Reland: Settings People: Add /signOut route for Disconnect dialog.

chrome/test/data/webui/settings/people_page_test.js

Index: chrome/test/data/webui/settings/people_page_test.js
diff --git a/chrome/test/data/webui/settings/people_page_test.js b/chrome/test/data/webui/settings/people_page_test.js
index f00d12a50f9bd6e8b185cac04dbc476247ca0d92..5187d62fe10160cd66b5f69633b1da07f87333b1 100644
--- a/chrome/test/data/webui/settings/people_page_test.js
+++ b/chrome/test/data/webui/settings/people_page_test.js
@@ -164,12 +164,19 @@ cr.define('settings_people_page', function() {
           var disconnectConfirm = peoplePage.$.disconnectConfirm;
           assertTrue(!!disconnectConfirm);
           assertFalse(disconnectConfirm.hidden);
+
+          // Wait for exit of dialog route.
+          var dialogExitPromise = new Promise(function(resolve) {
+            window.addEventListener('popstate', function callback() {

[michaelpg, 2016/08/26 20:56:09]

i don't think it actually makes a difference when the listener is added

[tommycli, 2016/08/29 22:43:23]

I discussed this with Dan. I think we want to be sure we are listening to the
popstate event before we proceed to issue the mock interaction.

In practice I think there is another layer of asynchrony in MockInteracions.tap,
so it doesn't matter if we attach after, but I think it's good practice to
attach the listener before.

+              window.removeEventListener('popstate', callback);
+              resolve(browserProxy.whenCalled('signOut'));

[michaelpg, 2016/08/26 20:56:09]

so you're resolving this promise with... a promise argument?

[tommycli, 2016/08/26 20:57:51]

Yes my theory was:

Wait for the popstate and also wait for the signOut call on the test browser
proxy.

Since the signOut call has probably already occurred, it would just resolve
instantly.

[michaelpg, 2016/08/26 21:34:10]

I don't think it works like that; passing a promise to resolve() doesn't make
the resolve "wait" on that promise. As far as I know, this just means the
function inside the "then" will be called with that argument, i.e. whatever
browserProxy.whenCalled('signOut') returns...

[tommycli, 2016/08/29 22:43:23]

Done.

[michaelpg, 2016/08/30 02:01:40]

I lied! It does work like that! O_O

http://jsbin.com/bikocotezo/1/edit?js,console

source:
http://www.ecma-international.org/ecma-262/6.0/#sec-promise-resolve-functions

+            });
+          });
+
           MockInteractions.tap(disconnectConfirm);
 
-          return browserProxy.whenCalled('signOut');
+          return dialogExitPromise;
         }).then(function(deleteProfile) {

[Dan Beam, 2016/08/29 18:08:55]

/**
 * @param {string} eventName
 * @param {!EventTarget} target
 * @return {!Promise<!Event>}
 */

function listenOnce(eventName, target) {
  return new Promise(function(resolve) {
    target.addEventListener(eventName, function once() {
      target.removeEventListener(eventName, once);
      resolve(event);
    });
  });
}

test('GetProfileInfo', function() {
  ...
  return browserProxy.whenCalled('getSyncStatus').then(function() {
    ...
    return listenOnce(window, 'popstate');
  }).then(function() {
    return browserProxy.whenCalled('signOut');
  }).then(function() {
    ...
  });
});

[tommycli, 2016/08/29 18:13:36]

Yes I think that would be excellent.

-          Polymer.dom.flush();
-
           assertFalse(deleteProfile);

[michaelpg, 2016/08/26 21:34:10]

...but then how is this false? browserProxy.whenCalled('signOut') should return
a non-null Promise... so I'm missing something, because in my mind this should
fail locally.

[tommycli, 2016/08/29 22:43:23]

deleteProfile is an argument of signout. It's only true for managed profiles and
when the user checks the "delete profile" checkbox. In this case, it should be
false.

[michaelpg, 2016/08/30 02:01:40]

yeah, nevermind -- I didn't know that calling resolve(somePromise) was different
from resolve(someValue). Weirdly, I still can't find any documentation for that
on MDN or the blogosphere.

 
           cr.webUIListenerCallback('sync-status-changed', {