| 1 |
# `bytemuck` changelog |
| 2 |
|
| 3 |
## 1.21 |
| 4 |
|
| 5 |
* Implement `Pod` and `Zeroable` for `core::arch::{x86, x86_64}::__m512`, `__m512d` and `__m512i` without nightly. |
| 6 |
Requires Rust 1.72, and is gated through the `avx512_simd` cargo feature. |
| 7 |
* Allow the use of `must_cast_mut` and `must_cast_slice_mut` in const contexts. |
| 8 |
Requires Rust 1.83, and is gated through the `must_cast_extra` cargo feature. |
| 9 |
* internal: introduced the `maybe_const_fn` macro that allows defining some function |
| 10 |
to be const depending upon some `cfg` predicate. |
| 11 |
|
| 12 |
## 1.20 |
| 13 |
|
| 14 |
* New functions to allocate zeroed `Arc` and `Rc`. Requires Rust 1.82 |
| 15 |
* `TransparentWrapper` impls for `core::cmp::Reverse` and `core::num::Saturating`. |
| 16 |
* internal: Simplified the library's `fill_zeroes` calls to `write_bytes` |
| 17 |
|
| 18 |
## 1.19 |
| 19 |
|
| 20 |
* Adds the `#[track_caller]` attribute to functions which may panic. |
| 21 |
|
| 22 |
## 1.18 |
| 23 |
|
| 24 |
* Adds the `latest_stable_rust` cargo feature, which is a blanket feature that turns all other features on that are both sound and compatible with Stable rust. |
| 25 |
|
| 26 |
## 1.17.1 |
| 27 |
|
| 28 |
* Adds `#[repr(C)]` to the `union Transmute<A, B>` type that's used internally |
| 29 |
for most of the transmutations. |
| 30 |
|
| 31 |
## 1.17.0 |
| 32 |
|
| 33 |
* Makes the `must_cast` versions of the by-value and by-ref casts be `const`. |
| 34 |
The mut ref cast is unaffected for now (mut references aren't yet stable in `const fn`). |
| 35 |
This increases the MSRV of using that particular feature from 1.57 to 1.64. |
| 36 |
|
| 37 |
## 1.16.3 |
| 38 |
|
| 39 |
* Fully described in https://github.com/Lokathor/bytemuck/pull/256, This makes |
| 40 |
casting slices to/from ZST elements more consistent between the crate's core |
| 41 |
module and other modules. |
| 42 |
|
| 43 |
## 1.16.2 |
| 44 |
|
| 45 |
* Fixes potential UB where `BoxBytes` could attempt to free a dangling pointer |
| 46 |
if the `Layout` is zero sized. This type was introduced in 1.14.1, so that |
| 47 |
version and the others up to and including 1.16.1 are now yanked for safety. |
| 48 |
|
| 49 |
## 1.16.1 |
| 50 |
|
| 51 |
* **NOT SEMVER SUPPORTED:** Adds the `nightly_float` Cargo feature. This |
| 52 |
activates the `f16` and `f128` nightly features, and then provides `Zeroable` |
| 53 |
and `Pod` impls for those types. |
| 54 |
|
| 55 |
## 1.16.0 |
| 56 |
|
| 57 |
* Adds a `const_zeroed` feature (MSRV 1.75) which puts a `zeroed` fn at the crate root. |
| 58 |
This is just like the `Zeroable::zeroed` method, but as a `const fn`. |
| 59 |
|
| 60 |
## 1.15.0 |
| 61 |
|
| 62 |
This primarily relaxes the bounds on a `From` impl. |
| 63 |
|
| 64 |
Previously: |
| 65 |
|
| 66 |
> `impl<T: NoUninit> From<Box<T>> for BoxBytes` |
| 67 |
|
| 68 |
Now: |
| 69 |
|
| 70 |
> `impl<T: ?Sized + sealed::BoxBytesOf> From<Box<T>> for BoxBytes` |
| 71 |
|
| 72 |
All related functions and methods are similarly updated. |
| 73 |
|
| 74 |
We believe this to be backwards compatible with all previous uses, |
| 75 |
and now `BoxBytes` can be converted to/from more types than before. |
| 76 |
|
| 77 |
## 1.14.3 |
| 78 |
|
| 79 |
* The new std simd nightly features are apparently arch-specific. |
| 80 |
This adjusts the feature activation to be x86/ x86_64 only. |
| 81 |
|
| 82 |
## 1.14.2 |
| 83 |
|
| 84 |
* Changes the name of the Nightly feature activated by the crate's |
| 85 |
`nightly_stdsimd` feature. This is needed as of (approximately) Nightly |
