Allowed-by-default Lints

These lints are all set to the 'allow' level by default. As such, they won't show up unless you set them to a higher lint level with a flag or attribute.

absolute-paths-not-starting-with-crate

The absolute_paths_not_starting_with_crate lint detects fully qualified paths that start with a module name instead of crate, self, or an extern crate name

Example

#![deny(absolute_paths_not_starting_with_crate)]

mod foo {
    pub fn bar() {}
}

fn main() {
    ::foo::bar();
}

This will produce:

error: absolute paths must start with `self`, `super`, `crate`, or an external crate name in the 2018 edition
 --> lint_example.rs:8:5
  |
8 |     ::foo::bar();
  |     ^^^^^^^^^^ help: use `crate`: `crate::foo::bar`
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(absolute_paths_not_starting_with_crate)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  = warning: this is accepted in the current edition (Rust 2015) but is a hard error in Rust 2018!
  = note: for more information, see issue #53130 <https://github.com/rust-lang/rust/issues/53130>

Explanation

Rust editions allow the language to evolve without breaking backwards compatibility. This lint catches code that uses absolute paths in the style of the 2015 edition. In the 2015 edition, absolute paths (those starting with ::) refer to either the crate root or an external crate. In the 2018 edition it was changed so that they only refer to external crates. The path prefix crate:: should be used instead to reference items from the crate root.

If you switch the compiler from the 2015 to 2018 edition without updating the code, then it will fail to compile if the old style paths are used. You can manually change the paths to use the crate:: prefix to transition to the 2018 edition.

This lint solves the problem automatically. It is "allow" by default because the code is perfectly valid in the 2015 edition. The cargo fix tool with the --edition flag will switch this lint to "warn" and automatically apply the suggested fix from the compiler. This provides a completely automated way to update old code to the 2018 edition.

box-pointers

The box_pointers lints use of the Box type.

Example


#![allow(unused)]
#![deny(box_pointers)]
fn main() {
struct Foo {
    x: Box<isize>,
}
}

This will produce:

error: type uses owned (Box type) pointers: Box<isize>
 --> lint_example.rs:4:5
  |
4 |     x: Box<isize>,
  |     ^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(box_pointers)]
  |         ^^^^^^^^^^^^

Explanation

This lint is mostly historical, and not particularly useful. Box<T> used to be built into the language, and the only way to do heap allocation. Today's Rust can call into other allocators, etc.

elided-lifetimes-in-paths

The elided_lifetimes_in_paths lint detects the use of hidden lifetime parameters.

Example


#![allow(unused)]
#![deny(elided_lifetimes_in_paths)]
fn main() {
struct Foo<'a> {
    x: &'a u32
}

fn foo(x: &Foo) {
}
}

This will produce:

error: hidden lifetime parameters in types are deprecated
 --> lint_example.rs:7:12
  |
7 | fn foo(x: &Foo) {
  |            ^^^ expected lifetime parameter
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(elided_lifetimes_in_paths)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^
help: indicate the anonymous lifetime
  |
7 | fn foo(x: &Foo<'_>) {
  |               ++++

Explanation

Elided lifetime parameters can make it difficult to see at a glance that borrowing is occurring. This lint ensures that lifetime parameters are always explicitly stated, even if it is the '_ placeholder lifetime.

This lint is "allow" by default because it has some known issues, and may require a significant transition for old code.

explicit-outlives-requirements

The explicit_outlives_requirements lint detects unnecessary lifetime bounds that can be inferred.

Example


#![allow(unused)]
fn main() {
#![allow(unused)]
#![deny(explicit_outlives_requirements)]

struct SharedRef<'a, T>
where
    T: 'a,
{
    data: &'a T,
}
}

This will produce:

error: outlives requirements can be inferred
 --> lint_example.rs:5:24
  |
5 |   struct SharedRef<'a, T>
  |  ________________________^
6 | | where
7 | |     T: 'a,
  | |__________^ help: remove this bound
  |
note: the lint level is defined here
 --> lint_example.rs:2:9
  |
2 | #![deny(explicit_outlives_requirements)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Explanation

If a struct contains a reference, such as &'a T, the compiler requires that T outlives the lifetime 'a. This historically required writing an explicit lifetime bound to indicate this requirement. However, this can be overly explicit, causing clutter and unnecessary complexity. The language was changed to automatically infer the bound if it is not specified. Specifically, if the struct contains a reference, directly or indirectly, to T with lifetime 'x, then it will infer that T: 'x is a requirement.

This lint is "allow" by default because it can be noisy for existing code that already had these requirements. This is a stylistic choice, as it is still valid to explicitly state the bound. It also has some false positives that can cause confusion.

See RFC 2093 for more details.

ffi-unwind-calls

The ffi_unwind_calls lint detects calls to foreign functions or function pointers with C-unwind or other FFI-unwind ABIs.

Example

#![feature(ffi_unwind_calls)]
#![feature(c_unwind)]

mod impl {
    #[no_mangle]
    pub fn "C-unwind" fn foo() {}
}

extern "C-unwind" {
    fn foo();
}

fn bar() {
    unsafe { foo(); }
    let ptr: unsafe extern "C-unwind" fn() = foo;
    unsafe { ptr(); }
}

{{produces}}

Explanation

For crates containing such calls, if they are compiled with -C panic=unwind then the produced library cannot be linked with crates compiled with -C panic=abort. For crates that desire this ability it is therefore necessary to avoid such calls.

fuzzy-provenance-casts

The fuzzy_provenance_casts lint detects an as cast between an integer and a pointer.

