Make `need_type_info_err` more conservative #73027

doctorn · 2020-06-05T11:30:56Z

Makes sure arg patterns we are going to suggest on are actually contained within the span of the obligation that caused the inference error (credit to @lcnr for suggesting this fix).

There's a subtle trade-off regarding the handling of local patterns which I've left a comment about.

Resolves #72690

rust-highfive · 2020-06-05T11:30:59Z

r? @varkor

(rust_highfive has picked a reviewer for you, use r? to override)

doctorn · 2020-06-05T11:39:02Z

r? @estebank

lcnr

LGTM.

src/librustc_infer/infer/error_reporting/need_type_info.rs

Co-authored-by: Bastian Kauschke <bastian_kauschke@hotmail.de>

estebank

I need to look a bit deeper into this PR to confirm it is ok to merge. It certainly fixes some cases, but others are still present. I think we might be able to come up with a more robust way of handling this.

Ideally we would also be able to handle let _ = "x".as_ref(); with a suggestion to give _ a type (which we don't today).

I think that we might be able to fix this by changing how the visitor works to try and keep into account some kind of inference chain, but again, I'm not sure how that would work.

estebank · 2020-06-05T16:45:16Z

src/test/ui/closure-expected-type/expect-two-infer-vars-supply-ty-with-bound-region.stderr

 error[E0282]: type annotations needed
-  --> $DIR/expect-two-infer-vars-supply-ty-with-bound-region.rs:8:27
+  --> $DIR/expect-two-infer-vars-supply-ty-with-bound-region.rs:8:5
   |
 LL |     with_closure(|x: u32, y| {});
-   |                           ^ consider giving this closure parameter a type
+   |     ^^^^^^^^^^^^ cannot infer type for type parameter `B` declared on the function `with_closure`


This is technically a regression, right?

I think this is the cost of making the suggestion more conservative. The point is at the moment we're struggling with things like this playground which urgently requires attention

estebank · 2020-06-05T16:48:19Z

src/test/ui/issues/issue-23046.stderr

+error[E0282]: type annotations needed for the closure `fn(Expr<'_, _>) -> Expr<'_, _>`
+  --> $DIR/issue-23046.rs:18:9
   |
-LL |     let ex = |x| {
-   |               ^ consider giving this closure parameter the explicit type `Expr<'_, VAR>`, where the type parameter `VAR` is specified
+LL |         let_(add(x,x), |y| {
+   |         ^^^^ cannot infer type for type parameter `VAR` declared on the function `let_`
+   |
+help: give this closure an explicit return type without `_` placeholders
+   |
+LL |             let_(add(x, x), |x|-> Expr<'_, _> { x })})};
+   |                                ^^^^^^^^^^^^^^^^   ^


...so it could be argued is this (but inference works with either suggestion, so this is acceptable).

estebank · 2020-06-05T16:51:23Z

src/test/ui/issues/issue-72690.stderr

+error[E0283]: type annotations needed for `std::string::String`
+  --> $DIR/issue-72690.rs:41:5
+   |
+LL |     String::from("x".as_ref());
+   |     ^^^^^^^^^^^^ cannot infer type for struct `std::string::String`
+LL |     let _ = String::from("x");
+   |         - consider giving this pattern a type
+   |
+   = note: cannot satisfy `std::string::String: std::convert::From<&_>`
+   = note: required by `std::convert::From::from`


This seems to still give an incorrect suggestion.

For these two incorrect suggestions the immediate solution is to require that local.span contains target_span, but doing so causes a huge number of other regressions - hence the trade-off discussed in the comment I added.

(There's some discussion in this thread)

Just looked at the thread. I would feel better about merging this PR as is if we had someone focusing on adding better tracking to ObligationCause. I already did a lot there for obligations in the type namespace but won't have time to look at it in value namespace. Would you be interested in digging into doing that?

Yeah definitely! If you send me some details about what needs doing/where to start I can get on that right away

estebank · 2020-06-05T16:53:48Z

src/test/ui/issues/issue-72690.stderr

+error[E0283]: type annotations needed for `std::string::String`
+  --> $DIR/issue-72690.rs:47:5
+   |
+LL |     let _ = String::from("x");
+   |         - consider giving this pattern a type
+LL |     String::from("x".as_ref());
+   |     ^^^^^^^^^^^^ cannot infer type for struct `std::string::String`
+   |
+   = note: cannot satisfy `std::string::String: std::convert::From<&_>`
+   = note: required by `std::convert::From::from`


...and this.

