Allow Tuple.Head and Tuple.Tail to work with EmptyTuple

[Decel]

Fixes #17186.

I think this is correct behaviour since we already allow things like

val a = List(1).tail
val b = Tuple1(1).tail

[dwijnand]

head and tail are partial functions, so I think it makes more sense as they stand, to have them as "partial match types". I'd say this change is akin to making Nil.tail == Nil, which I think is surprising behaviour.

[Decel]

head and tail are partial functions, so I think it makes more sense as they stand, to have them as "partial match types". I'd say this change is akin to making Nil.tail == Nil, which I think is surprising behaviour.

I definitely agree, but that seems like the only way to make it work (and I think it's important to make it work since we want type-level programming with tuples), without changing the logic of match types. And I think trying to change it to fit the logic here is a very bad idea, -- There are, basically, no cases, where we should allow type parameters to violate type bounds, except here.

But I'd offer the following compromise, -- we shouldn't return EmptyTuple in case of EmptyTuple.tail, but we should allow Tuple.Tail to take EmptyTuple. It would solve the issue and preserve errors.

[dwijnand]

OK, makes sense. So we're making it more partial, rather than less - which is fine. And it allows for irreducible Tuple.Tail/Head types to nest, without running into type parameter bound checking issues.

Now the problem is when can we change this, in terms of tasty compatibility? I assume next minor version.

[sjrd]

Now the problem is when can we change this, in terms of tasty compatibility? I assume next minor version.

Yes that's a minor version change. It is backward compatible, but not forward compatible.

[nicolasstucki]

I also rebased to solve some conflicts with the blacklist.

[dwijnand]

Your rebase, somehow, threw away my comments in 4983ebb