Example

#![feature(strict_provenance)]
#![warn(fuzzy_provenance_casts)]

fn main() {
    let _dangling = 16_usize as *const u8;
}

This will produce:

warning: strict provenance disallows casting integer `usize` to pointer `*const u8`
 --> lint_example.rs:5:21
  |
5 |     let _dangling = 16_usize as *const u8;
  |                     ^^^^^^^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:2:9
  |
2 | #![warn(fuzzy_provenance_casts)]
  |         ^^^^^^^^^^^^^^^^^^^^^^
  = help: if you can't comply with strict provenance and don't have a pointer with the correct provenance you can use `std::ptr::from_exposed_addr()` instead
help: use `.with_addr()` to adjust a valid pointer in the same allocation, to this address
  |
5 |     let _dangling = (...).with_addr(16_usize);
  |                     ++++++++++++++++        ~

Explanation

This lint is part of the strict provenance effort, see issue #95228. Casting an integer to a pointer is considered bad style, as a pointer contains, besides the address also a provenance, indicating what memory the pointer is allowed to read/write. Casting an integer, which doesn't have provenance, to a pointer requires the compiler to assign (guess) provenance. The compiler assigns "all exposed valid" (see the docs of ptr::from_exposed_addr for more information about this "exposing"). This penalizes the optimiser and is not well suited for dynamic analysis/dynamic program verification (e.g. Miri or CHERI platforms).

It is much better to use ptr::with_addr instead to specify the provenance you want. If using this function is not possible because the code relies on exposed provenance then there is as an escape hatch ptr::from_exposed_addr.

keyword-idents

The keyword_idents lint detects edition keywords being used as an identifier.

Example


#![allow(unused)]
#![deny(keyword_idents)]
fn main() {
// edition 2015
fn dyn() {}
}

This will produce:

error: `dyn` is a keyword in the 2018 edition
 --> lint_example.rs:4:4
  |
4 | fn dyn() {}
  |    ^^^ help: you can use a raw identifier to stay compatible: `r#dyn`
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(keyword_idents)]
  |         ^^^^^^^^^^^^^^
  = warning: this is accepted in the current edition (Rust 2015) but is a hard error in Rust 2018!
  = note: for more information, see issue #49716 <https://github.com/rust-lang/rust/issues/49716>

Explanation

Rust editions allow the language to evolve without breaking backwards compatibility. This lint catches code that uses new keywords that are added to the language that are used as identifiers (such as a variable name, function name, etc.). If you switch the compiler to a new edition without updating the code, then it will fail to compile if you are using a new keyword as an identifier.

You can manually change the identifiers to a non-keyword, or use a raw identifier, for example r#dyn, to transition to a new edition.

This lint solves the problem automatically. It is "allow" by default because the code is perfectly valid in older editions. The cargo fix tool with the --edition flag will switch this lint to "warn" and automatically apply the suggested fix from the compiler (which is to use a raw identifier). This provides a completely automated way to update old code for a new edition.

lossy-provenance-casts

The lossy_provenance_casts lint detects an as cast between a pointer and an integer.

Example

#![feature(strict_provenance)]
#![warn(lossy_provenance_casts)]

fn main() {
    let x: u8 = 37;
    let _addr: usize = &x as *const u8 as usize;
}

This will produce:

warning: under strict provenance it is considered bad style to cast pointer `*const u8` to integer `usize`
 --> lint_example.rs:6:24
  |
6 |     let _addr: usize = &x as *const u8 as usize;
  |                        ^^^^^^^^^^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:2:9
  |
2 | #![warn(lossy_provenance_casts)]
  |         ^^^^^^^^^^^^^^^^^^^^^^
  = help: if you can't comply with strict provenance and need to expose the pointer provenance you can use `.expose_addr()` instead
help: use `.addr()` to obtain the address of a pointer
  |
6 |     let _addr: usize = (&x as *const u8).addr();
  |                        +               ~~~~~~~~

Explanation

This lint is part of the strict provenance effort, see issue #95228. Casting a pointer to an integer is a lossy operation, because beyond just an address a pointer may be associated with a particular provenance. This information is used by the optimiser and for dynamic analysis/dynamic program verification (e.g. Miri or CHERI platforms).

Since this cast is lossy, it is considered good style to use the ptr::addr method instead, which has a similar effect, but doesn't "expose" the pointer provenance. This improves optimisation potential. See the docs of ptr::addr and ptr::expose_addr for more information about exposing pointer provenance.

If your code can't comply with strict provenance and needs to expose the provenance, then there is ptr::expose_addr as an escape hatch, which preserves the behaviour of as usize casts while being explicit about the semantics.

macro-use-extern-crate

The macro_use_extern_crate lint detects the use of the macro_use attribute.

