I'm working on a CL that adds a text button to the status area when feature SingleProcessMash. See  issue 890400  and CL https://crrev.com/c/1252632. This changes the shelf view's available width. Strangely, this causes ExtendedDesktopTest.PassiveGrab to fail.

The test verifies that ui::test::EventGenerator correctly "grabs" the mouse during a click-and-drag across multiple monitors. Oshima added it in https://chromiumcodereview.appspot.com/11691010/

I notice it contains this assertion:

  UpdateDisplay("300x300,200x200");
  views::Widget* widget = CreateTestWidget(gfx::Rect(50, 50, 200, 200));
  widget->Show();
  ASSERT_EQ("50,44 200x200", widget->GetWindowBoundsInScreen().ToString());
                ^^
This is clearly wrong - the widget's Y position should not be 44 immediately after creation.

There is also a later assertion about the mouse position after it moves to the second display:

  generator->MoveMouseTo(400, 150);
  EXPECT_EQ("100,6 100,150", event_handler.GetLocationsAndReset());
                 ^
This is also wrong. The mouse's Y position should be 150.

These assertions used to contain the values I expect. However, manucornet changed the assertions in this CL that enabled the new shelf:
https://chromium-review.googlesource.com/c/chromium/src/+/1188848

Something in the new shelf code is breaking either:

* Widget size
* Display work area
* Event targeting

We should either figure out what's broken in the new shelf. In the short term we need to make this test less brittle such that unrelated changes don't break the hard-coded values.

 

I'm working on a CL that adds a text button to the status area when feature SingleProcessMash. See  issue 890400  and CL https://crrev.com/c/1252632. This changes the shelf view's available width. Strangely, this causes ExtendedDesktopTest.PassiveGrab to fail.

The test verifies that ui::test::EventGenerator correctly "grabs" the mouse during a click-and-drag across multiple monitors. Oshima added it in https://chromiumcodereview.appspot.com/11691010/

I notice it contains this assertion:

  UpdateDisplay("300x300,200x200");
  views::Widget* widget = CreateTestWidget(gfx::Rect(50, 50, 200, 200));
  widget->Show();
  ASSERT_EQ("50,44 200x200", widget->GetWindowBoundsInScreen().ToString());
                ^^
This is clearly wrong - the widget's Y position should not be 44 immediately after creation.

There is also a later assertion about the mouse position after it moves to the second display:

  generator->MoveMouseTo(400, 150);
  EXPECT_EQ("100,6 100,150", event_handler.GetLocationsAndReset());
                 ^
This is also wrong. The mouse's Y position should be 100.

These assertions used to contain the values I expect. However, manucornet changed the assertions in this CL that enabled the new shelf:
https://chromium-review.googlesource.com/c/chromium/src/+/1188848

Something in the new shelf code is breaking either:

* Widget size
* Display work area
* Event targeting

We should either figure out what's broken in the new shelf. In the short term we need to make this test less brittle such that unrelated changes don't break the hard-coded values.