[soap] fix #75747 correct detect namespace for attributeGroup

[iRipVanWinkle]

Corrects detect namespace for attributeGroup when parse xml file with absented targetNamespace or xmlns

On the step of parsing schema the empty namespace replace into target namespace. But on the step of fix up types the empty namespace stay empty so cannot find type in temporary hash table with attribute groups.
Detecting namespace in other tags are correct and similar to mine fix.

[DaveRandom]

I haven't dug into this fully (yet) but this looks to me like a sticking plaster for a specific issue caused by a wider problem. Based on what the test does and looking at the patch code, it seems like the real problem is that the targetNamespace attribute is not correctly take into account when processing type definitions that came from an <xsd:include>. If so, this most likely affects many things in the included definitions, and not just <attributeGroup>.

For example, consider the following code for include.xml:

<?xml version="1.0" encoding="UTF-8"?>
<xsd:schema xmlns:xsd="http://www.w3.org/2001/XMLSchema">
  <xsd:complexType name="Inner">
    <xsd:sequence>
      <xsd:element name="el" type="xsd:string" />
    </xsd:sequence>
  </xsd:complexType>
	<xsd:complexType name="User">
    <xsd:sequence>
      <xsd:element name="inner" type="Inner" />
    </xsd:sequence>
  </xsd:complexType>
</xsd:schema>

When I run the test in this PR with that code in the included file, the defined types come out as:

struct Inner {
 string el;
}
struct User {
 anyType inner;
}

When I declare the namespaces explicitly - with prefixed namespace declarations and prefixes on the type names - in the included file, the inner member has a type of Test. I'm not 100% clear that this is the correct behaviour either (I think that an included file should always inherit the tns of the parent and ignore explicit declarations in the child?) but it's certainly indicative of a higher-level problem.

Ping @ThomasWeinert thoughts please?

[krakjoe]

ping @DaveRandom can I ask that you look at this again (and in full) and relay an opinion please ?

[iRipVanWinkle]

ping

[krakjoe]

This appears to have gone stale, so closing.