* Reduce duplicate impls; show only the `fn (T)` and include a sentence
saying that there exists up to twelve of them.
* Show `Copy` and `Clone`.
* Show auto traits like `Send` and `Sync`, and blanket impls like `Any`.
Use constant eval to do strict mem::uninit/zeroed validity checks
I'm not sure about the code organisation here, I just dumped the check in rustc_const_eval at the root. Not hard to move it elsewhere, in any case.
Also, this means cranelift codegen intrinsics lose the strict checks, since they don't seem to depend on rustc_const_eval, and I didn't see a point in keeping around two copies.
I also left comments in the is_zero_valid methods about "uhhh help how do i do this", those apply to both methods equally.
Also rustc_codegen_ssa now depends on rustc_const_eval... is this okay?
Pinging `@RalfJung` since you were the one who mentioned this to me, so I'm assuming you're interested.
Haven't had a chance to run full tests on this since it's really warm, and it's 1AM, I'll check out any failures/comments in the morning :)
Add a special case for align_offset /w stride != 1
This generalizes the previous `stride == 1` special case to apply to any
situation where the requested alignment is divisible by the stride. This
in turn allows the test case from #98809 produce ideal assembly, along
the lines of:
leaq 15(%rdi), %rax
andq $-16, %rax
This also produces pretty high quality code for situations where the
alignment of the input pointer isn’t known:
pub unsafe fn ptr_u32(slice: *const u32) -> *const u32 {
slice.offset(slice.align_offset(16) as isize)
}
// =>
movl %edi, %eax
andl $3, %eax
leaq 15(%rdi), %rcx
andq $-16, %rcx
subq %rdi, %rcx
shrq $2, %rcx
negq %rax
sbbq %rax, %rax
orq %rcx, %rax
leaq (%rdi,%rax,4), %rax
Here LLVM is smart enough to replace the `usize::MAX` special case with
a branch-less bitwise-OR approach, where the mask is constructed using
the neg and sbb instructions. This appears to work across various
architectures I’ve tried.
This change ends up introducing more branches and code in situations
where there is less knowledge of the arguments. For example when the
requested alignment is entirely unknown. This use-case was never really
a focus of this function, so I’m not particularly worried, especially
since llvm-mca is saying that the new code is still appreciably faster,
despite all the new branching.
Fixes#98809.
Sadly, this does not help with #72356.
This generalizes the previous `stride == 1` special case to apply to any
situation where the requested alignment is divisible by the stride. This
in turn allows the test case from #98809 produce ideal assembly, along
the lines of:
leaq 15(%rdi), %rax
andq $-16, %rax
This also produces pretty high quality code for situations where the
alignment of the input pointer isn’t known:
pub unsafe fn ptr_u32(slice: *const u32) -> *const u32 {
slice.offset(slice.align_offset(16) as isize)
}
// =>
movl %edi, %eax
andl $3, %eax
leaq 15(%rdi), %rcx
andq $-16, %rcx
subq %rdi, %rcx
shrq $2, %rcx
negq %rax
sbbq %rax, %rax
orq %rcx, %rax
leaq (%rdi,%rax,4), %rax
Here LLVM is smart enough to replace the `usize::MAX` special case with
a branch-less bitwise-OR approach, where the mask is constructed using
the neg and sbb instructions. This appears to work across various
architectures I’ve tried.
This change ends up introducing more branches and code in situations
where there is less knowledge of the arguments. For example when the
requested alignment is entirely unknown. This use-case was never really
a focus of this function, so I’m not particularly worried, especially
since llvm-mca is saying that the new code is still appreciably faster,
despite all the new branching.
Fixes#98809.
Sadly, this does not help with #72356.
Fix flakyness of GUI tests
Fixes#98163.
All flaky tests seemed to be linked to the search. Since the search JS is loaded when we focus the search input, I think it's possible that we enter faster than the JS is actually loaded. The solution for that would be to do it in two steps: first we write into the search input (`browser-ui-test` adds a small sleep time after such commands) and then we press enter to be sure that it wasn't missed.
cc `@JohnTitor`
r? `@Dylan-DPC`
Revert "Highlight conflicting param-env candidates"
This reverts #98794, commit 08135254dc.
Seems to have caused an incremental compilation bug. The root cause of the incr comp bug is somewhat unrelated but is triggered by this PR, so I don't feel comfortable with having this PR in the codebase until it can be investigated further. Fixes#99233.
[rustdoc-json] JSON no longer inlines
Fixes#98007.
Fixes#96161.
Fixes https://github.com/rust-lang/rust/issues/83057.
Fixes https://github.com/rust-lang/rust/issues/83720.
I took over #93518 and applied the comments and added more tests.
There was one thing missing (which is in the second commit): if a non-exported item was used in a public API but not reexported, it was still missing.
cc `@CraftSpider` `@Urgau` `@Enselic`
r? `@notriddle`
migrate some of `rustc_passes::check_attr`'s diagnostics and derive improvements
- Implements `IntoDiagnosticArg` for `char` using its `Debug` implementation and introduces a macro for those types which just delegate the implementation to `ToString`.
- Apply the `#[rustc_lint_diagnostics]` attribute to `LintDiagnosticBuilder::build` so that diagnostic migration lints will trigger for it - some diagnostics in `rustc_privacy` need updated after this since the lints apply to that crate.
