Fix #14240: Fix off-by-1 when emitting Scala.js IR Positions.

[sjrd]

scalac positions are 1-based, while IR positions are 0-based. There is therefore a - 1 in the nsc plugin, that was ported over to dotc without questioning. It turns out that dotc uses 0-based positions as well (as documented in SourceFile), and so we should not make any adaptation there.

---

I'm not sure how to write a test for this.

[sjrd]

@anatoliykmetyuk Hum, why did you assign me to my own PR? Do I need to change anything here?

[anatoliykmetyuk]

@sjrd is this one review-ready or did you intend to also write the test?

[sjrd]

Unless someone has any idea how we can write that programmatically calls the compiler, with the -scalajs flag, extracts the IR that it produces, and checks positions on it ... hum, no, I don't plan to write any test about it.

[smarter]

Could this be done as an sbt scripted test? We have one for scalajs already: https://github.com/lampepfl/dotty/tree/master/sbt-test/scalajs

[sjrd]

It's possible, sure. If you think that's not too much overhead.

[smarter]

The scripted tests run reasonably fast nowadays so I think so, you could even replace the existing "basic" test.

[sjrd]

All right. I updated the PR with a test in the existing scripted test.