odersky

Yes I agree. If we fix that then we can also drop the exclude for i15922.scala. It would be great if someone could take a closer look here.

There are several issues at play here; For explicitly given types in `TypeApply`s we don't convert cap to fresh if the type in question is boxed. here's the relevant code...

The other issue is that we now create a separate fresh each time we see a `^` or `=>`. We should be able to unify these, but it seems there's...

> I think the freshness level is coming into play here. One thing that seems a bit interesting to me is that for scoped capabilities, the freshness level seems to...

> The other issue is that we now create a separate fresh each time we see a ^ or =>. We should be able to unify these, but it seems...

I think that's worth a try, yes. It would mean we also replace caps in argument position, which normally would be replaced later. But maybe that's OK.

Ah yes, since simple argument caps are not result caps, so they won't be replaced.

I tried this version of the old `subsumes` and compared its outcome with the new `subsumes` for all the tests in `testCompilation`. ```scala final def subsumesOld(y: CaptureRef)(using Context): Boolean =...

I would have expected more changes in the recheckSelect rule. Indeed, recheckSelect seems to not make use of path types. Here is a simple test program: ```scala class IO class...

Also, I tink in light of path types, there's scope for a simplification of `recheckSelect`. A lot of the contortions there were made to make up for the lack of...