[XERCESC-1659] Order sensitivity in schemaLocation and noNamespaceSchemaLocation

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.1.0
• Fix Version/s: 3.1.0
• Component/s: Validating Parser (XML Schema)
• Environment: all
• Assignee: Boris Kolpackov
• Reporter: Boris Kolpackov
• Created: 27/Dec/06 09:21
• Updated: 19/Nov/09 12:12
• Resolved: 19/Nov/09 12:12

Description

I am attaching two test cases (each consists of 3 schemas plus an XML instance). If you try to run domprint on the first test case, you will get the following error:

$ domprint -v=always -n -s -f test-users.xml

Error at file "test-users.xml", line 6, column 78
Message: Unknown element 'b:UserDatabase'

If you change the order of the schemaLocation and noNamespaceSchemaLocation attributes in test-users.xml then the error disappears.

The second test case is a slight modification of the first test case with the only difference being the schemas with targetNamespace are now do not have a namespace, and the schema that used to be without a namespace (derived-user-config.xsd) now is in a namespace. If you run domprint on this test case, you will get the following error:

$ domprint -v=always -n -s -f test-users.xml
Error at file "test-users.xml", line 6, column 55
Message: Unknown element 'UserDatabase'

This seems to prove that for Xerces-C++, for some reason, it is important that the schema that declares the root element is mentioned in the first *Location attribute (nor matter whether schemaLocation or noNamespaceSchemaLocation). Now comes the surprise: if we reverse the order of the two attributes in the second test case, domprint terminates with segmentation fault. Examination of the core points to the IGXMLScanner.cpp, line 2288:

elemDecl = fGrammar->getElemDecl(
uriId, nameRawBuf, qnameRawBuf, currentScope
);