Stabilize `min_exhaustive_patterns`
## Stabilisation report
I propose we stabilize the [`min_exhaustive_patterns`](https://github.com/rust-lang/rust/issues/119612) language feature.
With this feature, patterns of empty types are considered unreachable when matched by-value. This allows:
```rust
enum Void {}
fn foo() -> Result<u32, Void>;
fn main() {
let Ok(x) = foo();
// also
match foo() {
Ok(x) => ...,
}
}
```
This is a subset of the long-unstable [`exhaustive_patterns`](https://github.com/rust-lang/rust/issues/51085) feature. That feature is blocked because omitting empty patterns is tricky when *not* matched by-value. This PR stabilizes the by-value case, which is not tricky.
The not-by-value cases (behind references, pointers, and unions) stay as they are today, e.g.
```rust
enum Void {}
fn foo() -> Result<u32, &Void>;
fn main() {
let Ok(x) = foo(); // ERROR: missing `Err(_)`
}
```
The consequence on existing code is some extra "unreachable pattern" warnings. This is fully backwards-compatible.
### Comparison with today's rust
This proposal only affects match checking of empty types (i.e. types with no valid values). Non-empty types behave the same with or without this feature. Note that everything below is phrased in terms of `match` but applies equallly to `if let` and other pattern-matching expressions.
To be precise, a visibly empty type is:
- an enum with no variants;
- the never type `!`;
- a struct with a *visible* field of a visibly empty type (and no #[non_exhaustive] annotation);
- a tuple where one of the types is visibly empty;
- en enum with all variants visibly empty (and no `#[non_exhaustive]` annotation);
- a `[T; N]` with `N != 0` and `T` visibly empty;
- all other types are nonempty.
(An extra change was proposed below: that we ignore #[non_exhaustive] for structs since adding fields cannot turn an empty struct into a non-empty one)
For normal types, exhaustiveness checking requires that we list all variants (or use a wildcard). For empty types it's more subtle: in some cases we require a `_` pattern even though there are no valid values that can match it. This is where the difference lies regarding this feature.
#### Today's rust
Under today's rust, a `_` is required for all empty types, except specifically: if the matched expression is of type `!` (the never type) or `EmptyEnum` (where `EmptyEnum` is an enum with no variants), then the `_` is not required.
```rust
let foo: Result<u32, !> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required
}
let foo: Result<u32, &!> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required
}
let foo: &! = ...;
match foo {
_ => ..., // required
}
fn blah(foo: (u32, !)) {
match foo {
_ => ..., // required
}
}
unsafe {
let ptr: *const ! = ...;
match *ptr {} // allowed
let ptr: *const (u32, !) = ...;
match *ptr {
(x, _) => { ... } // required
}
let ptr: *const Result<u32, !> = ...;
match *ptr {
Ok(x) => { ... }
Err(_) => { ... } // required
}
}
```
#### After this PR
After this PR, a pattern of an empty type can be omitted if (and only if):
- the match scrutinee expression has type `!` or `EmptyEnum` (like before);
- *or* the empty type is matched by value (that's the new behavior).
In all other cases, a `_` is required to match on an empty type.
```rust
let foo: Result<u32, !> = ...;
match foo {
Ok(x) => ..., // `Err` not required
}
let foo: Result<u32, &!> = ...;
match foo {
Ok(x) => ...,
Err(_) => ..., // required because `!` is under a dereference
}
let foo: &! = ...;
match foo {
_ => ..., // required because `!` is under a dereference
}
fn blah(foo: (u32, !)) {
match foo {} // allowed
}
unsafe {
let ptr: *const ! = ...;
match *ptr {} // allowed
let ptr: *const (u32, !) = ...;
match *ptr {
(x, _) => { ... } // required because the matched place is under a (pointer) dereference
}
let ptr: *const Result<u32, !> = ...;
match *ptr {
Ok(x) => { ... }
Err(_) => { ... } // required because the matched place is under a (pointer) dereference
}
}
```
### Documentation
The reference does not say anything specific about exhaustiveness checking, hence there is nothing to update there. The nomicon does, I opened https://github.com/rust-lang/nomicon/pull/445 to reflect the changes.
### Tests
The relevant tests are in `tests/ui/pattern/usefulness/empty-types.rs`.
### Unresolved Questions
None that I know of.
try-job: dist-aarch64-apple
Cache supertrait outlives of impl header for soundness check
This caches the results of computing the transitive supertraits of an impl and filtering it to its outlives obligations. This is purely an optimization to improve https://github.com/rust-lang/rust/pull/124336.
