Migrate ast lowering to session diagnostic
I migrated the whole rustc_ast_lowering crate to session diagnostic *except* the for the use of `span_fatal` at /compiler/rustc_ast_lowering/src/expr.rs#L1268 because `#[fatal(...)]` is not yet supported (see https://github.com/rust-lang/rust/pull/100694).
In some places we use `Vec<Attribute>` and some places we use
`ThinVec<Attribute>` (a.k.a. `AttrVec`). This results in various points
where we have to convert between `Vec` and `ThinVec`.
This commit changes the places that use `Vec<Attribute>` to use
`AttrVec`. A lot of this is mechanical and boring, but there are
some interesting parts:
- It adds a few new methods to `ThinVec`.
- It implements `MapInPlace` for `ThinVec`, and introduces a macro to
avoid the repetition of this trait for `Vec`, `SmallVec`, and
`ThinVec`.
Overall, it makes the code a little nicer, and has little effect on
performance. But it is a precursor to removing
`rustc_data_structures::thin_vec::ThinVec` and replacing it with
`thin_vec::ThinVec`, which is implemented more efficiently.
Rollup of 6 pull requests
Successful merges:
- #100338 (when there are 3 or more return statements in the loop)
- #100384 (Add support for generating unique profraw files by default when using `-C instrument-coverage`)
- #100460 (Update the minimum external LLVM to 13)
- #100567 (Add missing closing quote)
- #100590 (Suggest adding an array length if possible)
- #100600 (Rename Machine memory hooks to suggest when they run)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
- Rename `ast::Lit::token` as `ast::Lit::token_lit`, because its type is
`token::Lit`, which is not a token. (This has been confusing me for a
long time.)
reasonable because we have an `ast::token::Lit` inside an `ast::Lit`.
- Rename `LitKind::{from,to}_lit_token` as
`LitKind::{from,to}_token_lit`, to match the above change and
`token::Lit`.
Remove manual implementations of HashStable for hir::Expr and hir::Ty.
We do not need to force hashing HIR bodies inside those nodes. The contents of bodies are not accessible from the `hir_owner` query which used `hash_without_bodies`. When the content of a body is required, the access is still done using `hir_owner_nodes`, which continues hashing HIR bodies.
Visit attributes in more places.
This adds 3 loosely related changes (I can split PRs if desired):
- Attribute checking on pattern struct fields.
- Attribute checking on struct expression fields.
- Lint level visiting on pattern struct fields, struct expression fields, and generic parameters.
There are still some lints which ignore lint levels in various positions. This is a consequence of how the lints themselves are implemented. For example, lint levels on associated consts don't work with `unused_braces`.
This was incorrectly inserting the ExprField as a sibling of the struct
expression.
This required adjusting various parts which were looking at parent node
of a field expression to find the struct.
This helps simplify the code. It also fixes it to use the correct parent
when lowering. One consequence is the `non_snake_case` lint needed
to change the way it looked for parent nodes in a struct pattern.
This also includes a small fix to use the correct `Target` for
expression field attribute validation.
Attributes on struct expression fields were not being checked for
validity. This adds the fields as HIR nodes so that `CheckAttrVisitor`
can visit those nodes to check their attributes.
Attributes on pattern struct fields were not being checked for validity.
This adds the fields as HIR nodes so that the `CheckAttrVisitor` can
visit those nodes to check their attributes.
Resolve function lifetime elision on the AST
~Based on https://github.com/rust-lang/rust/pull/97720~
Lifetime elision for functions is purely syntactic in nature, so can be resolved on the AST.
This PR replicates the elision logic and diagnostics on the AST, and replaces HIR-based resolution by a `delay_span_bug`.
This refactor allows for more consistent diagnostics, which don't have to guess the original code from HIR.
r? `@petrochenkov`
PR #5956 started checking the stability of path segments.
However, this was not applied to 'use tree' items
(e.g. 'use some::path::{ItemOne, ItemTwo}') due to the way
that we desugar these items in HIR lowering.
This PR modifies 'use tree' lowering to preserve resolution
information, which is needed by stability checking.
