Bug 11726 – SVG Image do not take into account clip/overflow when hit testing

• Status: RESOLVED
• Resolution: FIXED
• Component: SVG

[Rob Buis]

SVG Images can be quite large and one might want to clip them. For instance 
http://www.carto.net/papers/svg/gui/scrollbar/index.svg show a nice example of large images
that are clipped to a container. However it doesnt work well for ToT, whereas html images do not seem
to have a problem.

[Rob Buis]

Created attachment 11690 [details]
Testcase showing the problem

This testcase clearly shows the problem. Please place in LayoutTests/svg/W3C-SVG-1.1/.
Cheers,

Rob.

[Rob Buis]

Created attachment 11708 [details]
first attempt

This patch fixes the testcase. I worked it into a nice layout test.
Cheers,

Rob.

[Oliver Hunt]

Comment on attachment 11708 [details]
first attempt

SVGContainer can be <svg> <use> or <g> all of which can have transforms.  As far as i can make out you don't account for this.

Other than that everything looks fine, so if my concern is unfounded this can be r+

[Rob Buis]

Created attachment 11832 [details]
Improved patch

Make it compile against ToT. Also make it work for <svg>s that set x and y to something other than 0. The testscase now tests this.
Cheers,

Rob.

[Oliver Hunt]

Comment on attachment 11832 [details]
Improved patch

RenderSVGContainer::nodeAtPoint you're still doing a check to ensure only svg elements get transforms applied, but to my knowledge <g> and <use> can both have transforms which you need to honour.

[Rob Buis]

Hi Oliver,

(In reply to comment #5)
> (From update of attachment 11832 [details] [edit])
> RenderSVGContainer::nodeAtPoint you're still doing a check to ensure only svg
> elements get transforms applied, but to my knowledge <g> and <use> can both
> have transforms which you need to honour.

Hmm, that is true. But note that this particular bug is about clip/overflow, something which <g> can't have. <use> doesnt really "exist" right now, it just creates a RenderSVGContainer or RenderSVGImage IIRC. So IMHO it may be worthwhile to (a) check whether there is a problem hit detecting a transformed <g> and contents and (b) if so, create a new bugreport for that, What do you think?
Cheers,

Rob.

[Oliver Hunt]

Comment on attachment 11832 [details]
Improved patch

my bad.  File a bug on nodeAtPoint however.  I don't think that will be correct in all cases

[David Kilzer (:ddkilzer)]

Commit by rwlbuis in r18210.