Bug 29937 – caretRangeFromPoint() returns wrong result for text in <textarea>

• Status: RESOLVED
• Resolution: INVALID
• Component: DOM

[Xiaomei Ji]

Created attachment 40403 [details]
test case

1. Open the attached HTML page.
2. Type in "this is a test" in the textarea.
3. click anywhere in above text. observe the result
4. nodeType is always "element" node, and offset is always equals 3.


The following code in Document::caretRangeFromPoint() changed the node type from
TextNode to ElementNode, and the offset is set as the node index.

    Node* shadowAncestorNode = node->shadowAncestorNode();
    if (shadowAncestorNode != node) {
        unsigned offset = shadowAncestorNode->nodeIndex();
        Node* container = shadowAncestorNode->parentNode();
        return Range::create(this, container, offset, container, offset);
    }

What is the purpose of shadowAncestorNode()?
Should the above code piece be changed to the following? (related to issue 29249)

    Node::NodeType type = node->nodeType();
    if (type != Node::CDATA_SECTION_NODE && type != Node::COMMENT_NODE && type != Node::TEXT_NODE)
        return NULL;

[Sam Weinig]

I don't believe this can work.  Typing into a text area does not alter the DOM, so this would return shadow content which is an implementation detail.

[Xiaomei Ji]

(In reply to comment #1)
> I don't believe this can work.  Typing into a text area does not alter the DOM,
> so this would return shadow content which is an implementation detail.

I tried to change the above code piece, after I type in "this is a test" in the textarea, mouse clicking shows the same/similar result as mouse clicking in the paragraph above the textarea (the actual containing nodes are different though).

Typing into textarea does not alter DOM, but why need shadowAncestorNode() to change the hit-test node (and offset) if hit in textarea node? Sorry I did not get much idea on what shadowAncestorNode() does and why it is needed.

[Sam Weinig]

shadowAncestorNode() tells us if we are in a shadow tree, which is an implementation detail of how we represent certain form controls.  To the user, clicking in a textarea is just like clicking on an opaque replaced element.  The important take away is that stuff in the textarea cannot be made visible in this fashion.