Add message format checking
Checks the indentation of diagnostic messages, it checks that everything is at least as indented as the attribute that contains it
r? @jdonszelmann (Anyone else, also feel free to review, just assigning to Jana because she's been reviewing the other PRs)
Implement MVP for opaque generic const arguments
This is meant to be the interim successor to generic const expressions.
Essentially, const item RHS's will be allowed to do arbitrary const
operations using generics. The limitation is that these const items will
be treated opaquely, like ADTs in nominal typing, such that uses of them
will only be equal if the same const item is referenced. In other words,
two const items with the exact same RHS will not be considered equal.
I also added some logic to check feature gates that depend on others
being enabled (like oGCA depending on mGCA).
### Coherence
During coherence, OGCA consts should be normalized ambiguously because
they are opaque but eventually resolved to a real value. We don't want
two OGCAs that have the same value to be treated as distinct for
coherence purposes. (Just like opaque types.)
This actually doesn't work yet because there are pre-existing
fundamental issues with equate relations involving consts that need to
be normalized. The problem is that we normalize only one layer of the
const item and don't actually process the resulting anon const. Normally
the created inference variable should be handled, which in this case
would cause us to hit the anon const, but that's not happening.
Specifically, `visit_const` on `Generalizer` should be updated to be
similar to `visit_ty`.
r? @BoxyUwU
Fix a few diagnostics
When working on the inline diagnostics conversion (https://github.com/rust-lang/rust/issues/151366), I noticed that my script sometimes took the wrong message.
Because it didn't happen very often, I just fixed it manually when a uitest fails.
However I got paranoid that the script changed messages that were not covered by uitests, so I checked for all messages in the previous `messages.ftl` files, whether they occured at least once in the codebase. I found 3 messages that indeed were wrongly replaced by my script, fixed them, and added uitests to make sure this doesn't happen again :)
r? @jdonszelmann (Anyone else, also feel free to review, just assigning to Jana because she's been reviewing the other PRs)
This is meant to be the interim successor to generic const expressions.
Essentially, const item RHS's will be allowed to do arbitrary const
operations using generics. The limitation is that these const items will
be treated opaquely, like ADTs in nominal typing, such that uses of them
will only be equal if the same const item is referenced. In other words,
two const items with the exact same RHS will not be considered equal.
I also added some logic to check feature gates that depend on others
being enabled (like oGCA depending on mGCA).
= Coherence =
During coherence, OGCA consts should be normalized ambiguously because
they are opaque but eventually resolved to a real value. We don't want
two OGCAs that have the same value to be treated as distinct for
coherence purposes. (Just like opaque types.)
This actually doesn't work yet because there are pre-existing
fundamental issues with equate relations involving consts that need to
be normalized. The problem is that we normalize only one layer of the
const item and don't actually process the resulting anon const. Normally
the created inference variable should be handled, which in this case
would cause us to hit the anon const, but that's not happening.
Specifically, `visit_const` on `Generalizer` should be updated to be
similar to `visit_ty`.
Support long diff conflict markers
git can be configured to use more than 7 characters for conflict markers, and jj automatically uses longer conflict markers when the text contains any char sequence that could be confused with conflict markers. Ensure that we only point at markers that are consistent with the start marker's length.
Ensure that we only consider char sequences at the beginning of a line as a diff marker.
Fix https://github.com/rust-lang/rust/issues/150352.
Feature-gate `mut ref` patterns in struct pattern field shorthand
Tracking issue for `mut_ref` (and other parts of Match Ergonomics 2024): https://github.com/rust-lang/rust/issues/123076https://github.com/rust-lang/rust/pull/123080 introduced `mut ref`[^1] patterns (for by-reference bindings where the binding itself is mutable), feature-gated behind the `mut_ref` feature, except for in struct pattern shorthand, where the feature gating was missing. Thus, `mut ref` patterns in struct pattern shorthand has been unintentionally stable for ~18 months (since 1.79.0 ([compiler explorer](https://rust.godbolt.org/z/4WTrvhboT))).
This PR adds feature-gating for `mut ref` patterns in struct pattern shorthand. Since this is reverting an accidental stabilization, this probably needs a crater run and a T-lang FCP?
Some alternative possibilities:
* Do nothing (let the inconsistency exist until `feature(mut_ref)` is stabilized)
* Document the existing behavior
* Do a FCW instead of fully feature-gating
* Stabilize `feature(mut_ref)`
CC https://github.com/rust-lang/rust/pull/123080#issuecomment-3746793632
CC @Nadrieril
[^1]: everything in this description also applies analogously to `mut ref mut` patterns.
fix(parser): Disallow CR in frontmatter
T-lang came back on the stabilization PR (rust-lang/rust#148051) asking for CR to be disallowed
to leave room for all stray CRs to be rejected in the future.
At that point, the test can remain but the implementation can be
removed.
