On implicit `Sized` bound on fn argument, point at type instead of pattern
Instead of
```
error[E0277]: the size for values of type `(dyn ThriftService<(), AssocType = _> + 'static)` cannot be known at compilation time
--> $DIR/issue-59324.rs:23:20
|
LL | fn with_factory<H>(factory: dyn ThriftService<()>) {}
| ^^^^^^^ doesn't have a size known at compile-time
```
output
```
error[E0277]: the size for values of type `(dyn ThriftService<(), AssocType = _> + 'static)` cannot be known at compilation time
--> $DIR/issue-59324.rs:23:29
|
LL | fn with_factory<H>(factory: dyn ThriftService<()>) {}
| ^^^^^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
```
When the template string passed to asm!() is produced by
a macro call like concat!() we were producing wrong error
spans. Now in the case of a macro call we just use the entire
arg to asm!(), macro call and all, as the error span.
Like #130865 did for the standard library, we can use `&raw` in the
compiler now that stage0 supports it. Also like the other issue, I did
not make any doc or test changes at this time.
Instead of
```
error[E0277]: the size for values of type `(dyn ThriftService<(), AssocType = _> + 'static)` cannot be known at compilation time
--> $DIR/issue-59324.rs:23:20
|
LL | fn with_factory<H>(factory: dyn ThriftService<()>) {}
| ^^^^^^^ doesn't have a size known at compile-time
```
output
```
error[E0277]: the size for values of type `(dyn ThriftService<(), AssocType = _> + 'static)` cannot be known at compilation time
--> $DIR/issue-59324.rs:23:29
|
LL | fn with_factory<H>(factory: dyn ThriftService<()>) {}
| ^^^^^^^^^^^^^^^^^^^^^ doesn't have a size known at compile-time
```
Pass Module Analysis Manager to Standard Instrumentations
This PR introduces changes related to llvm::PassInstrumentationCallbacks. Now, we pass Module Analysis Manager to StandardInstrumentations::registerCallbacks, so it can take advantage of such instrumentations as IR verifier or preserved CFG checker. So basically this is NFC PR.
Fix diagnostics for coroutines with () as input.
This may be a more real-life example to trigger the diagnostic:
```rust
#![features(try_blocks, coroutine_trait, coroutines)]
use std::ops::Coroutine;
struct Request;
struct Response;
fn get_args() -> Result<String, String> { todo!() }
fn build_request(_arg: String) -> Request { todo!() }
fn work() -> impl Coroutine<Option<Response>, Yield = Request> {
#[coroutine]
|_| {
let r: Result<(), String> = try {
let req = get_args()?;
yield build_request(req)
};
if let Err(msg) = r {
eprintln!("Error: {msg}");
}
}
}
```
Reorder stack spills so that constants come later.
Currently constants are "pulled forward" and have their stack spills emitted first. This confuses LLVM as to where to place breakpoints at function entry, and results in argument values being wrong in the debugger. It's straightforward to avoid emitting the stack spills for constants until arguments/etc have been introduced in debug_introduce_locals, so do that.
Example LLVM IR (irrelevant IR elided):
Before:
```
define internal void `@_ZN11rust_1289457binding17h2c78f956ba4bd2c3E(i64` %a, i64 %b, double %c) unnamed_addr #0 !dbg !178 { start:
%c.dbg.spill = alloca [8 x i8], align 8
%b.dbg.spill = alloca [8 x i8], align 8
%a.dbg.spill = alloca [8 x i8], align 8
%x.dbg.spill = alloca [4 x i8], align 4
store i32 0, ptr %x.dbg.spill, align 4, !dbg !192 ; LLVM places breakpoint here.