Example

#![deny(macro_use_extern_crate)]

#[macro_use]
extern crate serde_json;

fn main() {
    let _ = json!{{}};
}

This will produce:

error: deprecated `#[macro_use]` attribute used to import macros should be replaced at use sites with a `use` item to import the macro instead
 --> src/main.rs:3:1
  |
3 | #[macro_use]
  | ^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> src/main.rs:1:9
  |
1 | #![deny(macro_use_extern_crate)]
  |         ^^^^^^^^^^^^^^^^^^^^^^

Explanation

The macro_use attribute on an extern crate item causes macros in that external crate to be brought into the prelude of the crate, making the macros in scope everywhere. As part of the efforts to simplify handling of dependencies in the 2018 edition, the use of extern crate is being phased out. To bring macros from extern crates into scope, it is recommended to use a use import.

This lint is "allow" by default because this is a stylistic choice that has not been settled, see issue #52043 for more information.

meta-variable-misuse

The meta_variable_misuse lint detects possible meta-variable misuse in macro definitions.

Example

#![deny(meta_variable_misuse)]

macro_rules! foo {
    () => {};
    ($( $i:ident = $($j:ident),+ );*) => { $( $( $i = $k; )+ )* };
}

fn main() {
    foo!();
}

This will produce:

error: unknown macro variable `k`
 --> lint_example.rs:5:55
  |
5 |     ($( $i:ident = $($j:ident),+ );*) => { $( $( $i = $k; )+ )* };
  |                                                       ^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(meta_variable_misuse)]
  |         ^^^^^^^^^^^^^^^^^^^^

Explanation

There are quite a few different ways a macro_rules macro can be improperly defined. Many of these errors were previously only detected when the macro was expanded or not at all. This lint is an attempt to catch some of these problems when the macro is defined.

This lint is "allow" by default because it may have false positives and other issues. See issue #61053 for more details.

missing-abi

The missing_abi lint detects cases where the ABI is omitted from extern declarations.

Example


#![allow(unused)]
#![deny(missing_abi)]

fn main() {
extern fn foo() {}
}

This will produce:

error: extern declarations without an explicit ABI are deprecated
 --> lint_example.rs:4:1
  |
4 | extern fn foo() {}
  | ^^^^^^^^^^^^^^^ ABI should be specified here
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(missing_abi)]
  |         ^^^^^^^^^^^
  = help: the default ABI is C

Explanation

Historically, Rust implicitly selected C as the ABI for extern declarations. We expect to add new ABIs, like C-unwind, in the future, though this has not yet happened, and especially with their addition seeing the ABI easily will make code review easier.

missing-copy-implementations

The missing_copy_implementations lint detects potentially-forgotten implementations of Copy.

Example

#![deny(missing_copy_implementations)]
pub struct Foo {
    pub field: i32
}
fn main() {}

This will produce:

error: type could implement `Copy`; consider adding `impl Copy`
 --> lint_example.rs:2:1
  |
2 | / pub struct Foo {
3 | |     pub field: i32
4 | | }
  | |_^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(missing_copy_implementations)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Explanation

Historically (before 1.0), types were automatically marked as Copy if possible. This was changed so that it required an explicit opt-in by implementing the Copy trait. As part of this change, a lint was added to alert if a copyable type was not marked Copy.

This lint is "allow" by default because this code isn't bad; it is common to write newtypes like this specifically so that a Copy type is no longer Copy. Copy types can result in unintended copies of large data which can impact performance.

missing-debug-implementations

The missing_debug_implementations lint detects missing implementations of fmt::Debug.

Example

#![deny(missing_debug_implementations)]
pub struct Foo;
fn main() {}

This will produce:

error: type does not implement `Debug`; consider adding `#[derive(Debug)]` or a manual implementation
 --> lint_example.rs:2:1
  |
2 | pub struct Foo;
  | ^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(missing_debug_implementations)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Explanation

Having a Debug implementation on all types can assist with debugging, as it provides a convenient way to format and display a value. Using the #[derive(Debug)] attribute will automatically generate a typical implementation, or a custom implementation can be added by manually implementing the Debug trait.

This lint is "allow" by default because adding Debug to all types can have a negative impact on compile time and code size. It also requires boilerplate to be added to every type, which can be an impediment.

missing-docs

The missing_docs lint detects missing documentation for public items.

Example


#![allow(unused)]
#![deny(missing_docs)]
fn main() {
pub fn foo() {}
}

This will produce:

error: missing documentation for the crate
 --> lint_example.rs:1:1
  |
1 | / #![deny(missing_docs)]
2 | | fn main() {
3 | | pub fn foo() {}
4 | | }
  | |_^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(missing_docs)]
  |         ^^^^^^^^^^^^

Explanation

This lint is intended to ensure that a library is well-documented. Items without documentation can be difficult for users to understand how to use properly.

