Orphanck: Reject uncovered opaque types #135910
Conversation
fmease
added
S-waiting-on-author
rustbot
added
S-waiting-on-review
| // FIXME(fmease): This is a temporary HACK; rework it. In the next solver, | ||
| // we normalize structurally meaning opaque types "remain behind" weak | ||
| // aliases. I kind of want to expand_weak_alias_tys this but that's | ||
| // pretty hacky, too. |
There was a problem hiding this comment.
The HACK mentioned in the PR description.
There was a problem hiding this comment.
Yeah 🤔 I don't really know what to do here. I feel like ideally we potentially don't need to recompute things here and report an error using only the original uncovered type as this folder is somewhat :/
I am personally fine with slightly regressing diagnostics here if necessary for that, but don't have a clear idea of how bad this would be
fmease
removed
the
S-waiting-on-review
| } | ||
| } else { | ||
| // Regarding *opaque types* specifically, we choose to treat them as non-local, |
There was a problem hiding this comment.
We probably want to bring back some of these nicely written points.
|
☔ The latest upstream changes (presumably #136135) made this pull request unmergeable. Please resolve the merge conflicts. |
| if let Some(local_ty) = self.local_ty { | ||
| diag.arg("local_ty", local_ty); | ||
| } | ||
| } |
There was a problem hiding this comment.
if you've changing this anyways, do you mind changing this back into non-translateable diagnostics? 🤔 it feels somewhat hard to review otherwise cc #132181
Fixes #130978.
TODO: Get rid of a HACK.
TODO: Add new tests.
Feel free to review already though!
r? lcnr