Merge of needsLayout and childNeedsLayout from Webcore/125

Germain Garand germain at ebooksfrance.org
Wed Oct 13 15:23:59 BST 2004


Le Mardi 12 Octobre 2004 23:21, Allan Sandfeld Jensen a écrit :
> Hi!
>
> I am trying to merge various bits from Safari while learning KHTML
> (low-hanging fruit and all).
>
> One of the larger, is the change from layouted() to needsLayout() and
> relatives in renderObject; it touches almost every file in the rendering
> directory.  Most of the changes are fortunatly trivial (some of the them
> involving satisfying removal of "// #HACK!" comments). There are however a
> few I have doubts about, which is why I post it here.

Hi Allan,
great patch!

>
> During the merge I found I few places in our KHTML where we called
> layouted(false), but where the equivalent calls in Webcore was missing. I
> am guessing these removals are optimizations from Apples side, but I
> haven't quite wringed my head around the rendering-callgraph enough to be
> sure. These have been converted but commented-out in the attached patch.
> There are a few places in render_block.cpp, render_container.cpp and one in
> render_inline.cpp.

those in RenderBlock::{removeChild, makeChildrendNonInline,  addChildToFlow, 
removeChild} and RenderContainer::addChild looks OK, since the patch changes
RenderContainer::appendChildNode/insertChildNode to dirty the relevant bits on 
'this' and the new child,
but for the one in RenderContainer::removeChildNode, I'm not sure at all... 
I don't see where 'this' coud be dirtied. Might be dangerous.

Does it pass test regression?

Greetings,
Germain




More information about the kfm-devel mailing list