We use a dummy document there to match the processing model in the spec. If we used the real document, we'd have to add a bunch of branches to the HTMLTreeBuilder that aren't in the spec to get the correct answer when parsing. How is this document getting associated with a request object? That seems like real bug. The dummy document has a NULL frame, which means it shouldn't be kicking off any requests. Adam On Mon, Aug 30, 2010 at 4:12 AM, Roland Steiner <rolandsteiner@google.com> wrote:
Hi knowledgeable WebKit folks! I'm currently trying to debug an annoying loader/resource request bug that seems to originate with the dummy Document used in HTMLTreeBuilder::FragmentParsingContext. It seems that in some cases the DocLoader that is created with this dummy is assigned to a Request object, which seems to be the underlying reason of the bug that I'm seeing. Instead of just throwing additional RefPtr at the problem to make it go away, I wondered why this dummy document is necessary in the first place. Wouldn't it be more correct to just use the document assigned to the passed-in DocumentFragment (or context element)? I.e., would bad things happen if I removed this dummy?
- Roland _______________________________________________ webkit-dev mailing list webkit-dev@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev