57999 - XML that is NOT well formed is displayed when a certain stylesheet is applied.

Status: VERIFIED FIXED

(Core :: XML, defect, P3)

Description

[Chris Petersen]

Build: 2000102508
Platform: All
Expected Results: The parser error should occur.
What I got: The document's content is rendered (with style applied).

Steps to reproduce:

1) Save attached xml and css documents:

2) Open the xml document in N6.

3) The document's content is rendered.

4) Since the document is not well-formed, a parser error should be displayed.

Comment 1

[Chris Petersen]

Attachment: A xml document that's not well formed

Comment 2

[Chris Petersen]

Attachment: a stylesheet

Comment 3

[Chris Petersen]

This problem occurs if the stylesheet is referenced in the xml file (whether or
not the css href can be found). If you remove the xml-stylement declaration, the
parser will display the error properly.

Comment 4

[Chris Petersen]

Basically, the xml document must not be well-formed and have a stylesheet
applied to reproduce this bug.

Comment 5

[Chris Petersen]

Nominating this bug.

Comment 6

[Chris Petersen]

IE 5 (Mac) displays a parser error when attempting this xml document. The 
following tags are not closed: main. 

Comment 7

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Updated the severity to major. XML applications MUST report well-formedness errors.

Comment 8

[Chris Petersen]

Looking at this issue further, the problem only occurs if the root element is 
missing the end tag. If a child element (inside the well-formed root element) is 
missing a end tag, you get a parser error.

Comment 9

[Nisheeth Ranjan]

Marking this rtm need info.  About to look at it.

Comment 10

[Nisheeth Ranjan]

The problem here is that when the stylesheet PI is included in the attached XML 
document, the expat tokenizer's DidTokenize() method is never called with the 
aIsFinalChunk parameter set to true.  Because of this, expat keeps thinking that 
more content is about to come down the pipe and does not throw an error.

When the stylesheet PI is removed, the DidTokenize() method *is* called with the 
aIsFinalChunk parameter set to true and expat throws an error.  This is the 
right behavior.

We need to figure out why, when the stylesheet PI is included, DidTokenize() is 
never called with aIsFinalChunk set to true.

Over to Harish.

Comment 11

[harishd]

Attachment: Proposed fix.

Comment 12

[Jason Eager]

This needs a r= and a sr= to make it in.

Changing the summary to make the severity more important.

Comment 13

[Jason Eager]

umm /s/important/apparant/g

Comment 14

[harishd]

Nisheeth and I went thro' the patch and we're not completely comfortable with 
it. Will work on it.

Comment 15

[vidur (gone)]

The fix is correct. It passes on the same state that we'd get through other
codepaths when the stream has ended. I've asked Harish to run the XML test suite
and also see whether we hit this case while loading up chrome. I do have
reservations about its importance for rtm, but if it precludes us from showing
well-formedness errors at all for most XML files, it should be fixed.

sr=vidur

Comment 16

[Nisheeth Ranjan]

r=nisheeth.  Marking rtm+.  This is important because it prevents us from
displaying an XML error for certain classes of non well formed XML pages.  One
of the major design goals of XML is to avoid the proliferation of erroneous XML
content on the web by ensuring that every XML processor rejects such content
outright.  We will be doing the web and the XML community a disservice if we
don't fix this bug.  Please accept Harish's patch.

Comment 17

[Nisheeth Ranjan]

Updating status whiteboard.

Comment 18

[selmer (gone)]

rtm-, not ship stopper.

Comment 19

[harishd]

Fix landed. Marking FIXED.

Comment 20

[Chris Petersen]

Fixed in the Feb 26th build.