| 86 |
2024-02-06 and later, because the Nightly feature was changed. |
| 87 |
|
| 88 |
## 1.14.1 |
| 89 |
|
| 90 |
* docs clarifications. |
| 91 |
|
| 92 |
## 1.14 |
| 93 |
|
| 94 |
* `write_zeroes` and `fill_zeroes` functions: Writes (to one) or fills (a slice) |
| 95 |
zero bytes to all bytes covered by the provided reference. If your type has |
| 96 |
padding, this will even zero out the padding bytes. |
| 97 |
* `align_offset` feature: causes pointer alignment checks to use the |
| 98 |
`align_offset` pointer method rather than as-casting the pointer to `usize`. |
| 99 |
This *may* improve codegen, if the compiler would have otherwise thought that |
| 100 |
the pointer address escaped. No formal benchmarks have been done either way. |
| 101 |
* `must_cast` feature: Adds `must_*` family of functions. These functions will |
| 102 |
fail to compile if the cast requested can't be statically known to succeed. |
| 103 |
The error messages can be kinda bad when this happens, but eliminating the |
| 104 |
possibility of a runtime error might be worth it to you. |
| 105 |
|
| 106 |
## 1.13.1 |
| 107 |
|
| 108 |
* Remove the requirement for the *source* data type to be `AnyBitPattern` on |
| 109 |
`pod_collect_to_vec`, allowing you to pod collect vecs of `char` into vecs of |
| 110 |
`u32`, or whatever. |
| 111 |
|
| 112 |
## 1.13 |
| 113 |
|
| 114 |
* Now depends on `bytemuck_derive-1.4.0` |
| 115 |
* Various small enhancements that would have been patch version updates, but |
| 116 |
which have been rolled into this minor version update. |
| 117 |
|
| 118 |
## 1.12.4 |
| 119 |
|
| 120 |
* This has additional impls for existing traits and cleans up some internal code, |
| 121 |
but there's no new functions so I guess it counts as just a patch release. |
| 122 |
|
| 123 |
## 1.12.3 |
| 124 |
|
| 125 |
* This bugfix makes the crate do stuff with `Arc` or not based on the |
| 126 |
`target_has_atomic` config. Previously, some targets that have allocation but |
| 127 |
not atomics were getting errors. This raises the MSRV of the |
| 128 |
`extern_crate_alloc` feature to 1.60, but opt-in features are *not* considered |
| 129 |
to be hard locked to 1.34 like the basic build of the crate is. |
| 130 |
|
| 131 |
## 1.12.2 |
| 132 |
|
| 133 |
* Fixes `try_pod_read_unaligned` bug that made it always fail unless the target |
| 134 |
type was exactly pointer sized in which case UB *could* happen. The |
| 135 |
`CheckedBitPattern::is_valid_bit_pattern` was being asked to check that a |
| 136 |
*reference* to the `pod` value was a valid bit pattern, rather than the actual |
| 137 |
bit pattern itself, and so the check could in some cases be illegally |
| 138 |
bypassed. |
| 139 |
|
| 140 |
## 1.12.1 |
| 141 |
|
| 142 |
* Patch bumped the required `bytemuck_derive` version because of a regression in |
| 143 |
how it handled `align(N)` attributes. |
| 144 |
|
| 145 |
## 1.12 |
| 146 |
|
| 147 |
* This minor version bump is caused by a version bump in our `bytemuck_derive` |
| 148 |
dependency, which is in turn caused by a mixup in the minimum version of `syn` |
| 149 |
that `bytemuck_derive` uses. See [Issue |
| 150 |
122](https://github.com/Lokathor/bytemuck/issues/122). There's not any |
| 151 |
specific "new" API as you might normally expect from a minor version bump. |
| 152 |
* [pali](https://github.com/pali6) fixed a problem with SPIR-V builds being |
| 153 |
broken. The error handling functions were trying to be generic over `Display`, |
| 154 |
which the error types normally support, except on SPIR-V targets (which run on |
| 155 |
