Downgrade trivially_copy_pass_by_ref to pedantic
The rationale for this lint is documented as:
> In many calling conventions instances of structs will be passed through registers if they fit into two or less general purpose registers.
I think the purported performance benefits of clippy's recommendation are overstated. This isn't worth asking people to sprinkle code with more `*``*``&``*``&` to chase the alleged performance.
This should be a pedantic lint that is disabled by default and opted in if some specific performance sensitive codebase determines that it is worthwhile.
As a reminder, a typical place that a reference to a primitive would come up is if the function is used as a filter. Triggering a performance-oriented lint on this type of code is the definition of pedantic.
```rust
fn filter(_n: &i32) -> bool {
true
}
fn main() {
let v = vec![1, 2, 3];
v.iter().copied().filter(filter).for_each(drop);
}
```
```console
warning: this argument (4 byte) is passed by reference, but would be more efficient if passed by value (limit: 8 byte)
--> src/main.rs:1:15
|
1 | fn filter(_n: &i32) -> bool {
| ^^^^ help: consider passing by value instead: `i32`
```
changelog: Remove trivially_copy_pass_by_ref from default set of enabled lints
Downgrade let_unit_value to pedantic
Given that the false positive in #1502 is marked E-hard and I don't have much hope of it getting fixed, I think it would be wise to disable this lint by default. I have had to suppress this lint in every substantial codebase (\>100k line) I have worked in. Any time this lint is being triggered, it's always the false positive case.
The motivation for this lint is documented as:
> A unit value cannot usefully be used anywhere. So binding one is kind of pointless.
with this example:
> ```rust
> let x = {
> 1;
> };
> ```
Sure, but the author would find this out via an unused_variable warning or from `x` not being the type that they need further down. If there ends up being a type error on `x`, clippy's advice isn't going to help get the code compiling because it can only run if the code already compiles.
changelog: Remove let_unit_value from default set of enabled lints
Fix update_lints
This fixes a bug in update_lints, where `internal` lints were not registered properly. This also cleans up some code. For example: The code generation functions no longer filter the lints the are given. This is now the task of the caller. This way, it is more obvious in the `replace_in_file` calls which lints will be included in which part of a file.
This also turns the lint modules private. There is no need for them to be public, since shared code should be in the utils module anyway.
And last but not least, this fixes the `register_lints` code generation, so also internal lints get registered.
changelog: none
Result<T, E> has an `ok()` method that adapts a Result<T,E> into an Option<T>.
It's possible to get around this adapter by writing Result<T,E>.map_or(None, Some).
This lint is implemented as a new variant of the existing
[`option_map_none` lint](https://github.com/rust-lang/rust-clippy/pull/2128)
useless Rc<Rc<T>>, Rc<Box<T>>, Rc<&T>, Box<&T>
refers to #2394
changelog: Add lints for Rc<Rc<T>> and Rc<Box<T>> and Rc<&T>, Box<&T>
this is based on top of another change #5310 so probably should go after that one.
Downgrade option_option to pedantic
Based on a search of my work codebase (\>500k lines) for `Option<Option<`, it looks like a bunch of reasonable uses to me. The documented motivation for this lint is:
> an optional optional value is logically the same thing as an optional value but has an unneeded extra level of wrapping
which seems a bit bogus in practice. For example a typical usage would look like:
```rust
let mut host: Option<String> = None;
let mut port: Option<i32> = None;
let mut payload: Option<Option<String>> = None;
for each field {
match field.name {
"host" => host = Some(...),
"port" => port = Some(...),
"payload" => payload = Some(...), // can be null or string
_ => return error,
}
}
let host = host.ok_or(...)?;
let port = port.ok_or(...)?;
let payload = payload.ok_or(...)?;
do_thing(host, port, payload)
```
This lint seems to fit right in with the pedantic group; I don't think linting on occurrences of `Option<Option<T>>` by default is justified.
---
changelog: Remove option_option from default set of enabled lints
Lint unnamed address comparisons
Functions and vtables have an insignificant address. Attempts to compare such addresses will lead to very surprising behaviour. For example: addresses of different functions could compare equal; two trait object pointers representing the same object and the same type could be unequal.