#dbg_declare(ptr %x.dbg.spill, !190, !DIExpression(), !192)
store i64 %a, ptr %a.dbg.spill, align 8
#dbg_declare(ptr %a.dbg.spill, !187, !DIExpression(), !193)
store i64 %b, ptr %b.dbg.spill, align 8
#dbg_declare(ptr %b.dbg.spill, !188, !DIExpression(), !194)
store double %c, ptr %c.dbg.spill, align 8
#dbg_declare(ptr %c.dbg.spill, !189, !DIExpression(), !195)
ret void, !dbg !196
}
```
After:
```
define internal void `@_ZN11rust_1289457binding17h2c78f956ba4bd2c3E(i64` %a, i64 %b, double %c) unnamed_addr #0 !dbg !178 { start:
%x.dbg.spill = alloca [4 x i8], align 4
%c.dbg.spill = alloca [8 x i8], align 8
%b.dbg.spill = alloca [8 x i8], align 8
%a.dbg.spill = alloca [8 x i8], align 8
store i64 %a, ptr %a.dbg.spill, align 8
#dbg_declare(ptr %a.dbg.spill, !187, !DIExpression(), !192)
store i64 %b, ptr %b.dbg.spill, align 8
#dbg_declare(ptr %b.dbg.spill, !188, !DIExpression(), !193)
store double %c, ptr %c.dbg.spill, align 8
#dbg_declare(ptr %c.dbg.spill, !189, !DIExpression(), !194)
store i32 0, ptr %x.dbg.spill, align 4, !dbg !195 ; LLVM places breakpoint here.
#dbg_declare(ptr %x.dbg.spill, !190, !DIExpression(), !195)
ret void, !dbg !196
}
```
Note in particular the position of the "LLVM places breakpoint here" comment relative to the stack spills for the function arguments. LLVM assumes that the first instruction with with a debug location is the end of the prologue. As LLVM does not currently offer front ends any direct control over the placement of the prologue end reordering the IR is the only mechanism available to fix argument values at function entry in the presence of MIR optimizations like SingleUseConsts. Fixes#128945
r? `@michaelwoerister`
Collect relevant item bounds from trait clauses for nested rigid projections
Rust currently considers trait where-clauses that bound the trait's *own* associated types to act like an item bound:
```rust
trait Foo where Self::Assoc: Bar { type Assoc; }
// acts as if:
trait Foo { type Assoc: Bar; }
```
### Background
This behavior has existed since essentially forever (i.e. before Rust 1.0), since we originally started out by literally looking at the where clauses written on the trait when assembling `SelectionCandidate::ProjectionCandidate` for projections. However, looking at the predicates of the associated type themselves was not sound, since it was unclear which predicates were *assumed* and which predicates were *implied*, and therefore this was reworked in #72788 (which added a query for the predicates we consider for `ProjectionCandidate`s), and then finally item bounds and predicates were split in #73905.
### Problem 1: GATs don't uplift bounds correctly
All the while, we've still had logic to uplift associated type bounds from a trait's where clauses. However, with the introduction of GATs, this logic was never really generalized correctly for them, since we were using simple equality to test if the self type of a trait where clause is a projection. This leads to shortcomings, such as:
```rust
trait Foo
where
for<'a> Self::Gat<'a>: Debug,
{
type Gat<'a>;
}
fn test<T: Foo>(x: T::Gat<'static>) {
//~^ ERROR `<T as Foo>::Gat<'a>` doesn't implement `Debug`
println!("{:?}", x);
}
```
### Problem 2: Nested associated type bounds are not uplifted
We also don't attempt to uplift bounds on nested associated types, something that we couldn't really support until #120584. This can be demonstrated best with an example:
```rust
trait A
where Self::Assoc: B,
where <Self::Assoc as B>::Assoc2: C,
{
type Assoc; // <~ The compiler *should* treat this like it has an item bound `B<Assoc2: C>`.
}
trait B { type Assoc2; }
trait C {}
fn is_c<T: C>() {}
fn test<T: A>() {
is_c::<<Self::Assoc as B>::Assoc2>();
//~^ ERROR the trait bound `<<T as A>::Assoc as B>::Assoc2: C` is not satisfied
}
```
Why does this matter?
Well, generalizing this behavior bridges a gap between the associated type bounds (ATB) feature and trait where clauses. Currently, all bounds that can be stably written on associated types can also be expressed as where clauses on traits; however, with the stabilization of ATB, there are now bounds that can't be desugared in the same way. This fixes that.
## How does this PR fix things?
First, when scraping item bounds from the trait's where clauses, given a trait predicate, we'll loop of the self type of the predicate as long as it's a projection. If we find a projection whose trait ref matches, we'll uplift the bound. This allows us to uplift, for example `<Self as Trait>::Assoc: Bound` (pre-existing), but also `<<Self as Trait>::Assoc as Iterator>::Item: Bound` (new).