This lint is "allow" by default because it can be noisy, and not all projects may want to enforce everything to be documented.

must-not-suspend

The must_not_suspend lint guards against values that shouldn't be held across suspend points (.await)

Example


#![allow(unused)]
#![feature(must_not_suspend)]
#![warn(must_not_suspend)]

fn main() {
#[must_not_suspend]
struct SyncThing {}

async fn yield_now() {}

pub async fn uhoh() {
    let guard = SyncThing {};
    yield_now().await;
}
}

This will produce:

warning: `SyncThing` held across a suspend point, but should not be
  --> lint_example.rs:11:9
   |
11 |     let guard = SyncThing {};
   |         ^^^^^
12 |     yield_now().await;
   |                ------ the value is held across this suspend point
   |
note: the lint level is defined here
  --> lint_example.rs:2:9
   |
2  | #![warn(must_not_suspend)]
   |         ^^^^^^^^^^^^^^^^
help: consider using a block (`{ ... }`) to shrink the value's scope, ending before the suspend point
  --> lint_example.rs:11:9
   |
11 |     let guard = SyncThing {};
   |         ^^^^^

Explanation

The must_not_suspend lint detects values that are marked with the #[must_not_suspend] attribute being held across suspend points. A "suspend" point is usually a .await in an async function.

This attribute can be used to mark values that are semantically incorrect across suspends (like certain types of timers), values that have async alternatives, and values that regularly cause problems with the Send-ness of async fn's returned futures (like MutexGuard's)

non-ascii-idents

The non_ascii_idents lint detects non-ASCII identifiers.

Example

#![allow(unused)]
#![deny(non_ascii_idents)]
fn main() {
    let föö = 1;
}

This will produce:

error: identifier contains non-ASCII characters
 --> lint_example.rs:4:9
  |
4 |     let föö = 1;
  |         ^^^
  |
note: the lint level is defined here
 --> lint_example.rs:2:9
  |
2 | #![deny(non_ascii_idents)]
  |         ^^^^^^^^^^^^^^^^

Explanation

This lint allows projects that wish to retain the limit of only using ASCII characters to switch this lint to "forbid" (for example to ease collaboration or for security reasons). See RFC 2457 for more details.

non-exhaustive-omitted-patterns

The non_exhaustive_omitted_patterns lint detects when a wildcard (_ or ..) in a pattern for a #[non_exhaustive] struct or enum is reachable.

Example

// crate A
#[non_exhaustive]
pub enum Bar {
    A,
    B, // added variant in non breaking change
}

// in crate B
#![feature(non_exhaustive_omitted_patterns_lint)]

match Bar::A {
    Bar::A => {},
    #[warn(non_exhaustive_omitted_patterns)]
    _ => {},
}

This will produce:

warning: reachable patterns not covered of non exhaustive enum
   --> $DIR/reachable-patterns.rs:70:9
   |
LL |         _ => {}
   |         ^ pattern `B` not covered
   |
 note: the lint level is defined here
  --> $DIR/reachable-patterns.rs:69:16
   |
LL |         #[warn(non_exhaustive_omitted_patterns)]
   |                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   = help: ensure that all possible cases are being handled by adding the suggested match arms
   = note: the matched value is of type `Bar` and the `non_exhaustive_omitted_patterns` attribute was found

Explanation

Structs and enums tagged with #[non_exhaustive] force the user to add a (potentially redundant) wildcard when pattern-matching, to allow for future addition of fields or variants. The non_exhaustive_omitted_patterns lint detects when such a wildcard happens to actually catch some fields/variants. In other words, when the match without the wildcard would not be exhaustive. This lets the user be informed if new fields/variants were added.

noop-method-call

The noop_method_call lint detects specific calls to noop methods such as a calling <&T as Clone>::clone where T: !Clone.

Example


#![allow(unused)]
fn main() {
#![allow(unused)]
#![warn(noop_method_call)]
struct Foo;
let foo = &Foo;
let clone: &Foo = foo.clone();
}

This will produce:

warning: call to `.clone()` on a reference in this situation does nothing
 --> lint_example.rs:6:22
  |
6 | let clone: &Foo = foo.clone();
  |                      ^^^^^^^^ unnecessary method call
  |
note: the lint level is defined here
 --> lint_example.rs:2:9
  |
2 | #![warn(noop_method_call)]
  |         ^^^^^^^^^^^^^^^^
  = note: the type `&Foo` which `clone` is being called on is the same as the type returned from `clone`, so the method call does not do anything and can be removed

Explanation

Some method calls are noops meaning that they do nothing. Usually such methods are the result of blanket implementations that happen to create some method invocations that end up not doing anything. For instance, Clone is implemented on all &T, but calling clone on a &T where T does not implement clone, actually doesn't do anything as references are copy. This lint detects these calls and warns the user about them.

pointer-structural-match

The pointer_structural_match lint detects pointers used in patterns whose behaviour cannot be relied upon across compiler versions and optimization levels.

Example

#![deny(pointer_structural_match)]
fn foo(a: usize, b: usize) -> usize { a + b }
const FOO: fn(usize, usize) -> usize = foo;
fn main() {
    match FOO {
        FOO => {},
        _ => {},
    }
}

This will produce:

error: function pointers and unsized pointers in patterns behave unpredictably and should not be relied upon. See https://github.com/rust-lang/rust/issues/70861 for details.
 --> lint_example.rs:6:9
  |
6 |         FOO => {},
  |         ^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(pointer_structural_match)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^
  = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
  = note: for more information, see issue #62411 <https://github.com/rust-lang/rust/issues/70861>

Explanation

Previous versions of Rust allowed function pointers and wide raw pointers in patterns. While these work in many cases as expected by users, it is possible that due to optimizations pointers are "not equal to themselves" or pointers to different functions compare as equal during runtime. This is because LLVM optimizations can deduplicate functions if their bodies are the same, thus also making pointers to these functions point to the same location. Additionally functions may get duplicated if they are instantiated in different crates and not deduplicated again via LTO.

rust-2021-incompatible-closure-captures

The rust_2021_incompatible_closure_captures lint detects variables that aren't completely captured in Rust 2021, such that the Drop order of their fields may differ between Rust 2018 and 2021.

