[Transforms] event.offsetX and offsetY are wrong for transformed elements

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Simon Fraser (smfr)]

event.offsetX and offsetY (which are supposed to be relative to the target element are not computed correctly for transformed elements.

[Simon Fraser (smfr)]

Created attachment 24653 [details]
Testcase

[Simon Fraser (smfr)]

Note that this breaks some event handling on transformed sliders.

[Simon Fraser (smfr)]

Created attachment 24668 [details]
WIP patch

Here's a first cut that adds two new methods to RenderObject:

+    virtual bool objectPointToAbsoluteCoords(FloatPoint&, bool fixed = false) const;
+    virtual bool absolutePointToObjectCoords(FloatPoint&, bool fixed = false) const;

to map points to and from absolute coords, taking transforms into account.

Dave, does this look like a reasonable approach?

I should note that with this patch, a rotated <input type="range"> works perfectly!

[Simon Fraser (smfr)]

Created attachment 24670 [details]
Patch

Add two new methods to RenderObject:
+    virtual bool objectPointToAbsoluteCoords(FloatPoint&, bool fixed = false) const;
+    virtual bool absolutePointToObjectCoords(FloatPoint&, bool fixed = false) const;

which map points between RenderObject and absolute coords taking transforms into account.

Factor some RenderBox code into a new offsetFromContainer() method, now called in 3 places.

Call absolutePointToObjectCoords() when setting offsetX, offsetY in mouse events.

Add some comments, and a convenience expand() method on IntSize.

[Simon Fraser (smfr)]

Hmm, I think offsetX/offsetY will be wrong when there is shadow content

[Simon Fraser (smfr)]

(In reply to comment #5)
> Hmm, I think offsetX/offsetY will be wrong when there is shadow content

No, it's OK. MouseRelatedEvent::receivedTarget() gets the renderer from the target element.

[Dave Hyatt]

Comment on attachment 24670 [details]
Patch

It seems like absolutePosition could just be patched rather than adding new code...

[Simon Fraser (smfr)]

Created attachment 24928 [details]
Implement absoluteToLocal, fix event.offsetX/offsetY

[Simon Fraser (smfr)]

Created attachment 24929 [details]
Testcase in the patch

[Dave Hyatt]

Comment on attachment 24928 [details]
Implement absoluteToLocal, fix event.offsetX/offsetY

r=me (comments given on IRC).

[Simon Fraser (smfr)]

TODO: assert about LayoutState in absoluteToLocal()

[Simon Fraser (smfr)]

Committed r38186
	M	WebCore/rendering/RenderObject.cpp
	M	WebCore/rendering/RenderTableCell.cpp
	M	WebCore/rendering/RenderBox.h
	M	WebCore/rendering/RenderObject.h
	M	WebCore/rendering/RenderBox.cpp
	M	WebCore/rendering/RenderView.cpp
	M	WebCore/rendering/RenderView.h
	M	WebCore/rendering/RenderTableCell.h
	M	WebCore/dom/MouseRelatedEvent.cpp
	M	WebCore/ChangeLog
	M	LayoutTests/ChangeLog
	A	LayoutTests/fast/events/offsetX-offsetY.html
	A	LayoutTests/fast/events/offsetX-offsetY-expected.txt
r38186 = 845c2633f46b1eaecab8595b53b2a1cdfb06b7e2 (trunk)