const-eval: fix and re-enable pointer fragment support

The pointer fragment support from rust-lang#144081 got disabled due to rust-lang#146291. This brings it back. To fix the issue, the per-byte provenance fragment tracking tracks *both* the provenance and raw address of the full pointer, so we can ensure that only fragments that are truly part of the same pointer are being merged.

r? `@oli-obk`
Cc `@theemathas`
Fixes rust-lang/const-eval#72 again.
Also fixes rust-lang#147959.

`@traviscross` I assume this won't need another t-lang FCP since it already got FCP'd in rust-lang#144081?

This MR contains the following updates:

| Package | Update | Change |
|---|---|---|
| [rust](https://github.com/rust-lang/rust) | minor | `1.92.0` → `1.93.0` |

MR created with the help of [el-capitano/tools/renovate-bot](https://gitlab.com/el-capitano/tools/renovate-bot).

**Proposed changes to behavior should be submitted there as MRs.**

---

### Release Notes

<details>
<summary>rust-lang/rust (rust)</summary>

### [`v1.93.0`](https://github.com/rust-lang/rust/blob/HEAD/RELEASES.md#Version-1930-2026-01-22)

[Compare Source](rust-lang/rust@1.92.0...1.93.0)

\==========================

<a id="1.93.0-Language"></a>

## Language

- [Stabilize several s390x `vector`-related target features and the `is_s390x_feature_detected!` macro](rust-lang/rust#145656)
- [Stabilize declaration of C-style variadic functions for the `system` ABI](rust-lang/rust#145954)
- [Emit error when using some keyword as a `cfg` predicate](rust-lang/rust#146978)
- [Stabilize `asm_cfg`](rust-lang/rust#147736)
- [During const-evaluation, support copying pointers byte-by-byte](rust-lang/rust#148259)
- [LUB coercions now correctly handle function item types, and functions with differing safeties](rust-lang/rust#148602)
- [Allow `const` items that contain mutable references to `static` (which is *very* unsafe, but not *always* UB)](rust-lang/rust#148746)
- [Add warn-by-default `const_item_interior_mutations` lint to warn against calls which mutate interior mutable `const` items](rust-lang/rust#148407)
- [Add warn-by-default `function_casts_as_integer` lint](rust-lang/rust#141470)

<a id="1.93.0-Compiler"></a>

## Compiler

- [Stabilize `-Cjump-tables=bool`](rust-lang/rust#145974). The flag was previously called `-Zno-jump-tables`.

<a id="1.93.0-Platform-Support"></a>

## Platform Support

- [Promote `riscv64a23-unknown-linux-gnu` to Tier 2 (without host tools)](rust-lang/rust#148435)

Refer to Rust's [platform support page][platform-support-doc]
for more information on Rust's tiered platform support.

[platform-support-doc]: https://doc.rust-lang.org/rustc/platform-support.html

<a id="1.93.0-Libraries"></a>

## Libraries

- [Stop internally using `specialization` on the `Copy` trait as it is unsound in the presence of lifetime dependent `Copy` implementations. This may result in some performance regressions as some standard library APIs may now call `Clone::clone` instead of performing bitwise copies](rust-lang/rust#135634)
- [Allow the global allocator to use thread-local storage and `std::thread::current()`](rust-lang/rust#144465)
- [Make `BTree::append` not update existing keys when appending an entry which already exists](rust-lang/rust#145628)
- [Don't require `T: RefUnwindSafe` for `vec::IntoIter<T>: UnwindSafe`](rust-lang/rust#145665)

