**What it does:** Checks for generics with `std::ops::Drop` as bounds.
**Why is this bad?** `Drop` bounds do not really accomplish anything.
A type may have compiler-generated drop glue without implementing the
`Drop` trait itself. The `Drop` trait also only has one method,
`Drop::drop`, and that function is by fiat not callable in user code.
So there is really no use case for using `Drop` in trait bounds.
**Known problems:** None.
**Example:**
```rust
fn foo<T: Drop>() {}
```
Fix ICE in needless_pass_by_value lint
If I understand it correctly, we were first creating a type with a
`RegionKind::ReErased` region and then deleted it again in
`util::implements_trait` with:
cx.tcx.erase_regions(&ty);
causing the type query to fail.
It looks like using `ReEmpty` works around that deletion.
Fixes#3144
Macro check for assertion_on_constants lint
The `assertion_on_constants` lint currently has following output for this code [Playground](https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=6f2c9df6fc50baf847212d3b5136ee97):
```rust
macro_rules! assert_const {
($len:expr) => {
assert!($len > 0);
}
}
fn main() {
assert_const!(3);
assert_const!(-1);
}
```
```
warning: assert!(const: true) will be optimized out by the compiler
--> src/main.rs:3:9
|
3 | assert!($len > 0);
| ^^^^^^^^^^^^^^^^^^
...
8 | assert_const!(3);
| ---------------- in this macro invocation
|
= note: #[warn(clippy::assertions_on_constants)] on by default
= help: remove it
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#assertions_on_constants
warning: assert!(const: false) should probably be replaced
--> src/main.rs:3:9
|
3 | assert!($len > 0);
| ^^^^^^^^^^^^^^^^^^
...
9 | assert_const!(-1);
| ----------------- in this macro invocation
|
= help: use panic!() or unreachable!()
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#assertions_on_constants
```
This is contradictory. This lint should not trigger if the `assert!` is in a macro itself.
Add a uitest subcommand to simplify UI test invocation
This makes running single tests a lot easier.
It's now
`TESTNAME=xxx cargo uitest`
instead of
`TESTNAME=xxx cargo test --test compile-test`
Start making clippy easier to invoke in non-cargo contexts
Clippy (clippy-driver) currently has a couple of strong but unnecessary couplings with cargo. This series:
1. makes detection of check builds more robust, and
2. make clippy-driver use the --sysroot specified on the command line as its internal sysroot.
If I understand it correctly, we were first creating a type with a
`RegionKind::ReErased` region and then deleted it again in
`util::implements_trait` with:
cx.tcx.erase_regions(&ty);
causing the type query to fail.
It looks like using `ReEmpty` works around that deletion.
Fix `cast_sign_loss` false positive
This checks if the value is a non-negative constant before linting about
losing the sign.
Because the `constant` function doesn't handle const functions, we check if
the value is from a call to a `max_value` function directly. A utility method
called `get_def_path` was added to make checking for the function paths
easier.
Fixes#2728
The rustc change added HirId to a few nodes. As I understand it, the plan is
to remove the NodeId from these nodes eventually. Where the NodeId was
not being matched, I used `..` to try and avoid further breakage. Where it
was, I used `_` to make the fix easier when NodeId is removed.
Adding lint test for excessive LOC.
This is a WIP for #2377. Just wanted to pull in because I had a few questions:
1. Is it okay that I'm approaching this via counting by looking at each line in the snippet instead of looking at the AST tree? If there's another way to do it, I want to make sure I'm doing the correct way, but I wasn't sure since the output AST JSON doesn't seem to contain whitespace.
2. My function is definitely going to trigger the lint, so also wanted to see if there was something obvious I could do to reduce it.
3. Are the two tests fine, or is there something obvious I'm missing?
4. Obviously bigger question - am I approaching the line count correctly. Current strategy is count a line if it contains some code, so skip if it's just comments or empty.
`hir::Ty` doesn't seem to know anything about type bounds and
`cx.tcx.type_of(def_id)` caused an ICE when it was passed a generic type
with a bound:
```
src/librustc_typeck/collect.rs:1311: unexpected non-type Node::GenericParam: Type { default: None, synthetic: None }
```
Converting it to a proper `Ty` fixes the ICE and catches a few more
places where the lint applies.
