Bug 19033

Created attachment 21112 [details] The patch containing the proposed fix for this issue.

Comment on attachment 21112 [details] The patch containing the proposed fix for this issue. The patch has tabs in it. Will submit a new patch for review.

Created attachment 21113 [details] The patch containing the proposed fix for this issue.

Comment on attachment 21113 [details] The patch containing the proposed fix for this issue. For this to get checked in, we'll need a test case demonstrating the problem, in the form of a regression test. I can't tell from reading the patch alone if this is correct. The test case will both help me understand that there truly is a bug here, and also is required as part of our "all bug fixes must have a regression test" policy. To simulate a mouse click, you can use the eventSender feature of the regression test engine. The test in fast/events/mousedown_in_scrollbar.html is an example of one that tests a mouse event in a scrollbar, and could be used as a starting point in making a test.

We still have this diff in our tree, and I'm not sure why. http://www.corp.google.com/~dglazkov/merge/WebCore-page-EventHandler.cpp-before.html#difflib_chg_to58__1

My apologies. This is the external URL: http://build.chromium.org/merge/WebCore-page-EventHandler.cpp-before.html#difflib_chg_to106__2

jQuery UI Dialog is affected by this issue: http://dev.jqueryui.com/ticket/4671

I think the change looks ok (I am not a reviewer) since MouseEventWithHitTestResults does not set "scrollbar" when hit-testing a frame scrollbar, but only in-frame scrollbars (like scrollable <div>, <textarea>, etc). (In reply to comment #3) > Created an attachment (id=21113) [details] > The patch containing the proposed fix for this issue. However I could not reproduce the problem on Safari (webkit - trunk) on Snow Leonard. Given PLATFORM "WIN XP", I think it is Windows specific. Could you confirm?

(In reply to comment #0) > Launch Safari 3.1 and navigate to http://cn.msn.com > Wait for the page to load fully. > Now click on the vertical/horizontal scrollbars. > Observe that the page does not scroll. It does scroll > with the mousewheel. Safari seems to scroll the page ok now. Please provide a (reduced) test case or a new URL that shows the bug or it is not actionable. > Debugged this and found that the mouse down event is handled by the > node returned as a result of the hit test. The hit test does not have > a scrollbar associated with it. This causes the page to not scroll. > The code should also check if there is a scrollbar under the mouse and > pass it off if yes. The webkit code already does this in the case when > the mouse down event is not handled. >

Created attachment 209805 [details] Reduced test case This bug is occurring on http://acko.net/ anywhere at the top. Debugging confirms a document mousedown handler fires when clicking the scrollbar. Emptying out the entire DOM does not fix this. This is on Mac OS X, both with and without permanent scrollbars enabled. When using temporary scrollbars, you can also click the Gear button in the top right through the scrollbar. I've attached a reduced test case. Note: 1) The scrollbar is unusable with the mouse 2) The grey div punches through the scrollbar for hit testing Bug appears on Safari 6.0.5, Chrome 29, Chrome 31.

Reproduced on Linux (webkit-gtk) and Mac/Safari. Taking..

Created attachment 273008 [details] Patch, no tests yet. For EWK

Comment on attachment 273008 [details] Patch, no tests yet. For EWK View in context: https://bugs.webkit.org/attachment.cgi?id=273008&action=review I am not familiar with this code but have a couple of minor comments. > Source/WebCore/page/EventHandler.cpp:1580 > +Scrollbar* scrollbarForMouseEvent(const MouseEventWithHitTestResults mouseEvent, FrameView* view) const MouseEventWithHitTestResults& > Source/WebCore/page/EventHandler.cpp:1582 > + Scrollbar* scrollbar = nullptr; Scrollbar* scrollbar = view ? view->scrollbarAtPoint(mouseEvent.event().position()) : nullptr;

Comment on attachment 273008 [details] Patch, no tests yet. For EWK View in context: https://bugs.webkit.org/attachment.cgi?id=273008&action=review >> Source/WebCore/page/EventHandler.cpp:1580 >> +Scrollbar* scrollbarForMouseEvent(const MouseEventWithHitTestResults mouseEvent, FrameView* view) > > const MouseEventWithHitTestResults& also should be static.

