Remove differences in diagnostic printing in the REPL

[dwijnand]

No description provided.

[dwijnand]

Who's the closest to being the REPL maintainer?

[SethTisue]

can you explain the build.sbt changes a little?

[dwijnand]

can you explain the build.sbt changes a little?

+  val repl = taskKey[Unit]("spawns a repl with the correct classpath") // (1)
..
-    // Spawns a repl with the correct classpath
-    addCommandAlias("repl", "scala3-compiler-bootstrapped/console") // (2)
..
+    repl := (Compile / console).value,        // (3)
+    Compile / console / scalacOptions := Nil, // (4)
..
+        repl := (`scala3-compiler-bootstrapped` / repl).value, // (5)

Sure.

(1), (2), and (3): I was thinking (2) did the double-compilation of bootstrapping and then building a bootstrapped compiler for the REPL, so I was looking to copy how scala and scalac are set up, as tasks, so you can run either scala3-compiler/scalac or scala3-compiler-bootstrapped/scalac. But scala3-compiler/repl didn't work as desired, because it's based on the project's scalaVersion. So I just left the converting to a task: (1) and (3).

(5): But then the root project doesn't aggregate scala3-compiler-bootstrapped, so I need to manually set up that delegation.

(4): Finally, I don't want the scalacOptions used in compiling the (bootstrapped) compiler from tainting the out-the-box experience of using the REPL, e.g. by enabling -deprecated and -feature, so I reset the scalacOptions.

[dwijnand]

Rebased because this conflicted with #13000. Notice from the output diff how this PR removes any remaining differences.

[SethTisue]

The user-facing changes look very good to me. This is a nice improvement in usability.

I'm satisfied by the build.sbt explanation (to the extent that I understand the dotty build, anyway).

And, Dale walked me through the implementation just now and I see nothing to object to. A few more comments might help future maintainers.

@bjornregnell any thoughts?

[som-snytt]

Footnote that Scala 2 REPL enables -deprecation -feature unless those flags are explicitly set. (Actually in MainGenericRunner, the test is now that Wconf and Xlint are not set.) You only see extended warning once:

scala
Welcome to Scala 2.13.6 (OpenJDK 64-Bit Server VM, Java 17-ea).
Type in expressions for evaluation. Or try :help.

scala> implicit def f(s: String) = s.toDouble
                    ^
       warning: implicit conversion method f should be enabled
       by making the implicit value scala.language.implicitConversions visible.
       This can be achieved by adding the import clause 'import scala.language.implicitConversions'
       or by setting the compiler option -language:implicitConversions.
       See the Scaladoc for value scala.language.implicitConversions for a discussion
       why the feature should be explicitly enabled.
def f(s: String): Double

scala> implicit def f(s: String) = s.toDouble
                    ^
       warning: implicit conversion method f should be enabled
       by making the implicit value scala.language.implicitConversions visible.
def f(s: String): Double

scala>

[SethTisue]

@som-snytt True, but regardless, the sbt task should have the same defaults as what users get. I would be in favor of a separate PR that added -deprecation -feature in both contexts.

[nicolasstucki]

@dwijnand the CI failed on the merge commit if this PR. It then failed on nightly.

[dwijnand]

Does GitHub Actions run on the branch then, rather than a merge commit? I'll fix up master asap.