[Webkit-unassigned] [Bug 29966] Assertion failure in RenderBlock::positionForPointWithInlineChildren when running fast/inline/relative-positioned-overflow.html

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=29966





--- Comment #9 from mitz at webkit.org  2009-11-26 10:02:33 PST ---
(From update of attachment 43916)
Thanks for working on this bug!

> +    eventSender.mouseMoveTo(x, y + 50);
> +    eventSender.mouseDown();
> +    eventSender.mouseMoveTo(x, y + 40);
> +    eventSender.mouseMoveTo(x, y + 30);
> +    eventSender.mouseMoveTo(x, y + 20);
> +    eventSender.mouseMoveTo(x, y + 10);
> +    eventSender.mouseMoveTo(x, y);
> +    eventSender.mouseUp();

Is there a reason to visit all the intermediate positions? I think you need
just two mouseMoveTo()s, and even that is only because the selection start
position is determined not on mouse down but on the first mouse move after that
(which is arguably a bug). I think mouseMoveTo(x, y + 50); mouseDown();
mouseMoveTo(x, y + 50); mouseMoveTo(x, y); mouseUp(); will work.

I don’t understand why this fix is correct. Just using a non-leaf child (which
is not even necessarily a left in the line box tree, it may have nested inline
boxes) in this case seems inconsistent. What if there is a line in the middle
that has no leaf boxes. What if the first line that has children has no leaf
boxes? It seems like we may end up calling positionForBox(null, …) in such
cases because we consider lines that have no leaf boxes and then call
{first,last,closest}LeafChild on them. Perhaps the right thing to do is in the
initial vertical scan loop, change
        if (!root->firstChild())
            continue;
to
        if (!root->firstLeafChild())
            continue;

so that only lines with leaf child boxes are considered. Did you try that?

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.