Implement `for<>` lifetime binder for closures
This PR implements RFC 3216 ([TI](https://github.com/rust-lang/rust/issues/97362)) and allows code like the following:
```rust
let _f = for<'a, 'b> |a: &'a A, b: &'b B| -> &'b C { b.c(a) };
// ^^^^^^^^^^^--- new!
```
cc ``@Aaron1011`` ``@cjgillot``
Create fresh lifetime parameters for bare fn trait too
The current code fails to account for the equivalence between `dyn FnMut(&mut u8)` and bare `FnMut(&mut u8)`, and treated them differently.
This PR introduces a special case for `Fn` traits, which are always fully resolved.
Fixes#98616Fixes#98726
This will require a beta-backport, as beta contains that bug.
r? `@petrochenkov`
Make lowering a query
Split from https://github.com/rust-lang/rust/pull/88186.
This PR refactors the relationship between lowering and the resolver outputs in order to make lowering itself a query.
In a first part, lowering is changed to avoid modifying resolver outputs, by maintaining its own data structures for creating new `NodeId`s and so.
Then, the `TyCtxt` is modified to allow creating new `LocalDefId`s from inside it. This is done by:
- enclosing `Definitions` in a lock, so as to allow modification;
- creating a query `register_def` whose purpose is to declare a `LocalDefId` to the query system.
See `TyCtxt::create_def` and `TyCtxt::iter_local_def_id` for more detailed explanations of the design.
Refactor path segment parameter error
This PR attempts to rewrite the error handling for an unexpected parenthesised type parameters to:
- Use provided data instead of re-parsing the whole span
- Add a multipart suggestion to reflect on the changes with an underline
- Remove the unnecessary "if" nesting
Make `ExprKind::Closure` a struct variant.
Simple refactor since we both need it to introduce additional fields in `ExprKind::Closure`.
r? ``@Aaron1011``
Remove label/lifetime shadowing warnings
This PR removes some pre-1.0 shadowing warnings for labels and lifetimes.
The current behaviour of the compiler is to warn
* labels that shadow unrelated labels in the same function --> removed
```rust
'a: loop {}
'a: loop {} // STOP WARNING
```
* labels that shadow enclosing labels --> kept, but only if shadowing is hygienic
```rust
'a: loop {
'a: loop {} // KEEP WARNING
}
```
* labels that shadow lifetime --> removed
```rust
fn foo<'a>() {
'a: loop {} // STOP WARNING
}
```
* lifetimes that shadow labels --> removed
```rust
'a: loop {
let b = Box::new(|x: &i8| *x) as Box<dyn for <'a> Fn(&'a i8) -> i8>; // STOP WARNING
}
```
* lifetimes that shadow lifetimes --> kept
```rust
fn foo<'a>() {
let b = Box::new(|x: &i8| *x) as Box<dyn for <'a> Fn(&'a i8) -> i8>; // KEEP WARNING
}
```
Closes https://github.com/rust-lang/rust/issues/31745.
-----
From `@petrochenkov` in https://github.com/rust-lang/rust/pull/95781#issuecomment-1105199014
> I think we should remove these silly checks entirely.
> They were introduced long time ago in case some new language features appear and require this space.
> Now we have another mechanism for such language changes - editions, and if "lifetimes in expressions" or something like that needs to be introduced it could be introduced as an edition change.
> However, there was no plans to introduce anything like for years, so it's unlikely that even the edition mechanism will be necessary.
r? rust-lang/lang
This commit adds an alternative content boxing syntax,
and uses it inside alloc.
The usage inside the very performance relevant code in
liballoc is the only remaining relevant usage of box syntax
in the compiler (outside of tests, which are comparatively
easy to port).
box syntax was originally designed to be used by all Rust
developers. This introduces a replacement syntax more tailored
to only being used inside the Rust compiler, and with it,
lays the groundwork for eventually removing box syntax.
Track if a where bound comes from a impl Trait desugar
With https://github.com/rust-lang/rust/pull/93803 `impl Trait` function arguments get desugared to hidden where bounds. However, Clippy needs to know if a bound was originally a `impl Trait` or an actual bound. This adds a field to the `WhereBoundPredicate` struct to keep track of this information during AST->HIR lowering.
r? `@cjgillot`
cc `@estebank` (as the reviewer of #93803)
With #93803 `impl Trait` function arguments get desugared to hidden
where bounds. However, Clippy needs to know if a bound was originally a
impl Trait or an actual bound. This adds a field to the
`WhereBoundPredicate` struct to keep track of this information during
HIR lowering.
