`-Yexplicit-nulls` breaks macros expecting `Type` context bound

[WojciechMazur]

Based on OpenCB failure in https://github.com/tabdulradi/nullable - OpenCB logs

Compiler version

Last good release: 3.8.0-RC1-bin-20250822-658c8bd-NIGHTLY
First bad release: 3.8.0-RC1-bin-20250823-712d5bc-NIGHTLY

Bisect pointed to 712d5bc but it does not make sense - this change was not touching the compiler.

Minimized code

//> using options -Yexplicit-nulls
import scala.quoted.*

private def typeNameMacro[A: Type](using Quotes) = Expr(Type.show[A])
transparent inline def typeName[A]: String = ${typeNameMacro[A]}

Output

-- [E007] Type Mismatch Error: /Users/wmazur/projects/scala/community-build3/example.scala:4:47 ---------------------------------------------------------------------------------------------
4 |transparent inline def typeName[A]: String = ${typeNameMacro[A]}
  |                                               ^^^^^^^^^^^^^^^^
  |                                               Found:    (scala.quoted.Type[A], scala.quoted.Quotes) ?=> (scala.quoted.Expr[(String)?])?
  |                                               Required: (((scala.quoted.Quotes)?) ?=> (scala.quoted.Expr[(String)?])?)?
  |
  | longer explanation available when compiling with `-explain`

Expectation

Should compile

[noti0na1]

Looks like related to #23386? But I couldn't reproduce the error locally on the latest main brach

cc @olhotak @HarrisL2

[Gedochao]

But I couldn't reproduce the error locally on the latest main brach

@noti0na1 I reproduced with the latest nightly (3.8.0-RC1-bin-20250901-ca400bd-NIGHTLY).

Note that to use the Scala 3 nightlies with Scala CLI, you temporarily must use the Scala CLI nightly (brewing a stable version as we speak).

 scala-cli --cli-version nightly compile repro.scala --server=false -S 3.nightly                             
# Downloading Scala 3.8.0-RC1-bin-20250901-ca400bd-NIGHTLY compiler
# -- [E007] Type Mismatch Error: /Users/pchabelski/IdeaProjects/scala-cli-tests-3/scala-cli-tests-3/compiler-repro/repro.scala:5:47 
# 5 |transparent inline def typeName[A]: String = ${typeNameMacro[A]}
#   |                                               ^^^^^^^^^^^^^^^^
#   |Found:    (scala.quoted.Type[A], scala.quoted.Quotes) ?=> (scala.quoted.Expr[(String)?])?
#   |Required: (((scala.quoted.Quotes)?) ?=> (scala.quoted.Expr[(String)?])?)?
#   |
#   | longer explanation available when compiling with `-explain`
# 1 error found
# Compilation failed

[Gedochao]

@jchyb any clue about this one? 🤔

[Gedochao]

Note: this is indeed a regression, but as it's also tied with -Yexplicit-nulls (which is a private option and has no real guarantees), so it's not a top priority (although it'd be great if we figured out what's wrong, still).

[jchyb]

So far I am also unable to reproduce this :( . It compiles fine when running scalac through the compiler repo and when publishing locally and using the new version in scala-cli. When running scala-cli --cli-version nightly compile repro.scala --server=false -S 3.nightly it uses an older version (3.8.0-RC1-bin-20250814-0cf7a18-NIGHTLY). When passing a newer version explicitly (.8.0-RC1-bin-20250823-712d5bc-NIGHTLY ), it claims it cannot find it.

EDIT: nvm, looks like I did have an older version of scala-cli, as it didn't upgrade successfully before and I didn't notice. I'll should be able to reproduce this now.

[jchyb]

I checked contents of Expr.tasty from locally compiled out/bootstrap/scala3-library-bootstrapped/scala-3.8.0-RC1-bin-SNAPSHOT-nonbootstrapped/classes/scala/quoted/Expr.tasty and compared that to the one published with nightlies in https/repo.scala-lang.org/artifactory/maven-nightlies/org/scala-lang/scala-library/3.8.0-RC1-bin-20250901-ca400bd-NIGHTLY/scala-library-3.8.0-RC1-bin-20250901-ca400bd-NIGHTLY.jar.
The locally published one had EXPLICITNULLSattr attribute set, and the one from Nightly did not - so it seems like we have different options set for publishing nightlies and we do not set -Yexplicit-nulls there (I don't know if that is on purpose).

I feel like that explains what is going on, as it seems that after #23386 we insert flexible types into types returned from symbols from files compiled without -Yexplicit-nulls, which is what we see happen in the error output. If we could switch to publishing nightlies compiled with that -Yexplicit-nulls option set, that would fix the issue I think

[noti0na1]

That's weird, the scala-library should not have EXPLICITNULLSattr in any case at this moment. We should investigate the flag and fix the error as soon as possible.