Warn that implicit should have explicit type [ci: last-only]

[som-snytt]

Implicit definitions are required to have an explicit type ascribed in Scala 3. Now Scala 2 warns if the type is missing (and therefore inferred). Under -Xsource:3, the warning becomes an error.

Local implicits are exempt because they are not available before they are defined.

The warnings may be suppressed globally with -Wconf:cat=other-implicit-type:s or locally using @annotation.nowarn("cat=other-implicit-type").

Note that a whitebox macro may be declared of type Any and arbitrarily narrowed by its expansion.

Fixes scala/bug#5265

[som-snytt]

-Werror FTW.

[error] /home/jenkins/workspace/scala-2.13.x-validate-main/src/reflect/scala/reflect/api/Internals.scala:1101:18: Implicit definition should have explicit type
[error]     implicit val token = new CompatToken
[error]                  ^
[error] one error found

[som-snytt]

@dwijnand inviting your thoughts on your example: implicits should have explicit types, but it is willing to infer a type that can't be written explicitly. In this case, the default arg is not allowed in the refinement. (One could introduce a trait, but that would negate the test.) Probably you will say: "Rules are made to be broken."

[dwijnand]

Nice find!

[som-snytt]

Fields are not marked implicit, and can't access the accessor without cycles, so mark it and check when typing the getter.

[som-snytt]

I didn't switch to draft after handling val, and still doesn't handle val in traits. Plus cycle of test breakage.

[som-snytt]

Hung again. I don't see the red ex for canceling jenkins job.

Travis:

[info] + IndexScript.allPackages: OK, proved property.
[info] Elapsed time: 0.432 sec 
failing seed for mutable.TreeMapProjection.get, contains is WGrJJZh_6eL-dRMg_fX5Komy1UboU3eTd2SlGrJ3d9G=
[info] ! mutable.TreeMapProjection.get, contains: Exception raised on property evaluation.
[info] > ARG_0: HashMap( -> 0)
[info] > ARG_1: None
[info] > ARG_2: Some()
[info] > ARG_0_ORIGINAL: HashMap(պ -> -2147483648)
[info] > ARG_1_ORIGINAL: Some()
[info] > Exception: java.lang.NullPointerException: null
[info] scala.collection.mutable.TreeMap$TreeMapProjection.isInsideViewBounds(TreeMap.scala:193)
[info] scala.collection.mutable.TreeMap$TreeMapProjection.contains(TreeMap.scala:211)
[info] scala.collection.mutable.MutableTreeMapProjectionProperties$.$anonfun$new$125(MutableTreeMap.scala:238)
[info] scala.collection.mutable.MutableTreeMapProjectionProperties$.$anonfun$new$125$adapted(MutableTreeMap.scala:237)
[info] scala.collection.IterableOnceOps.forall(IterableOnce.scala:589)
[info] scala.collection.IterableOnceOps.forall$(IterableOnce.scala:586)
[info] scala.collection.AbstractIterable.forall(Iterable.scala:933)
[info] scala.collection.mutable.MutableTreeMapProjectionProperties$.$anonfun$new$124(MutableTreeMap.scala:237)
[info] scala.collection.mutable.MutableTreeMapProjectionProperties$.$anonfun$new$124$adapted(MutableTreeMap.scala:230)

Locally,

[info] + Range inclusiveByOne.foreach.visited.size: OK, passed 100 tests.
[info] Elapsed time: 0.002 sec
#
# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007fad5f9c03fc, pid=30176, tid=30206
#
# JRE version: OpenJDK Runtime Environment (18.0.1.1+2) (build 18.0.1.1+2-6)
# Java VM: OpenJDK 64-Bit Server VM (18.0.1.1+2-6, mixed mode, sharing, tiered, compressed oops, compressed class ptrs, g1 gc, linux-amd64)
# Problematic frame:
# V  [libjvm.so+0xd263fc][info] + Range inclusiveByOne.sum: OK, passed 100 tests.
[info] Elapsed time: 0.002 sec
[info] + Range inclusiveByOne.sumCustomNumeric: OK, passed 100 tests.
[info] Elapsed time: 0.004 sec

Travis says both

export ADOPTOPENJDK=8
$ java -Xmx32m -version
openjdk version "11.0.2" 2019-01-15
OpenJDK Runtime Environment 18.9 (build 11.0.2+9)
OpenJDK 64-Bit Server VM 18.9 (build 11.0.2+9, mixed mode)
$ javac -J-Xmx32m -version
javac 11.0.2

[som-snytt]

I should have @SethTisue about cancel culture.

[lrytz]

LGTM, thank you!

@SethTisue what's your view on making the warning on by default with -Wconf / @nowarn opt-out?

[SethTisue]

Maybe if we introduce a new warning category and publish the opt-out flag in the release notes, it's OK?

Yes, I think that's sufficient.

[SethTisue]

Note that this impacts macro authors too; see scala/bug#12798