Improve reporting errors and suggestions for trait bounds
Fix#66802
- When printing errors for unsized function parameter, properly point at the parameter instead of function's body.
- Improve `consider further restricting this bound` (and related) messages by separating human-oriented hints from the machine-oriented ones.
When a type error involves a `dyn Trait` as the return type, do not emit
the type error, as the "return type is not `Sized`" error will provide
enough information to the user.
When we normalize the types in a function signature, we may end up
resolving a projection to an opaque type (e.g. `Self::MyType` when
we have `type MyType = impl SomeTrait`). When the projection is
resolved, we will instantiate the generic parameters into fresh
inference variables.
While we do want to normalize projections to opaque types, we don't want
to replace the explicit generic parameters (e.g. `T` in `impl
MyTrait<T>`) with inference variables. We want the opaque type in the
function signature to be eligible to be a defining use of that opaque
type - adding inference variables prevents this, since the opaque type
substs now appears to refer to some specific type, rather than a generic
type.
To resolve this issue, we inspect the opaque types in the function
signature after normalization. Any inference variables in the substs are
replaced with the corresponding generic parameter in the identity substs
(e.g. `T` in `impl MyTrait<T>`). Note that normalization is the only way
that we can end up with inference variables in opaque substs in a
function signature - users have no way of getting inference variables
into a function signature.
Note that all of this refers to the opaque type (ty::Opaque) and its
subst - *not* to the underlying type.
Fixes#59342
Fix 'type annotations needed' error with opaque types
Related: #66426
This commit adds handling for opaque types during inference variable
fallback. Type variables generated from the instantiation of opaque
types now fallback to the opaque type itself.
Normally, the type variable for an instantiated opaque type is either
unified with the concrete type, or with the opaque type itself (e.g when
a function returns an opaque type by calling another function).
However, it's possible for the type variable to be left completely
unconstrained. This can occur in code like this:
```rust
pub type Foo = impl Copy;
fn produce() -> Option<Foo> {
None
}
```
Here, we'll instantatiate the `Foo` in `Option<Foo>` to a fresh type
variable, but we will never unify it with anything due to the fact
that we return a `None`.
This results in the error message:
```
type annotations needed: cannot resolve `_: std::marker::Copy
```
pointing at `pub type Foo = impl Copy`.
This message is not only confusing, it's incorrect. When an opaque type
inference variable is completely unconstrained, we can always fall back
to using the opaque type itself. This effectively turns that particular
use of the opaque type into a non-defining use, even if it appears in a
defining scope.
Add outlives suggestions for some lifetime errors
This PR implements suggestion diagnostics for some lifetime mismatch errors. When the borrow checker finds that some lifetime 'a doesn't outlive some other lifetime 'b that it should outlive, then in addition to the current lifetime error, we also emit a suggestion for how to fix the problem by adding a bound:
- If a and b are normal named regions, suggest to add the bound `'a: 'b`
- If b is static, suggest to replace a with static
- If b also needs to outlive a, they must be the same, so suggest unifying them
We start with a simpler implementation that avoids diagnostic regression or implementation complexity:
- We only makes suggestions for lifetimes the user can already name (eg not closure regions or elided regions)
- For now, we only emit a help note, not an actually suggestion because it is significantly easier.
Finally, there is one hack: it seems that implicit regions in async fn are given the name '_ incorrectly. To avoid suggesting '_: 'x, we simply filter out such lifetimes by name.
For more info, see this internals thread:
https://internals.rust-lang.org/t/mechanical-suggestions-for-some-borrow-checker-errors/9049/3
TL;DR Make suggestions to add a `where 'a: 'b` constraint for some lifetime errors. Details are in the paper linked from the internals thread above.
r? @estebank
TODO
- [x] Clean up code
- [x] Only make idiomatic suggestions
- [x] don't suggest naming `&'a self`
- [x] rather than `'a: 'static`, suggest replacing `'a` with `'static`
- [x] rather than `'a: 'b, 'b: 'a`, suggest replacing `'a` with `'b` or vice versa
- [x] Performance (maybe need a perf run when this is closer to the finish line?)
- perf run was clean...
- EDIT: perf run seems to only check non-error performance... How do we check that error performance didn't regress?
- [x] Needs ui tests
- [x] Integrate the `help` message into the main lifetime `error`
Related: #66426
This commit adds handling for opaque types during inference variable
fallback. Type variables generated from the instantiatino of opaque
types now fallback to the opque type itself.
Normally, the type variable for an instantiated opaque type is either
unified with the concrete type, or with the opaque type itself (e.g when
a function returns an opaque type by calling another function).
However, it's possible for the type variable to be left completely
unconstrained. This can occur in code like this:
```rust
pub type Foo = impl Copy;
fn produce() -> Option<Foo> {
None
}
```
Here, we'll instantatiate the `Foo` in `Option<Foo>` to a fresh type
variable, but we will never unify it with anything due to the fact
that we return a `None`.
