update sbt versions

[larsrh]

Subsumes #304 and #301

[larsrh]

The fact that the tests are failing here is very unsettling 😰

[SethTisue]

Partial investigation:

• the problem is reproducible locally, it's not some CI-only glitch
• it's the sbt version bump, not the sbt-dotty bump
• 2.13 passes, 2.11 and 2.12 fail

[SethTisue]

oh, I found a ticket on this: sbt/sbt#5934

[gabro]

I admit I've never bothered to upgrade Sbt because I didn't want to get into that issue 😅

Wild guess: something related to escaping of output? It seems that the value names are not being printed

[mzuehlke]

Updates didn't worked since sbt version 1.4.0.
see #213

[SethTisue]

something related to escaping of output? It seems that the value names are not being printed

It's not a printing problem. In MacroCompatScala2.scala, in clueImpl, if I println(text) after val text = ..., it's printing empty strings (on 2.11 and 2.12 only). So the Clue values themselves incorrectly contain empty strings.

[gabro]

Wow. The fact that this has something to do with sbt is deeply unnerving 😳

I'll take a look when I have a minute

[larsrh]

It looks like -Yrangepos is not passed to the scalacOptions in test. However, I failed to fix it locally. Maybe a useful hint, though.

[gabro]

@larsrh Nice find, that would explain the differences between 2.11/2.12 and 2.13

[mpilquist]

It's definitely the -Yrangepos setting. On current head, we see it on testsJVM project when running +show Test/scalacOptions. Doing same with this PR shows -Yrangepos missing. Not sure what changed upstream to cause this but we could patch easily by setting Test / scalacOptions := (Compile / scalacOptions).value

[gabro]

I can confirm the workaround proposed by @mpilquist works locally.

[gabro]

Code LGTM, just a minor comment 👍

[gabro]

🙏 thank you @larsrh and everyone who contributed to this distributed bug hunt!