Don't report numeric inference ambiguity when we have previous errors #95751
Conversation
rustbot
added
the
T-compiler
|
r? @jackh726 (rust-highfive has picked a reviewer for you, use r? to override) |
rust-highfive
added
the
S-waiting-on-review
| @@ -79,35 +79,7 @@ LL | trait bar { fn dup(&self) -> Self; fn blah<X>(&self); } | |||
| = note: required because of the requirements on the impl of `CoerceUnsized<Box<dyn bar>>` for `Box<{integer}>` | |||
| = note: required by cast to type `Box<dyn bar>` | |||
|
|
|||
| error[E0283]: type annotations needed | |||
There was a problem hiding this comment.
hehe, wonderful to see this issue getting better too
|
Not sure this is the best fix. It would be better, I think to fallback even under error. |
|
@jackh726, like in general? that would have a bunch of unintended side-effects i think -- the choice to not fallback when infcx is tainted by errors is intentional apparently: https://doc.rust-lang.org/stable/nightly-rustc/src/rustc_typeck/check/fallback.rs.html#92 |
|
If we're falling back to |
So when we |
|
I'll investigate modifying the fallback logic to unconditionally apply float and integer fallback even when infcx is tainted, but i'm assuming that'll have additional diagnostic fallout that may need discussion on its merit, or extra patching-up to suppress, so i will keep that separate from this PR. |
| && crate_names.len() == 1 | ||
| && ["`core`", "`alloc`", "`std`"].contains(&crate_names[0].as_str()) | ||
| && spans.len() == 0 | ||
| && (crate_names.len() == 1 |
There was a problem hiding this comment.
Could this check just be simplified to "if the self type is a numeric infer"?
There was a problem hiding this comment.
So we need to check for int/float variables not just in self, but in the trait's substs as well, so we don't report the things that implement ConcreteType: Trait<{integer}>.
Also removing the core/alloc/std crate name check leads to some regressions in errors reported with simd stuff and when we're unable to infer const variables due to other errors :/
|
@bors r+ rollup |
|
📌 Commit 39bff4b has been approved by |
bors
added
S-waiting-on-bors
Don't report numeric inference ambiguity when we have previous errors Fixes rust-lang#95648
Don't report numeric inference ambiguity when we have previous errors Fixes rust-lang#95648
Don't report numeric inference ambiguity when we have previous errors Fixes rust-lang#95648
Don't report numeric inference ambiguity when we have previous errors Fixes rust-lang#95648
Don't report numeric inference ambiguity when we have previous errors Fixes rust-lang#95648
Rollup of 8 pull requests Successful merges: - rust-lang#90066 (Add new ThinBox type for 1 stack pointer wide heap allocated trait objects) - rust-lang#95374 (assert_uninit_valid: ensure we detect at least arrays of uninhabited types) - rust-lang#95599 (Strict provenance lints) - rust-lang#95751 (Don't report numeric inference ambiguity when we have previous errors) - rust-lang#95764 ([macro_metavar_expr] Add tests to ensure the feature requirement) - rust-lang#95787 (reword panic vs result section to remove recoverable vs unrecoverable framing) - rust-lang#95797 (Remove explicit delimiter token trees from `Delimited`.) - rust-lang#95804 (rustdoc: Fix empty doc comment with backline ICE) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup
Fixes #95648