Lint against unnamed address comparisons to avoid issues like those in rust-lang/rust#69757 and rust-lang/rust#54685.
changelog: New lints: [`fn_address_comparisons`] [#5294](https://github.com/rust-lang/rust-clippy/pull/5294), [`vtable_address_comparisons`] [#5294](https://github.com/rust-lang/rust-clippy/pull/5294)
`unused_self` false positive
fixes#5351
Remove the for loop in `unused_self` so that lint enabled for one method doesn't trigger on another method.
changelog: Fix false positive in `unused_self` around lint gates on impl items
Lint for `pub(crate)` items that are not crate visible due to the visibility of the module that contains them
changelog: Add `redundant_pub_crate` lint
Closes#5274.
Improvement: Don't show function body in needless_lifetimes
Changes the span on which the lint is reported to point to only the
function return type instead of the entire function body.
Fixes#5284
changelog: none
add lint on File::read_to_string and File::read_to_end
Adds lint `verbose_file_reads` which checks for use of File::read_to_end and File::read_to_string.
Closes https://github.com/rust-lang/rust-clippy/issues/4916
changelog: add lint on File::{read_to_end, read_to_string}
Use `edition:2018` flag more widely
Now we recommend using `// edition:2018`, so let's use it more widely.
Also replace a too old example with new one in the docs.
changelog: none
redundant_pattern: take binding (ref, ref mut) into account in suggestion
fixes#5271
changelog: redundant_pattern: take binding (ref, ref mut) into account in suggestion (#5271)
Fix match single binding when in a let stmt
Fix bad suggestion when `match_single_binding` lints when inside a local (let) statement.
Fixes#5267
changelog: none
Resolve false positives of unnecessary_cast for non-decimal integers
This PR resolves false positives of `unnecessary_cast` for hexadecimal integers to floats and adds a corresponding test case.
Fixes: #5220
changelog: none
Add lint for .. use in fully binded struct
This PR adds the lint `match-wild-in-fully-binded-struct` to prevent the use of the `..` pattern when all fields of the struct are already binded.
Fixes: #638
changelog: Add [`rest_pat_in_fully_bound_structs`] lint to warn against the use of `..` in fully binded struct
Detect usage of custom floating-point abs implementation
Closes#5224
changelog: Enhance [`suboptimal_flops`] lint to detect manual implementations of the `abs` method
Whitelist unused attribute for use items.
This PR whitelists the `unused` attribute with `use` items and adds a corresponding test case.
Fixes: #5229
changelog: none
Add lint to detect floating point operations that can be computed more
accurately at the cost of performance. `cbrt`, `ln_1p` and `exp_m1`
library functions call their equivalent cmath implementations which is
slower but more accurate so moving checks for these under this new lint.
Merge the accuracy and efficiency lints into a single lint that
checks for improvements to accuracy, efficiency and readability
of floating-point expressions.
Move check for lossy whole-number floats out of `excessive_precision`
changelog: Add new lint `lossy_float_literal` to detect lossy whole number float literals and move it out of `excessive_precision` again.
Fixes#5201
Make it possible to correctly indent snippet_block snippets
This adds a `indent_relative_to` arg to the `{snippet,expr}_block` functions. This makes it possible to keep the correct indentation of block like suggestions.
In addition, this makes the `trim_multiline` function private and adds a `indent_of` function, to get the indentation of the first line of a span.
The suggestion of `needless_continue` cannot be made auto applicable, since it would be also necessary to remove code following the linted expression. (Well, maybe it is possible, but I don't know how to do it. Expanding the suggestion span to the last expression, that should be removed didn't work)
changelog: Improve suggestions, when blocks of code are involved
Port mitsuhiko's excessive bools lints
Closes#4 .
changelog: add `struct_excessive_bools` and `fn_params_excessive_bools` lints.
I moved is_trait_impl_item check because at first I implemented it as a late pass for some reason but then I realized it's actually an early lint. But it's a useful function to have, should I move it into a separate pr?
Do not lint `unnecessary_unwrap` in external macros
Fixes#5131
I think we shouldn't lint `{panicking, unnecessary}_unwrap` in macros, not just `assert!`.
changelog: Fix false positive in `unnecessary_unwrap`
Deprecate util/dev in favor of cargo alias
This means one less shell script and a bit more cross-platform support
for contributors.
If you've been using `./util/dev` before, this now becomes `cargo dev`.
The key part of this change is found in `.cargo/config` where an alias for calling the `clippy_dev` binary is defined.
changelog: none
If you've been using `./util/dev` before, this now becomes `cargo dev`.
The key part of this change is found in `.cargo/config`.