Created attachment 273012 [details] Patch, no tests yet. For EWK (v2)

Attachment 273012 [details] did not pass style-queue: ERROR: Source/WebCore/ChangeLog:8: You should remove the 'No new tests' and either add and list tests, or explain why no new tests were possible. [changelog/nonewtests] [5] Total errors found: 1 in 2 files If any of these errors are false positives, please file a bug against check-webkit-style.

Created attachment 273032 [details] Patch with tests (v1)

Comment on attachment 273032 [details] Patch with tests (v1) Attachment 273032 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-queues.webkit.org/results/923600 New failing tests: fast/events/prevent-default-prevents-interaction-with-scrollbars.html

Created attachment 273040 [details] Archive of layout-test-results from ews106 for mac-yosemite-wk2 The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews. Bot: ews106 Port: mac-yosemite-wk2 Platform: Mac OS X 10.10.5

Created attachment 273084 [details] Patch with tests (v2) Attempt to make the test run on EWK-Mac2.

Comment on attachment 273084 [details] Patch with tests (v2) Attachment 273084 [details] did not pass mac-wk2-ews (mac-wk2): Output: http://webkit-queues.webkit.org/results/927157 New failing tests: fast/events/prevent-default-prevents-interaction-with-scrollbars.html

Created attachment 273085 [details] Archive of layout-test-results from ews105 for mac-yosemite-wk2 The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews. Bot: ews105 Port: mac-yosemite-wk2 Platform: Mac OS X 10.10.5

Created attachment 273196 [details] Patch with tests (v3)

Adding Simon/Beth/Darin, which I believe are familiar with this code. This is a simple fix, that affect real world websites, and differ WebKit from Gecko/Blink these days.

Created attachment 273295 [details] Patch with tests (v3.1) Rebased against ToT, fixed typos on WebCore/ChangeLog.

Comment on attachment 273295 [details] Patch with tests (v3.1) View in context: https://bugs.webkit.org/attachment.cgi?id=273295&action=review > Source/WebCore/page/EventHandler.cpp:1585 > + Scrollbar* scrollbar = view ? view->scrollbarAtPoint(mouseEvent.event().position()) : nullptr; > + if (!scrollbar) > + scrollbar = mouseEvent.scrollbar(); > + return scrollbar; Now that this is factored out into a function, it can be refactored to make the logic slightly clearer. Like this, perhaps. if (view) { if (auto* scrollbar = view->scrollbarAtPoint(mouseEvent.event().position())) return scrollbar; } return mouseEvent.scrollbar();

Created attachment 273303 [details] Patch for landing.

(In reply to comment #26) > Comment on attachment 273295 [details] > Patch with tests (v3.1) > > View in context: > https://bugs.webkit.org/attachment.cgi?id=273295&action=review > > > Source/WebCore/page/EventHandler.cpp:1585 > > + Scrollbar* scrollbar = view ? view->scrollbarAtPoint(mouseEvent.event().position()) : nullptr; > > + if (!scrollbar) > > + scrollbar = mouseEvent.scrollbar(); > > + return scrollbar; > > Now that this is factored out into a function, it can be refactored to make > the logic slightly clearer. Like this, perhaps. > > if (view) { > if (auto* scrollbar = > view->scrollbarAtPoint(mouseEvent.event().position())) > return scrollbar; > } > return mouseEvent.scrollbar(); Sounds good, done.

The commit-queue encountered the following flaky tests while processing attachment 273303 [details]: http/tests/xmlhttprequest/workers/xmlhttprequest-file-not-found.html bug 155176 (authors: atwilson@chromium.org and levin@chromium.org) The commit-queue is continuing to process your patch.

Comment on attachment 273303 [details] Patch for landing. Clearing flags on attachment: 273303 Committed r197784: <http://trac.webkit.org/changeset/197784>

All reviewed patches have been landed. Closing bug.