Context bound behaves differently from a using-clause

[dkz]

Compiler version

3.3.1, 3.4.1

Minimized code

trait F[A]
given F[Int] = new F {}
def f[A: F] = { (x: A) => x }
println(f[Int](1))

Output

  |f[Int](1)
  |       ^
  |       Found:    (1 : Int)
  |       Required: F[Int]
  |
  | longer explanation available when compiling with `-explain`

Expectation

Scala 3 Book suggests that context bound de-sugaring mirrors the following code:

def f[A](using F[A]) = { (x: A) => x }

Yet, the latter doesn't lead to a compile error in f[Int](1).

[som-snytt]

It compiles under -source:future. (I'm not following it closely enough.)

A footnote is that 3.4 also warns and offers a rewrite, but I don't think the rewrite works while there is an error, so that is an interesting conundrum.

This code can be rewritten automatically under -rewrite -source 3.4-migration.

[reid-spencer]

A footnote is that 3.4 also warns and offers a rewrite, but I don't think the rewrite works while there is an error, so that is an interesting conundrum.

There's a related issue with context bounds that I reported a while back; where a warning generates a notice of an available automatic -rewrite that appears to not be implemented (i.e. specifying -rewrite -source 3.4-migration does not actually get the warned-about code rewritten).

#19872

[Gedochao]

It compiles under -source:future. (I'm not following it closely enough.)

This is very weird. And indeed, true.

[som-snytt]

This was fixed in 3.6 (last summer, released at end of year), probably just the "release the hounds" PR to "switch mapping", i.e., flags. That is why just the application was confused.