<a id="1.93.0-Stabilized-APIs"></a>

## Stabilized APIs

- [`<[MaybeUninit<T>]>::assume_init_drop`](https://doc.rust-lang.org/stable/core/primitive.slice.html#method.assume_init_drop)
- [`<[MaybeUninit<T>]>::assume_init_ref`](https://doc.rust-lang.org/stable/core/primitive.slice.html#method.assume_init_ref)
- [`<[MaybeUninit<T>]>::assume_init_mut`](https://doc.rust-lang.org/stable/core/primitive.slice.html#method.assume_init_mut)
- [`<[MaybeUninit<T>]>::write_copy_of_slice`](https://doc.rust-lang.org/stable/std/primitive.slice.html#method.write_copy_of_slice)
- [`<[MaybeUninit<T>]>::write_clone_of_slice`](https://doc.rust-lang.org/stable/std/primitive.slice.html#method.write_clone_of_slice)
- [`String::into_raw_parts`](https://doc.rust-lang.org/stable/std/string/struct.String.html#method.into_raw_parts)
- [`Vec::into_raw_parts`](https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.into_raw_parts)
- [`<iN>::unchecked_neg`](https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_neg)
- [`<iN>::unchecked_shl`](https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_shl)
- [`<iN>::unchecked_shr`](https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_shr)
- [`<uN>::unchecked_shl`](https://doc.rust-lang.org/stable/std/primitive.usize.html#method.unchecked_shl)
- [`<uN>::unchecked_shr`](https://doc.rust-lang.org/stable/std/primitive.usize.html#method.unchecked_shr)
- [`<[T]>::as_array`](https://doc.rust-lang.org/stable/std/primitive.slice.html#method.as_array)
- [`<[T]>::as_array_mut`](https://doc.rust-lang.org/stable/std/primitive.slice.html#method.as_mut_array)
- [`<*const [T]>::as_array`](https://doc.rust-lang.org/stable/std/primitive.pointer.html#method.as_array)
- [`<*mut [T]>::as_array_mut`](https://doc.rust-lang.org/stable/std/primitive.pointer.html#method.as_mut_array)
- [`VecDeque::pop_front_if`](https://doc.rust-lang.org/stable/std/collections/struct.VecDeque.html#method.pop_front_if)
- [`VecDeque::pop_back_if`](https://doc.rust-lang.org/stable/std/collections/struct.VecDeque.html#method.pop_back_if)
- [`Duration::from_nanos_u128`](https://doc.rust-lang.org/stable/std/time/struct.Duration.html#method.from_nanos_u128)
- [`char::MAX_LEN_UTF8`](https://doc.rust-lang.org/stable/std/primitive.char.html#associatedconstant.MAX_LEN_UTF8)
- [`char::MAX_LEN_UTF16`](https://doc.rust-lang.org/stable/std/primitive.char.html#associatedconstant.MAX_LEN_UTF16)
- [`std::fmt::from_fn`](https://doc.rust-lang.org/stable/std/fmt/fn.from_fn.html)
- [`std::fmt::FromFn`](https://doc.rust-lang.org/stable/std/fmt/struct.FromFn.html)

<a id="1.93.0-Cargo"></a>

## Cargo

- [Enable CARGO\_CFG\_DEBUG\_ASSERTIONS in build scripts based on profile](rust-lang/cargo#16160)
- [In `cargo tree`, support long forms for `--format` variables](rust-lang/cargo#16204)
- [Add `--workspace` to `cargo clean`](rust-lang/cargo#16263)

<a id="1.93.0-Rustdoc"></a>

## Rustdoc

- [Remove `#![doc(document_private_items)]`](rust-lang/rust#146495)
- [Include attribute and derive macros in search filters for "macros"](rust-lang/rust#148176)
- [Include extern crates in search filters for `import`](rust-lang/rust#148301)
- [Validate usage of crate-level doc attributes](rust-lang/rust#149197).  This means if any of `html_favicon_url`, `html_logo_url`, `html_playground_url`, `issue_tracker_base_url`, or `html_no_source` either has a missing value, an unexpected value, or a value of the wrong type, rustdoc will emit the deny-by-default lint `rustdoc::invalid_doc_attributes`.

