1.38 release post #405
1.38 release post #405
Conversation
|
(rust_highfive has picked a reviewer for you, use r? to override) |
posts/2019-09-26-Rust-1.38.0.md
Outdated
|
|
||
| ### Linting some incorrect uses of `mem::{uninitialized, zeroed}` | ||
|
|
||
| As [previously announced](https://blog.rust-lang.org/2019/07/04/Rust-1.36.0.html#maybeuninitt%3E-instead-of-mem::uninitialized), `std::mem::uninitialized` is hard to use soundly. Instead, `MaybeUninit<T>` should be used. |
There was a problem hiding this comment.
The 1.36 announcement says that it is "essentially impossible" to use soundly. But it's not unsound to use to initialize plain-old-data (in the C++ sense), is it? If that's correct, then "hard to use soundly" seems reasonable; if not, then maybe we should go back to using stronger language.
There was a problem hiding this comment.
I'd be fine with "essentially impossible" as well -- I tweaked this because it sounded 100% certain but there are cases where it is OK... for example: mem::uninitialized::<MaybeUninit<u8>>(). ;)
Actually, let's change back to "essentially impossible".
There was a problem hiding this comment.
| As [previously announced](https://blog.rust-lang.org/2019/07/04/Rust-1.36.0.html#maybeuninitt%3E-instead-of-mem::uninitialized), `std::mem::uninitialized` is hard to use soundly. Instead, `MaybeUninit<T>` should be used. | |
| As [previously announced](https://blog.rust-lang.org/2019/07/04/Rust-1.36.0.html#maybeuninitt%3E-instead-of-mem::uninitialized), `std::mem::uninitialized` is essentially impossible to use safely. Instead, `MaybeUninit<T>` should be used. |
Mark-Simulacrum
force-pushed
the
1.38
branch
from
7931db4 to
38216cb
Compare
|
This should be ready for the release, and will be merged in ~2 hours. |
cc @rust-lang/release @rust-lang/core