Stop unnecessarily taking GenericPredicates by `&self`
This results in overcapturing in edition 2024, and is unnecessary since `GenericPredicates: Copy`.
WF-check struct field types at construction site
Fixes#126272.
Fixes#127299.
Rustc of course already WF-checked the field types at the definition
site, but for error tainting of consts to work properly, there needs to
be an error emitted at the use site. Previously, with no use-site error,
we proceeded with CTFE and ran into ICEs since we are running code with
type errors.
Emitting use-site errors also brings struct-like constructors more in
line with fn-like constructors since they already emit use-site errors
for WF issues.
r? `@BoxyUwU`
Enable zstd for debug compression.
Set LLVM_ENABLE_ZSTD alongside LLVM_ENABLE_ZLIB so that --compress-debug-sections=zstd is an option.
See #120953
try-job: x86_64-gnu-tools
Subtree sync for rustc_codegen_cranelift
The main highlight this time is support for raw-dylib on Windows thanks to `@dpaoliello.` Compiling the ring crate for arm64 macOS has been fixed too.
r? `@ghost`
`@rustbot` label +A-codegen +A-cranelift +T-compiler
Ensure let stmt compound assignment removal suggestion respect codepoint boundaries
Previously we would try to issue a suggestion for `let x <op>= 1`, i.e.
a compound assignment within a `let` binding, to remove the `<op>`. The
suggestion code unfortunately incorrectly assumed that the `<op>` is an
exactly-1-byte ASCII character, but this assumption is incorrect because
we also recover Unicode-confusables like `➖=` as `-=`. In this example,
the suggestion code used a `+ BytePos(1)` to calculate the span of the
`<op>` codepoint that looks like `-` but the mult-byte Unicode
look-alike would cause the suggested removal span to be inside a
multi-byte codepoint boundary, triggering a codepoint boundary
assertion.
The fix is to use `SourceMap::start_point(token_span)` which properly accounts for codepoint boundaries.
Fixes#128845.
cc #128790
r? ````@fmease````
Use `SourceMap::end_point` instead of `- BytePos(1)` in arg removal suggestion
Previously, we tried to remove extra arg commas when providing extra arg removal suggestions. One of
the edge cases is having to account for an arg that has a closing delimiter `)` following it.
However, the previous suggestion code assumed that the delimiter is in fact exactly the 1-byte `)`
character. This assumption was proven incorrect, because we recover from Unicode-confusable
delimiters in the parser, which means that the ending delimiter could be a multi-byte codepoint
that looks *like* a `)`. Subtracing 1 byte could land us in the middle of a codepoint, triggering a
codepoint boundary assertion.
This is fixed by using `SourceMap::end_point` which properly accounts for codepoint boundaries.
Fixes#128717.
cc ````@fmease```` and #128790
use stable sort to sort multipart diagnostics
I think a stable sort should be used to sort the different parts of a multipart selection. The current unstable sort uses the text of the suggestion as a tie-breaker. That just doesn't seem right, and the order of the input is a better choice I think, because it gives the diagnostic author more control.
This came up when I was building a suggestion where
```rust
fn foo() {}
```
must be turned into an unsafe function, and an attribute must be added
```rust
#[target_feature(enable = "...")]
unsafe fn foo() {}
```
In this example, the two suggestions occur at the same position, but the order is extremely important: unsafe must come after the attribute. But the situation changes if there is a pub/pub(crate), and if the unsafe is already present. It just out that because of the suggestion text, there is no way for me to order the suggestions correctly.
This change probably should be tested, but are there tests of the diagnostics code itself in the tests?
r? ```@estebank```
Add `Steal::is_stolen()`
Writers of rustc drivers (such as myself) often encounter stealing issues. It is currently impossible to gracefully handle them. This PR adds a `Steal::is_stolen()` function for that purpose.
Set LLVM_ENABLE_ZSTD alongside LLVM_ENABLE_ZLIB so that --compress-debug-sections=zstd is an option.
Use static linking to avoid a new runtime dependency. Add an llvm.libzstd bootstrap option for LLVM
with zstd. Set it off by default except for the dist builder. Handle llvm-config --system-libs output
that contains static libraries.