This results in the error message:
`type annotations needed: cannot resolve `_: std::marker::Copy``
pointing at `pub type Foo = impl Copy`.
This message is not only confusing, it's incorrect. When an opaque type
inference variable is completely unconstrained, we can always fall back
to using the opaque type itself. This effectively turns that particular
use of the opaque type into a non-defining use, even if it appears in a
defining scope.
consistent handling of missing sysroot spans
Due to https://github.com/rust-lang/rust/issues/53081, sysroot spans (pointing to code in libcore/libstd/...) fails to print on some x86 runners. This consolidates the ignore directives for that and references the relevant issue.
I also did that for the generated derive-error-span tests -- but there the script and the tests were not entirely in sync any more since https://github.com/rust-lang/rust/pull/64151. Cc @estebank @varkor
Type parameters are referenced in the error message after the previous
few commits (using span label). But when the main error message already
references the very same type parameter it becomes clumsy. Do not show
the additional label in this case as per code review comment by
@estebank.
Also this contains a small style fix.
Don't ICE for completely unexpandable `impl Trait` types
Save the resolution of these types (to themselves) to the typeck tables so that they will eventually reach E0720.
closes#65561
Increase spacing for suggestions in diagnostics
Make the spacing between the code snippet and verbose structured
suggestions consistent with note and help messages.
r? @Centril
Remove blanket silencing of "type annotation needed" errors
Remove blanket check for existence of other errors before emitting "type annotation needed" errors, and add some eager checks to avoid adding obligations when they refer to types that reference `[type error]` in order to reduce unneeded errors.
Fix#64084.
Remove blanket check for existence of other errors before emitting
"type annotation needed" errors, and add some eager checks to avoid
adding obligations when they refer to types that reference
`[type error]` in order to reduce unneded errors.
Currently the default is "inherited" from context, so e.g. `&impl
Foo<Item = dyn Bar>` would default to `&'x impl Foo<Item = dyn Bar +
'x>`, but this triggers an ICE and is not very consistent.
This patch doesn't implement what I expect would be the correct
semantics, because those are likely too complex. Instead, it handles
what I'd expect to be the common case -- where the trait has no
lifetime parameters.
Object-lifetime-default elision is distinct from other forms of
elision; it always refers to some enclosing lifetime *present in the
surrounding type* (e.g., `&dyn Bar` expands to `&'a (dyn Bar + 'a)`.
If there is no enclosing lifetime, then it expands to `'static`.
Therefore, in an `impl Trait<Item = dyn Bar>` setting, we don't expand
to create a lifetime parameter for the `dyn Bar + 'X` bound. It will
just be resolved to `'static`.
Annoyingly, the responsibility for this resolution is spread across
multiple bits of code right now (`middle::resolve_lifetimes`,
`lowering`). The lowering code knows that the default is for an object
lifetime, but it doesn't know what the correct result would be.
Probably this should be fixed, but what we do now is a surgical fix:
we have it generate a different result for elided lifetimes in a
object context, and then we can ignore those results when figuring out
the lifetimes that are captured in the opaque type.
This commit prohibits return position `impl Trait` types that "inherit
lifetimes" from the parent scope. The intent is to forbid cases that are
challenging until they can be addressed properly.
Cleanup some surrounding code.
Support resolution of intra doc links in unnamed block scopes.
(Paths from rustdoc now use early resolution and no longer need results of late resolution like all the built ribs.)
Fix one test hitting file path limits on Windows.
resolve: Improve candidate search for unresolved macro suggestions
Use same scope visiting machinery for both collecting suggestion candidates and actually resolving the names.
The PR is better read in per-commit fashion with whitespace changes ignored (the first commit in particular moves some code around).
This should be the last pre-requisite for https://github.com/rust-lang/rust/pull/62086.
r? @davidtwco
Normalize projections appearing in `impl Trait`
Fixes#60414
This does not try to do the same for `existential type`s (which have the same bug), since that always seems to lead to cycle errors.
Migrate `compile-pass` annotations to `build-pass`
This is a part of #62277.
As a first step, the `compile-pass` tests are migrated to `build-pass`.
r? @cramertj
cc @Centril
This commit merges the logic used for opaque type type inference for
impl Trait and non-impl Trait cases. This fixes an ICE where
existential types used in the return types of functions would be allowed
to have an out-of-scope generic type parameter.
Esteban Küber
Matthew Jasper
bors
Matthias Prechtl
Patryk Wychowaniec
Mazdak Farrokhzad
Tyler Lanphear
Jonas Schievink
csmoe
Yuki Okushi
Vadim Petrochenkov
Ohad Ravid
Remy Rakic
Aaron Hill
Guillaume Gomez
Tomasz Miąsko
Ralf Jung
Dmitry Kadashev
Mark Mansi
Sydney Acksman
varkor
Alexander Regueiro
Mark Rousskov
Niko Matsakis
sd234678
David Wood
Samy Kacimi
Giles Cope
Pulkit Goyal