It can also detect when a variable implements a trait like Send, but one of its fields does not, and the field is captured by a closure and used with the assumption that said field implements the same trait as the root variable.

Example of drop reorder

#![deny(rust_2021_incompatible_closure_captures)]
#![allow(unused)]

struct FancyInteger(i32);

impl Drop for FancyInteger {
    fn drop(&mut self) {
        println!("Just dropped {}", self.0);
    }
}

struct Point { x: FancyInteger, y: FancyInteger }

fn main() {
  let p = Point { x: FancyInteger(10), y: FancyInteger(20) };

  let c = || {
     let x = p.x;
  };

  c();

  // ... More code ...
}

This will produce:

error: changes to closure capture in Rust 2021 will affect drop order
  --> lint_example.rs:17:11
   |
17 |   let c = || {
   |           ^^
18 |      let x = p.x;
   |              --- in Rust 2018, this closure captures all of `p`, but in Rust 2021, it will only capture `p.x`
...
24 | }
   | - in Rust 2018, `p` is dropped here, but in Rust 2021, only `p.x` will be dropped here as part of the closure
   |
note: the lint level is defined here
  --> lint_example.rs:1:9
   |
1  | #![deny(rust_2021_incompatible_closure_captures)]
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   = note: for more information, see <https://doc.rust-lang.org/nightly/edition-guide/rust-2021/disjoint-capture-in-closures.html>
help: add a dummy let to cause `p` to be fully captured
   |
17 ~   let c = || {
18 +      let _ = &p;
   |

Explanation

In the above example, p.y will be dropped at the end of f instead of with c in Rust 2021.

Example of auto-trait

#![deny(rust_2021_incompatible_closure_captures)]
use std::thread;

struct Pointer(*mut i32);
unsafe impl Send for Pointer {}

fn main() {
    let mut f = 10;
    let fptr = Pointer(&mut f as *mut i32);
    thread::spawn(move || unsafe {
        *fptr.0 = 20;
    });
}

This will produce:

error: changes to closure capture in Rust 2021 will affect which traits the closure implements
  --> lint_example.rs:10:19
   |
10 |     thread::spawn(move || unsafe {
   |                   ^^^^^^^ in Rust 2018, this closure implements `Send` as `fptr` implements `Send`, but in Rust 2021, this closure will no longer implement `Send` because `fptr` is not fully captured and `fptr.0` does not implement `Send`
11 |         *fptr.0 = 20;
   |         ------- in Rust 2018, this closure captures all of `fptr`, but in Rust 2021, it will only capture `fptr.0`
   |
note: the lint level is defined here
  --> lint_example.rs:1:9
   |
1  | #![deny(rust_2021_incompatible_closure_captures)]
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   = note: for more information, see <https://doc.rust-lang.org/nightly/edition-guide/rust-2021/disjoint-capture-in-closures.html>
help: add a dummy let to cause `fptr` to be fully captured
   |
10 ~     thread::spawn(move || { let _ = &fptr; unsafe {
11 |         *fptr.0 = 20;
12 ~     } });
   |

Explanation

In the above example, only fptr.0 is captured in Rust 2021. The field is of type *mut i32, which doesn't implement Send, making the code invalid as the field cannot be sent between threads safely.

rust-2021-incompatible-or-patterns

The rust_2021_incompatible_or_patterns lint detects usage of old versions of or-patterns.

Example

#![deny(rust_2021_incompatible_or_patterns)]

macro_rules! match_any {
    ( $expr:expr , $( $( $pat:pat )|+ => $expr_arm:expr ),+ ) => {
        match $expr {
            $(
                $( $pat => $expr_arm, )+
            )+
        }
    };
}

fn main() {
    let result: Result<i64, i32> = Err(42);
    let int: i64 = match_any!(result, Ok(i) | Err(i) => i.into());
    assert_eq!(int, 42);
}

This will produce:

error: the meaning of the `pat` fragment specifier is changing in Rust 2021, which may affect this macro
 --> lint_example.rs:4:26
  |
4 |     ( $expr:expr , $( $( $pat:pat )|+ => $expr_arm:expr ),+ ) => {
  |                          ^^^^^^^^ help: use pat_param to preserve semantics: `$pat:pat_param`
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(rust_2021_incompatible_or_patterns)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  = warning: this is accepted in the current edition (Rust 2018) but is a hard error in Rust 2021!
  = note: for more information, see <https://doc.rust-lang.org/nightly/edition-guide/rust-2021/or-patterns-macro-rules.html>

Explanation

In Rust 2021, the pat matcher will match additional patterns, which include the | character.

rust-2021-prefixes-incompatible-syntax

The rust_2021_prefixes_incompatible_syntax lint detects identifiers that will be parsed as a prefix instead in Rust 2021.