This means one less shell script and a bit more cross-platform support
for contributors.
add `let_underscore_lock` lint
closes#1574
changelog: add `let_underscore_lock` lint
I am not entirely sure about my docs/messages wording here, improvements are welcome
Clean up `span_lint` in `methods/mod.rs`
Uses `span_help_and_lint` instead of `span_lint` and `span_lint_and_sugg` instead of `span_lint_and_then`.
changelog: none
dont fire `possible_missing_comma` if intendation is present
Closes#4399
changelog: dont fire `possible_missing_comma` if intendation is present
I suspect there is already a utils function for indentation (but searching indent didn't yield a function for that), and my solution is certainly not universal, but it's probably the best we can do.
Detect usage of invalid atomic ordering in memory fences
Detect usage of `core::sync::atomic::{fence, compiler_fence}` with `Ordering::Relaxed` and suggest valid alternatives.
changelog: Extend `invalid_atomic_ordering` to lint memory fences
Fixes#5026
Avoid mut_key on types of unknown layout
This fixes#5020 by requiring a known layout for the key type before linting. Edit: This fixes#5043, too.
changelog: none
Keep the ordering in `nonminimal_bool` lint
I believe it shouldn't cause any regression but feel free to let me know if you have a doubtful example.
Also, splits up `booleans` ui test.
Fixes#5045
changelog: none
Autogenerate new lints
Add option in clippy_dev to automatically generate boilerplate code for adding new lints
example:
`./util/dev new_lint --name=iter_nth_zero --type=late`
Fixes#4942
changelog: none
Match wild err arm improvements
This lint should trigger on other identifiers which have `_` prefix (such as `_e`) and only if they are unused in the panic block.
_Note_: the `is_unused` function is greatly inspired from `pat_is_wild` function in [loops lints](43ac9416d9/clippy_lints/src/loops.rs (L1689)).
I've been considering doing some refactoring, maybe in utils. Maybe this PR or a new one. What do you think ?
fixes#5024
changelog: none
New lint: pats_with_wild_match_arm
Wildcard use with other pattern in same match arm.
The wildcard covers other(s) pattern(s) as it will match anyway.
changelog: add new lint when multiple patterns (including wildcard) are used in a match arm.
Fixes#4640.
Fix redundant_clone lint not working with PathBuf and OsString
https://github.com/rust-lang/rust-clippy/pull/4825 diabled MIR optimization in clippy, including `rustc_mir::transform::InstCombine` which reduces `&(*x)` to `x`. This PR tries to unwrap `&*` when looking into `mir::Rvalue`s.
Fixes#5014.
---
changelog: fixed `redundant_clone` lint not working with `PathBuf` and `OsString`
Detect usage of invalid atomic ordering modes such as
`Ordering::{Release, AcqRel}` in atomic loads and
`Ordering::{Acquire, AcqRel}` in atomic stores.
Use `mem::take` instead of `mem::replace` when applicable
`std::mem::take` can be used to replace a value of type `T` with `T::default()` instead of `std::mem::replace`.
Fixes issue #4871
changelog: Added lint for [`mem_replace_with_default`]
Fix ICE on `unsound_collection_transmute`
Fixes#4968
Check if `Ty`s are normalizable. It might show hidden false negative, I'm not sure.
Also, the regression tests are placed on two dirs, so move them to `/crashes`. I think it will be easier to find the right place.
changelog: Fix ICE on `unsound_collection_transmute`
Currently `replace_consts` lint applies within match patterns but
the suggestion is incorrect as function calls are disallowed in
them. To fix this we prevent the lint from firing within patterns.
new lint: mutable_key_type
This fixes#732 - well, partly, it doesn't adress `Hash` impls, but the use of mutable types as map keys or set members
changelog: add `mutable_key_type` lint
r? @flip1995
Add `let_underscore_must_use` lint
changelog: closes#4812 , added a new `let_underscore_must_use` lint, moved `is_must_use_ty` to utils, added `is_must_use_fn` util function
Fix `map_clone` false positive
Don't lint when the item type is not a reference. `copied` only applies to references.
changelog: Fix `map_clone` false positive
Currently the cognitive complexity lint spans the entire function
body making it really difficult to read and refactor the code in
editors. To fix this we reduce the lint span to the function name.
Move `iterator_step_by_zero` and correct the documentation
Move `iterator_step_by_zero` and correct the documentation.
changelog: Corrected `iterator_step_by_zero` documentation
Currently `cmp_nan` lint doesn't detect comparisons with NaN's
if the operands are consts variables so to fix this we evaluate
the const variables first before testing for NaN.
