Some current inconsistency.

scala> 42.ceilval res0: Float = 42.0scala> def i = 42def i: Intscala> i.ceil       ^       warning: Widening conversion from Int to Float is deprecated because it loses precision. Write `.toFloat` instead.val res1: Float = 42.0scala> def c = 'c'def c: Charscala> c.ceilval res2: Float = 99.0

wherec.ceil is pronounced "Cécile" and is even less likely to be intended.

With the new warning, the current warning is less correct.

It warns in two spots that specializeArray[Unit].

The other thought is that (as with the current warning), any promotion is dubious, it doesn't matter which member is selected.

Then just needdefinitions.RichFloatClass.

The check is only at typer, which gives a pass to

case Literal(Constant(())) =>

Not sure why it was deemed bad to allowcase _: () => as the unit value ought to be a singleton. (It is notUnit with Singleton.) I guess it doesn't matter because the test is always boxed unit equality. There's no advantage tocase _: Unit.

Something became().asInstanceOf[Unit] in fields, a private lazy val after init. (That was thetest execution will be non-parallel inPartestDefaults.)

Rebased, and enhancedpartest --realeasy to skip tests that run on today's JDK.

-- 1 - run/blank.scala                           [skipped on Java 19, compiling against JDK8 but must run on 11+]

It would be ideal if--realeasy did not skip in case of explicitscalac: --release:8 but that was too subtle for me.

The option also addsrun/t6240-universe-code-gen.scala which is annoying to forget to check. (I had actually fixed it in the wrong commit.)

Selected 1810 tests drawn from 1 named test categories, 2 tests modified on this branch, and other tests you might have forgotten

In case of selection failure,

sbt:root> partest junkDiscarding 1 invalid test pathsSelected 0 tests drawn from the well of despair. I see you're not in a testing mood.[info] Passed: Total 0, Failed 0, Errors 0, Passed 0