CI test run fails but CI passes and missed a compiler regression

[soronpo]

The following issue code already exists in the test CI, but it does not cause the CI to fail.
For example:

2022-07-20T11:31:07.9036071Z [info] Test dotty.tools.dotc.FromTastyTests.runTestFromTasty started
2022-07-20T11:33:09.2040111Z No tests available under target - erroneous test?                               
2022-07-20T11:33:09.8939791Z [info] Test dotty.tools.dotc.FromTastyTests.posTestFromTasty started
2022-07-20T11:35:11.2843408Z No tests available under target - erroneous test? 

As a result, we missed a regression.
See tests/run/backwardsCompat-implicitParens/A_1_c3.0.2.scala

Related to the work done in #14840 and followed by #15679 (the last probably introduced the regression)

Compiler version

v3.2.0-RC4

Minimized code

class Bar(using x: Int)(y: String):
  override def toString = "Bar"
object Bar:
  given Int = 1
  inline def foo =
    println(new Bar()(""))
    println(Bar()(""))

Output

missing argument for parameter y of constructor Bar in class Bar: (y: String): Bar

Expectation

No error, and the CI should be fixed that this error reflects in the underlying CI fail/success indication.

[odersky]

I am not sure about the CI. I think that one file in the test should be run with version 3.0.2, that's why it has this funny name. Anyway, under 3.2 this code does not compile (nor should it).

[soronpo]

I am not sure about the CI. I think that one file in the test should be run with version 3.0.2, that's why it has this funny name. Anyway, under 3.2 this code does not compile (nor should it).

Oh, I thought it should, and I was not aware of that file name convention. Now it makes sense.

[odersky]

I think this is as should be. Maybe add a mock using to tests/run/backwardsCompat-implicitParens/A_1_c3.0.2.scala so that it's clear it will be run under 3.0.2?