If that projection is a GAT, we will check if all of the GAT's *own* args are all unique late-bound vars. We then map the late-bound vars to early-bound vars from the GAT -- this allows us to uplift `for<'a, 'b> Self::Assoc<'a, 'b>: Trait` into an item bound, but we will leave `for<'a> Self::Assoc<'a, 'a>: Trait` and `Self::Assoc<'static, 'static>: Trait` alone.
### Okay, but does this *really* matter?
I consider this to be an improvement of the status quo because it makes GATs a bit less magical, and makes rigid projections a bit more expressive.
Fix up setting strip = true in Cargo.toml makes build scripts fail in…
Fix issue: https://github.com/rust-lang/rust/issues/110536
Strip binary is PATH dependent which breaks builds in MacOS.
For example, on my Mac, the output of 'which strip' is '/opt/homebrew/opt/binutils/bin/strip', which leads to incorrect 'strip' results. Therefore, just like on other systems, it is also necessary to specify 'stripcmd' on macOS. However, it seems that there is a bug in binutils [bugzilla-Bug 31571](https://sourceware.org/bugzilla/show_bug.cgi?id=31571), which leads to the problem mentioned above.
Rollup of 6 pull requests
Successful merges:
- #130549 (Add RISC-V vxworks targets)
- #130595 (Initial std library support for NuttX)
- #130734 (Fix: ices on virtual-function-elimination about principal trait)
- #130787 (Ban combination of GCE and new solver)
- #130809 (Update llvm triple for OpenHarmony targets)
- #130810 (Don't trap into the debugger on panics under Linux)
r? `@ghost`
`@rustbot` modify labels: rollup
Ban combination of GCE and new solver
These do not work together. I don't want anyone to have the impression that they do.
I reused the conflicting features diagnostic but I guess I could make it more tailored to the new solver? OTOH I don't really about the presentation of diagnostics here; these are nightly features after all.
r? `@BoxyUwU` thoughts on this?
Fix: ices on virtual-function-elimination about principal trait
Extract `load_vtable` function to ensure the `virtual_function_elimination` option is always checked.
It's okay not to use `llvm.type.checked.load` to load the vtable if there is no principal trait.
Fixes#123955Fixes#124092
Add `File` constructors that return files wrapped with a buffer
In addition to the light convenience, these are intended to raise visibility that buffering is something you should consider when opening a file, since unbuffered I/O is a common performance footgun to Rust newcomers.
ACP: https://github.com/rust-lang/libs-team/issues/446
Tracking Issue: #130804
rustdoc: inherit parent's stability where applicable
It is currently not possible for a re-export to have a different stability (https://github.com/rust-lang/rust/issues/30827). Therefore the standard library uses a hack when moving items like `std::error::Error` or `std::net::IpAddr` into `core` by marking the containing module (`core::error` / `core::net`) as unstable or stable in a later version than the items the module contains.
Previously, rustdoc would always show the *stability as declared* for an item rather than the *stability as publicly reachable* (i.e. the features required to actually access the item), which could be confusing when viewing the docs. This PR changes it so that we show the stability of the first unstable parent or the most recently stabilized parent instead, to hopefully make things less confusing.
fixes https://github.com/rust-lang/rust/issues/130765
screenshots:
 
Pin memchr to 2.5.0 in the library rather than rustc_ast
The latest versions of `memchr` experience LTO-related issues when compiling for windows-gnu [1], so needs to be pinned. The issue is present in the standard library.
`memchr` has been pinned in `rustc_ast`, but since the workspace was recently split, this pin no longer has any effect on library crates.
Resolve this by adding `memchr` as an _unused_ dependency in `std`, pinned to 2.5. Additionally, remove the pin in `rustc_ast` to allow non-library crates to upgrade to the latest version.
Link: https://github.com/rust-lang/rust/issues/127890 [1]
try-job: x86_64-mingw
try-job: x86_64-msvc
Separate collection of crate-local inherent impls from error tracking
#119895 changed the return type of the `crate_inherent_impls` query from `CrateInherentImpls` to `Result<CrateInherentImpls, ErrorGuaranteed>` to avoid needing to use the non-parallel-friendly `track_errors()` to track if an error was reporting from within the query... This was mostly fine until #121113, which stopped halting compilation when we hit an `Err(ErrorGuaranteed)` in the `crate_inherent_impls` query.
