I'm getting constant off-by-one and off-by-two errors with these tests in bug 384336. The cause seems to be that bug 384336 maximizes the browser window the tests run in, which in theory shouldn't bother these tests at all. I suspect there's a win32 widget bug somewhere, however I can't actually reproduce the problem by just using window.open in a maximized window on Win10. Maybe this is Win7-specific. For now I'm going to disable these tests on Windows. CC'ing bz and dolske as the original authors. TEST-UNEXPECTED-FAIL | docshell/test/navigation/test_bug344861.html | window.open has correct height dimensions - got 199, expected 200 TEST-UNEXPECTED-FAIL | docshell/test/navigation/test_bug344861.html | window.open has correct width dimensions - got 198, expected 200 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | Checking doc width - got 398, expected 400 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | Checking doc height - got 299, expected 300 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image width - got 398, expected 400 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image height - got 298, expected 300 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image width - got 398, expected 400 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image height - got 298, expected 300 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | Checking scrollLeft - got 402, expected 400 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | Checking scrollTop - got 301, expected 300 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image width - got 398, expected 400 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug369370.html | image height - got 298, expected 300 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug741266.html | Popup height should be 100 when opened with window.open - got 99, expected 100 TEST-UNEXPECTED-FAIL | dom/html/test/test_bug741266.html | Popup height should be 100 when opened with document.open - got 99, expected 100

test_bug344861.html screenshot from https://treeherder.mozilla.org/#/jobs?repo=try&revision=1d8fb66fdbcb&selectedJob=14618598 In that try push I maximized the window regardless of the screen size to see if the failure surfaced on Windows only because we might be using larger screens for other platforms. Other platforms stayed green regardless.

Comment on attachment 8701015 [details] mozilla-test-fail-screenshot_xsteoc.png Note that the popup's content area is 198x199 here just like measured by the test, so this failure seems real (rather than a mistake in the test, which seems unlikely anyway since we have three tests failing and they're all pretty basic).

Dolske, any idea what's up here?

(In reply to Boris Zbarsky [:bz] from comment #5) > Dolske, any idea what's up here? Hmm, I don't really know. But I'd concur that this sounds like a widget bug. Is it possible these are the only tests opening a popup window and testing the expected size? (ie, I'd suspect the actual guts of the tests don't really matter). Maybe easy enough to verify by writing a trivial test that opens a popup window and checks the size. If I slide out a little further on this wild-guess branch, I'd wonder if it's something to do with the size of native window decorations on children of maximized windows. I think widget/windows/nsWindow.cpp does some funky stuff around this, and could see there being an off-by-one kind of bug there. Further, the top window border is "special" so that would explain why the width is off by 2 (1px left + 1px right), while the height is only off by 1px (1px bottom but not top, of vice versa). Probably needs someone to step through how the requested chrome window size and native window size interact? And see if this can be reproduced locally on a Win7 VM?

I just checked this on a Windows 7 machine and could indeed reproduce it.

FWIW, this seems eerily related to the test failures that appeared when trying to land bug 1005098...

I will try to get to this at some point, but right now I do not have the time.

This is causing some tests to be disabled in e10s mode in dom/html: https://dxr.mozilla.org/mozilla-central/source/dom/html/test/mochitest.ini

(Bug 384336 may have exposed this in testing more, but it sounds like a pre-existing issue, so removing to help tracking.)

(In reply to Andrew Overholt [:overholt] from comment #10) > This is causing some tests to be disabled in e10s mode in dom/html: > https://dxr.mozilla.org/mozilla-central/source/dom/html/test/mochitest.ini No, this is independent from e10s.

Pushed by VYV03354@nifty.ne.jp: https://hg.mozilla.org/integration/autoland/rev/b69e628e7733 Re-enable test_bug344861.html, test_bug369370.html, test_bug741266.html on Windows. r=dao