If the source html file has a newline after the body (which I would assume that most do), the parser inserts that newline as a text node inside, rather than outside, the body node in the content tree. The easiest way to see this is to run the editor on the file htmlparser/tests/outsinks/simple.html, then do a Debug->Dump Content Tree (or do the equivalent debug action in viewer). Note the three adjacent newlines near the end but inside the body node; from the source file, they should be two adjacent newlines inside the body plus another newline outside the body. For extra credit, it would be nice to get the right number of newlines after the </html>, but I can live without that and special-case it in the output sink; I could special-case a newline after the </body> tag, too, if necessary. But I can't detect that the extra newline node shouldn't be there inside the body, so the output will show an extra newline before the </body> tag where there shouldn't be one.
I'm guessing this is the same problem as the longstanding problem of newlines in the head getting inserted into the body node. For instance, the sample attachment to bug 15674: <!doctype HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html40/loose.dtd"> <html> <head> <title>An Example</title> </head> <body> <h1>Test</h1> </body> </html> gets parsed into this tree: html refcount=4< head refcount=3< title refcount=3< Text refcount=3<An Example> > > Text refcount=4< \n\n\n\n\n\n> body refcount=4< Text refcount=4<\n> h1 refcount=4< Text refcount=8<Test> > Text refcount=4<\n\n\n\n> > > and there's no way for the output routines to know which of these are valid body newlines and which are not, so they all go into the output.
This seems like bug 12998 at the opposite end...
Yes, it does look like a dup of 12998. That one is marked as fixed, but I'm still seeing it.
This one is for whitespace at both ends (see my second comment to the bug report); and as of Friday, I was still seeing whitespace at both ends being inserted in the wrong place.
OK... but at the beginning that extra whitespace is outside of body (it was once inside) whereas at the end it's inside. It's still not exactly where it was, though...
"where it was" == "where it was in the document"
Akkana, looks like there is nothing much that could done on the parser end. Giving the bug back to you.
IMO, this should be fixed in the parser, just like bug 12998 was. It's a bug in conformance with the W3C DOM Level 1 spec.
If the parser inserts newline nodes in the wrong place in the DOM tree, there's absolutely nothing the output system can do to distinguish these nodes from nodes which really were in the original source.
Rick and I discussed this at some length. He says that the whitespace at the end of the document (after the /body and which aren't associated with any other text or markup besides /html) should be trivial to fix. For the whitespace inside the head, at least some of it should also be easy to leave in place instead of coalescing it into one text node between the head and the body; he's going to look into that.
It turns out that this also happens for tables: all newlines occurring inside table nodes are coalesced into a couple of text nodes outside the table boundaries.
Fixed by updating CNavDTD.
updated qa contact.
Verified on: build: 2001-04-02-09-Mtrunk platform: WinNT Marking it verified as per the developer comments. Do not know how I can view the content tree using release builds?