I think the "magic" was just the general murk around implicit scopes. I'd move it to the companion and see what breaks.

There are examples of the wild of people referring to these things with the Predef qualifier.

In this case, unimporting Predef.conforms is done at the suggestion of a Stack Overflow answer.

https://github.com/joune/dependencymanager-scala/blob/master/org.apache.felix.dependencymanager.scala.itest/src/main/scala/org.apache.felix.dm.scala.test/Activator.scala#L3-L6

I've opened a separate PR on top of this PR where we can discuss the fate of conforms: #7369

There are examples of the wild of people referring to these things with the Predef qualifier

I'm okay with that, on the balance. Anyone opposed to just merging this...?

It seems not :-)

Shouldn't there be type aliases in Predef?

Edit: Ok I see they are in the scala package. The problem is with macros, where it's common to specify the full path from _root_ in quasi-quotes to avoid hygiene problems.

Example: https://github.com/mpilquist/simulacrum/blob/662aa50f0cbeb30bc651f1998a20863dc5d349d4/core/src/main/scala/simulacrum/typeclass.scala#L213

why? (and would that even work?)

why? (and would that even work?)

Sorry, see my edit:

Edit: Ok I see they are in the scala package. The problem is with macros, where it's common to specify the full path from root in quasi-quotes to avoid hygiene problems.

That's the change which broke simulacrum. I don't know how common it is to refer <:< and =:= in macros but it's not that far fetched. In most cases it should be possible to splice directly the type symbol:

val subType = typeOf[Any <:< Any].typeConstructor
q"val x: $A $subType $B = ..."

I'm just not sure it always works. E.g. how about macro annotations (like simulacrum)?

simulacrum is green now in the 2.13 community build thanks to your fix 👍