Thus we proceed onwards to typeck, and since a return type of `Result<CrateInherentImpls, ErrorGuaranteed>` means that the query can *either* return one of "the list inherent impls" or "error has been reported", later on when we want to assemble method or associated item candidates for inherent impls, we were just treating any `Err(ErrorGuaranteed)` return value as if Rust had no inherent impls defined anywhere at all! This leads to basically every inherent method call failing with an error, lol, which was reported in #127798.
This PR changes the `crate_inherent_impls` query to return `(CrateInherentImpls, Result<(), ErrorGuaranteed>)`, i.e. returning the inherent impls collected *and* whether an error was reported in the query itself. It firewalls the latter part of that query into a new `crate_inherent_impls_validity_check` just for the `ensure()` call.
This fixes#127798.
This changes the remaining span for the cast, because the new `Cast`
category has a higher priority (lower `Ord`) than the old `Coercion`
category, so we no longer report the region error for the "unsizing"
coercion from `*const Trait` to itself.
The latest versions of `memchr` experience LTO-related issues when
compiling for windows-gnu [1], so needs to be pinned. The issue is
present in the standard library.
`memchr` has been pinned in `rustc_ast`, but since the workspace was
recently split, this pin no longer has any effect on library crates.
Resolve this by adding `memchr` as an _unused_ dependency in `std`,
pinned to 2.5. Additionally, remove the pin in `rustc_ast` to allow
non-library crates to upgrade to the latest version.
Link: https://github.com/rust-lang/rust/issues/127890 [1]
Revert "Apply EarlyOtherwiseBranch to scalar value #129047"
This reverts PR #129047, commit a772336fb3, reversing changes made to 702987f75b.
cc `@DianQK` and `@cjgillot` as the PR author and reviewer of #129047 respectively.
It seems [Apply EarlyOtherwiseBranch to scalar value #129047](https://github.com/rust-lang/rust/pull/129047) may have lead to several nightly regressions:
- https://github.com/rust-lang/rust/issues/130769
- https://github.com/rust-lang/rust/issues/130774
- https://github.com/rust-lang/rust/issues/130771
Example test that would ICE with changes in #129047 (this test is included in this PR):
```rs
//@ compile-flags: -C opt-level=3
//@ check-pass
use std::task::Poll;
pub fn poll(val: Poll<Result<Option<Vec<u8>>, u8>>) {
match val {
Poll::Ready(Ok(Some(_trailers))) => {}
Poll::Ready(Err(_err)) => {}
Poll::Ready(Ok(None)) => {}
Poll::Pending => {}
}
}
```
Since this is a mir-opt ICE that seems to quite easy to trigger with real-world crates being affected, let's revert for now and reland the mir-opt after these are fixed.
llvm: replace some deprecated functions
`LLVMMDStringInContext` and `LLVMMDNodeInContext` are deprecated, replace them with `LLVMMDStringInContext2` and `LLVMMDNodeInContext2`.
Also replace `Value` with `Metadata` in some function signatures for better consistency.
Revert "Add recursion limit to FFI safety lint"
It's not necessarily clear if warning when we hit the recursion limit is the right thing to do, first of all.
**More importantly**, this PR was implemented incorrectly in the first place; it was not decrementing the recursion limit when stepping out of a type, so it would trigger when a ctype has more than RECURSION_LIMIT fields *anywhere* in the type's set of recursively reachable fields.
Reverts #130598Reopens#130310Fixes#130757
Check vtable projections for validity in miri
Currently, miri does not catch when we transmute `dyn Trait<Assoc = A>` to `dyn Trait<Assoc = B>`. This PR implements such a check, and fixes https://github.com/rust-lang/miri/issues/3905.
To do this, we modify `GlobalAlloc::VTable` to contain the *whole* list of `PolyExistentialPredicate`, and then modify `check_vtable_for_type` to validate the `PolyExistentialProjection`s of the vtable, along with the principal trait that was already being validated.
cc ``@RalfJung``
r? ``@lcnr`` or types
I also tweaked the diagnostics a bit.