Example


#![allow(unused)]
#![deny(rust_2021_prefixes_incompatible_syntax)]

fn main() {
macro_rules! m {
    (z $x:expr) => ();
}

m!(z"hey");
}

This will produce:

error: prefix `z` is unknown
 --> lint_example.rs:8:4
  |
8 | m!(z"hey");
  |    ^ unknown prefix
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(rust_2021_prefixes_incompatible_syntax)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  = warning: this is accepted in the current edition (Rust 2018) but is a hard error in Rust 2021!
  = note: for more information, see <https://doc.rust-lang.org/nightly/edition-guide/rust-2021/reserving-syntax.html>
help: insert whitespace here to avoid this being parsed as a prefix in Rust 2021
  |
8 - m!(z"hey");
8 + m!(z "hey");
  |

Explanation

In Rust 2015 and 2018, z"hey" is two tokens: the identifier z followed by the string literal "hey". In Rust 2021, the z is considered a prefix for "hey".

This lint suggests to add whitespace between the z and "hey" tokens to keep them separated in Rust 2021.

rust-2021-prelude-collisions

The rust_2021_prelude_collisions lint detects the usage of trait methods which are ambiguous with traits added to the prelude in future editions.

Example

#![deny(rust_2021_prelude_collisions)]

trait Foo {
    fn try_into(self) -> Result<String, !>;
}

impl Foo for &str {
    fn try_into(self) -> Result<String, !> {
        Ok(String::from(self))
    }
}

fn main() {
    let x: String = "3".try_into().unwrap();
    //                  ^^^^^^^^
    // This call to try_into matches both Foo:try_into and TryInto::try_into as
    // `TryInto` has been added to the Rust prelude in 2021 edition.
    println!("{x}");
}

This will produce:

error: trait method `try_into` will become ambiguous in Rust 2021
  --> lint_example.rs:14:21
   |
14 |     let x: String = "3".try_into().unwrap();
   |                     ^^^^^^^^^^^^^^ help: disambiguate the associated function: `Foo::try_into(&*"3")`
   |
note: the lint level is defined here
  --> lint_example.rs:1:9
   |
1  | #![deny(rust_2021_prelude_collisions)]
   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   = warning: this is accepted in the current edition (Rust 2018) but is a hard error in Rust 2021!
   = note: for more information, see <https://doc.rust-lang.org/nightly/edition-guide/rust-2021/prelude.html>

Explanation

In Rust 2021, one of the important introductions is the prelude changes, which add TryFrom, TryInto, and FromIterator into the standard library's prelude. Since this results in an ambiguity as to which method/function to call when an existing try_into method is called via dot-call syntax or a try_from/from_iter associated function is called directly on a type.

single-use-lifetimes

The single_use_lifetimes lint detects lifetimes that are only used once.

Example


#![allow(unused)]
#![deny(single_use_lifetimes)]

fn main() {
fn foo<'a>(x: &'a u32) {}
}

This will produce:

error: lifetime parameter `'a` only used once
 --> lint_example.rs:4:8
  |
4 | fn foo<'a>(x: &'a u32) {}
  |        ^^      -- ...is used only here
  |        |
  |        this lifetime...
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(single_use_lifetimes)]
  |         ^^^^^^^^^^^^^^^^^^^^
help: elide the single-use lifetime
  |
4 - fn foo<'a>(x: &'a u32) {}
4 + fn foo(x: &u32) {}
  |

Explanation

Specifying an explicit lifetime like 'a in a function or impl should only be used to link together two things. Otherwise, you should just use '_ to indicate that the lifetime is not linked to anything, or elide the lifetime altogether if possible.

This lint is "allow" by default because it was introduced at a time when '_ and elided lifetimes were first being introduced, and this lint would be too noisy. Also, there are some known false positives that it produces. See RFC 2115 for historical context, and issue #44752 for more details.

trivial-casts

The trivial_casts lint detects trivial casts which could be replaced with coercion, which may require type ascription or a temporary variable.

Example


#![allow(unused)]
#![deny(trivial_casts)]
fn main() {
let x: &u32 = &42;
let y = x as *const u32;
}

This will produce:

error: trivial cast: `&u32` as `*const u32`
 --> lint_example.rs:4:9
  |
4 | let y = x as *const u32;
  |         ^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(trivial_casts)]
  |         ^^^^^^^^^^^^^
  = help: cast can be replaced by coercion; this might require a temporary variable

Explanation

A trivial cast is a cast e as T where e has type U and U is a subtype of T. This type of cast is usually unnecessary, as it can be usually be inferred.

This lint is "allow" by default because there are situations, such as with FFI interfaces or complex type aliases, where it triggers incorrectly, or in situations where it will be more difficult to clearly express the intent. It may be possible that this will become a warning in the future, possibly with type ascription providing a convenient way to work around the current issues. See RFC 401 for historical context.

trivial-numeric-casts

The trivial_numeric_casts lint detects trivial numeric casts of types which could be removed.

Example


#![allow(unused)]
#![deny(trivial_numeric_casts)]
fn main() {
let x = 42_i32 as i32;
}

This will produce:

error: trivial numeric cast: `i32` as `i32`
 --> lint_example.rs:3:9
  |
3 | let x = 42_i32 as i32;
  |         ^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(trivial_numeric_casts)]
  |         ^^^^^^^^^^^^^^^^^^^^^
  = help: cast can be replaced by coercion; this might require a temporary variable

Explanation

A trivial numeric cast is a cast of a numeric type to the same numeric type. This type of cast is usually unnecessary.

