Mark DefineOpaqueTypes::Yes use-sites uncovered by tests as FIXMEs
#117369
Conversation
rustbot
added
S-waiting-on-review
|
duplicate of #117348 ? 😅 |
There was a problem hiding this comment.
Hm... I actually wonder if we may just want to track these usage sites as issues, rather than FIXMEs in the code.
Ideally, I would like to have interested individuals (or maybe just myself) study the way that the code is being used and create tests from that (or conclude that tests cannot be created and switch them to DefineOpaqueTypes::Yes).
Also, each usage site alone is a bit too granular. Except for candidate assembly and selection in the new trait solver, we probably want to group (e.g.) all of the method probe usages together, and all of the coercion usages together, etc.
| @@ -165,6 +165,7 @@ fn visit_implementation_of_dispatch_from_dyn(tcx: TyCtxt<'_>, impl_did: LocalDef | |||
| use rustc_type_ir::TyKind::*; | |||
| match (source.kind(), target.kind()) { | |||
| (&Ref(r_a, _, mutbl_a), Ref(r_b, _, mutbl_b)) | |||
| // FIXME(DefineOpaqueTypes): no test exercizes using `DefineOpaqueTypes::Yes` below. | |||
There was a problem hiding this comment.
| // FIXME(DefineOpaqueTypes): no test exercizes using `DefineOpaqueTypes::Yes` below. | |
| // FIXME(DefineOpaqueTypes): no test exercises using `DefineOpaqueTypes::Yes` below. |
I saw that after starting and it's more complementary: it flipped uses to Some of the new |
|
And yeah, this work is kind-of underway in #117348. I actually slightly regret the approach suggested #116652. Seems a bit too much like busy work -- creating 36 commits that all just add a single Ideally we'd be working incrementally towards our goal of actually getting rid of |
|
Ok then let's close this. |
I noticed #116652 earlier, and this PR marks all the
DefineOpaqueTypes::Nouse-sites that didn't make any test fail when flipped toDefineOpaqueTypes::Yes(each into their dedicated commit as requested in that issue).r? @compiler-errors