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
refactor(rustc_expand::mbe): Don't require full ExtCtxt when not necessary
Refactor `mbe::diagnostics::failed_to_match_macro()` to not require a full `ExtCtxt`, but only a `&ParseSess`. It hard-required the `ExtCtxt` only for a call to `cx.trace_macros_diag()`, which we move instead to the only call-site of the function.
Note: This could be a potential change in observed behavior, because a call to `cx.trace_macros_diag()` now always happens after `failed_to_match_macro()` was called, where before it was only called at the end of the main return path of the function. But since `trace_macros_diag()` "flushes" out any not-yet-reported errors, it should be ok to call it for all paths, since there shouldn't be any on the non-main paths I think. However, I don't know the rest of the codebase well enough to say that with 100% confidence, but `tests/ui` still pass, which gives at least some confidence in the change.
Also concretize the return type from `Box<dyn MacResult>` to `(Span, ErrorGuaranteed)`, because this function will _always_ return an error, and never any other kind of result.
Was part of #128605 and #128747, but is a standalone refactoring.
r? ``@petrochenkov``
Emit an error for invalid use of the `#[no_sanitize]` attribute
fixes#128487.
Currently, the use of the `#[no_sanitize]` attribute for Mod, Impl,... is incorrectly permitted. This PR will correct this issue by generating errors, and I've also added some UI test cases for it.
Referenced #128458. As far as I know, the `#[no_sanitize]` attribute can only be used with functions, so I changed that part to `Fn` and `Method` using `check_applied_to_fn_or_method`. However, I couldn't find explicit documentation on this, so I could be mistaken...
Skip over args when determining if async-closure's inner coroutine consumes its upvars
#125306 implements a strategy for when we have an `async move ||` async-closure that is inferred to be `async FnOnce`, it will force the inner coroutine to also be `move`, since we cannot borrow any upvars from the parent async-closure (since `FnOnce` is not lending):
8e86c95671/compiler/rustc_hir_typeck/src/upvar.rs (L211-L229)
However, when this strategy was implemented, it reused the `ExprUseVisitor` data from visiting the whole coroutine, which includes additional statements due to `async`-specific argument desugaring:
8e86c95671/compiler/rustc_ast_lowering/src/item.rs (L1197-L1228)
Well, it turns out that we don't care about these argument desugaring parameters, because arguments to the async-closure are not the *async-closure*'s captures -- they exist for only one invocation of the closure, and they're always consumed by construction (see the argument desugaring above), so they will force the coroutine's inferred kind to `FnOnce`. (Unless they're `Copy`, because we never consider `Copy` types to be consumed):
8e86c95671/compiler/rustc_hir_typeck/src/expr_use_visitor.rs (L60-L66)
However, since we *were* visiting these arg exprs, this resulted in us too-aggressively applying `move` to the inner coroutine, resulting in regressions. For example, this PR fixes#128516. Fascinatingly, the note above about how we never consume `Copy` types is why this only regressed when the argument types weren't all `Copy`.
I tried to leave some comments inline to make this more clear :)
Only walk ribs to collect possibly shadowed params if we are adding params in our new rib
No need to collect params from parent ribs if we literally have no params to declare in this new rib.
Attempt to win back some of the perf in https://github.com/rust-lang/rust/pull/128357#issuecomment-2262677031.