This checks if the value is a non-negative constant before linting about
losing the sign.
Because the `constant` function doesn't handle const functions, we check if
the value is from a call to a `max_value` function directly. A utility method
called `get_def_path` was added to make checking for the function paths
easier.
Fixes#2728
Add initial version of const_fn lint
This adds an initial version of a lint that can tell if a function could be `const`.
TODO:
- [x] Finish up the docs
- [x] Fix the ICE
cc #2440
Prevent incorrect cast_lossless suggestion in const_fn
`::from` is not a const fn, so applying the suggestion of
`cast_lossless` would fail to compile. The fix is to skip the lint if
the cast is found inside a const fn.
Fixes#3656
Fix documentation for `slow_vector_initialization`
This PR fixes the documentation for the lint `slow_vector_initialization`. The documentation recommended writing `vec![len; 0]` but the correct solution is `vec![0; len]`.
for file in `fd \.rs$` ; do sed -i s/span_suggestion_with_applicability/span_suggestion/g $file ; done
for file in `fd \.rs$` ; do sed -i s/span_suggestion_short_with_applicability/span_suggestion_short/g $file ; done
for file in `fd \.rs$` ; do sed -i s/span_suggestions_with_applicability/span_suggestions/g $file ; done
Fix `expect_fun_call` lint suggestions
This commit corrects some bad suggestions produced by the
`expect_fun_call` lint and enables `rust-fix` checking on the tests.
Addresses #3630
`::from` is not a const fn, so applying the suggestion of
`cast_lossless` would fail to compile. The fix is to skip the lint if
the cast is found inside a const fn.
* master: (58 commits)
Rustfmt all the things
Don't make decisions on values that don't represent the decision
Improving comments.
Rustup
Added rustfix to the test.
Improve span shortening.
Added "make_return" and "blockify" convenience methods in Sugg and used them in "needless_bool".
Actually check for constants.
Fixed potential mistakes with nesting. Added tests.
formatting fix
Update clippy_lints/src/needless_bool.rs
formatting fix
Fixing typo in CONTRIBUTING.md
Fix breakage due to rust-lang/rust#57651
needless bool lint suggestion is wrapped in brackets if it is an "else" clause of an "if-else" statement
Fix automatic suggestion on `use_self`.
Remove negative integer literal checks.
Fix `implicit_return` false positives.
Run rustfmt
Fixed breakage due to rust-lang/rust#57489
...
Fix automatic suggestion on `use_self`.
In an example like this:
```rust
impl Example {
fn fun_1() { }
fn fun_2() {
Example::fun_1();
}
}
```
Clippy tries to replace `Example::fun_1` with `Self`, loosing `::fun_1` in the process, it should rather try to replace `Example` with `Self`.
**Question**
- There may be other paths that need the same treatment, but I'm not sure I understand them fully:
- e648adf086/clippy_lints/src/use_self.rs (L94-L96)
- e648adf086/clippy_lints/src/use_self.rs (L225-L229)
Fix `implicit_return` false positives.
Fixes the following false positives:
- linting on `if let` without `else` in a `loop` even with a present `return`
- linting on `unreachable!()`
Catch up with `format_args` change
Catches up with a change in rust-lang/rust#57537. (Since the optimization is optional, this clippy PR can be merged before the rustc PR.)
Happened to fix a bug in `expect_fun_call`, that is the lint ignores more than
one arguments to `format`.
```
warning: use of `expect` followed by a function call
--> src/main.rs:2:17
|
2 | Some("foo").expect(format!("{} {}", 1, 2).as_ref());
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ help: try this: `unwrap_or_else(|| panic!("{} {}", 1))`
|
```
Catches up with a change in rust-lang/rust#57537
Happened to fix a bug in `expect_fun_call`, that is the lint ignores more than
one arguments to `format`.
`cloned` requires that the elements of the iterator must be references. This
change determines if that is the case by examining the type of the closure
argument and suggesting `.cloned` only if it is a reference. When the closure
argument is not a reference, it suggests removing the `map` call instead.
