Fixes scala/bug#12851

Does it? It looks to me like you're avoiding the crash, but the spurious warning is still emitted.

(Which is an appropriate ambition level for 2.13.12, but I want to double check our understanding.)

Oh I didn't read carefully, I missed that there's a spurious warning. I'll take a quick look.

It's not spurious!

trait T1 {
  def f: Int
}
trait T2 {
  def f() = 42
}
class C extends T1 with T2

The warning is at C.

Okay, I see that you've now amended the PR to include two different test cases, t12851 (in which T2 extends T1, as in the case in the ticket) and t12851b (in which T2 does not extend T1, but C mixes in both traits).

t12851b makes sense to me.

t12851 still puzzles me. Shouldn't the argument be issued when T2 is compiled, rather than when C is compiled?

Added a commit that removes the check file for the test that was moved to pos.

If I had accidentally moved the check to pos, it would have told me!

if (checkFile.exists) genFail("unexpected check file for pos test (use -Werror with neg test to verify warnings)")

Still needs the improvement for valdefs. That came up on an odd ticket so I keep forgetting about it.

Only squashed commits someone may not have yet rebased upon.

@som-snytt Thanks for doing such a thorough job addressing the underlying issue!

@lrytz Given that we're already in the RC phase (and we don't want to slip the release date by resetting the testing clock unless truly necessary), do you think this is sufficiently safe to merge for 2.13.12, or would it be better to merge #10509 for 2.13.12 and hold this one for 2.13.13...?

I approved the other PR, as it is of notably smaller scope. It also now comes with a test! which I'm afraid has bloated the PR, sorry!

Re-submitted in #10520

Except for the namePos change, which is in https://github.com/lrytz/scala/compare/pr10512...lrytz:scala:pr10512a?expand=1

@lrytz will you handle further uptake of the namePos or whatever? That is, you don't need me to rework something?

Yeah, I'll re-submit the namePos change (for 2.13.13) as a new PR once #10520 is merged.