---
**Open question:** We don't validate the auto traits. You can transmute `dyn Foo` into `dyn Foo + Send`. Should we check that? We currently have a test that *exercises* this as not being UB:
6c6d210089/src/tools/miri/tests/pass/dyn-upcast.rs (L14-L20)
I'm not actually sure if we ever decided that's actually UB or not 🤔
We could perhaps still check that the underlying type of the object (i.e. the concrete type that was unsized) implements the auto traits, to catch UB like:
```rust
fn main() {
let x: &dyn Trait = &std::ptr::null_mut::<()>();
let _: &(dyn Trait + Send) = std::mem::transmute(x);
//~^ this vtable is not allocated for a type that is `Send`!
}
```
Skip query in get_parent_item when possible.
For HirIds with a non-zero item local id, `self.parent_owner_iter(hir_id).next()` just returns the same HirId with the item local id set to 0, but also does a query to retrieve the Node which is ignored here, which seems wasteful.
Rework `non_local_definitions` lint to only use a syntactic heuristic
This PR reworks the `non_local_definitions` lint to only use a syntactic heuristic, i.e. not use a type-system logic for whenever an `impl` is local or not.
Instead the new logic wanted by T-lang in https://github.com/rust-lang/rust/issues/126768#issuecomment-2192634762, which is to consider every paths in `Self` and `Trait` and to no longer use the type-system inference trick.
`@rustbot` labels +L-non_local_definitions
Fixes#126768
add unqualified_local_imports lint
This lint helps deal with https://github.com/rust-lang/rustfmt/issues/4709 by having the compiler detect imports of local items that are not syntactically distinguishable from imports from other cates. Making them syntactically distinguishable ensures rustfmt can consistently apply the desired import grouping.
Rollup of 7 pull requests
Successful merges:
- #129201 (std: implement the `random` feature (alternative version))
- #130536 (bootstrap: Set the dylib path when building books with rustdoc)
- #130551 (Fix `break_last_token`.)
- #130657 (Remove x86_64-fuchsia and aarch64-fuchsia target aliases)
- #130721 (Add more test cases for block-no-opening-brace)
- #130736 (Add rustfmt 2024 reformatting to git blame ignore)
- #130746 (readd `@tgross35` and `@joboet` to the review rotation)
r? `@ghost`
`@rustbot` modify labels: rollup
Fix `break_last_token`.
It currently doesn't handle the three-char tokens `>>=` and `<<=` correctly. These can be broken twice, resulting in three individual tokens. This is a latent bug that currently doesn't cause any problems, but does cause problems for #124141, because that PR increases the usage of lazy token streams.
r? `@petrochenkov`
Apply `EarlyOtherwiseBranch` to scalar value
In the future, I'm thinking of hoisting discriminant via GVN so that we only need to write very little code here.
r? `@cjgillot`
Replace calls to `ty::Const::{try_}eval` in mir build/pattern analysis
We normalize consts in writeback: #130645. This means that consts are gonna be as normalized as they're ever gonna get in MIR building and pattern analysis. Therefore we can just use `try_to_target_usize` rather than calling `eval_target_usize`.
Regarding the `.expect` calls, I'm not totally certain whether they're correct given rigid unevaluated consts. But this PR shouldn't make *more* ICEs occur; we may have to squash these ICEs when mGCE comes around, tho 😺
Introduce `structurally_normalize_const`, use it in `rustc_hir_typeck`
Introduces `structurally_normalize_const` to typecking to separate the "eval a const" step from the "try to turn a valtree into a target usize" in HIR typeck, where we may still have infer vars and stuff around.
I also changed `check_expr_repeat` to move a double evaluation of a const into a single one. I'll leave inline comments.
r? ```@BoxyUwU```
I hesitated to really test this on the new solver where it probably matters for unevaluated consts. If you're worried about the side-effects, I'd be happy to craft some more tests 😄
Don't call `ty::Const::normalize` in error reporting
We do this to ensure that trait refs with unevaluated consts have those consts simplified to their evaluated forms. Instead, use `try_normalize_erasing_regions`.
**NOTE:** This has the side-effect of erasing regions from all of our trait refs. If this is too much to review or you think it's too opinionated of a diagnostics change, then I could split out the effective change (i.e. erasing regions from this impl suggestion) into another PR and have someone else review it.
