fast/events/selectstart-by-drag.html has been failing on Windows ever since it was checked in: Initial state: PASS -Mouse down: PASS -Moving slightly to the right: PASS -Moving slightly to the left: PASS +Mouse down: FAIL - expected selection to be caret but was range +Moving slightly to the right: FAIL - expected selection to be caret but was range +Moving slightly to the left: FAIL - expected selection to be caret but was range Moving to the right: PASS Moving further to the right: PASS -Moving back to the left: PASS +Moving back to the left: FAIL - expected selection to be caret but was range Moving to the right again: PASS Mouse down on the right: PASS Moving to the left: PASS
Ryosuke checked in expected failure results in r87105 (thanks!).
Ryosuke, do you think this represents a bug in WebKit, or just a bug in the test or test harness? If the latter, we should move this bug to the Tools/Tests component.
(In reply to comment #2) > Ryosuke, do you think this represents a bug in WebKit, or just a bug in the test or test harness? If the latter, we should move this bug to the Tools/Tests component. It's hard to tell. It's very odd that the first click on a page creates a range selection though. There might be a bug in Windows port or Windows port's DRT because Chromium Windows passed the test.
Maybe a click from a previous test is causing the click in this test to be counted as a double-click? We've seen that before. See bug 56347.
(In reply to comment #4) > Maybe a click from a previous test is causing the click in this test to be counted as a double-click? We've seen that before. See bug 56347. That is an annoying bug. But I'm confused as to how that happens. m_clickCount is stored in EventHandler but that class exists per frame. If we're loading a new page, we'll surely create a new frame, right?
(In reply to comment #5) > (In reply to comment #4) > > Maybe a click from a previous test is causing the click in this test to be counted as a double-click? We've seen that before. See bug 56347. > > That is an annoying bug. But I'm confused as to how that happens. m_clickCount is stored in EventHandler but that class exists per frame. If we're loading a new page, we'll surely create a new frame, right? EventHandler::m_clickCount is set based on the PlatformMouseEvent that's passed in. The PlatformMouseEvent has a click count of 2 (or more) due to WebKit not clearing its global click count on navigation.
(In reply to comment #6) > EventHandler::m_clickCount is set based on the PlatformMouseEvent that's passed in. The PlatformMouseEvent has a click count of 2 (or more) due to WebKit not clearing its global click count on navigation. Where is this global counter stored? It seems like we should just fix that.
(In reply to comment #7) > (In reply to comment #6) > > EventHandler::m_clickCount is set based on the PlatformMouseEvent that's passed in. The PlatformMouseEvent has a click count of 2 (or more) due to WebKit not clearing its global click count on navigation. > > Where is this global counter stored? It seems like we should just fix that. I'll put the info in bug 56347.