Implement diagnostics in all places left: generics (predicates, defaults, const params' types), fields, and type aliases.
Unfortunately this results in a 20mb addition in `analysis-stats .` due to many type methods returning an addition diagnostics result now (even if it's `None` in most cases). I'm not sure if this can be improved.
An alternative strategy that can prevent the memory usage growth is to never produce diagnostics in hir-ty methods. Instead, lower all types in the hir crate when computing diagnostics from scratch (with diagnostics this time). But this has two serious disadvantages:
1. This can cause code duplication (although it can probably be not that bad, it will still mean a lot more code).
2. I believe we eventually want to compute diagnostics for the *entire* workspace (either on-type or on-save or something alike), so users can know when they have diagnostics even in inactive files. Choosing this approach will mean we lose all precomputed salsa queries. For one file this is fine, for the whole workspace this will be very slow.
The diagnostic implemented is a simple one (E0109). It serves as a test for the new foundation.
This commit only implements diagnostics for type in bodies and body-carrying signatures; the next commit will include diagnostics in the rest of the things.
Also fix one weird bug that was detected when implementing this that caused `Fn::(A, B) -> C` (which is a valid, if bizarre, alternative syntax to `Fn(A, B) -> C` to lower incorrectly.
And also fix a maybe-bug where parentheses were sneaked into a code string needlessly; this was not detected until now because the parentheses were removed (by the make-AST family API), but with a change in this commit they are now inserted. So fix that too.
We add union fields access (in both expressions and patterns) and inline assembly.
That completes the unsafe check (there are some other unsafe things but they are unstable), and so also opens the door to reporting unused unsafe without annoying people about their not-unused unsafe blocks.
Only in calls, because to support them in bounds we need support from Chalk. However we don't yet report error from bounds anyway, so this is less severe.
The returned future is shown in its name within inlay hints instead of as a nicer `impl Future`, but that can wait for another PR.
I.e. the following situation:
```
fn foo() {
mod bar {
fn qux() {
// Prelude path here (e.g. macro use prelude or extern prelude).
}
}
}
```
Those were previously unresolved, because, in order to support `self` and `super` properly, since #15148 we do not ascend block paths when there is a module in between, but only crate def maps register preludes, not block def maps, and we can't change this because block def map prelude can always be overridden by another block. E.g.
```
fn foo() {
struct WithTheSameNameAsPreludeItem;
{
WithTheSameNameAsPreludeItem
}
}
```
Here `WithTheSameNameAsPreludeItem` refer to the item from the top block, but if we would register prelude items in each block the child block would overwrite it incorrectly.
E.g.:
```rust
let v;
macro_rules! m { () => { v }; }
```
This was an existing bug, but it was less severe because unless the variable was shadowed it would be correctly resolved. With hygiene however, without this fix the variable is never resolved.
And few more fixups.
I was worried this will lead to more memory usage since `ExprOrPatId` is double the size of `ExprId`, but this does not regress `analysis-stats .`. If this turns out to be a problem, we can easily use the high bit to encode this information.
Instead of lowering them to `<expr> = <expr>`, then hacking on-demand to resolve them, we lower them to `<pat> = <expr>`, and use the pattern infrastructure to handle them. It turns out, destructuring assignments are surprisingly similar to pattern bindings, and so only minor modifications are needed.
This fixes few bugs that arose because of the non-uniform handling (for example, MIR lowering not handling slice and record patterns, and closure capture calculation not handling destructuring assignments at all), and furthermore, guarantees we won't have such bugs in the future, since the programmer will always have to explicitly handle `Expr::Assignment`.
Tests don't pass yet; that's because the generated patterns do not exist in the source map. The next commit will fix that.
Because our lint infra *can* handle allows from within macro expansions!
(Also, what did this reason have to do with something that is a hard error and not a lint? I'm puzzled).
I wonder how many such diagnostics we have...
Maybe that also mean we can change `unused_mut` to no-longer-experimental? But this is a change I'm afraid to do without checking.
fix: Do not consider mutable usage of deref to `*mut T` as deref_mut
Fixes#15799
We are doing some heuristics for deciding whether the given deref is deref or deref_mut here;
5982d9c420/crates/hir-ty/src/infer/mutability.rs (L182-L200)
But this heuristic is erroneous if we are dereferencing to a mut ptr and normally those cases are filtered out here as builtin;
5982d9c420/crates/hir-ty/src/mir/lower/as_place.rs (L165-L177)
Howerver, this works not so well if the given dereferencing is double dereferencings like the case in the #15799.
```rust
struct WrapPtr(*mut u32);
impl core::ops::Deref for WrapPtr {
type Target = *mut u32;
fn deref(&self) -> &Self::Target {
&self.0
}
}
fn main() {
let mut x = 0u32;
let wrap = WrapPtr(&mut x);
unsafe {
**wrap = 6;
}
}
```
Here are two - outer and inner - dereferences here, and the outer dereference is marked as deref_mut because there is an assignment operation.
And this deref_mut marking is propagated into the inner dereferencing.
