warn about unnecessary uses of .nn

[olhotak]

Issue warnings when .nn is called on a receiver that is already not null or when the expected type of the result of .nn admits null.

cc @noti0na1

[olhotak]

I can't figure out the bootstrapped test failures. It seems to be a Heisenbug.

Locally, if I do scala3-compiler-bootstrapped/testCompilation after rm -rf out; sbt clean, sometimes it fully passes and sometimes every test fails.

[noti0na1]

Otherwise LGTM!

[noti0na1]

I'd like to delete the case tp: TypeProxy, since it will essentially widen a type.
Let's also move it next to isNotNull because of the similarity.

def admitsNull(using Context): Boolean =
  self.isNullType || self.isAny || self.dealias match
    case OrType(l, r) => r.admitsNull || l.admitsNull
    case AndType(l, r) => r.admitsNull && l.admitsNull
    case tp: TypeBounds => tp.lo.admitsNull
    case _: FlexibleType => true
    case _ => false

[olhotak]

Without the TypeProxy case, the following test fails (no warning):

def f6[T >: String|Null](s: String|Null): T = s.nn // warn

Since the test is approximately NullType <:< tp, what's the downside of widening tp if the original narrow tp fails the test. (I can see that widening eagerly before the match would cause us to miss some cases.)

[noti0na1]

Will it be enough if we add a case for TypeRef and recur on its info?

[olhotak]

That does cover this f6 test case, but would miss other cases such as AnnotatedType.

Do you have a particular example in mind where TypeProxy would do the wrong thing?

[noti0na1]

I'm ok with the current rule and can be merged, given a singleton type wouldn't be inferred at type argument places of nn.