Imported From: https://issues.scala-lang.org/browse/SI-9763?orig=1
Reporter: @etorreborre
See #7411

@som-snytt said:
Related:

milessabin/shapeless@312b405

in case it didn't get an issue.

@milessabin said:
It didn't get an issue because it wasn't clear at the time whether this was a shapeless bug, a compiler bug or "in a macro, so undefined behaviour". Given that we can tickle the problem without either shapeless or macros involved blaming the compiler seems a lot more reasonable.

@paulp said:
This is #7411.

@som-snytt said:
Slight progressive mitigation:

$ scala -Xlint
Welcome to Scala 2.11.8 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_60).
Type in expressions for evaluation. Or try :help.

scala> import annotation._
import annotation._

scala> @implicitNotFound("No ${A}" + " at all") class C[A]
<console>:14: warning: possible missing interpolator: detected an interpolated expression
       @implicitNotFound("No ${A}" + " at all") class C[A]
                         ^
defined class C

scala> :quit
$ scalam -Xlint
Welcome to Scala 2.12.0-M4 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_60).
Type in expressions for evaluation. Or try :help.

scala> import annotation._
import annotation._

scala> @implicitNotFound("No ${A}" + " at all") class C[A]
defined class C

scala> :pa
// Entering paste mode (ctrl-D to finish)

object A
@implicitNotFound("No ${A}" + " at all") class C[A]

// Exiting paste mode, now interpreting.

<console>:15: warning: possible missing interpolator: detected an interpolated expression
       @implicitNotFound("No ${A}" + " at all") class C[A]
                         ^
defined object A
defined class C

It still warns if there is a value with the name of your type param.

@milessabin said:
@SethTisue While this is related to #7411 I don't think it's a duplicate. That ticket is calling for a general cleanup in the area, but isn't reporting a bug as such. This ticket on the other hand is reporting a very specific bug.

@som-snytt said:
I was going to comment similarly: current state of the art tries to avoid the undue warning. Apologies to the word "art".

It occurs to me that one fix for the other ticket is, if the "expected type" is an interpolated string (there is no such notion now), then apply an appropriate interpolator. A macro interpolator should be allowed to guide interpretation of holes (such as type args) (no such notion either).

@milessabin said:
Seeing as it's not just me I'll reopen and unduplicate. @SethTisue reclose if you disagree.

@etorreborre said:
Thanks Miles, I agree too.

@SethTisue said:
just testing y'all ;-)

@paulp said:
I guess I meant "this is SI-7411" in the sort of grandiose this-bug-wouldn't-even-be-possible-if-we-did-that way you've come to expect.

Closing with the caveat there are two issues with the annotation:

1. the lint is limited in detecting that it is in an ImplicitNotFound (what paulp said)
2. the other lint for checking well-formedness is limited by ability to fetch the msg arg, see AnnotationInfos.

The interpolation here is legal and resonable but warns for reason 2:

// Missing argument `msg` on implicitNotFound annotation.
class C {
  final val A = "thing"
  @implicitNotFound(s"not found ${A}") trait T[B]
}
class D {
  final val A = "thing"
  @implicitNotFound("not found " + A) trait T[B]
}