No longer mark RTN as incomplete
The RFC is accepted and the feature is basically fully implemented. This doesn't mean it's necesarily *ready* for stabiliation; there's probably some diagnostic improvements to be made, and as always, users uncover the most creative bugs.
But marking this feature as incomplete no longer serves any purpose, so let's fix that.
Handle unsized consts with type `str` in v0 symbol mangling
This PR fixes#116303 by handling consts with type `str` in v0 symbol mangling as partial support for unsized consts.
This PR is related to `#![feature(adt_const_params)]` (#95174) and `#![feature(unsized_const_params)]` (#128028).
r? ``@BoxyUwU``
Add str.as_str() for easy Deref to string slices
Working with `Box<str>` is cumbersome, because in places like `iter.filter()` it can end up being `&Box<str>` or even `&&Box<str>`, and such type doesn't always get auto-dereferenced as expected.
Dereferencing such box to `&str` requires ugly syntax like `&**boxed_str` or `&***boxed_str`, with the exact amount of `*`s.
`Box<str>` is [not easily comparable with other string types](https://github.com/rust-lang/rust/pull/129852) via `PartialEq`. `Box<str>` won't work for lookups in types like `HashSet<String>`, because `Borrow<String>` won't take types like `&Box<str>`. OTOH `set.contains(s.as_str())` works nicely regardless of levels of indirection.
`String` has a simple solution for this: the `as_str()` method, and `Box<str>` should too.
Bump stage0 to beta-2024-09-22 and rustfmt to nightly-2024-09-22
I'm doing this to apply the changes to version sorting (https://github.com/rust-lang/rustfmt/pull/6284) that have occurred since rustfmt last upgraded (and a few other miscellaneous changes, like changes to expression overflowing: https://github.com/rust-lang/rustfmt/pull/6260). Eagerly updating rustfmt and formatting-the-world will ideally move some of the pressure off of the beta bump which will happen at the beginning of the next release cycle.
You can verify this is correct by checking out the changes, reverting the last commit, reapplying them, and diffing the changes:
```
git fetch git@github.com:compiler-errors/rust.git bump
git checkout -b bump FETCH_HEAD
git reset --hard HEAD~5
./x.py fmt --all
git diff FETCH_HEAD
# ignore the changes to stage0, and rustfmt.toml,
# and test file changes in rustdoc-js-std, run-make.
```
Or just take my word for it? Up to the reviewer.
r? release
It currently doesn't handle the three-char tokens `>>=` and `<<=`
correctly. These can be broken twice, resulting in three individual
tokens. This is a latent bug that currently doesn't cause any problems,
but does cause problems for #124141, because that PR increases the usage
of lazy token streams.
interpret: remove outdated FIXME
The rule about `repr(C)` types with compatible fields got removed from the ABI compat docs before they landed, so this FIXME here is no longer correct. (So this is basically a follow-up to https://github.com/rust-lang/rust/pull/130185, doing some more cleanup around deciding not to guarantee ABI compatibility for structurally compatible `repr(C)` types.)
fix rustc_nonnull_optimization_guaranteed docs
As far as I can tell, even back when this was [added](https://github.com/rust-lang/rust/pull/60300) it never *enabled* any optimizations. It just indicates that the FFI compat lint should accept those types for NPO.
Rollup of 5 pull requests
Successful merges:
- #130648 (move enzyme flags from general cargo to rustc-specific cargo)
- #130650 (Fixup Apple target's description strings)
- #130664 (Generate line numbers for non-rust code examples as well)
- #130665 (Prevent Deduplication of `LongRunningWarn`)
- #130669 (tests: Test that `extern "C" fn` ptrs lint on slices)
r? `@ghost`
`@rustbot` modify labels: rollup
Prevent Deduplication of `LongRunningWarn`
Fixes#118612
As mention in the issue, `LongRunningWarn` is meant to be repeated multiple times.
Therefore, this PR stores a unique number in every instance of `LongRunningWarn` so that it's not hashed into the same value and omitted by the deduplication mechanism.
