Tests with non-BMP characters in regexes

[nwellnhof]

There are some MS tests with non-BMP characters in regexes which are marked as invalid, although they don't seem to be.

Previous discussion: https://www.w3.org/Bugs/Public/show_bug.cgi?id=4113

[michaelhkay]

This one doesn't appear to have been resolved by the WG.

There's a problem with tests like this in some cases, where Unicode has changed the character classes of individual characters.

The citation to the Unicode Database in the XSD spec says "The edition cited is the one current at the date of publication of XSD 1.0. Implementations may follow the edition cited and/or any later edition(s); it is implementation-defined which."

It's hard to reflect that decision in the test suite without making all tests that rely on Unicode character classes implementation-defined. As an implementor, I think you need to keep your own record of tests where you produce a different result and believe it to be correct for legitimate reasons.

In case it's of any interest, I attach copies of the current exceptions lists for Saxon 13 against XSD 1.0 and XSD 1.1

xsd.xml
xsd11.xml

[nwellnhof]

There already are version tokens Unicode_4.0.0 and Unicode_6.0.0 used in the test suite and the latter should also work with the latest Unicode standard. I don't see a problem here.

Regarding the tests, it seems like the MS implementation simply couldn't work with non-BMP characters in regexes. It would be nice if this could be addressed.