The fix for bug 11524 broke this test as it caused DOMNode::valueOf() to be used when comparing [meta] to a string, rather than DOMNode::toString(). In Firefox DOMNode::valueOf() and DOMNode::toString() return the same value, but we return a string based on nodeName for the former, and className for the latter.
Created attachment 11396 [details] Bring consistency to DOMNode::toString/valueOf
Comment on attachment 11396 [details] Bring consistency to DOMNode::toString/valueOf r=me
+FAILURE: expected "[object META]", actual "[object HTMLMetaElement]" Perhaps the test can adjust its expectations.
Landed in r17615, fast/dom/HTMLMetaElement/meta-attributes.html results updated in r17616. Good catch :)