On Wed, Aug 4, 2010 at 4:05 PM, Yuxiang Luo <luoyx@google.com> wrote:
Hi,
I'm currently looking into a bug (42799), and find that class CSSComputedStyleDeclaration has a property 'RefPtr<Node> m_node', which means this class takes ownership of the Node to which the computedStyle object belongs back to. Does this mean that if the computedStyle object get released, the owner Node will be released too?
I have figured out the issue, there is no problem for the two objects to Ref the Node. Actually the usage case is the just purpose the RefPtr...
For example, the class Position owns the node object 'RefPtr<Node> m_anchorNode', and in method 'Position::computedStyle', it passed the raw pointer m_anchorNode.get() to CSSComputedStyleDeclaration, after that, both 'Position' and 'CSSComputedStyleDeclaration' own the Node pointer. Is this a mistake in CSSComputedStyleDeclaration which incorrectly uses RefPtr to wrap the Node? (another similar class CSSMutableStyleDeclaration only has the raw pointer)
Another question is, as the bug 42799 shows, when getting 'pixel*' or 'pos*' properties, if the unit type is 'em' or 'ex', the result is not multiplied with factor. The factor for these two unit types depends on the RenderStyle of corresponding element, but the class CSSStyleDeclaration does't provide any access to corresponding element/node. I find that both CSSComputedStyleDeclaration and CSSMutableStyleDeclaration have the field 'm_node’. So to fix the bug, we might need to put the field up to the base class CSSStyleDeclaration. But I'm not very familiar with the code base yet, and not sure whether the change will break other logic? Thanks in advance for your help.
Best Regards.
-- Yuxiang Luo Engineer@Google <luoyx@google.com>
-- Yuxiang Luo Engineer@Google <luoyx@google.com>