the GPU and don't have text formatting). |
| 156 |
|
| 157 |
## 1.11 |
| 158 |
|
| 159 |
* [WaffleLapkin](https://github.com/WaffleLapkin) added `wrap_box` and `peel_box` |
| 160 |
to the `TransparentWrapperAlloc` trait. Default impls of these functions are |
| 161 |
provided, and (as usual with the transparent trait stuff) you should not override |
| 162 |
the default versions. |
| 163 |
|
| 164 |
## 1.10 |
| 165 |
|
| 166 |
* [TheEdward162](https://github.com/TheEdward162) added the `ZeroableInOption` |
| 167 |
and `PodInOption` traits. These are for types that are `Zeroable` or `Pod` |
| 168 |
*when in an option*, but not on their own. We provide impls for the various |
| 169 |
"NonZeroINTEGER" types in `core`, and if you need to newtype a NonZero value |
| 170 |
then you can impl these traits when you use `repr(transparent)`. |
| 171 |
|
| 172 |
## 1.9.1 |
| 173 |
|
| 174 |
* Bumped the minimum `bytemuck_derive` dependency version from `1.0` to `1.1`. |
| 175 |
The fact that `bytemuck` and `bytemuck_derive` are separate crates at all is |
| 176 |
an unfortunate technical limit of current Rust, woe and calamity. |
| 177 |
|
| 178 |
## 1.9.0 |
| 179 |
|
| 180 |
* [fu5ha](https://github.com/fu5ha) added the `NoUninit`, `AnyBitPattern`, and |
| 181 |
`CheckedBitPattern` traits. This allows for a more fine-grained level of |
| 182 |
detail in what casting operations are allowed for a type. Types that already |
| 183 |
implement `Zeroable` and `Pod` will have a blanket impl for these new traits. |
| 184 |
This is a "preview" of the direction that the crate will probably go in the |
| 185 |
eventual 2.0 version. We're still waiting on [Project Safe |
| 186 |
Transmute](https://github.com/rust-lang/project-safe-transmute) for an actual |
| 187 |
2.0 version of the crate, but until then please enjoy this preview. |
| 188 |
* Also Fusha added better support for `union` types in the derive macros. I |
| 189 |
still don't know how any of the proc-macro stuff works at all, so please |
| 190 |
direct questions to her. |
| 191 |
|
| 192 |
## 1.8.0 |
| 193 |
|
| 194 |
* `try_pod_read_unaligned` and `pod_read_unaligned` let you go from `&[u8]` to |
| 195 |
`T:Pod` without worrying about alignment. |
| 196 |
|
| 197 |
## 1.7.3 |
| 198 |
|
| 199 |
* Experimental support for the `portable_simd` language extension under the |
| 200 |
`nightly_portable_simd` cargo feature. As the name implies, this is an |
| 201 |
experimental crate feature and it's **not** part of the semver contract. All |
| 202 |
it does is add the appropriate `Zeroable` and `Pod` impls. |
| 203 |
|
| 204 |
## 1.7.2 |
| 205 |
|
| 206 |
* Why does this repo keep being hit with publishing problems? What did I do to |
| 207 |
deserve this curse, Ferris? This doesn't ever happen with tinyvec or fermium, |
| 208 |
only bytemuck. |
| 209 |
|
| 210 |
## 1.7.1 |
| 211 |
|
| 212 |
* **Soundness Fix:** The wrap/peel methods for owned value conversion, added to |
| 213 |
`TransparentWrapper` in 1.6, can cause a double-drop if used with types that |
| 214 |
impl `Drop`. The fix was simply to add a `ManuallyDrop` layer around the value |
| 215 |
before doing the `transmute_copy` that is used to wrap/peel. While this fix |
| 216 |
could technically be backported to the 1.6 series, since 1.7 is semver |
| 217 |
compatible anyway the 1.6 series has simply been yanked. |
| 218 |
|
| 219 |
## 1.7 |
| 220 |
|
| 221 |
* In response to [Unsafe Code Guidelines Issue |
| 222 |
#286](https://github.com/rust-lang/unsafe-code-guidelines/issues/286), this |
| 223 |