Fixup Apple target's description strings
Noticed this inconsistency in how the Apple target's had their new descriptions written while looking at https://github.com/rust-lang/rust/pull/130614, and figured it was easy enough to fixup shortly. I think prefixing every OS with `Apple` is clearer, especially for less known ones like `visionOS` and `watchOS`; so that's what was done here along with making the architecture names more consistent and then some other small tweaks.
~~r? `@thomcc~~`
cc `@madsmtm`
rustc_llvm: adapt to flattened CLI args in LLVM
This changed in
llvm/llvm-project@e190d074a0. I decided to stick with more duplication between the ifdef blocks to make the code easier to read for the next two years before we can plausibly drop LLVM 19.
`@rustbot` label: +llvm-main
try-job: x86_64-msvc
rustc_expand: remember module `#[path]`s during expansion
During invocation collection, if a module item parsed from a `#[path]` attribute needed a second pass after parsing, its path wouldn't get added to the file path stack, so cycle detection broke. This checks the `#[path]` in such cases, so that it gets added appropriately. I think it should work identically to the case for external modules that don't need a second pass, but I'm not 100% sure.
Fixes#97589
Fix anon const def-creation when macros are involved take 2
Fixes#130321
There were two cases that #129137 did not handle correctly:
- Given a const argument `Foo<{ bar!() }>` in which `bar!()` expands to `N`, we would visit the anon const and then visit the `{ bar() }` expression instead of visiting the macro call. This meant that we would build a def for the anon const as `{ bar!() }` is not a trivial const argument as `bar!()` is not a path.
- Given a const argument `Foo<{ bar!() }>` is which `bar!()` expands to `{ qux!() }` in which `qux!()` expands to `N`, it should not be considered a trivial const argument as `{{ N }}` has two pairs of braces. If we only looked at `qux`'s expansion it would *look* like a trivial const argument even though it is not. We have to track whether we have "unwrapped" a brace already when recursing into the expansions of `bar`/`qux`/any macro
r? `@camelid`
Assert that `explicit_super_predicates_of` and `explicit_item_super_predicates` truly only contains bounds for the type itself
We distinguish _implied_ predicates (anything that is implied from elaborating a trait bound) from _super_ predicates, which are are the subset of implied predicates that share the same self type as the trait predicate we're elaborating. This was originally done in #107614, which fixed a large class of ICEs and strange errors where the compiler expected the self type of a trait predicate not to change when elaborating super predicates.
Specifically, super predicates are special for various reasons: they're the valid candidates for trait upcasting, are the only predicates we elaborate when doing closure signature inference, etc. So making sure that we get this list correct and don't accidentally "leak" any other predicates into this list is quite important.
This PR adds some debug assertions that we're in fact not doing so, and it fixes an oversight in the effect desugaring rework.
Implement Return Type Notation (RTN)'s path form in where clauses
Implement return type notation (RTN) in path position for where clauses. We already had RTN in associated type position ([e.g.](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=627a4fb8e2cb334863fbd08ed3722c09)), but per [the RFC](https://rust-lang.github.io/rfcs/3654-return-type-notation.html#where-rtn-can-be-used-for-now):
> As a standalone type, RTN can only be used as the Self type of a where-clause [...]
Specifically, in order to enable code like:
```rust
trait Foo {
fn bar() -> impl Sized;
}
fn is_send(_: impl Send) {}
fn test<T>()
where
T: Foo,
T::bar(..): Send,
{
is_send(T::bar());
}
```
* In the resolver, when we see a `TyKind::Path` whose final segment is `GenericArgs::ParenthesizedElided` (i.e. `(..)`), resolve that path in the *value* namespace, since we're looking for a method.
* When lowering where clauses in HIR lowering, we first try to intercept an RTN self type via `lower_ty_maybe_return_type_notation`. If we find an RTN type, we lower it manually in a way that respects its higher-ranked-ness (see below) and resolves to the corresponding RPITIT. Anywhere else, we'll emit the same "return type notation not allowed in this position yet" error we do when writing RTN in every other position.
* In `resolve_bound_vars`, we add some special treatment for RTN types in where clauses. Specifically, we need to add new lifetime variables to our binders for the early- and late-bound vars we encounter on the method. This implements the higher-ranked desugaring [laid out in the RFC](https://rust-lang.github.io/rfcs/3654-return-type-notation.html#converting-to-higher-ranked-trait-bounds).
