Incorrect Level 3 core tests

Hi,

I've been running the latest release of the DOM level 3 core tests (20040405) on our DOM
implementation and there are several tests that I think are incorrect. Below are some comments
pasted from my wrapper code. I hope this is of use.

Regards,
Bas de Bakker
X-Hive Corporation


     // The test creates an attribute with entity reference 'gamma', which according to the DTD has
     // the value γ and expects it to contain 'Texas'.
     "documentadoptnode16",
     // The spec allows returning null if adopting a node is not supported, but this test uses the
     // return value without checking.
     "documentadoptnode24",
     // This test sets a text node in an element that is not declared as having element content to a
     // string of spaces and then expects isElementContentWhitespace to become true. There is no
     // indication whatsoever in the spec that this should happen and it seems wrong, as the element
     // is not declared as having element content.
     "domconfigurationcansetparameter06",
     // These tests call cloneNode on a document, which the spec says is implementation dependent.
     "documentgetinputencoding04",
     "documentgetxmlencoding05",
     // Test attempts to add a second document element.
     "documentnormalizedocument10",
     // Test retrieves an element by tagname "acronym" and expects its nodename to be "address".
     "documentnormalizedocument11",
     // This test assumes that adoptNode() succeeds (does not return null), which is not required.
     "documentrenamenode09",
     // This test assumes that using a qualified name of ::0; throws a NAMESPACE_ERR, while we throw
     // an INVALID_CHARACTER_ERR due to the semicolon as in my reading of the spec.
     "documentrenamenode19",
     // The spec says that typeinfo gives the member type definition if it exists (as it does in the
     // tests), but the tests expect the (non-member) type definition.
     "elementgetschematypeinfo02",
     "elementgetschematypeinfo07",
     // This test seems to assume that setAttributeNS and setIdAttributeNS can be used with "*" as
     // the uri to use any uri, but the spec says no such thing. Actually, the test creates two
     // attributes with the same name but different URIs and then retrieves an attribute by simple
     // name. It assumes it will get one of the attributes, but finds and then tests another.
     "elementsetidattributens05",
     // The test checks that the length of the value of the first 'code' element is 18, while it is
     // only 15.
     "infoset03",
     // Test creates a level 1 node, while namespaces is (by default) true, so normalizeDocument
     // signals an error that the test does not expect.
     "infoset07",
     // Test says that documents with only different encoding should be unequal according to
     // Node.isEqualNode(), but the spec says to ignore unmentioned attributes of which this is one.
     "nodeisequalnode03",
     // These tests have DocumentBuilderSettings that include (first) schemaValidating and (later)
     // validating. LSDocumentBuilderFactory therefore sets the schema-type to XML Schema and then
     // overwrites it with DTD. Consequently, the tests load a document without XML Schema validation
     // and getting the TypeInfo information fails.
     "typeinfogettypename03",
     "typeinfogettypename04",
     "typeinfogettypenamespace01",
     "typeinfogettypenamespace03",
     "typeinfogettypenamespace04",
     // The spec says that the TypeInfo interface gives the namespace and name of the _member_ type
     // definition if it exists. These tests test that it equals the (non-member) type definition.
     "typeinfoisderivedfrom15",
     "typeinfoisderivedfrom16",
     "typeinfoisderivedfrom17",
     "typeinfoisderivedfrom18",
     "typeinfoisderivedfrom19",
     "typeinfoisderivedfrom21",
     "typeinfoisderivedfrom40",
     "typeinfoisderivedfrom41",
     // This test casts a return value of Element.getSchemaTypeInfo (a TypeInfo) to Element.
     "typeinfoisderivedfrom65",

Received on Tuesday, 11 October 2005 14:04:49 UTC