From @usev6

Since rakudo commit fe2be65806 there are 5 failing test on rakudo-j in S04-declarations/smiley.t and S12-attributes/smiley.t, each. (I'll fudge them as 'todo').

As it looks said rakudo commit shows a formerly hidden difference between rakudo-m and rakudo-j (or nqp-m and nqp-j). I checked with a Rakudo version 2015.12-60-g37e998d and got exactly the same results for the above evaluations.

$ perl6-j -e 'say { my Int​:U $a }().^name'
?​:?
$ perl6-m -e 'say { my Int​:U $a }().^name'
Int​:U

The reason for this seems to be a different behavior of nqp​::typeparameterized (which is called via src/Perl6/Metamodel/DefiniteHOW.nqp)​:

$ perl6-j -e 'use nqp; my Int​:U $foo; say nqp​::isnull(nqp​::typeparameterized($foo))'
1
$ perl6-m -e 'use nqp; my Int​:U $foo; say nqp​::isnull(nqp​::typeparameterized($foo))'
0

If I call nqp​::typeparameterized with Int​:U directly, JVM and Moar agree​:

$ perl6-j -e 'use nqp; say nqp​::isnull(nqp​::typeparameterized(Int​:U))'
0
perl6-m -e 'use nqp; say nqp​::isnull(nqp​::typeparameterized(Int​:U))'
0

From @usev6

The tests in question are passing again.

I'm closing this ticket as resolved.

@usev6 - Status changed from 'new' to 'resolved'