With https://bugs.webkit.org/show_bug.cgi?id=71490 and http://trac.webkit.org/changeset/99493 I added support for uiStateTransitionProgress on scrollbars. To properly animate uiStateTransitionProgress, we keep track of whenever the mouse enters or exits a scrollbar. However, to fully support the feature, we have to keep track of that information even when the window is not key. Currently, we only track mouseMoved events when the window is key, so this needs to change to fully support uiStateTransitionProgress. <rdar://problem/10409328>
Created attachment 115209 [details] Patch
Comment on attachment 115209 [details] Patch Attachment 115209 [details] did not pass gtk-ews (gtk): Output: http://queues.webkit.org/results/10372632
Created attachment 115235 [details] Patch that probably won't break builds
Created attachment 115279 [details] Smaller patch that does less I decided to break this up into two bugs so that the patches will be less intimidating. This patch makes is so we track the mouse all of the time when (at the time the WKView is constructed) the recommended scrollbar style is the legacy style. The rest of the work in my initial patch was to update this dynamically whenever the scrollbar style changes, but I will do that in a second patch instead.
Created attachment 115280 [details] Now with WebCore Oops, forgot to include WebCore changes in that one.
Comment on attachment 115280 [details] Now with WebCore View in context: https://bugs.webkit.org/attachment.cgi?id=115280&action=review > Source/WebCore/platform/Scrollbar.h:97 > + bool mouseEntered(); Why does this have a return value? The only caller doesn’t use it.
(In reply to comment #6) > (From update of attachment 115280 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=115280&action=review > > > Source/WebCore/platform/Scrollbar.h:97 > > + bool mouseEntered(); > > Why does this have a return value? The only caller doesn’t use it. Thanks for the review! I gave it a return value so that it would be like mouseExited() which has a return value. But ultimately, I think that you are right; it's silly to give something a totally unused return value. I'll remove it.
Committed change with revision 100483.
This is in the blame list for this API test failure: http://build.webkit.org/builders/SnowLeopard%20Intel%20Release%20%28Tests%29/builds/34756/steps/run-api-tests/logs/stdio