Fix false positive with cast_sign_loss lint
`cast_sign_loss` lint incorrectly suggests that the result of `checked_abs`, `rem_euclid` and `checked_rem_euclid` cannot be casted to an unsigned integer without loss.
Fixes#4818#4764#4743
changelog: Fix false positives in `cast_sign_loss` lint
The Rust Book recommends that functions that return a `Result` type have
a doc comment with an `# Errors` section describing the kind of errors
that can be returned
(https://doc.rust-lang.org/book/ch14-02-publishing-to-crates-io.html#commonly-used-sections).
This change adds a lint to enforce this. The lint is allow by default;
it can be enabled with `#![warn(clippy::missing_errors_doc)]`.
Closes#4854.
`cast_sign_loss` lint incorrectly suggests that the result of
`checked_abs`, `rem_euclid` and `checked_rem_euclid` cannot
be casted to an unsigned integer without loss.
Fix FP in manual_swap lint with slice-like types
Fixes#4853
changelog: Fix FP in [`manual_swap`] lint with slice-like types and make it auto applicable
This utilizes https://github.com/rust-lang/rust/pull/60584 by setting
our own `panic_hook` and pointing to our own issue tracker instead of
the rustc issue tracker.
This also adds a new internal lint to test the ICE message.
**Potential downsides**
* This essentially copies rustc's `report_ice` function as
`report_clippy_ice`. I think that's how it's meant to be implemented, but
maybe @jonas-schievink could have a look as well =)
The downside of more-or-less copying this function is that we have to
maintain it as well now.
The original function can be found [here][original].
* `driver` now depends directly on `rustc` and `rustc_errors`
Closes#2734
[original]: 59367b074f/src/librustc_driver/lib.rs (L1185)
Literal Representation Restructure
This pull request restructures the literal_representation module to be easier to understand and maintain. I split the changes into a lot of commits to make reviewing easier.
changelog: none
Fix false positive in derive_hash_xor_eq
This fixes a false positive in derive_hash_xor_eq where the lint was
triggering on user-defined traits called `Hash`.
changelog: Fix false positive in `derive_hash_xor_eq`
Fixes#4658
Fix check_infinite_loop (while_immutable_condition) by checking for break or return inside loop body
changelog: Fix check_infinite_loop (while_immutable_condition) by checking for break or return inside loop body
fixes#4648
Run ./util/dev
Revert changelog entry
Rename lint to same_functions_in_if_condition and add a doc example
Add testcases with different arg in fn invocation
added documentation
minor style fix
change as to ::from
add ignore to doc
include threshold in lint message/make suggestion more apparent/use Scalar api instead of matching
style fix
shange snippet_opt to snippet
Fix false positive in explicit_counter_loop lint
When the counter was used in a closure after the loop the lint didn't detect the
usage of the counter correctly.
changelog: Fix false positive in `explicit_counter_loop`
Fixes#4732
restriction lint for `std::process::exit`
Addition to #4655 - adds the lint checking for `std::process::exit`
changelog: add restriction lint for `std::process::exit`
build: improve script and travis config
* fix a diff failure on windows
See https://travis-ci.com/rust-lang/rust-clippy/jobs/245971932#L1625
for an example.
* use cmp instead of diff > /dev/null
* clone single branch instead of clone then checking out
* do not decrypt key if have no diff change
changelog: none
Add lint for debug_assert_with_mut_call
closes#1526
**What does not work:**
* detecting a mut call in the format string itself, e.g. `debug_assert!(false, "{}", vec![1].pop())`
* detecting `*mut T` usage (pointer)
---
changelog: add new lint `debug_assert_with_mut_call`
Fix suggestion of `explicit_counter_loop`
changelog: In the suggestion of `explicit_counter_loop`, if the `for` loop argument doesn't implement `Iterator`, then we suggest `x.into_iter().enumerate()` (or `x.iter{_mut}()` as appropriate). Also, the span of the suggestion has been corrected.
Fixes#4678
This lint will complain when you put a mutable function/method call
inside a `debug_assert` macro, because it will not be executed in
release mode, therefore it will change the execution flow, which is not
wanted.
`must_use_unit` lints unit-returning functions with a `#[must_use]`
attribute, suggesting to remove it.