version of Bytemuck has a ***Soundness-Required Breaking Change***. This is |
| 224 |
"allowed" under Rust's backwards-compatibility guidelines, but it's still |
| 225 |
annoying of course so we're trying to keep the damage minimal. |
| 226 |
* **The Reason:** It turns out that pointer values should not have been `Pod`. More |
| 227 |
specifically, `ptr as usize` is *not* the same operation as calling |
| 228 |
`transmute::<_, usize>(ptr)`. |
| 229 |
* LLVM has yet to fully sort out their story, but until they do, transmuting |
| 230 |
pointers can cause miscompilations. They may fix things up in the future, |
| 231 |
but we're not gonna just wait and have broken code in the mean time. |
| 232 |
* **The Fix:** The breaking change is that the `Pod` impls for `*const T`, |
| 233 |
`*mut T`, and `Option<NonNull<T>` are now gated behind the |
| 234 |
`unsound_ptr_pod_impl` feature, which is off by default. |
| 235 |
* You are *strongly discouraged* from using this feature, but if a dependency |
| 236 |
of yours doesn't work when you upgrade to 1.7 because it relied on pointer |
| 237 |
casting, then you might wish to temporarily enable the feature just to get |
| 238 |
that dependency to build. Enabled features are global across all users of a |
| 239 |
given semver compatible version, so if you enable the feature in your own |
| 240 |
crate, your dependency will also end up getting the feature too, and then |
| 241 |
it'll be able to compile. |
| 242 |
* Please move away from using this feature as soon as you can. Consider it to |
| 243 |
*already* be deprecated. |
| 244 |
* [PR 65](https://github.com/Lokathor/bytemuck/pull/65) |
| 245 |
|
| 246 |
## 1.6.3 |
| 247 |
|
| 248 |
* Small goof with an errant `;`, so [PR 69](https://github.com/Lokathor/bytemuck/pull/69) |
| 249 |
*actually* got things working on SPIR-V. |
| 250 |
|
| 251 |
## 1.6.2 |
| 252 |
|
| 253 |
cargo upload goof! ignore this one. |
| 254 |
|
| 255 |
## 1.6.1 |
| 256 |
|
| 257 |
* [DJMcNab](https://github.com/DJMcNab) did a fix so that the crate can build for SPIR-V |
| 258 |
[PR 67](https://github.com/Lokathor/bytemuck/pull/67) |
| 259 |
|
| 260 |
## 1.6 |
| 261 |
|
| 262 |
* The `TransparentWrapper` trait now has more methods. More ways to wrap, and |
| 263 |
now you can "peel" too! Note that we don't call it "unwrap" because that name |
| 264 |
is too strongly associated with the Option/Result methods. |
| 265 |
Thanks to [LU15W1R7H](https://github.com/LU15W1R7H) for doing |
| 266 |
[PR 58](https://github.com/Lokathor/bytemuck/pull/58) |
| 267 |
* Min Const Generics! Now there's Pod and Zeroable for arrays of any size when |
| 268 |
you turn on the `min_const_generics` crate feature. |
| 269 |
[zakarumych](https://github.com/zakarumych) got the work started in |
| 270 |
[PR 59](https://github.com/Lokathor/bytemuck/pull/59), |
| 271 |
and [chorman0773](https://github.com/chorman0773) finished off the task in |
| 272 |
[PR 63](https://github.com/Lokathor/bytemuck/pull/63) |
| 273 |
|
| 274 |
## 1.5.1 |
| 275 |
|
| 276 |
* Fix `bytes_of` failing on zero sized types. |
| 277 |
[PR 53](https://github.com/Lokathor/bytemuck/pull/53) |
| 278 |
|
| 279 |
## 1.5 |
| 280 |
|
| 281 |
* Added `pod_collect_to_vec`, which will gather a slice into a vec, |
| 282 |
allowing you to change the pod type while also safely ignoring alignment. |
| 283 |
[PR 50](https://github.com/Lokathor/bytemuck/pull/50) |
| 284 |
|
| 285 |
## 1.4.2 |
| 286 |
|
| 287 |
* [Kimundi](https://github.com/Kimundi) fixed an issue that could make `try_zeroed_box` |
| 288 |
stack overflow for large values at low optimization levels. |
| 289 |
