Re: [xml] libxml .xsd validation problem

[Csaba Raduly <rcsaba gmail com>]

Hi Patrick,

On Tue, Sep 7, 2010 at 10:35 PM, Patrick McClory wrote:
> Hello,
>
> I'm working on a project which requires validation of xml documents against .xsd schemas.  We both create 
> xml documents from scratch, and create xml docs from char * buffers read from a socket.  I've run into 
> trouble validating the docs created from buffers, even when the buffers are generated from a document that 
> already validated successfully.
>
> For example I have the following schema in a file called "example.xsd":
>
> <?xml version="1.0" encoding="UTF-8"?>
> <!--
>  A simple test schema
> -->
> <xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema";
>    targetNamespace="http://localhost/test_namespace";
>    xmlns="http://localhost/test_namespace";>
>
>    <xs:complexType name="testType">
>      <xs:sequence minOccurs="1" maxOccurs="1">
>        <xs:element name="element1" type="xs:string"/>
>        <xs:element name="element2" type="xs:int"/>
>      </xs:sequence>
>    </xs:complexType>
>
>    <xs:element name="testInstance" type="testType"/>
> </xs:schema>
>
> The following code creates an xml doc from scratch, which validates.  It then dumps that doc into a buffer, 
> reads that buffer into a new doc, and tries to validate that, but the second validation fails:
(snip C code)
> The output generated when this runs is:
>
> generated doc is valid
> element element1: Schemas validity error : Element '{http://localhost/test_namespace}element1': This 
> element is not expected. Expected is ( element1 ).
> doc from buffer is invalid
>
> I dump both docs to output files (generated.xml and buffer.xml), and I confirmed that they're the same on 
> disk using diff.
>
> The problem seems to be that when libxml reads from the buffer it attaches the parent namespace to the 
> children (if it isn't specified), which later causes validation to fail.  Is this a common problem?  Is 
> there a standard workaround for this?

If you run "xmllint --schema example.xsd generated.xml", you'll get
the same error. Same with buffer.xml too.

In generated.xml, the default namespace of
"http://localhost/test_namespace"; applies to all elements, including
<element1> and <element2>. However, in the xsd the targetNamespace
applies only to the top-level <element name="testInstance"> (I think)

So either the in-memory validation is incorrect or the writing and
re-reading of the XML document is not a null operation.
xmlDocDumpFormatMemory may be at fault. To validate, element1 and
element2 would need a xmlns="" each.

The following validates with examle.xsd:

<testInstance xmlns="http://localhost/test_namespace";>
  <element1 xmlns="">foo</element1>
  <element2 xmlns="">1</element2>
</testInstance>

Running under the debugger:
(gdb) p *root_node
$1 = {_private = 0x0, type = XML_ELEMENT_NODE, name = 0xdca3e8
"testInstance", children = 0xdca468, last = 0xdca508, parent =
0xdca330, next = 0x0, prev = 0x0, doc = 0xdca330, ns = 0xdca400, ...}
(gdb) p *root_node->children
$2 = {_private = 0x0, type = XML_ELEMENT_NODE, name = 0xdca4a8
"element1", children = 0xdca4b8, last = 0xdca4b8, parent = 0xdca3a8,
next = 0xdca508, prev = 0x0, doc = 0xdca330, ns = 0xdca448,  ...}

Note the different ns for the root and the first child.

(gdb) p *new_doc ->children
$5 = {_private = 0x0, type = XML_ELEMENT_NODE, name = 0xdcaa2b
"testInstance", children = 0xdcaef0, last = 0xdcaf88, parent =
0xdcadf0, next = 0x0, prev = 0x0, doc = 0xdcadf0, ns = 0xdcaea8, ...}
(gdb) p *new_doc ->children->children
$6 = {_private = 0x0, type = XML_ELEMENT_NODE, name = 0xdcaa58
"element1", children = 0xdcaf30, last = 0xdcaf30, parent = 0xdcc9b8,
next = 0xdcaf88, prev = 0x0, doc = 0xdcadf0, ns = 0xdcaea8,  ...}

This time, the ns member has the same value for the root element and its child.


-- 
Life is complex, with real and imaginary parts.
"Ok, it boots. Which means it must be bug-free and perfect. " -- Linus Torvalds
"People disagree with me. I just ignore them." -- Linus Torvalds