Please review with whitespace *off*, the diff should be like 2 lines.
r? petrochenkov
Rollup of 8 pull requests
Successful merges:
- #128221 (Add implied target features to target_feature attribute)
- #128261 (impl `Default` for collection iterators that don't already have it)
- #128353 (Change generate-copyright to generate HTML, with cargo dependencies included)
- #128679 (codegen: better centralize function declaration attribute computation)
- #128732 (make `import.vis` is immutable)
- #128755 (Integrate crlf directly into related test file instead via of .gitattributes)
- #128772 (rustc_codegen_ssa: Set architecture for object crate for 32-bit SPARC)
- #128782 (unused_parens: do not lint against parens around &raw)
r? `@ghost`
`@rustbot` modify labels: rollup
unused_parens: do not lint against parens around &raw
Requested by `@tmandry` in https://github.com/rust-lang/rust/pull/127679: with `&raw` one somewhat regularly has to write code like `(&raw const (*myptr).field).method()`, so parentheses around the expression are often required. To avoid churn between adding and removing parentheses as method calls appear and disappear, the proposal was made to silence the lint for unnecessary parentheses around `&raw` expressions. This PR implements that.
rustc_codegen_ssa: Set architecture for object crate for 32-bit SPARC
The `object` crate was recently updated to recognize the 32-bit SPARC ELF targets `EM_SPARC` and `EM_SPARC32PLUS`, so the proper architecture for 32-bit SPARC can now be set in `rustc_codegen_ssa`.
r? nagisa
codegen: better centralize function declaration attribute computation
For some reason, the codegen backend has two functions that compute which attributes a function declaration gets: `apply_attrs_llfn` and `attributes::from_fn_attrs`. They are called in different places, on entirely different layers of abstraction.
To me the code seems cleaner if we centralize this entirely in `apply_attrs_llfn`, so that's what this PR does.
Disallow setting some built-in cfg via set the command-line
This PR disallow users from setting some built-in cfg via set the command-line in order to prevent incoherent state, eg. `windows` cfg active but target is Linux based.
This implements MCP https://github.com/rust-lang/compiler-team/issues/610, with the caveat that we disallow cfgs no matter if they make sense or not, since I don't think it's useful to allow users to set a cfg that will be set anyway. It also complicates the implementation.
------
The `explicit_builtin_cfgs_in_flags` lint detects builtin cfgs set via the `--cfg` flag.
*(deny-by-default)*
### Example
```text
rustc --cfg unix
```
```rust,ignore (needs command line option)
fn main() {}
```
This will produce:
```text
error: unexpected `--cfg unix` flag
|
= note: config `unix` is only supposed to be controlled by `--target`
= note: manually setting a built-in cfg can and does create incoherent behaviours
= note: `#[deny(explicit_builtin_cfgs_in_flags)]` on by default
```
### Explanation
Setting builtin cfgs can and does produce incoherent behaviour, it's better to the use the appropriate `rustc` flag that controls the config. For example setting the `windows` cfg but on Linux based target.
-----
r? `@petrochenkov`
cc `@jyn514`
try-job: aarch64-apple
try-job: test-various
try-job: armhf-gnu
try-job: x86_64-msvc
try-job: x86_64-mingw
try-job: i686-msvc
try-job: i686-mingw
try-job: x86_64-gnu-llvm-17
try-job: dist-various-1
Don't arbitrarily choose one upper bound for hidden captured region error message
You could argue that the error message is objectively worse, even though it's more accurate. I guess we could also add a note explaining like "cannot capture the intersection of two regions" or something, though I'm not sure if that is confusing due to being totally technical jargon.
This addresses the fact that #128752 says "add `+ 'b`" even though it does nothing to fix the issue. It doesn't fix the issue's root cause, though.
r? `@spastorino`
More information for fully-qualified suggestion when there are multiple impls
```
error[E0790]: cannot call associated function on trait without specifying the corresponding `impl` type
--> $DIR/E0283.rs:30:21
|
LL | fn create() -> u32;
| ------------------- `Coroutine::create` defined here
...
