I don't think these tests are wrong, according to B.2 Operator Mapping, if the operands are of the same type (where type is xs:integer, xs:decimal, xs:float, xs:double, in that order), the result type after the calculation is applied must also be of that same type.

For this test, I believe the only possible outcome is "integer decimal float double double". What other outcome do you suggest is also valid and why?

Since integer is a subtype of decimal, every integer has the type decimal. So everywhere where a decimal occurs, you can just put an integer.

(It could be used as optimization if there is a faster evaluation path for integers)

> Since integer is a subtype of decimal, every integer has the type decimal
I agree, but I'm not sure how that influences the result in this test, I'm afraid I'm missing the point here.

Abel, the point is that if $X is a decimal, then we require that -$X and +$X return a decimal, but this does not preclude them returning an integer, because an integer is a decimal and therefore satisfies the requirement. But the test is written to fail if the returned value is an integer.

I have committed a change to the test to use xs:decimal(2.5) rather than xs:decimal(2) so that the question does not arise. Leaving the bug open in case others disagree.

I am marking this bug as resolved as we have reached our resolution