Re: [xml] Re: Access Violation

["William M. Brack" <wbrack mmm com hk>]

Daniel Veillard wrote:
> > On Mon, Jul 28, 2003 at 03:15:09PM +0100, Vincent Finn wrote:
> >
> > > I am using it as follows
> > > void fn(xmlNodePtr pOperationElement)
> > > {
> > >     xmlNodePtr pEvent = pOperationElement->children;
> > >     while (pEvent != NULL)
> > >     {
> > >             xmlNodePtr pValueText = pValue->children;
> > >             const char* bstrValueText = reinterpret_cast<char
> > > const*>(pValueText->content);
> > >             // etc...
> > >             pEvent = pEvent->next;
> > >     }
> > > }
> > >
> > > I assume since it looks like heap corruption that I am misusing
> > > the library
> > > in some way
> > > Is there anything here that looks wrong to anyone?
> >
> >
> >   accessing ->content without checking the type of the node first
> > is 100% wrong. You're making assumption which is very likely to be
> > false.
>
> In my case it won't ever be false.
> Unless I have misunderstood content is for getting the text from a
> text
> node, in which case I would have thought what I am doing is fine
>
> I know with 100% certainty that the node I am accessing is of the
> form
> <E><V>48471740</V><T>01c354fc14bb8030</T></E>
> so I use content to access the text values
>
> As far as I could tell it seemed the same as using
> xmlNodeListGetString() but without the need to free the xmlChar
> which
> seemed to be a better solution for me since I am purely reading and
> want
> to avoid any data copying I can
>
> Is using content inherently unsafe?
>
>       Vin

As Daniel mentioned, it's a little difficult making guesses with a
code snippet, but you have gotten me a little confused.  First,
could you confirm that the line

   xmlNodePtr pValueText = pValue->children;

really should be

   xmlNodePtr pValueText = pEvent->children;

or else give me a hint what pValue is pointing to.

Next, in your original mail you said the structure of your data was
  <root>
     <firstChild>
        <E>....</E>
        ...
     </firstChild>
  </root>
so I am assuming that your pOperationElement is pointing to
<firstChild>.  If that assumption is true, xmllint shows me that I
should expect a structure looking like this:-

bill billsuper work $ cat checkit.xml
<root>
    <firstChild>
        <E><V>48471740</V><T>01c354fc14bb8030</T></E>
    </firstChild>
</root>

bill billsuper work $ xmllint --debug checkit.xml
DOCUMENT
version=1.0
URL=checkit.xml
standalone=true
  ELEMENT root
    TEXT
      content=
    ELEMENT firstChild
      TEXT
        content=
      ELEMENT E
        ELEMENT V
          TEXT
            content=48471740
        ELEMENT T
          TEXT
            content=01c354fc14bb8030
      TEXT
        content=
    TEXT
      content=
bill billsuper work $

In other words, pEvent will be pointing to the TEXT node, which is
the first child of firstChild, and not to the ELEMENT E which you
expect.  This is what Daniel is talking about when he advises you to
check the node type.

If I have misunderstood, or if one of my assumptions is incorrect,
please clarify.

Regards,

Bill