LL | let cont: u32 = Coroutine::create();
| ^^^^^^^^^^^^^^^^^^^ cannot call associated function of trait
|
help: use a fully-qualified path to a specific available implementation
|
LL | let cont: u32 = <Impl as Coroutine>::create();
| ++++++++ +
LL | let cont: u32 = <AnotherImpl as Coroutine>::create();
| +++++++++++++++ +
```
minor `effects` cleanups
* remove the fixme comment about not needing defaults because it turns out we do need defaults (if I made it None instead it would ice a bunch of tests)
* remove the part that special cased trait args when lowering them. This is now historical because effects doesn't add host args to traits anymore (we use associated types now)
Make create_dll_import_lib easier to implement
This will make it easier to implement raw-dylib support in cg_clif and cg_gcc. This PR doesn't yet include an create_dll_import_lib implementation for cg_clif as I need to correctly implement dllimport in cg_clif first before raw-dylib can work at all with cg_clif.
Required for https://github.com/rust-lang/rustc_codegen_cranelift/issues/1345
We can replace some tricky iterator-mutation code with a much simpler version
that uses `while let` to shrink a slice.
We also check whether a subpattern would be a wildcard _before_ hoisting it,
which will be very useful when trying to get rid of `print::PatKind` later.
Version 0.3.1 has added support for writing import libraries. Version
0.3.2 fixed creating archives containing members of import libraries.
Version 0.3.3 fixed building on big-endian systems.
The object crate was recently updated to recognize the 32-bit SPARC
ELF targets EM_SPARC and EM_SPARC32PLUS, so the proper architecture
for 32-bit SPARC can now be set in rustc_codegen_ssa.
Add `f16` and `f128` math functions
This adds intrinsics and math functions for `f16` and `f128` floating point types. Support is quite limited and some things are broken so tests don't run on many platforms, but this provides a starting point.
Don't ICE when getting an input file name's stem fails
Fixes#128681
The file stem is only used as a user-friendly prefix on intermediary files. While nice to have, it's not the end of the world if it fails so there's no real reason to emit an error here. We can continue with a fixed name as we do when an anonymous string is used.
interpret: refactor function call handling to be better-abstracted
Add a new function `init_stack_frame` that pushes a stack frame and passes the arguments, and use that basically everywhere that the raw underlying `push_stack_frame` used to be called. This splits the previous monster function `eval_fn_call` into two parts: figuring out the MIR to call and the arguments to pass, and then actually setting up the stack frame.
Also re-organize the files a bit:
- The previous `terminator.rs` is split into a new `call.rs` with all the argument-passing logic, and the rest goes into `step.rs` where the other main dispatcher functions already live (in particular, `eval_statement`).
- All the stack frame handling from `eval_context.rs` is moved to a new `stack.rs`.
Pass the right `ParamEnv` to `might_permit_raw_init_strict`
Fixes#119620
`might_permit_raw_init_strict` currently passes an empty `ParamEnv` to the `InterpCx`, instead of the actual `ParamEnv` that was passed in to `check_validity_requirement` at callsite.
This leads to ICEs such as the linked issue where for `UnsafeCell<*mut T>` we initially get the layout with the right `ParamEnv` (which suceeds because it can prove that `T: Sized` and therefore `UnsafeCell<*mut T>` has a known layout) but then do the rest with an empty `ParamEnv` where `T: Sized` is not known to hold so getting the layout for `*mut T` later fails.
This runs into an assertion in other layout code where it's making the (valid) assumption that, when we already have a layout for a struct (`UnsafeCell<*mut T>`), getting the layout of one of its fields (`*mut T`) should also succeed, which wasn't the case here due to using the wrong `ParamEnv`.
So, this PR changes it to just use the same `ParamEnv` all the way throughout.
Fix ICE Caused by Incorrectly Delaying E0107
Fixes #128249
For the following code:
```rust
trait Foo<T> {}
impl Foo<T: Default> for u8 {}
```
#126054 added some logic to delay emitting E0107 as the names of associated type `T` in the impl header and generic parameter `T` in `trait Foo` match.
But it failed to ensure whether such unexpected associated type bounds are coming from a impl block header. This caused an ICE as the compiler was delaying E0107 for code like:
```rust
trait Trait<Type> {
type Type;
fn method(&self) -> impl Trait<Type: '_>;
}
```
because it assumed the associated type bound `Type: '_` is for the generic parameter `Type` in `trait Trait` since the names are same.
