602745 - GetScreenBounds returns size too large at first paint of new window

Status: NEW

(Core :: Widget: Win32, defect)

Description

[Karl Tomlinson (:karlt)]

The test in http://hg.mozilla.org/try/rev/33d1dfb0c7ae creates a new window
with outerWidth and outerHeight features, waits for the first paint and then
checks the window size.

> 15038 ERROR TEST-UNEXPECTED-FAIL | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | initial outerHeight - got 185, expected 170
> 15039 ERROR TEST-UNEXPECTED-FAIL | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | initial outerWidth - got 218, expected 200
> 15040 INFO TEST-PASS | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | resized outerWidth - 180 should equal 180
> 15041 INFO TEST-PASS | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | resized window top should not change - 18 should equal 18
> 15042 INFO TEST-PASS | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | restored outerWidth - 200 should equal 200
> 15043 INFO TEST-PASS | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | restored window left should not change - 28 should equal 28
> 15044 INFO TEST-PASS | chrome://mochitests/content/chrome/widget/tests/test_position_on_resize.xul | restored window top should not change - 18 should equal 18

toolkit/content/tests/chrome/test_screenPersistence.xul is similar but waits
for focus before passing when checking the size, suggesting that this might be just a timing issue.

Comment 1

[Brian R. Bondy [:bbondy]]

I'm going to try working on this, it bugs me on buildbot try results :)

Comment 2

[Jonathan Kew [:jfkthame]]

It looks like the window-management patches in bug 794038 may have the side-effect of fixing this - it's showing up as an "unexpected pass" in tryserver results there.

Comment 3

[Jonathan Kew [:jfkthame]]

It was attachment 670755 [details] [diff] [review] in bug 794038 that caused this test to start passing in tryserver builds, but that version unfortunately led to other test failures instead, so it's not being used there after all. Still, it might at least give some kind of a clue...