Implicits are resolved with empty parameter lists

[hamzaremmal]

Compiler version

Since 3.4.2

Minimized code

implicit val x: Int = 0
def foo(implicit x: Int): Int = x
foo()

Output

Complies

Expectation

Behaves as Scala < 3.4.2 and as in the specification.

Note

This is the error message that was generated before 3.4.2

-- [E171] Type Error: ----------------------------------------------------------
1 |foo()
  |^^^^^
  |missing argument for parameter x of method foo: (implicit x: Int): Int
1 error found

and in scala 2.13.14

scala> foo()
          ^
       error: not enough arguments for method foo: (implicit x: Int): Int.
       Unspecified value parameter x.

[Gedochao]

Last good stable release: 3.4.1
Last good nightly release: 3.4.2-RC1-bin-20240222-98efdab-NIGHTLY
First bad nightly release: 3.4.2-RC1-bin-20240226-e0cb1e7-NIGHTLY
First bad stable release: 3.4.2

e0cb1e7 is the first bad commit
cc @SethTisue

[Gedochao]

Looks similar to:

• Possibly incorrect suppression of parameter list #22380

Potentially also relevant:

• Overloading and implicit argument result in inconsistent outcomes #22421
• Overloading or implicit bug in 3.4.1 #20053
• Overloading with context function bug #15163

[sjrd]

e0cb1e7 is the first bad commit
cc @SethTisue

That ... does not look likely. 😅 Is it possible that bisect ended up in a "local minimum" of the build not even functioning at all, rather than the actual typechecking regression?

[som-snytt]

Seems to be 1e89016

A quarter-line tweak in #19737

[Gedochao]

Is it possible that bisect ended up in a "local minimum" of the build not even functioning at all, rather than the actual typechecking regression?

Yeah... that's likely what happened 😅 I need to think about tweaking the bisect script for cases such as this...

Anyway, as @som-snytt suggested, the correct commit seems to be 1e89016
cc @EugeneFlesselle

[SethTisue]

whew

[som-snytt]

I took a look. It turns on that an implicit parameter list is supplied only when it is missing, whereas with "use-site using" or explicit using you can supply just some args. (In matchArgs, it already has an arg list.) (I remember that subtlety came up a few years ago in documentation for scala 2.)

All this time, I thought I knew how to call a method.

I'll submit a PR when I see where to toss the monkey wrench.

[som-snytt]

Removing myself as assignee because the PR requires review. There's nothing on the "board" that indicates "requires review" or who is available.

[Gedochao]

Removing myself as assignee because the PR requires review.

@som-snytt Hey, would you mind actually staying assigned in this case? Your PR contains a solution and it's awaiting review, so in so many words, you currently "own" the solution. As in, there is no reason for anybody else to jump into this for the time being.
Asking, since it makes it easier for me to track things 😅

[som-snytt]

@Gedochao Sure, I just don't know how the process works.