Miscellaneous improvements to struct tail normalization
1. Make checks for foreign tails more accurate by normalizing the struct tail. I didn't write a test for this one.
2. Normalize when computing struct tail for `offset_of` for slice/str. This fixes the new solver only.
3. Normalizing when computing tails for disaligned reference check. This fixes both solvers.
r? lcnr
Previously we would try to issue a suggestion for `let x <op>= 1`, i.e.
a compound assignment within a `let` binding, to remove the `<op>`. The
suggestion code unfortunately incorrectly assumed that the `<op>` is an
exactly-1-byte ASCII character, but this assumption is incorrect because
we also recover Unicode-confusables like `➖=` as `-=`. In this example,
the suggestion code used a `+ BytePos(1)` to calculate the span of the
`<op>` codepoint that looks like `-` but the mult-byte Unicode
look-alike would cause the suggested removal span to be inside a
multi-byte codepoint boundary, triggering a codepoint boundary
assertion.
Issue: <https://github.com/rust-lang/rust/issues/128845>
Parser has error recovery for Unicode-confusables, which includes the
right parentheses `)`. If a multi-byte right parentheses look-alike
reaches the argument removal suggestion diagnostics, it would trigger an
assertion because the diagnostics used `- BytePos(1)` which can land
within a multi-byte codepoint.
This is fixed by using `SourceMap::end_point` to find the final right
delimiter codepoint, which correctly respects codepoint boundaries.
Don't inline tainted MIR bodies
Don't inline MIR bodies that are tainted, since they're not necessarily well-formed.
Fixes#128601 (I didn't add a new test, just copied one from the crashes, since they're the same root cause).
Fixes#122909.
Add comment that bors did not see pushed before it merged
In #128612, bors merged 470ada2de0 instead of 1e07c19.
This means it dropped a useful comment I added, and a stage rename that is more descriptive.
Don't implement `AsyncFn` for `FnDef`/`FnPtr` that wouldnt implement `Fn`
Due to unsafety, ABI, or the presence of target features, some `FnDef`/`FnPtr` types don't implement `Fn*`. Do the same for `AsyncFn*`.
Noticed this due to #128764, but this isn't really related to that ICE, which is fixed in #128792.
Update E0517 message to reflect RFC 2195.
E0517 occurs when a `#[repr(..)]` attribute is placed on an unsupported item. Currently, the explanation of the error implies that `#[repr(u*/i*)]` cannot be placed on fieldful enums, which is no longer the case since [RFC 2195](https://github.com/rust-lang/rfcs/pull/2195) was [stabilized](https://github.com/rust-lang/rust/issues/60553), which allows placing `#[repr(u*/i*)]` and/or `#[repr(C)]` on fieldful enums to produce a defined layout.
This PR doesn't (currently) add a description of the semantics of placing `#[repr(u*/i*)]` on a fieldful enum to the error explanation, it just removes the claims/implications that it is not allowed.
Make `validate_mir` ensure the final MIR for all bodies
A lot of the crashes tests use `-Zpolymorphize` or `-Zdump-mir` for their side effect of computing the `optimized_mir` for all bodies, which will uncover bugs with late MIR passes like the inliner. I don't like having all these tests depend on `-Zpolymorphize` (or other hacky ways) for no reason, so this PR extends the `-Zvalidate-mir` flag to ensure `optimized_mir`/`mir_for_ctfe` for all body owners during the analysis phase.
Two thoughts:
1. This could be moved later in the compilation pipeline I guess? I don't really think it matters, though.
1. This could alternatively be expressed using a new flag, though I don't necessarily see much value in separating these.
For example, #128171 could have used this flag, in the `tests/ui/polymorphization/inline-incorrect-early-bound.rs`.
r? mir
Some `const { }` asserts for #128200
The correctness of code in #128200 relies on an array being sorted (so that it can be used in binary search later), which is currently enforced with `// tidy-alphabetical` (and characters being written in `\u{XXXX}` form), as well as lack of duplicate entries with conflicting keys, which is not currently enforced.
This PR changes it to using a `const{ }` assertion (and also checks for duplicate entries). Sadly, we cannot use the recently-stabilized `is_sorted_by_key` here, because it is not const (but it would not allow us to check for uniqueness anyways). Instead, let's write a manual loop.
Alternative approach (perfect hash function): #128463
r? `@ghost`
Rollup of 7 pull requests
Successful merges:
- #128520 (Skip over args when determining if async-closure's inner coroutine consumes its upvars)
- #128552 (Emit an error for invalid use of the `#[no_sanitize]` attribute)
- #128691 (Update `compiler-builtins` to 0.1.117)
- #128702 (Add -Zmetrics-dir=PATH to save diagnostic metadata to disk)
- #128797 (Fuchsia Test Runner: enable ffx repository server)
- #128798 (refactor(rustc_expand::mbe): Don't require full ExtCtxt when not necessary)
- #128800 (Add tracking issue to core-pattern-type)
r? `@ghost`
`@rustbot` modify labels: rollup