Begin fixing all the broken doctests in `compiler/`
Begins to fix#95994.
All of them pass now but 24 of them I've marked with `ignore HELP (<explanation>)` (asking for help) as I'm unsure how to get them to work / if we should leave them as they are.
There are also a few that I marked `ignore` that could maybe be made to work but seem less important.
Each `ignore` has a rough "reason" for ignoring after it parentheses, with
- `(pseudo-rust)` meaning "mostly rust-like but contains foreign syntax"
- `(illustrative)` a somewhat catchall for either a fragment of rust that doesn't stand on its own (like a lone type), or abbreviated rust with ellipses and undeclared types that would get too cluttered if made compile-worthy.
- `(not-rust)` stuff that isn't rust but benefits from the syntax highlighting, like MIR.
- `(internal)` uses `rustc_*` code which would be difficult to make work with the testing setup.
Those reason notes are a bit inconsistently applied and messy though. If that's important I can go through them again and try a more principled approach. When I run `rg '```ignore \(' .` on the repo, there look to be lots of different conventions other people have used for this sort of thing. I could try unifying them all if that would be helpful.
I'm not sure if there was a better existing way to do this but I wrote my own script to help me run all the doctests and wade through the output. If that would be useful to anyone else, I put it here: https://github.com/Elliot-Roberts/rust_doctest_fixing_tool
Overhaul `MacArgs`
Motivation:
- Clarify some code that I found hard to understand.
- Eliminate one use of three places where `TokenKind::Interpolated` values are created.
r? `@petrochenkov`
The value in `MacArgs::Eq` is currently represented as a `Token`.
Because of `TokenKind::Interpolated`, `Token` can be either a token or
an arbitrary AST fragment. In practice, a `MacArgs::Eq` starts out as a
literal or macro call AST fragment, and then is later lowered to a
literal token. But this is very non-obvious. `Token` is a much more
general type than what is needed.
This commit restricts things, by introducing a new type `MacArgsEqKind`
that is either an AST expression (pre-lowering) or an AST literal
(post-lowering). The downside is that the code is a bit more verbose in
a few places. The benefit is that makes it much clearer what the
possibilities are (though also shorter in some other places). Also, it
removes one use of `TokenKind::Interpolated`, taking us a step closer to
removing that variant, which will let us make `Token` impl `Copy` and
remove many "handle Interpolated" code paths in the parser.
Things to note:
- Error messages have improved. Messages like this:
```
unexpected token: `"bug" + "found"`
```
now say "unexpected expression", which makes more sense. Although
arbitrary expressions can exist within tokens thanks to
`TokenKind::Interpolated`, that's not obvious to anyone who doesn't
know compiler internals.
- In `parse_mac_args_common`, we no longer need to collect tokens for
the value expression.
Using an obviously-placeholder syntax. An RFC would still be needed before this could have any chance at stabilization, and it might be removed at any point.
But I'd really like to have it in nightly at least to ensure it works well with try_trait_v2, especially as we refactor the traits.
The `token` is always an interpolated non-terminal expression, and
always a literal in valid code. This commit simplifies the processing
accordingly, by directly extracting and using the literal.
Perform lifetime resolution on the AST for lowering
Lifetime resolution is currently implemented several times. Once during lowering in order to introduce in-band lifetimes, and once in the resolve_lifetimes query. However, due to the global nature of lifetime resolution and how it interferes with hygiene, it is better suited on the AST.
This PR implements a first draft of lifetime resolution on the AST. For now, we specifically target named lifetimes and everything we need to remove lifetime resolution from lowering. Some diagnostics have already been ported, and sometimes made more precise using available hygiene information. Follow-up PRs will address in particular the resolution of anonymous lifetimes on the AST.