A minor problem with this change is that the new check sometimes overlaps
with the `clone_on_copy` lint.
Fixes#498
Add several run rustfix annotations
Adds `run-rustfix` to 18 of the tests from the tracking issue #3630.
Each test has its own commit, to make reviewing easier (hopefully this is easier to review than 18 separate PRs).
## Changes
- `cfg_attr_rustfmt`: Custom inner attributes are unstable. Let's disable the lint for inner attributes until [#54726](https://github.com/rust-lang/rust/issues/54726) stabilizes
- `collapsible_if`: unrelated cyclomatic_complexity warning that can be ignored
- `duration_subsec`: Simply needed `#![allow(dead_code)]`
- `excessive_precision`: Fixed by `#!allow(dead_code,unused_variables)`
- `explicit_write`: Fixed by `#![allow(unused_imports)]`
- `inconsistent_digit_grouping`: Avoid triggering `clippy::excessive_precision` lint
- `infallible_destructuring_match`: Fixed by `#![allow(dead_code, unreachable_code, unused_variables)]`
- `into_iter_on_ref`: Triggered unrelated `clippy::useless_vec` lint
- `large_digit_groups`: Avoid triggering `clippy::excessive_precision` lint
- `map_clone`: Fixed by `#![allow(clippy::iter_cloned_collect)]`
- `mem_replace`: Suggestion causes import to be unused, fixed by `#![allow(unused_imports)]`
- `precedence`: Allow some unrelated lints, and change out-of-range `0b1111_1111i8` literal
- `redundant_field_names`: Allow dead code, and remove stabilized feature toggles
- `replace_consts`: Fixed by `#![allow(unused_variables)]`
- `starts_ends_with`: Fixed by `#![allow(unused_must_use)]`
- `types`: Fixed by `#![allow(dead_code, unused_variables)]`
- `unit_arg`: Fixed by `#[allow(unused_must_use)]`
- `unnecessary_fold`: Fixed by adding type annotations and adding `#![allow(dead_code)]`
Restrict `use_self` on nested items
Fixes#3637Fixes#3463
These changes make it so that nested items aren't visited any more by the `use_self` lint.
I think visiting nested items should be possible (so that it uses a different `item_path` for the nested item), but I'm not sure whether it's viable and what the best approach would be.
- Can `item_path` be changed to a new `Self` path before visiting the item, and then changing it back afterwards?
- Alternatively, could a new visitor be created, re-using `check_trait_method_impl_decl`?
cast_ref_to_mut lint
I see this pattern way too often, and it's completely wrong. In fact, due to how common this incorrect pattern is, [the Rustonomicon specifically points this out](https://doc.rust-lang.org/nomicon/transmutes.html).
> - Transmuting an & to &mut is UB
> - Transmuting an & to &mut is always UB
> - No you can't do it
> - No you're not special
This is my first lint.
Trigger `use_self` lint in local macros
Closes#2098
The test currently only covers local macros. #2098 suggested this:
> You could add the macro in question into the `mini_macro` subcrate
But that doesn't work for a `macro_rules`:
```
error: cannot export macro_rules! macros from a `proc-macro` crate type currently
```
So I suggest leaving out the test for external macros, as using `in_external_macro` seems straigtforward enough. Alternatives would be to use to add an additional crate (overkill if you ask me), or test with a `proc-macro`.
Integrate rustfix into Clippy test suite
Once the [PR to compiletest-rs](https://github.com/laumann/compiletest-rs/pull/151) is reviewed and merged this fixes#2376.
I will create a separate tracking issue for adding `run-rustfix` to all tests.
Only trigger `infinite_iter` lint for infinitely allocating `collect()` calls
Fixes #3538
~Oh, I guess this should actually check other methods like `count` as well, not only `collect()`.~
Never mind, `collect` is the only of these functions that allocates a data structure.
Fix if_same_then_else false positive
This fixes false positive in #3559.
The problem was that `SpanlessEq` does not check patterns in declarations. So this two blocks considered equal.
