This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
In my opinion test "s3_4_2_4ii08s" and "s3_4_2_4ii10s" of set "defaultAttributesApply" should be invalid (or their second ts:schemaDocument element should be removed). Both of them have two ts:schemaDocument elements. The schema referenced by the second one is overridden by an xs:override in the first one. So listing both of them as ts:schemaDocument in the ts:schemaTest element causes duplicate components in the resulting schema (the overridden component clashes with the original "imported" by the ts:schemaDocument element). See the last note in http://www.w3.org/TR/xmlschema11-1/#src-override: "Note: It is a consequence of the semantics of inclusion, as defined in Inclusion Constraints and Semantics (§4.2.3) (in particular clause 3.1.2 and clause 3.2.2); redefinition, as defined in Including modified component definitions (<redefine>) (§4.2.4); import, as defined in References to schema components across namespaces (<import>) (§4.2.6); and overriding, as defined in this section, that if the same schema document is both (a) included, imported, or redefined, and (b) non-vacuously overridden, or if the same schema document overridden twice in different ways, then the resulting schema will have duplicate and conflicting versions of some components and will not be conforming, just as if two different schema documents had been included, with different declarations for the same ·named· components." Best regards, Andreas Meissl
The WG recognizes this problem with the test set metadata (and has done so for a while); this bug report may spur us to make the effort to resolve the situation. For what it's worth, this is what the Saxon test driver does: It loads the first schema document listed. Then for each subsequent schema document in turn, it reads the targetNamespace of the schema document, and then loads the schema in this schema document if and only if the targetNamespace is "new", that is, if the loaded schema does not already contain components for this namespace. This is a very pragmatic approach, and not justified by anything in the test specs, but it works for me.