This PR also adds a bunch of tests, mostly negative ones (testing error messages).
In a follow-up PR, I'm going to mark RTN as no longer incomplete, since this PR basically finishes the impl surface that we should initially stabilize, and the RFC was accepted.
cc [RFC 3654](https://github.com/rust-lang/rfcs/pull/3654) and https://github.com/rust-lang/rust/issues/109417
add `extern "C-cmse-nonsecure-entry" fn`
tracking issue #75835
in https://github.com/rust-lang/rust/issues/75835#issuecomment-1183517255 it was decided that using an abi, rather than an attribute, was the right way to go for this feature.
This PR adds that ABI and removes the `#[cmse_nonsecure_entry]` attribute. All relevant tests have been updated, some are now obsolete and have been removed.
Error 0775 is no longer generated. It contains the list of targets that support the CMSE feature, and maybe we want to still use this? right now a generic "this abi is not supported on this platform" error is returned when this abi is used on an unsupported platform. On the other hand, users of this abi are likely to be experienced rust users, so maybe the generic error is good enough.
Correct outdated object size limit
The comment here about 48 bit addresses being enough was written in 2016 but was made incorrect in 2019 by 5-level paging, and then persisted for another 5 years before being noticed and corrected.
The bolding of the "exclusive" part is merely to call attention to something I missed when reading it and doublechecking the math.
try-job: i686-msvc
try-job: test-various
Don't alloca for unused locals
We already have a concept of mono-unreachable basic blocks; this is primarily useful for ensuring that we do not compile code under an `if false`. But since we never gave locals the same analysis, a large local only used under an `if false` will still have stack space allocated for it.
There are 3 places we traverse MIR during monomorphization: Inside the collector, `non_ssa_locals`, and the walk to generate code. Unfortunately, https://github.com/rust-lang/rust/pull/129283#issuecomment-2297925578 indicates that we cannot afford the expense of tracking reachable locals during the collector's traversal, so we do need at least two mono-reachable traversals. And of course caching is of no help here because the benchmarks that regress are incr-unchanged; they don't do any codegen.
This fixes the second problem in https://github.com/rust-lang/rust/issues/129282, and brings us anther step toward `const if` at home.
Explain why `non_snake_case` is skipped for binary crates and cleanup tests
- Explain `non_snake_case` lint is skipped for bin crate names because binaries are not intended to be distributed or consumed like library crates (#45127).
- Coalesce the bunch of tests into a single one but with revisions, which is easier to compare the differences for `non_snake_case` behavior with respect to crate types.
Follow-up to #121749 with some more comments and test cleanup.
cc `@saethlin` who bumped into one of the tests and was confused why it was `only-x86_64-unknown-linux-gnu`.
try-job: dist-i586-gnu-i586-i686-musl
compiler: factor out `OVERFLOWING_LITERALS` impl
This puts it into `rustc_lint/src/types/literal.rs`. It then uses the fact that it's easier to navigate the logic to identify something that can easily be factored out, as an instance of "why".
Normalize consts in writeback when GCE is enabled
GCE lazily normalizes its unevaluated consts. This PR ensures that, like the new solver with its lazy norm types, we can assume that the writeback results are fully normalized.
This is important since we're trying to eliminate unnecessary calls to `ty::Const::{eval,normalize}` since they won't work with mGCE. Previously, we'd keep those consts unnormalized in writeback all the way through MIR build, and they'd only get normalized if we explicitly called `ty::Const::{eval,normalize}`, or during codegen since that calls `normalize_erasing_regions` (which invokes the `QueryNormalizer`, which evaluates the const accordingly).
This hack can (hopefully obviously) be removed when mGCE is implemented and we yeet the old GCE; it's only reachable with the GCE flag anyways, so I'm not worried about the implications here.
r? `@BoxyUwU`
Only expect valtree consts in codegen
Turn a bunch of `Const::eval_*` calls into `Const::try_to_*` calls, which implicitly assert that we only have valtrees by the time we get to codegen.
r? `@BoxyUwU`
Add recursion limit to FFI safety lint
Fixes#130310
Now we check against `tcx.recursion_limit()` and raise an error if it the limit is reached instead of overflowing the stack.