[PR 43](https://github.com/Lokathor/bytemuck/pull/43) |
| 290 |
|
| 291 |
## 1.4.1 |
| 292 |
|
| 293 |
* [thomcc](https://github.com/thomcc) fixed up the CI and patched over a soundness hole in `offset_of!`. |
| 294 |
[PR 38](https://github.com/Lokathor/bytemuck/pull/38) |
| 295 |
|
| 296 |
## 1.4 |
| 297 |
|
| 298 |
* [icewind1991](https://github.com/icewind1991) has contributed the proc-macros |
| 299 |
for deriving impls of `Pod`, `TransparentWrapper`, `Zeroable`!! Everyone has |
| 300 |
been waiting for this one folks! It's a big deal. Just enable the `derive` |
| 301 |
cargo feature and then you'll be able to derive the traits on your types. It |
| 302 |
generates all the appropriate tests for you. |
| 303 |
* The `zeroable_maybe_uninit` feature now adds a `Zeroable` impl to the |
| 304 |
`MaybeUninit` type. This is only behind a feature flag because `MaybeUninit` |
| 305 |
didn't exist back in `1.34.0` (the minimum rust version of `bytemuck`). |
| 306 |
|
| 307 |
## 1.3.1 |
| 308 |
|
| 309 |
* The entire crate is now available under the `Apache-2.0 OR MIT` license as |
| 310 |
well as the previous `Zlib` license |
| 311 |
[#24](https://github.com/Lokathor/bytemuck/pull/24). |
| 312 |
* [HeroicKatora](https://github.com/HeroicKatora) added the |
| 313 |
`try_zeroed_slice_box` function |
| 314 |
[#10](https://github.com/Lokathor/bytemuck/pull/17). `zeroed_slice_box` is |
| 315 |
also available. |
| 316 |
* The `offset_of!` macro now supports a 2-arg version. For types that impl |
| 317 |
Default, it'll just make an instance using `default` and then call over to the |
| 318 |
3-arg version. |
| 319 |
* The `PodCastError` type now supports `Hash` and `Display`. Also if you enable |
| 320 |
the `extern_crate_std` feature then it will support `std::error::Error`. |
| 321 |
* We now provide a `TransparentWrapper<T>` impl for `core::num::Wrapper<T>`. |
| 322 |
* The error type of `try_from_bytes` and `try_from_bytes_mut` when the input |
| 323 |
isn't aligned has been corrected from being `AlignmentMismatch` (intended for |
| 324 |
allocation casting only) to `TargetAlignmentGreaterAndInputNotAligned`. |
| 325 |
|
| 326 |
## 1.3.0 |
| 327 |
|
| 328 |
* Had a bug because the CI was messed up! It wasn't soundness related, because |
| 329 |
it prevented the crate from building entirely if the `extern_crate_alloc` |
| 330 |
feature was used. Still, this is yanked, sorry. |
| 331 |
|
| 332 |
## 1.2.0 |
| 333 |
|
| 334 |
* [thomcc](https://github.com/thomcc) added many things: |
| 335 |
* A fully sound `offset_of!` macro |
| 336 |
[#10](https://github.com/Lokathor/bytemuck/pull/10) |
| 337 |
* A `Contiguous` trait for when you've got enums with declared values |
| 338 |
all in a row [#12](https://github.com/Lokathor/bytemuck/pull/12) |
| 339 |
* A `TransparentWrapper` marker trait for when you want to more clearly |
| 340 |
enable adding and removing a wrapper struct to its inner value |
| 341 |
[#15](https://github.com/Lokathor/bytemuck/pull/15) |
| 342 |
* Now MIRI is run on CI in every single push! |
| 343 |
[#16](https://github.com/Lokathor/bytemuck/pull/16) |
| 344 |
|
| 345 |
## 1.1.0 |
| 346 |
|
| 347 |
* [SimonSapin](https://github.com/SimonSapin) added `from_bytes`, |
| 348 |
`from_bytes_mut`, `try_from_bytes`, and `try_from_bytes_mut` ([PR |
| 349 |
Link](https://github.com/Lokathor/bytemuck/pull/8)) |
| 350 |
|
| 351 |
## 1.0.1 |
| 352 |
|
| 353 |
* Changed to the [zlib](https://opensource.org/licenses/Zlib) license. |
| 354 |
* Added much more proper documentation. |
| 355 |
* Reduced the minimum Rust version to 1.34 |