Use `black_box` on SIMD intrinsic inputs to prevent the compiler from
constant folding SIMD operations, ensuring the corresponding SIMD
instructions are actually emitted and covered by tests.
Update `mir-opt` 64-bit panic-abort tests for `Alignment` rename
These seem to have been missed when the PR originally merged.
cc https://github.com/rust-lang/rust/pull/154004, @GrigorenkoPV
delete several `ui/consts` tests
* `tests/ui/consts/const.rs`: meaningless
* `tests/ui/consts/const-bound.rs`: has nothing to do with const bounds or const at all (anymore?)
* `tests/ui/consts/const-enum-tuplestruct2.rs`, `tests/ui/consts/const-enum-tuple2.rs`, `tests/ui/consts/const-enum-tuple.rs`: duplicates of `tests/ui/consts/const-enum-tuplestruct.rs`
r? @Kivooeo
Skip suggestions pointing to macro def for assert_eq
Fixesrust-lang/rust#146204
It's better to suggest:
```console
help: consider borrowing here
--> src/main.rs:3:16
|
3| assert_ne!(&buf, b"----");
| +
```
but i don't want to give a too heuristic but not general enough fix, let suppress them.
Fix regression when dealing with generics/values with unresolved inference
Follow up for rust-lang/rust#151703, fixing regression caused in rollup rust-lang/rust#152825
Forgot to handle generics & unresolved inference variables (as in `get_safe_transmute_error_and_reason`) in my previous PR. This followup checks for them before trying to normalize.
I am not completely sure its right approach to have this check cloned but as `select_transmute_obligation_for_reporting` fn just chooses obligation and doesn't actually return an error, this check shouldn't be removed from `get_safe_transmute_error_and_reasnon`. If there is any better solution, let me kmow.
Fixes: rust-lang/rust#153755
r? @jdonszelmann
Avoid ICE when param-env normalization leaves unresolved inference variables
Fixesrust-lang/rust#153354
Because the impl is already ill-formed, those variables are not fully constrained, so zfully_resolve` fails. We previously treated that as an immediate compiler bug with `span_bug!`, which caused an ICE on invalid input.
compiler-builtins: Clean up features
Remove the `compiler-builtins` feature from default because it prevents
testing via the default `cargo test` command. It made more sense as a
default when `compiler-builtins` was a dependency that some crates added
via crates.io, but is no longer needed.
The `rustc-dep-of-std` feature is also removed since it doesn't do
anything beyond what the `compiler-builtins` feature already does.
Update cargo submodule
11 commits in e84cb639edfea2c42efd563b72a9be0cc5de6523..888f675344eb1cf2308fd53183e667bdd2c58e51
2026-03-21 01:27:07 +0000 to 2026-03-30 16:59:25 +0000
- chore: bump to curl-sys@0.4.87 (rust-lang/cargo#16808)
- fix(resolver): Better match rustc in error styling (rust-lang/cargo#16795)
- missing -Zjson-target-spec error: mention that this is a cargo flag (rust-lang/cargo#16793)
- Update changelog for 1.94.1 (rust-lang/cargo#16789)
- test(build): Make it easier to review `build.warnings` behavior (rust-lang/cargo#16788)
- test(build): Cover more behavior of `build.warnings` (rust-lang/cargo#16785)
- chore(deps): bump curl-sys from 0.4.83 to 0.4.86 (rust-lang/cargo#16786)
- Warn when installing with a non-default toolchain (rust-lang/cargo#16131)
- feat(cli): Add support for completing `--config` argument values with `native-completions` (rust-lang/cargo#16249)
- feat(cli): complete `--config` and `--color` before command (rust-lang/cargo#16780)
- chore(deps): bump rustls-webpki from 0.103.9 to 0.103.10 (rust-lang/cargo#16778)
Remove the `compiler-builtins` feature from default because it prevents
testing via the default `cargo test` command. It made more sense as a
default when `compiler-builtins` was a dependency that some crates added
via crates.io, but is no longer needed.
The `rustc-dep-of-std` feature is also removed since it doesn't do
anything beyond what the `compiler-builtins` feature already does.
Rollup of 4 pull requests
Successful merges:
- rust-lang/rust#154548 (Add regression test for TransmuteFrom ICE with min_generic_const_args)
- rust-lang/rust#154563 (Point at binop lhs and rhs when expression is multiline)
- rust-lang/rust#154564 (Tweak wording of E0275 WF errors)
- rust-lang/rust#154566 (loongarch: use "_mcount" as the default mcount symbol)
loongarch: use "_mcount" as the default mcount symbol
Set the `mcount` field to "_mcount" for all LoongArch targets to match Clang/GCC behavior [1]. This fixes linking failures when using `-Z instrument-mcount`, where the runtime expects "_mcount" instead of "mcount".
[1] https://github.com/llvm/llvm-project/blob/main/clang/lib/Basic/Targets/LoongArch.h#L60
Tweak wording of E0275 WF errors
Modify the main error message to read better:
```
error[E0275]: overflow evaluating whether `&'a mut Bar` is well-formed
```
Folkert de Vries