In the later MIR lowering, the outer dereference is filtered out as it's expr type is `*mut u32`, but the expr type in the inner dereference is an ADT, so this false-mutablility is not filtered out.
This PR cuts propagation of this false mutablilty chain if the expr type is mut ptr.
Since this happens before the resolve_all, it may have some limitations when the expr type is determined as mut ptr at the very end of inferencing, but I couldn't find simple fix for it 🤔
This check is incorrect when we have comments and whitespace in the text.
We can strip comments, but then we still have whitespace, which we cannot strip without changing meaning for the parser. So instead I opt to remove the check, and wrap the expression in parentheses (asserting what produced is a parenthesized expression) to strengthen verification.
fix: Handle errors and lints from external macros
Some lints should not be reported if they originate from an external macro, and quickfixes should be disabled (or they'll change library code).
Fixes#18122.
Closes#18124.
Don't lint names of #[no_mangle] extern fns
[Rust doesn't run the `non_snake_case_name` lint on `extern fn`s with the `#[no_mangle]` attribute](https://github.com/rust-lang/rust/pull/44966).
The conditions are:
- The function must be `extern` and have a `#[no_mangle]` attribute.
- The function's ABI must not be explicitly set to "Rust".
This PR replicates that logic here.
Use more correct handling of lint attributes
The previous analysis was top-down, and worked on a single file (expanding macros). The new analysis is bottom-up, starting from the diagnostics and climbing up the syntax and module tree.
While this is more efficient (and in fact, efficiency was the motivating reason to work on this), unfortunately the code was already fast enough. But luckily, it also fixes a correctness problem: outline parent modules' attributes were not respected for the previous analysis. Case lints specifically did their own analysis to accommodate that, but it was limited to only them. The new analysis works on all kinds of lints, present and future.
It was basically impossible to fix the old analysis without rewriting it because navigating the module hierarchy must come bottom-up, and if we already have a bottom-up analysis (including syntax analysis because modules can be nested in other syntax elements, including macros), it makes sense to use only this kind of analysis.
Few other bugs (not fundamental to the previous analysis) are also fixed, e.g. overwriting of lint levels (i.e. `#[allow(lint)] mod foo { #[warn(lint)] mod bar; }`.
After this PR is merged I intend to work on an editor command that does workspace-wide diagnostics analysis (that is, `rust-analyzer diagnostics` but from your editor and without having to spawn a new process, which will have to analyze the workspace from scratch). This can be useful to users who do not want to enable check on save because of its overhead, but want to see workspace wide diagnostics from r-a (or to maintainers of rust-analyzer).
Closes#18086.
Closes#18081.
Fixes#18056.
The previous analysis was top-down, and worked on a single file (expanding macros). The new analysis is bottom-up, starting from the diagnostics and climbing up the syntax and module tree.
While this is more efficient (and in fact, efficiency was the motivating reason to work on this), unfortunately the code was already fast enough. But luckily, it also fixes a correctness problem: outline parent modules' attributes were not respected for the previous analysis. Case lints specifically did their own analysis to accommodate that, but it was limited to only them. The new analysis works on all kinds of lints, present and future.
It was basically impossible to fix the old analysis without rewriting it because navigating the module hierarchy must come bottom-up, and if we already have a bottom-up analysis (including syntax analysis because modules can be nested in other syntax elements, including macros), it makes sense to use only this kind of analysis.
Few other bugs (not fundamental ti the previous analysis) are also fixed, e.g. overwriting of lint levels (i.e. `#[allow(lint)] mod foo { #[warn(lint)] mod bar; }`.
fix: Properly account for editions in names
This PR touches a lot of parts. But the main changes are changing `hir_expand::Name` to be raw edition-dependently and only when necessary (unrelated to how the user originally wrote the identifier), and changing `is_keyword()` and `is_raw_identifier()` to be edition-aware (this was done in #17896, but the FIXMEs were fixed here).
It is possible that I missed some cases, but most IDE parts should properly escape (or not escape) identifiers now.
The rules of thumb are:
- If we show the identifier to the user, its rawness should be determined by the edition of the edited crate. This is nice for IDE features, but really important for changes we insert to the source code.
- For tests, I chose `Edition::CURRENT` (so we only have to (maybe) update tests when an edition becomes stable, to avoid churn).
- For debugging tools (helper methods and logs), I used `Edition::LATEST`.
Reviewing notes:
This is a really big PR but most of it is mechanical translation. I changed `Name` displayers to require an edition, and followed the compiler errors. Most methods just propagate the edition requirement. The interesting cases are mostly in `ide-assists`, as sometimes the correct crate to fetch the edition from requires awareness (there may be two). `ide-completions` and `ide-diagnostics` were solved pretty easily by introducing an edition field to their context. `ide` contains many features, for most of them it was propagated to the top level function and there the edition was fetched based on the file.
I also fixed all FIXMEs from #17896. Some required introducing an edition parameter (usually not for many methods after the changes to `Name`), some were changed to a new method `is_any_identifier()` because they really want any possible keyword.
Fixes#17895.
Fixes#17774.