- Add support for `MultiSpan` with any of the attributes that work on a `Span` in the diagnostic derive (`SessionDiagnostic` + `LintDiagnostic`). Requires that diagnostic logic generated for these attributes are emitted in the by-move block rather than the by-ref block that they would normally have been generated in.
- Both diagnostic and subdiagnostic derives were missing the ability to add warnings to diagnostics - this is made more difficult by the `warn` attribute already existing, so this name being unavailable for the derives to use. `#[warn_]` is used instead, which requires special-casing so that `{span_,}warn` is called instead of `{span_,}warn_`.
- Migrate half of the `rustc_passes::check_attr` diagnostics to using diagnostic derives and being translatable. I got tired after a while. I modified some diagnostic output for consistency while doing this, nothing too crazy.
r? `@compiler-errors`
Allow destructuring opaque types in their defining scopes
fixes#96572
Before this PR, the following code snippet failed with an incomprehensible error, and similar code just ICEd in mir borrowck.
```rust
type T = impl Copy;
let foo: T = (1u32, 2u32);
let (a, b) = foo;
```
The problem was that the last line created MIR projections of the form `foo.0` and `foo.1`, but `foo`'s type is `T`, which doesn't have fields (only its hidden type does). But the pattern supplies enough type information (a tuple of two different inference types) to bind a hidden type.
Fix ICE in `named_arguments_used_positionally` lint
Fixes#99261Fixes#99289Fixes#99284Fixes#99273Fixes#99297Fixes#99271
This match pattern:
```
FormatSpec { width: Count::CountIsName(s, _), .. }
| FormatSpec { precision: Count::CountIsName(s, _), .. }
```
does not account for when both `width` and `precision` are both `Count::CountIsName`, so split the check for these two fields into two separate `if let`.
Also, remove any future potential for ICEs by removing the index operator altogether.
---
It is still suspicious that this indexing was broken and caused the ICE, as opposed to just causing a spurious lint message.
cc `@PrestonFrom,` who may be familiar with this code because of implementing the lint this touches, perhaps you'd like to look into why named arguments in `FormatSpec.precision` seem to have indices that don't correspond to a span in `Context.arg_spans`?
Edit: Opened #99265 to track a (related?) incorrect argument indexing issue.
Add Output = expected type trait obligation for known binary operators
This PR is a follow-on to #94034 that addresses #96442. That is, after replacing the trait-suggestion logic in `op.rs` with a more generic path that analyzes a general set of `Obligation`s, then we lost some specificity in the suggestions where the bounds on the associated type `Output=` would not get suggested.
This PR fixes this issue by changing `FnCtxt::construct_obligation_for_trait` to include a new `ProjectionPredicate` obligation for binary operators that obliges that `Output` is the same as the expected type of the expression. Additionally, to get the expected type of the expression, this PR threads the `Expectation<'tcx>` structure throughout several functions.
See src/test/ui/generic-associated-types/missing-bounds.stderr for an example of how this works.
One side effect of this change is it causes type-check failures with binops to include additional information. Specifically, many now say
```
error: type mismatch resolving `<Lhs as TheBinop>::Output == ExpectedTy`
```
It's up for discussion whether this added context is worth it to the user.
r? `@estebank`
Revert "Work around invalid DWARF bugs for fat LTO"
Since September, the toolchain has not been generating reliable DWARF
information for static variables when LTO is on. This has affected
projects in the embedded space where the use of LTO is typical. In our
case, it has kept us from bumping past the 2021-09-22 nightly toolchain
lest our debugger break. This has been a pretty dramatic regression for
people using debuggers and static variables. See #90357 for more info
and a repro case.
This commit is a mechanical revert of
d5de680e20 from PR #89041, which caused
the issue. (Note on that PR that the commit's author has requested it be
reverted.)
I have locally verified that this fixes#90357 by restoring the
functionality of both the repro case I posted on that bug, and debugger
behavior on real programs. There do not appear to be test cases for this
in the toolchain; if I've missed them, point me at 'em and I'll update
them.
Mark stabilized intrinsics with `rustc_allowed_through_unstable_modules`
Fixes#99286
PR #95956 accidentally made these intrinsics unstable when
accessed through the unstable path segment 'std::intrinsics'
We check that there's a single level of block nesting to ensure always
correct suggestions. If we don't, then we only provide a free-form
message to avoid misleading users in cases like
`src/test/ui/nll/borrowed-temporary-error.rs`.
We could expand the analysis to suggest hoising all of the relevant
parts of the users' code to make the code compile, but that could be
too much.
stop using `FnCtxt` outside of hir typeck
the requirements between **hir typeck**™, and **not hir typeck**™ are different enough to warrant a full split. with this PR `FnCtxt` is now only used for hir typeck (and for one clippy lint which seems to be emulating hir typeck).
Once this PR has landed I intend to move `FnCtxt` into a new crate. This should also allow some further general improvements here.
r? rust-lang/types
Both diagnostic and subdiagnostic derives were missing the ability to
add warnings to diagnostics - this is made more difficult by the `warn`
attribute already existing, so this name being unavailable for the
derives to use. `#[warn_]` is used instead, which requires
special-casing so that `{span_,}warn` is called instead of
`{span_,}warn_`.
Signed-off-by: David Wood <david.wood@huawei.com>
