Fix link highlights on links that contain images.

Fix link highlights on links that contain images.

Link highlights are now generated by walking down the render tree and
accumulating quads, rather than taking one bounding quad.

The real issue was with RenderInlines; absoluteQuads would return the
quads for the line box but only as calculated from line-height and font
metrics. This meant that replaced elements like <img> would overflow out of
the highlight. With this approach we get a highlight that hopefully
matches the actual hit testing region.

BUG=289764
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=165024

[wjmaclean, 2014-01-08 23:08:02]

lgtm - thanks David!

[aelias_OOO_until_Jul13, 2014-01-11 00:22:45]

https://codereview.chromium.org/129643002/diff/30001/Source/web/LinkHighlight...
File Source/web/LinkHighlight.cpp (right):

https://codereview.chromium.org/129643002/diff/30001/Source/web/LinkHighlight...
Source/web/LinkHighlight.cpp:199: for (unsigned i = 0; i < outQuads.size(); ++i)
{
Nit: use size_t instead of unsigned

https://codereview.chromium.org/129643002/diff/30001/Source/web/LinkHighlight...
Source/web/LinkHighlight.cpp:202: newQuads.remove(newIx - 1);
removal is an O(n) operation on a Vector.  In addition this is within two loops
inside a recursive call.  I'm concerned the running time of this algorithm may
become pathological.  Could you give some thought about how this could be done
more efficiently?

[bokan, 2014-01-13 15:52:00]

Alexandre, PTAL. I've removed the containsQuad checks and opted instead to not
add the quads from the RenderInline but only from its children.

[bokan, 2014-01-13 17:08:10]

On 2014/01/13 15:52:00, bokan wrote:
> Alexandre, PTAL. I've removed the containsQuad checks and opted instead to not
> add the quads from the RenderInline but only from its children.

Actually, ignore for now, I've found some issues with my approach. I'll update
again when its ready.

[aelias_OOO_until_Jul13, 2014-01-13 17:37:10]

lgtm, thanks.

[aelias_OOO_until_Jul13, 2014-01-13 17:37:59]

Ah okay, missed your last update.

[bokan, 2014-01-13 19:13:12]

On 2014/01/13 17:37:59, aelias wrote:
> Ah okay, missed your last update.

Sorry about that. The issue I was having was that the render subtree can be
broken up into anonymous renderers (as is the case when we have block-level
boxes mixed with text/inline boxes) that aren't all parented from our renderer.
I've changed the quad computation to descend down the DOM tree instead, I think
that matches our actual intent here.

Still lgtm?

[aelias_OOO_until_Jul13, 2014-01-13 22:19:13]

still lgtm

[commit-bot: I haz the power, 2014-01-13 22:40:17]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/bokan@chromium.org/129643002/200002

[commit-bot: I haz the power, 2014-01-14 04:27:41]

Change committed as 165024