This lint is "allow" by default because there are situations, such as with FFI interfaces or complex type aliases, where it triggers incorrectly, or in situations where it will be more difficult to clearly express the intent. It may be possible that this will become a warning in the future, possibly with type ascription providing a convenient way to work around the current issues. See RFC 401 for historical context.

unreachable-pub

The unreachable_pub lint triggers for pub items not reachable from the crate root.

Example


#![allow(unused)]
#![deny(unreachable_pub)]
fn main() {
mod foo {
    pub mod bar {

    }
}
}

This will produce:

error: unreachable `pub` item
 --> lint_example.rs:4:5
  |
4 |     pub mod bar {
  |     ---^^^^^^^^
  |     |
  |     help: consider restricting its visibility: `pub(crate)`
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unreachable_pub)]
  |         ^^^^^^^^^^^^^^^
  = help: or consider exporting it for use by other crates

Explanation

A bare pub visibility may be misleading if the item is not actually publicly exported from the crate. The pub(crate) visibility is recommended to be used instead, which more clearly expresses the intent that the item is only visible within its own crate.

This lint is "allow" by default because it will trigger for a large amount existing Rust code, and has some false-positives. Eventually it is desired for this to become warn-by-default.

unsafe-code

The unsafe_code lint catches usage of unsafe code.

Example

#![deny(unsafe_code)]
fn main() {
    unsafe {

    }
}

This will produce:

error: usage of an `unsafe` block
 --> lint_example.rs:3:5
  |
3 | /     unsafe {
4 | |
5 | |     }
  | |_____^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unsafe_code)]
  |         ^^^^^^^^^^^

Explanation

This lint is intended to restrict the usage of unsafe, which can be difficult to use correctly.

unsafe-op-in-unsafe-fn

The unsafe_op_in_unsafe_fn lint detects unsafe operations in unsafe functions without an explicit unsafe block.

Example

#![deny(unsafe_op_in_unsafe_fn)]

unsafe fn foo() {}

unsafe fn bar() {
    foo();
}

fn main() {}

This will produce:

error: call to unsafe function is unsafe and requires unsafe block (error E0133)
 --> lint_example.rs:6:5
  |
6 |     foo();
  |     ^^^^^ call to unsafe function
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unsafe_op_in_unsafe_fn)]
  |         ^^^^^^^^^^^^^^^^^^^^^^
  = note: consult the function's documentation for information on how to avoid undefined behavior

Explanation

Currently, an unsafe fn allows any unsafe operation within its body. However, this can increase the surface area of code that needs to be scrutinized for proper behavior. The unsafe block provides a convenient way to make it clear exactly which parts of the code are performing unsafe operations. In the future, it is desired to change it so that unsafe operations cannot be performed in an unsafe fn without an unsafe block.

The fix to this is to wrap the unsafe code in an unsafe block.

This lint is "allow" by default since this will affect a large amount of existing code, and the exact plan for increasing the severity is still being considered. See RFC #2585 and issue #71668 for more details.

unstable-features

The unstable_features is deprecated and should no longer be used.

unused-crate-dependencies

The unused_crate_dependencies lint detects crate dependencies that are never used.

Example

#![deny(unused_crate_dependencies)]

This will produce:

error: external crate `regex` unused in `lint_example`: remove the dependency or add `use regex as _;`
  |
note: the lint level is defined here
 --> src/lib.rs:1:9
  |
1 | #![deny(unused_crate_dependencies)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^^

Explanation

After removing the code that uses a dependency, this usually also requires removing the dependency from the build configuration. However, sometimes that step can be missed, which leads to time wasted building dependencies that are no longer used. This lint can be enabled to detect dependencies that are never used (more specifically, any dependency passed with the --extern command-line flag that is never referenced via use, extern crate, or in any path).

This lint is "allow" by default because it can provide false positives depending on how the build system is configured. For example, when using Cargo, a "package" consists of multiple crates (such as a library and a binary), but the dependencies are defined for the package as a whole. If there is a dependency that is only used in the binary, but not the library, then the lint will be incorrectly issued in the library.

unused-extern-crates

The unused_extern_crates lint guards against extern crate items that are never used.

Example


#![allow(unused)]
#![deny(unused_extern_crates)]
fn main() {
extern crate proc_macro;
}

This will produce:

error: unused extern crate
 --> lint_example.rs:3:1
  |
3 | extern crate proc_macro;
  | ^^^^^^^^^^^^^^^^^^^^^^^^ help: remove it
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unused_extern_crates)]
  |         ^^^^^^^^^^^^^^^^^^^^

Explanation

extern crate items that are unused have no effect and should be removed. Note that there are some cases where specifying an extern crate is desired for the side effect of ensuring the given crate is linked, even though it is not otherwise directly referenced. The lint can be silenced by aliasing the crate to an underscore, such as extern crate foo as _. Also note that it is no longer idiomatic to use extern crate in the 2018 edition, as extern crates are now automatically added in scope.

