Context for constructor error is outer context

[som-snytt]

Fixes #7709

[dottybot]

Hello, and thank you for opening this PR! 🎉

All contributors have signed the CLA, thank you! ❤️

Have an awesome day! ☀️

[liufengyun]

The code does not handle secondary constructors. Will something like the following work?

Suggested change

	val encl = if !ctx.owner.isConstructor then ctx else ctx.outersIterator.dropWhile(_.owner.isConstructor).next()
	val encl = if ctx.owner.isPrimaryConstructor then ctx.owner.enclosingClass.owner.enclosingClass else ctx

[som-snytt]

The expressions are the same, as the next outer context which is not owned by a ctor will be the correct enclosing class; also, both tests will discover an enclosing package symbol. I prefer your suggestion, which avoids an iterator -- although I almost talked myself out of it: the average view is that we seem to be skipping our class for the enclosing class; though from a context perspective, we're really just pushing out to the true enclosing class. (The context of expressions to super constructors is always confusing.) However, it seems isConstructor works here, also for secondaries; outersIterator would just plow through a couple more contexts in that case.

This change is a smallish tweak, but a good opportunity to take a look at scala 3 contexts...

[liufengyun]

Sorry, I wasn't clear. I meant that in a secondary constructor, we might not want to go to the outer class --- as the current class itself could extend the related class.

[som-snytt]

It's obvious when you put it that way.

[som-snytt]

Dotty CI claims the test output does not match the check file, but it looks the same to me. At this point, I wish the test rig would just display the diff (as partest learned to do, eventually).

Edit: maybe the way I naively changed the first line of the interpolation caused the VBARs to drop out? But it works locally...

[smarter]

Running testCompilation --update-checkfiles tests/neg/i7709.scala should fix the checkfile.

[som-snytt]

Thanks @smarter I've stuck up a post-it note.

The test passes locally. When I munge the test, the "actual output" in sbt doesn't show the | on the LHS of the message output, although it is present as usual in the .check.out. I did verify my sha is pushed.

Soon I'll push an update and see what happens! I can't wait!

[som-snytt]

Just --update-checkfiles and squash. Previous feedback had been addressed.

[liufengyun]

LGTM 🎉

[som-snytt]

I came back for --update-checkfiles because no autocomplete and partest --update-check.