`double_must_use` lints functions with a plain `#[must_use]`
attribute, but which return a type which is already `#[must_use]`,
so the attribute has no benefit.
`must_use_candidate` is a pedantic lint that lints functions and
methods that return some non-unit type that is not already
`#[must_use]` and suggests to add the annotation.
Lints when, on the RHS of a BinOp, there is a UnOp without a space
before the operator but with a space after (e.g. foo >- 1).
Signed-off-by: Nikos Filippakis <nikolaos.filippakis@cern.ch>
Add assert message to suggestion in lint assertions_on_constants
<!--
Thank you for making Clippy better!
We're collecting our changelog from pull request descriptions.
If your PR only updates to the latest nightly, you can leave the
`changelog` entry as `none`. Otherwise, please write a short comment
explaining your change.
If your PR fixes an issue, you can add "fixes #issue_number" into this
PR description. This way the issue will be automatically closed when
your PR is merged.
If you added a new lint, here's a checklist for things that will be
checked during review or continuous integration.
- [x] Followed [lint naming conventions][lint_naming]
- [x] Added passing UI tests (including committed `.stderr` file)
- [ ] `cargo test` passes locally
- [x] Executed `./util/dev update_lints`
- [ ] Added lint documentation
- [ ] Run `./util/dev fmt`
[lint_naming]: https://rust-lang.github.io/rfcs/0344-conventions-galore.html#lints
Note that you can skip the above if you are just opening a WIP PR in
order to get feedback.
Delete this line and everything above before opening your PR -->
- [x] suggest replacing `assert!(false, "msg")` with `panic!("msg")`
- [x] extend to allow ~~variables~~ any expression for `"msg"`
- ~~suggest replacing `assert!(false, "msg {}", "arg")` with `panic!("msg {}", "arg")`~~
changelog: add assert message to suggestion in lint assertions_on_constants
Work towards fixing: #3575
account for doc visibility
This fixes#4608.
Also I noticed that the lint failed to look at trait and impl items. There's a small bit of fallout in the code, too, but not enough to warrant its own commit.
changelog: check docs of trait items and impl items, also make `missing_safety_doc` account for visibility
Add check for assert_eq macros to unit_cmp lint
changelog: Add check for unit comparisons through `assert_eq!`, `debug_assert_eq!`, `assert_ne!` and `debug_assert_ne!` macros to unit_cmp lint.
fixes#4481
Add a new lint for comparison chains
changelog: Adds a new lint: `comparison_chain`.
`comparison_chain` lints all `if` conditional chains where all the conditions are binary comparisons on the same two operands and will suggest a rewrite with `match`.
Closes#4531.
#4542 remove machine applicable suggestion
This helps #4542 (but does not completely resolve) by removing the machine applicable suggestion (which was incorrect) for that case.
I would have preferred to fix the machine applicable suggestion to handle format strings, but that's a bit beyond my current understanding of the clippy codebase. I'd be happy to give it a try given some guidance.
changelog: only produce machine applicable suggestions on `explicit_write` lint
Changes cast-lossless to a pedantic lint
As discussed in #4528, this moves the cast-lossless lint from `all` to `pedantic`.
I couldn't tell from description alone if it should also be removed from the complexity category, so I left it as part of complexity for now. I didn't see any impact to the tests from this change, but I could be wrong (as this is my first PR).
fixes#4528
changelog: Moves cast-lossless from default to checking only as a `pedantic` lint.
tests: disable incremental compilation in dogfood to reduce target dir size
tests: execute dogfood tests with incremental compilation disabled reduces target/ dir size of when "cargo test"ing by around 2 gigs.
changelog: none
Fix incorrect swap suggestion
Clippy suggests using swap on fields belonging to the same owner causing two mutable borrows of the owner.
Disclosure: This is my first time working with clippy and rusts HIR. I'm very grateful for assistance and suggestions to improve the PR.
fixes#981
changelog: Fix false positive in `manual_swap` lint
New `is_integer_const` to check more const ints
This mostly affects loop checks and the modulo_one lint. Tests were also updated where applicable.
changelog: none
Allow block_in_if_{stmt,expr} in external macro
I found this by running `cargo fix --clippy` on quite a big codebase.
You could refactor this assert to
```rust
let block_expr = _;
assert!(block_expr);
```
but,
1. it doesn't increase the readability IMO
2. That isn't possible in a `debug_assert!`
I'm not sure though, if we should allow this for macros in general or just for external macros.
changelog: Allow `block_in_if_{stmt,expr}` in external macros
Fix missing_const_for_fn false positive
We don't want to lint if the type of the method implements drop.