This lint is "allow" by default because it can be noisy, and produce false-positives. If a dependency is being removed from a project, it is recommended to remove it from the build configuration (such as Cargo.toml) to ensure stale build entries aren't left behind.

unused-import-braces

The unused_import_braces lint catches unnecessary braces around an imported item.

Example

#![deny(unused_import_braces)]
use test::{A};

pub mod test {
    pub struct A;
}
fn main() {}

This will produce:

error: braces around A is unnecessary
 --> lint_example.rs:2:1
  |
2 | use test::{A};
  | ^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unused_import_braces)]
  |         ^^^^^^^^^^^^^^^^^^^^

Explanation

If there is only a single item, then remove the braces (use test::A; for example).

This lint is "allow" by default because it is only enforcing a stylistic choice.

unused-lifetimes

The unused_lifetimes lint detects lifetime parameters that are never used.

Example


#![allow(unused)]
fn main() {
#[deny(unused_lifetimes)]

pub fn foo<'a>() {}
}

This will produce:

error: lifetime parameter `'a` never used
 --> lint_example.rs:4:12
  |
4 | pub fn foo<'a>() {}
  |           -^^- help: elide the unused lifetime
  |
note: the lint level is defined here
 --> lint_example.rs:2:8
  |
2 | #[deny(unused_lifetimes)]
  |        ^^^^^^^^^^^^^^^^

Explanation

Unused lifetime parameters may signal a mistake or unfinished code. Consider removing the parameter.

unused-macro-rules

The unused_macro_rules lint detects macro rules that were not used.

Note that the lint is distinct from the unused_macros lint, which fires if the entire macro is never called, while this lint fires for single unused rules of the macro that is otherwise used. unused_macro_rules fires only if unused_macros wouldn't fire.

Example

#[warn(unused_macro_rules)]
macro_rules! unused_empty {
    (hello) => { println!("Hello, world!") }; // This rule is unused
    () => { println!("empty") }; // This rule is used
}

fn main() {
    unused_empty!(hello);
}

This will produce:

warning: 2nd rule of macro `unused_empty` is never used
 --> lint_example.rs:4:5
  |
4 |     () => { println!("empty") }; // This rule is used
  |     ^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:8
  |
1 | #[warn(unused_macro_rules)]
  |        ^^^^^^^^^^^^^^^^^^

Explanation

Unused macro rules may signal a mistake or unfinished code. Furthermore, they slow down compilation. Right now, silencing the warning is not supported on a single rule level, so you have to add an allow to the entire macro definition.

If you intended to export the macro to make it available outside of the crate, use the macro_export attribute.

unused-qualifications

The unused_qualifications lint detects unnecessarily qualified names.

Example

#![deny(unused_qualifications)]
mod foo {
    pub fn bar() {}
}

fn main() {
    use foo::bar;
    foo::bar();
}

This will produce:

error: unnecessary qualification
 --> lint_example.rs:8:5
  |
8 |     foo::bar();
  |     ^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unused_qualifications)]
  |         ^^^^^^^^^^^^^^^^^^^^^

Explanation

If an item from another module is already brought into scope, then there is no need to qualify it in this case. You can call bar() directly, without the foo::.

This lint is "allow" by default because it is somewhat pedantic, and doesn't indicate an actual problem, but rather a stylistic choice, and can be noisy when refactoring or moving around code.

unused-results

The unused_results lint checks for the unused result of an expression in a statement.

Example

#![deny(unused_results)]
fn foo<T>() -> T { panic!() }

fn main() {
    foo::<usize>();
}

This will produce:

error: unused result of type `usize`
 --> lint_example.rs:5:5
  |
5 |     foo::<usize>();
  |     ^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(unused_results)]
  |         ^^^^^^^^^^^^^^

Explanation

Ignoring the return value of a function may indicate a mistake. In cases were it is almost certain that the result should be used, it is recommended to annotate the function with the must_use attribute. Failure to use such a return value will trigger the unused_must_use lint which is warn-by-default. The unused_results lint is essentially the same, but triggers for all return values.

This lint is "allow" by default because it can be noisy, and may not be an actual problem. For example, calling the remove method of a Vec or HashMap returns the previous value, which you may not care about. Using this lint would require explicitly ignoring or discarding such values.

variant-size-differences

The variant_size_differences lint detects enums with widely varying variant sizes.

Example


#![allow(unused)]
#![deny(variant_size_differences)]
fn main() {
enum En {
    V0(u8),
    VBig([u8; 1024]),
}
}

This will produce:

error: enum variant is more than three times larger (1024 bytes) than the next largest
 --> lint_example.rs:5:5
  |
5 |     VBig([u8; 1024]),
  |     ^^^^^^^^^^^^^^^^
  |
note: the lint level is defined here
 --> lint_example.rs:1:9
  |
1 | #![deny(variant_size_differences)]
  |         ^^^^^^^^^^^^^^^^^^^^^^^^

Explanation

It can be a mistake to add a variant to an enum that is much larger than the other variants, bloating the overall size required for all variants. This can impact performance and memory usage. This is triggered if one variant is more than 3 times larger than the second-largest variant.

Consider placing the large variant's contents on the heap (for example via Box) to keep the overall size of the enum itself down.

This lint is "allow" by default because it can be noisy, and may not be an actual problem. Decisions about this should be guided with profiling and benchmarking.