FunctionCall-049

[michaelhkay]

I believe this test fails under the current 4.0 coercion rules.

We have clarified the rules: see the note in §3.4.3 starting "Function coercion applies even if the supplied function matches the required type.".

We are supplying a function of type

function(element(*, lu:approximateDate)) as xs:integer
where the expected type is

function(element(*, lu:restrictedUnion)) as xs:integer

and the schema says that restrictedUnion is a subtype of approximateDate

We are calling the function with an element of type element(*, approximateDate)

The new rules say that the function is coerced to the required type even though it already matches the required type, which means that the argument to the function must be an instance of element(*, restrictedUnion) - which this is not.

I believe that the spec change is merely a clarification; a careful reading of 3.1 gives the same interpretation. So the test is wrong for 3.1 as well.