```rust
if true {
let (x, y) = foo();
} else {
let (y, x) = foo();
}
```
Not sure if the proposed change is safe as `SpanlessEq` is used extensively in other lints, but I tried hard to come up with counterexample and failed.
Merge new_without_default_derive into new_without_default
Closes#3525, deprecating new_without_default_derive and moving both lints into new_without_default.
Fix false positives for `implicit_return` and `empty_loop` on macro expansion.
This PR only fixes `implicit_return` and `empty_loop`.
But I suspect this bug may affect a lot of other lints.
Teach `suspicious_else_formatting` about `if .. {..} {..}`
We essentially treat bare blocks `{..}` identically to `if .. {..}`, except for different lint messages.
Fixes#3044
Implements lint for order comparisons against bool (#3438)
As described on issue #3438, this change implements linting for `>` and `<` comparisons against both `boolean` literals and between expressions.
Adds lint for Vec<Box<T: Sized>>
This adds, and subsequently closes#3530. This is the first time I've ever worked with anything remotely close to internal Rust code, so I'm very much unsure about the if_chain! to figure this out!
I can't get rustfmt working on WSL with nightly 2018-12-07:
`error: component 'rustfmt' for target 'x86_64-unknown-linux-gnu' is unavailable for download`
Lint redundant clone of fields
Makes `redundant_clone` warn on unnecessary `foo.field.clone()` sometimes (it can detect an unnecessary clone only if the base of projection, `foo` in this case, is not used at all after that). This is enough for cases like `returns_tuple().0.clone()`.
If there are more than one such assignment, the last one may be
the one supplied to `clone` method.
Makes `find_stmt_assigns_to` internally reverses the iterator to make
the intent to "iterate statements backward" clear.
The path of `libc::c_void` has changes in 5c1a6b8a6d
The DefId path is now always platform specific like
`libc::windows::c_void`. This fixes our c_void detection to only check
the first and last elements.
This now only checks for wildcard_dependencies if the source is a
non-git source.
I tried adding a compiletest suite for the cargo lints, but I was unable
to override the `Cargo.toml` of the original executable.
I tested this manually by modifying the main `Cargo.toml`.
Fixes#3458
I believe if the user already decided to put underscores in their
literal, Clippy should be willing to believe that they put a number of
underscores that they felt was readable.
This lint looks for:
let mut vec = Vec::with_capacity(len);
vec.set_len(len);
The suggested replacement is `vec![0; len]`.
This is far too opinionated to be a deny-by-default lint because the performance
characteristics of the suggested replacement are totally different.
I am not convinced that this lint has value beyond what deny(unsafe_code) gives
you. Unsafe code is unsafe but please don't deny-by-default lint it if that's
the only reason.
Warning was:
warning: the feature `macro_at_most_once_rep` has been stable since 1.32.0 and no longer requires an attribute to enable
--> clippy_lints/src/lib.rs:19:12
|
19 | #![feature(macro_at_most_once_rep)]
| ^^^^^^^^^^^^^^^^^^^^^^
|
= note: #[warn(stable_features)] on by default
Some bugs and some documentation is unrelated to the Applicability change, but
these bugs were serious and the documentation was kind of required to
understand what's going on.
Instead of searching for all the successive expressions after a vector
allocation, check only the first expression.
This is done to minimize the amount of false positives of the lint.
Add lint to detect slow zero-filled vector initialization. It detects
when a vector is zero-filled with extended with `repeat(0).take(len)`
or `resize(len, 0)`.
This zero-fillings are usually slower than simply using `vec![0; len]`.
I noticed that I suppress this lint in many of my projects.
https://github.com/search?q=needless_pass_by_value+user%3Adtolnay&type=Codehttps://github.com/search?q=needless_pass_by_value+user%3Aserde-rs&type=Code
Upon further inspection, this lint has a *long* history of false
positives (and several remaining).
Generally I feel that this lint is the definition of pedantic and should
not be linted by default.
#[derive(Debug)]
enum How {
ThisWay,
ThatWay,
}
// Are we really better off forcing the call sites to write f(&_)...?
fn f(how: How) {
println!("You want to do it {:?}", how);
}
fn main() {
f(How::ThatWay);
}
