Use `join_with_double_colon` in `write_shared.rs`.
For consistency. Also, it's faster because `join_with_double_colon` does a better job estimating the allocation size than `join` from `itertools`.
r? `@camelid`
Inspired by <https://github.com/rust-lang/rust/pull/138759#discussion_r2156375342> where I noticed that we were nearly at this point, plus the comments I was writing in 143410 that reminded me a type-dependent `true` is fine.
This PR splits the `OperandRef::builder` logic out to a separate type, with the updates needed to handle SIMD as well. In doing so, that makes the existing `Aggregate` path in `codegen_rvalue_operand` capable of handing SIMD values just fine.
As a result, we no longer need to do layout calculations for aggregate result types when running the analysis to determine which things can be SSA in codegen.
The mips64-unknown-linux-muslabi64 target is currently rather broken,
but I'm working on getting it fixed so that it can at least be used
again.
While I can't commit to maintaining the LLVM side of this target, I
don't mind looking into any other MIPS or musl-related issues that arise
with this target.
Signed-off-by: Jens Reidel <adrian@travitia.xyz>
Fix in std::String docs
This PR removes the word “else” from the sentence ('something else similar') in the String documentation to improve clarity.
Fixesrust-lang/rust#143579.
const-block-as-pattern: do not refer to no-longer-existing nightly feature
Surely everyone who used this nightly feature has fixed their code by now. So let's not confused people on stable that try to use a const block as a pattern by referring to some dead nightly feature.
Fix wrong cache event query key
I messed this up in https://github.com/rust-lang/rust/pull/142978. It is only an issue if someone enables the event manually, which almost no-one does, so it could take a while before we found it :D Luckily I noticed it while re-reading the PR.
r? `@oli-obk`
Point to correct argument in Func Call when Self type fails trait bound
Fixesrust-lang/rust#143336
When a trait bound fails due to the `Self` type parameter, `adjust_fulfillment_errors` now correctly points to the corresponding function argument instead of incorrectly pointing to other arguments. This is because `Call` may also need to handle the `self` parameter, and not just `MethodCall` needs to be handled, as rust-lang/rust#143336.
r? compiler
Originally, we used an array of `riscv_hwprobe` directly and indexed
using raw numbers, making correspondence between the index and the query
key less obvious.
We also frequently used `out[idx].key != -1` to test whether the key is
supported by the `riscv_hwprobe` system call (on the Linux kernel
version we are testing) but we'd better to integrate with an operation
to retrieve the value.
This commit improves the ergonomics of feature querying by:
1. Utilizing macros to
a. enable indexing by identifier names and
b. encapsulate accesses to the `riscv_hwprobe` array to query and
2. New method `riscv_hwprobe::get()` returning `Option<u64>`,
integrating availability checking and value retrieval.
It also removes `has_ima` for now because it's redundant if we only need
to test for single base behavior.
Makes it more convenient to test rustdoc on non x86_64. I mainly care
about the aarch64 dev-desktops.
This works because rustdoc uses all target features, not just that of
the target.