If that plan does not go through, we'll need to re-evaluate
- whether this is more lint-like and should defer to the calling tool
that is managing the frontmatter
- how much Rust should treat the frontmatter as Rust and apply the same
grammar restrictions of "no stray CR" (like raw string literals)
Part of rust-lang/rust#136889
T-lang came back on the stabilization PR asking for CR to be disallowed
to leave room for all stray CRs to be rejected in the future.
At that point, the test can remain but the implementation can be
removed.
If that plan does not go through, we'll need to re-evaluate
- whether this is more lint-like and should defer to the calling tool
that is managing the frontmatter
- how much Rust should treat the frontmatter as Rust and apply the same
grammar restrictions of "no stray CR" (like raw string literals)
From the `parser` module to the `errors` module, which is where most of
its uses are.
This means the `errors` module no longer depends on the `parser` module,
removing a cyclic dependency between the two modules.
Every diagnostic struct in `rustc_parse` is in the `errors` module,
except for `ForbiddenLetReason` and `MisspelledKw`. There's no good
reason for this, and presumably it is just an accidental inconsistency.
This commit moves them into `errors`.
Based on earlier work by León Orell Valerian Liehr.
Co-authored-by: León Orell Valerian Liehr <me@fmease.dev>
Signed-off-by: Usman Akinyemi <uniqueusman@archlinux>
`const` blocks as a `mod` item
Tracking issue: rust-lang/rust#149226
This adds support for writing `const { ... }` as an item in a module. In the current implementation, this is a unique AST item that gets lowered to `const _: () = const { ... };` in HIR.
rustfmt support included.
TODO:
- `pub const { ... }` does not make sense (see rust-lang/rust#147136). Reject it. Should this be rejected by the parser or smth?
- Improve diagnostics (preferably they should not mention the fake `_` ident).
Do not recover from `Trait()` if generic list is unterminated
If we encounter `fn foo<T: Trait()`, the recovery logic would it as if `Trait` was intended to use the Fn-like trait syntax, but if we don't know for certain that we've parsed a full trait bound (`fn foo<T: Trait()>`), we bail from the recovery as more likely there could have been a missing closing `>` and the `(` corresponds to the start of the fn parameter list.
Fixrust-lang/rust#141436.
If we encounter `fn foo<T: Trait()`, the recovery logic would it as if `Trait` was intended to use the Fn-like trait syntax, but if we don't know for certain that we've parsed a full trait bound (`fn foo<T: Trait()>`), we bail from the recovery as more likely there could have been a missing closing `>` and the `(` corresponds to the start of the fn parameter list.
Recover parse gracefully from `<const N>`
When a const param doesn't have a `: Type`, recover the parser state and provide a structured suggestion. This not only provides guidance on what was missing, but it also makes subsuequent errors to be emitted that would otherwise be silenced.
```
error: expected `:`, found `>`
--> $DIR/incorrect-const-param.rs:26:16
|
LL | impl<T, const N> From<[T; N]> for VecWrapper<T>
| ^ expected `:`
|
help: you might have meant to write the type of the const parameter here
|
LL | impl<T, const N: /* Type */> From<[T; N]> for VecWrapper<T>
| ++++++++++++
```
r? @fmease
Follow up to rust-lang/rust#151077. Fixrust-lang/rust#84327.
Don't try to recover keyword as non-keyword identifier
Fixesrust-lang/rust#149692.
On beta after rust-lang/rust#146978, we ICE on
```rs
macro_rules! m {
($id:item()) => {}
}
m!(Self());
```
where `Self` in the macro invocation is a keyword not a "normal" identifier, while attempting to recover an missing keyword before an identifier. Except, `Self` *is* a keyword, so trying to parse that as a non-reserved identifier expectedly fails.
I suspect rust-lang/rust#146978 merely unmasked a possible code path to hit this case; this logic has been so for a good while. Previously, on stable, the error message looks something like
```rs
error: expected identifier, found keyword `Self`
--> src/lib.rs:5:4
|
5 | m!(Self());
| ^^^^ expected identifier, found keyword
error: missing `fn` or `struct` for function or struct definition
--> src/lib.rs:5:4
|
2 | ($id:item()) => {}
| -------- while parsing argument for this `item` macro fragment
...
5 | m!(Self());
| ^^^^
|
help: if you meant to call a macro, try
|
5 | m!(Self!());
| +
```
I considered restoring this diagnostic, but I'm not super convinced it's worth the complexity (and to me, it's not super clear what the user actually intended here).
When a const param doesn't have a `: Type`, recover the parser state and provide a structured suggestion. This not only provides guidance on what was missing, but it also makes subsuequent errors to be emitted that would otherwise be silenced.
```
error: expected `:`, found `>`
--> $DIR/incorrect-const-param.rs:26:16
|
LL | impl<T, const N> From<[T; N]> for VecWrapper<T>
| ^ expected `:`
|
help: you might have meant to write the type of the const parameter here
|
LL | impl<T, const N: /* Type */> From<[T; N]> for VecWrapper<T>
| ++++++++++++
```
feat: invisible character help string
I was playing around with zero width spaces in different programming languages and thought that this error message could be more helpful. Hopefully it's a good first contribution! :)