estebank · 2020-06-19T16:47:17Z

@bors r+

bors · 2020-06-19T16:47:19Z

📌 Commit b4ddd91 has been approved by estebank

@ghost

…arth Rollup of 16 pull requests Successful merges: - rust-lang#71420 (Specialization is unsound) - rust-lang#71899 (Refactor `try_find` a little) - rust-lang#72689 (add str to common types) - rust-lang#72791 (update coerce docs and unify relevant tests) - rust-lang#72934 (forbid mutable references in all constant contexts except for const-fns) - rust-lang#73027 (Make `need_type_info_err` more conservative) - rust-lang#73347 (Diagnose use of incompatible sanitizers) - rust-lang#73359 (shim.rs: avoid creating `Call` terminators calling `Self`) - rust-lang#73399 (Clean up E0668 explanation) - rust-lang#73436 (Clean up E0670 explanation) - rust-lang#73440 (Add src/librustdoc as an alias for src/tools/rustdoc) - rust-lang#73442 (pretty/mir: const value enums with no variants) - rust-lang#73452 (Unify region variables when projecting associated types) - rust-lang#73458 (Use alloc::Layout in DroplessArena API) - rust-lang#73484 (Update the doc for std::prelude to the correct behavior) - rust-lang#73506 (Bump Rustfmt and RLS) Failed merges: r? @ghost

Trait predicate ambiguities are not always in `Self` When reporting ambiguities in trait predicates, the compiler incorrectly assumed the ambiguity was always in the type the trait should be implemented on, and never the generic parameters of the trait. This caused silly suggestions for predicates like `<KnownType as Trait<_>>`, such as giving explicit types to completely unrelated variables that happened to be of type `KnownType`. This also reverts rust-lang#73027, which worked around this issue in some cases and does not appear to be necessary any more. fixes rust-lang#77982 fixes rust-lang#78055

Make need_type_info_err more conservative

df996e8

rust-highfive assigned varkor Jun 5, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jun 5, 2020

rust-highfive assigned estebank and unassigned varkor Jun 5, 2020

lcnr approved these changes Jun 5, 2020

View reviewed changes

src/librustc_infer/infer/error_reporting/need_type_info.rs Outdated Show resolved Hide resolved

Add FIXME to comment regarding local patterns

b4ddd91

Co-authored-by: Bastian Kauschke <bastian_kauschke@hotmail.de>

estebank reviewed Jun 5, 2020

View reviewed changes

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jun 19, 2020

Manishearth mentioned this pull request Jun 20, 2020

Rollup of 16 pull requests #73528

Merged

bors merged commit 2d1bd57 into rust-lang:master Jun 20, 2020

SNCPlay42 mentioned this pull request Oct 19, 2020

Trait predicate ambiguities are not always in Self #78111

Merged

SNCPlay42 added a commit to SNCPlay42/rust that referenced this pull request Oct 19, 2020

revert workaround rust-lang#73027

c146e8c

cuviper added this to the 1.46 milestone May 2, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make `need_type_info_err` more conservative #73027

Make `need_type_info_err` more conservative #73027

doctorn commented Jun 5, 2020 •

edited

Loading

rust-highfive commented Jun 5, 2020

doctorn commented Jun 5, 2020

lcnr left a comment

estebank left a comment

estebank Jun 5, 2020

doctorn Jun 5, 2020

estebank Jun 5, 2020

estebank Jun 5, 2020

doctorn Jun 5, 2020 •

edited

Loading

doctorn Jun 5, 2020

estebank Jun 10, 2020

doctorn Jun 10, 2020 •

edited

Loading

estebank Jun 5, 2020

estebank commented Jun 19, 2020

bors commented Jun 19, 2020

Make need_type_info_err more conservative #73027

Make need_type_info_err more conservative #73027

Conversation

doctorn commented Jun 5, 2020 • edited Loading

rust-highfive commented Jun 5, 2020

doctorn commented Jun 5, 2020

lcnr left a comment

Choose a reason for hiding this comment

estebank left a comment

Choose a reason for hiding this comment

estebank Jun 5, 2020

Choose a reason for hiding this comment

doctorn Jun 5, 2020

Choose a reason for hiding this comment

estebank Jun 5, 2020

Choose a reason for hiding this comment

estebank Jun 5, 2020

Choose a reason for hiding this comment

doctorn Jun 5, 2020 • edited Loading

Choose a reason for hiding this comment

doctorn Jun 5, 2020

Choose a reason for hiding this comment

estebank Jun 10, 2020

Choose a reason for hiding this comment

doctorn Jun 10, 2020 • edited Loading

Choose a reason for hiding this comment

estebank Jun 5, 2020

Choose a reason for hiding this comment

estebank commented Jun 19, 2020

bors commented Jun 19, 2020

Make `need_type_info_err` more conservative #73027

Make `need_type_info_err` more conservative #73027

doctorn commented Jun 5, 2020 •

edited

Loading

doctorn Jun 5, 2020 •

edited

Loading

doctorn Jun 10, 2020 •

edited

Loading