This PR adds a check to ensure that E0107 is delayed only in the context of impl block header.
When encountering an E0277, if the type and the trait both come from a crate with the same name but different crate number, we explain that there are multiple crate versions in the dependency tree.
If there's a type that fulfills the bound, and it has the same name as the passed in type and has the same crate name, we explain that the same type in two different versions of the same crate *are different*.
```
error[E0277]: the trait bound `Type: dependency::Trait` is not satisfied
--> src/main.rs:4:18
|
4 | do_something(Type);
| ------------ ^^^^ the trait `dependency::Trait` is not implemented for `Type`
| |
| required by a bound introduced by this call
|
help: you have multiple different versions of crate `dependency` in your dependency graph
--> src/main.rs:1:5
|
1 | use bar::do_something;
| ^^^ one version of crate `dependency` is used here, as a dependency of crate `bar`
2 | use dependency::Type;
| ^^^^^^^^^^ one version of crate `dependency` is used here, as a direct dependency of the current crate
note: two types coming from two different versions of the same crate are different types even if they look the same
--> /home/gh-estebank/crate_versions/baz-2/src/lib.rs:1:1
|
1 | pub struct Type;
| ^^^^^^^^^^^^^^^ this type doesn't implement the required trait
|
::: /home/gh-estebank/crate_versions/baz/src/lib.rs:1:1
|
1 | pub struct Type;
| ^^^^^^^^^^^^^^^ this type implements the required trait
2 | pub trait Trait {}
| --------------- this is the required trait
note: required by a bound in `bar::do_something`
--> /home/gh-estebank/crate_versions/baz/src/lib.rs:4:24
|
4 | pub fn do_something<X: Trait>(_: X) {}
| ^^^^^ required by this bound in `do_something`
```
Address #22750.
On short error format, append primary span label to message
The `error-format=short` output only displays the path, error code and main error message all in the same line. We now add the primary span label as well after the error message, to provide more context.
The `error-format=short` output only displays the path, error code and
main error message all in the same line. We now add the primary span label
as well after the error message, to provide more context.
Change output normalization logic to be linear against size of output
Modify the rendered output normalization routine to scan each character *once* and construct a `String` to be printed out to the terminal *once*, instead of using `String::replace` in a loop multiple times. The output doesn't change, but the time spent to prepare a diagnostic is now faster (or rather, closer to what it was before #127528).
Tweak type inference for `const` operands in inline asm
Previously these would be treated like integer literals and default to `i32` if a type could not be determined. To allow for forward-compatibility with `str` constants in the future, this PR changes type inference to use an unbound type variable instead.
The actual type checking is deferred until after typeck where we still ensure that the final type for the `const` operand is an integer type.
<!--
If this PR is related to an unstable feature or an otherwise tracked effort,
please link to the relevant tracking issue here. If you don't know of a related
tracking issue or there are none, feel free to ignore this.
This PR will get automatically assigned to a reviewer. In case you would like
a specific user to review your work, you can assign it to them by using
r? <reviewer name>
-->
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.
interpret: move nullary-op evaluation into operator.rs
We call it an operator, so we might as well treat it like one. :)
Also use more consistent naming for the "evaluate intrinsic" functions. "emulate" is really the wrong term, this *is* a genuine implementation of the intrinsic semantics after all.
Use `ParamEnv::reveal_all` in CFI
I left a huge comment for why this ICEs in the test I committed.
`typeid_for_instance` should only be called on monomorphic instances during codegen, and we should just be using `ParamEnv::reveal_all()` rather than the param-env of the instance itself. I added an assertion to ensure that we only do this for fully substituted instances (this may break with polymorphization, but I kinda don't care lol).
Fixes#114160
cc `@rcvalle`
