web_ui_browsertest* .
  • Test source files next to the implementation or in chrome/test/data/webui/ - see chrome/browser/ui/webui/options/options_browsertest.js and chrome/test/data/webui/print_preview.js for reference.

How to write a test

The best reference examples are chrome/test/data/webui/print_preview.js and chrome/browser/ui/webui/options/options_browsertest.js

  • [maybe] create a new test file

    • Create new file chrome/test/data/webui/mytest.js
    • Add chrome/test/data/webui/mytest.js to the sources for browser_tests in chrome/chrome_tests.gypi

  • Write a test fixture, defining the page to browse to:

 * TestFixture for OptionsPage WebUI testing.
 * @extends {testing.Test}
 * @constructor
function OptionsWebUITest() {}

OptionsWebUITest.prototype = {
  __proto__: testing.Test.prototype,

   * Browse to the options page & call our preLoad().
  browsePreload: 'chrome://settings-frame',

  // ...


  • Mock the Javascript handler:

OptionsWebUITest.prototype = {
   * Register a mock handler to ensure expectations are met and options pages
   * behave correctly.
  preLoad: function() {

      // Register stubs for methods expected to be called before/during tests.
      // Specific expectations can be made in the tests themselves.

  • Mock stubs which call a function:

        will(callFunction(function() {

  • Define a test using mock expectations:

TEST_F('OptionsWebUITest', 'testSetBooleanPrefTriggers', function() {
  var showHomeButton = $('toolbarShowHomeButton');
  var trueListValue = [
  // Note: this expectation is checked in testing::Test::TearDown.

  // Cause the handler to be called.;

  • Conditionally run a test using generated c++ ifdefs:

See Handling a failing test for more details on style and how/when to disable a test.

// Not meant to run on ChromeOS at this time.
// Not finishing in windows.
GEN('#if defined(OS_CHROMEOS) || defined(OS_MACOSX) || defined(OS_WIN) \\');
GEN('    || defined(TOUCH_UI)');
GEN('#define MAYBE_testRefreshStaysOnCurrentPage \\');
GEN('    DISABLED_testRefreshStaysOnCurrentPage');
GEN('#define MAYBE_testRefreshStaysOnCurrentPage ' +

TEST_F('OptionsWebUITest', 'MAYBE_testRefreshStaysOnCurrentPage', function() {
  var item = $('advancedPageNav');
  var pageInstance = AdvancedOptions.getInstance();
  var topPage = OptionsPage.getTopmostVisiblePage();
  var expectedTitle = pageInstance.title;
  var actualTitle = document.title;
  expectEquals("chrome://settings/advanced", document.location.href);
  expectEquals(expectedTitle, actualTitle);
  expectEquals(pageInstance, topPage);


  • Adding more goodies to generator - edit chrome/test/ui/webui/javascript2webui.js and tests in chrome/test/data/webui/.
  • Disabling a test - find tests in chrome/test/data/webui/ & mark FLAKY_, DISABLED_ or use the MAYBE_ trick shown above to conditionally decide.


  • Isn't mocking in javascript not testing the WebUI message passing?  True, but that should be tested as a unit test and then trusted.  Mocking in JS is easier, less flaky (always recursive - no synchronization challenges) and much better than not having any tests at all.
  • If you already have the page in question why don't you just run all tests without starting a new IN_PROCESS_BROWSER_TEST? You are more than welcome to group tons of expect* calls into a single test; all errors will be reported after the entire test runs. Having tests be separate IN_PROCESS_BROWSER_TEST calls ensures the state is exactly the same at the start of each tests with no pollution from previous tests.


  • [] The generator relies on d8 to be built for the host.  Currently the v8.gyp rules aren't correct for Arm as they have conditionals on the 'target_host' and don't heed the 'toolset". A gyp condition only runs the js2webui rule on non-arm platforms.
  • Use of MAYBE_ to ifdef will have a different run name from GTEST than from test_api.js - this is because MAYBE_xyz will be defined as either xyz or DISABLED_xyz in C++, but not changed in javascript.

Best practices

  • As described in the gtest docs, prefer expect* over assert*, as it will not halt the test, but will register the failure and allow other checks in that particular testcase to run.
  • Since the call is included in the failure error message, the optional message parameter should only include information not available:
// NO
TEST_F('FooTest', 'TestFoo', function() {
  expectEquals(foo, bar, 'foo != bar');
  expectEquals(foo, bar, foo + '!=' + bar);
  var i = ...;
  expectEquals(5, array[i], 'array[i] != 5');

// YES
TEST_F('FooTest', 'TestFoo', function() {
  expectEquals(foo, bar);
  expectEquals(foo, bar);
  var i = ...;
  expectEquals(5, array[i], 'i=' + i);