(constant functions cannot evaluate destructors)
changelog: Fix `missing_const_for_fn` false positive
Fixes#4449
More rustfix tests
<!--
Thank you for making Clippy better!
We're collecting our changelog from pull request descriptions.
If your PR only updates to the latest nightly, you can leave the
`changelog` entry as `none`. Otherwise, please write a short comment
explaining your change.
If your PR fixes an issue, you can add "fixes #issue_number" into this
PR description. This way the issue will be automatically closed when
your PR is merged.
If you added a new lint, here's a checklist for things that will be
checked during review or continuous integration.
- [ ] Followed [lint naming conventions][lint_naming]
- [ ] Added passing UI tests (including committed `.stderr` file)
- [ ] `cargo test` passes locally
- [ ] Executed `./util/dev update_lints`
- [ ] Added lint documentation
- [ ] Run `./util/dev fmt`
[lint_naming]: https://rust-lang.github.io/rfcs/0344-conventions-galore.html#lints
Note that you can skip the above if you are just opening a WIP PR in
order to get feedback.
Delete this line and everything above before opening your PR -->
cc #3630
This is probably easier reviewed per-commit.
changelog: none
Split up cmp_owned tests, add run-rustfix
Some of the cmp_owned tests emitted non-machine-applicable suggestions,
so I moved them to `tests/ui/cmp_owned/without_suggestion.rs` and added
`// run-rustfix` to the other half.
changelog: none
cc #3630
Fix unused_unit false positive
changelog: Fix `unused_unit` false positive
For some reason the `expr` of `stmt.node` didn't contain the expansion information, but the `stmt.span` does.
Fixes#4076
Account for trait alias when looking for defid
I hit the crash on the `expect` call when running clippy on rustc libcore.
Hopefully this will fix it.
changelog: none
Some of the cmp_owned tests emitted non-machine-applicable suggestions,
so I moved them to `tests/ui/cmp_owned/without_suggestion.rs` and added
`// run-rustfix` to the other half.
cc #3630
Fix some suggestions for redundant_pattern_matching
.. and change the Applicability to `MaybeIncorrect`.
Fixes the problem displayed in https://github.com/rust-lang/rust-clippy/issues/4344#issuecomment-519206388.
We now append `{}` to the suggestion so that the conditional has the
correct syntax again.
(If we were to _remove_ the `if` instead, it would trigger the
`unused_must_use` warning for `#[must_use]` types.)
changelog: Fix some suggestions for `redundant_pattern_matching`
Fixes the problem displayed in https://github.com/rust-lang/rust-clippy/issues/4344#issuecomment-519206388.
We now append `{}` to the suggestion so that the conditional has the
correct syntax again.
(If we were to _remove_ the `if` instead, it would trigger the
`unused_must_use` warning for `#[must_use]` types.
Add autofixable suggestion for unseparated integer literal suffixes
changelog: Add autofixable suggestion for unseparated integer literal suffixes
Somewhat WIP, since I haven't been able to get this working when adding `// run-rustfix` to `ui/literals.rs`. I think the issue is that there are multiple suggestions operating on one numerical literal, and I'm not sure what the best approach is to work around that.
Thanks
Update lint deprecation for tool lints
changelog: Allow tool lints (`clippy::*`) to be deprecated
Our lint deprecation previously didn't work for tool lints, because
`register_removed` was registering lints to be removed _without_ the
`clippy` prefix.
Fixes#4349
Our lint deprecation previously didn't work for tool lints, because
`register_removed` was registering lints to be removed _without_ the
`clippy` prefix.
new_ret_no_self: allow Self in inner type for impl Trait return types
Check the inner types of associated types of a trait when checking for Self in the return type of a `new` method. This means that the following will no longer warn:
```rust
trait Trait {
type Inner;
}
struct S;
impl S {
fn new() -> impl Trait<Inner = Option<Self>> {
struct TraitImpl;
impl Trait for TraitImpl {
type Inner = Option<S>;
}
TraitImpl
}
}
```
```rust
#![feature(async_await)]
struct Connection;
impl Connection {
async fn new() -> Result<Self, ()> {
Ok(S)
}
}
```
closes#4359
changelog: fix `new_ret_no_self` lint for async `new` functions.