We reuse the rib design of the current resolution framework. Specific `LifetimeRib` and `LifetimeRibKind` types are introduced. The most important variant is `LifetimeRibKind::Generics`, which happens each time we encounter something which may introduce generic lifetime parameters. It can be an item or a `for<...>` binder. The `LifetimeBinderKind` specifies how this rib behaves with respect to in-band lifetimes.
r? `@petrochenkov`
Implement sym operands for global_asm!
Tracking issue: #93333
This PR is pretty much a complete rewrite of `sym` operand support for inline assembly so that the same implementation can be shared by `asm!` and `global_asm!`. The main changes are:
- At the AST level, `sym` is represented as a special `InlineAsmSym` AST node containing a path instead of an `Expr`.
- At the HIR level, `sym` is split into `SymStatic` and `SymFn` depending on whether the path resolves to a static during AST lowering (defaults to `SynFn` if `get_early_res` fails).
- `SymFn` is just an `AnonConst`. It runs through typeck and we just collect the resulting type at the end. An error is emitted if the type is not a `FnDef`.
- `SymStatic` directly holds a path and the `DefId` of the `static` that it is pointing to.
- The representation at the MIR level is mostly unchanged. There is a minor change to THIR where `SymFn` is a constant instead of an expression.
- At the codegen level we need to apply the target's symbol mangling to the result of `tcx.symbol_name()` depending on the target. This is done by calling the LLVM name mangler, which handles all of the details.
- On Mach-O, all symbols have a leading underscore.
- On x86 Windows, different mangling is used for cdecl, stdcall, fastcall and vectorcall.
- No mangling is needed on other platforms.
r? `@nagisa`
cc `@eddyb`
remove find_use_placement
A more robust solution to finding where to place use suggestions was added in #94584.
The algorithm uses the AST to find the span for the suggestion so we pass this span
down to the HIR during lowering and use it instead of calling `find_use_placement`
Fixes#94941
A more robust solution to finding where to place use suggestions was added.
The algorithm uses the AST to find the span for the suggestion so we pass this span
down to the HIR during lowering and use it.
Signed-off-by: Miguel Guarniz <mi9uel9@gmail.com>
async: Give predictable name to binding generated from .await expressions.
This name makes it to debuginfo and allows debuggers to identify such bindings and their captured versions in suspended async fns.
This will be useful for async stack traces, as discussed in https://internals.rust-lang.org/t/async-debugging-logical-stack-traces-setting-goals-collecting-examples/15547.
I don't know if this needs some discussion by ````@rust-lang/compiler,```` e.g. about the name of the binding (`__awaitee`) or about the fact that this PR introduces a (soft) guarantee about a compiler generated name. Although, regarding the later, I think the same reasoning applies here as it does for debuginfo in general.
r? ````@tmandry````
There are a few places were we have to construct it, though, and a few
places that are more invasive to change. To do this, we create a
constructor with a long obvious name.
More robust fallback for `use` suggestion
Our old way to suggest where to add `use`s would first look for pre-existing `use`s in the relevant crate/module, and if there are *no* uses, it would fallback on trying to use another item as the basis for the suggestion.
But this was fragile, as illustrated in issue #87613
This PR instead identifies span of the first token after any inner attributes, and uses *that* as the fallback for the `use` suggestion.
Fix#87613
then we just suggest the first legal position where you could inject a use.
To do this, I added `inject_use_span` field to `ModSpans`, and populate it in
parser (it is the span of the first token found after inner attributes, if any).
Then I rewrote the use-suggestion code to utilize it, and threw out some stuff
that is now unnecessary with this in place. (I think the result is easier to
understand.)
Then I added a test of issue 87613.
The previous approach of checking for the reserve-r9 target feature
didn't actually work because LLVM only sets this feature very late when
initializing the per-function subtarget.
Adopt let else in more places
Continuation of #89933, #91018, #91481, #93046, #93590, #94011.
I have extended my clippy lint to also recognize tuple passing and match statements. The diff caused by fixing it is way above 1 thousand lines. Thus, I split it up into multiple pull requests to make reviewing easier. This is the biggest of these PRs and handles the changes outside of rustdoc, rustc_typeck, rustc_const_eval, rustc_trait_selection, which were handled in PRs #94139, #94142, #94143, #94144.
Nicholas Nethercote