<a id="1.93.0-Compatibility-Notes"></a>

## Compatibility Notes

- [Introduce `pin_v2` into the builtin attributes namespace](rust-lang/rust#139751)
- [Update bundled musl to 1.2.5](rust-lang/rust#142682)
- [On Emscripten, the unwinding ABI used when compiling with `panic=unwind` was changed from the JS exception handling ABI to the wasm exception handling ABI.](rust-lang/rust#147224) If linking C/C++ object files with Rust objects, `-fwasm-exceptions` must be passed to the linker now. On nightly Rust, it is possible to get the old behavior with `-Zwasm-emscripten-eh=false -Zbuild-std`, but it will be removed in a future release.
- The `#[test]` attribute, used to define tests, was previously ignored in various places where it had no meaning (e.g on trait methods or types). Putting the `#[test]` attribute in these places is no longer ignored, and will now result in an error; this may also result in errors when generating rustdoc. [Error when `test` attribute is applied to structs](rust-lang/rust#147841)
- Cargo now sets the `CARGO_CFG_DEBUG_ASSERTIONS` environment variable in more situations. This will cause crates depending on `static-init` versions 1.0.1 to 1.0.3 to fail compilation with "failed to resolve: use of unresolved module or unlinked crate `parking_lot`". See [the linked issue](rust-lang/rust#150646 (comment)) for details.
- [User written types in the `offset_of!` macro are now checked to be well formed.](rust-lang/rust#150465)
- `cargo publish` no longer emits `.crate` files as a final artifact for user access when the `build.build-dir` config is unset
- [Upgrade the `deref_nullptr` lint from warn-by-default to deny-by-default](rust-lang/rust#148122)
- [Add future-incompatibility warning for `...` function parameters without a pattern outside of `extern` blocks](rust-lang/rust#143619)
- [Introduce future-compatibility warning for `repr(C)` enums whose discriminant values do not fit into a `c_int` or `c_uint`](rust-lang/rust#147017)
- [Introduce future-compatibility warning against ignoring `repr(C)` types as part of `repr(transparent)`](rust-lang/rust#147185)

</details>

---

### Configuration

📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.

♻ **Rebasing**: Whenever MR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 **Ignore**: Close this MR and you won't be reminded about this update again.

---

 - [ ] <!-- rebase-check -->If you want to rebase/retry this MR, check this box

---

This MR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0Mi44OC4yIiwidXBkYXRlZEluVmVyIjoiNDIuODguMiIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOlsiUmVub3ZhdGUgQm90IiwiYXV0b21hdGlvbjpib3QtYXV0aG9yZWQiLCJkZXBlbmRlbmN5LXR5cGU6Om1pbm9yIl19-->

Pkgsrc changes:
 * Update version & checksums.
 * Adapt openssl-src patches to minor version update.

Noteable failures at the time of commit:
 * The cross-build for sparc64 fails, not yet reported.

Upstream changes relative to 1.92.0:

Version 1.93 (2026-01-22)
==========================

Language
--------
- [Add warn-by-default `function_casts_as_integer` lint]
  (rust-lang/rust#141470)
- [Add future-incompatibility warning for `...` function parameters
  without a pattern outside of `extern` blocks]
  (rust-lang/rust#143619)
- [Stabilize several s390x `vector`-related target features and
  the `is_s390x_feature_detected!` macro]
  (rust-lang/rust#145656)
- [Stabilize declaration of C-style variadic functions for the `system` ABI]
  (rust-lang/rust#145954)
- [Emit error when using some keyword as a `cfg` predicate]
  (rust-lang/rust#146978)
- [Introduce future-compatibility warning for `repr(C)` enums whose
  discriminant values do not fit into a `c_int` or `c_uint`]
  (rust-lang/rust#147017)
- [Introduce future-compatibility warning against ignoring `repr(C)`
  types as part of `repr(transparent)`]
  (rust-lang/rust#147185)
- [Stabilize `asm_cfg`]
  (rust-lang/rust#147736)
- [Upgrade the `deref_nullptr` lint from warn-by-default to deny-by-default]
  (rust-lang/rust#148122)
- [During const-evaluation, support copying pointers byte-by-byte]
  (rust-lang/rust#148259)
- [Add warn-by-default `const_item_interior_mutations` lint to warn
  against calls which mutate interior mutable `const` items]
  (rust-lang/rust#148407)
- [LUB coercions now correctly handle function item types, and
  functions with differing safeties]
  (rust-lang/rust#148602)
- [Allow `const` items that contain mutable references to `static`
  (which is *very* unsafe, but not *always* UB)]
  (rust-lang/rust#148746)

Compiler
--------
- [Stabilize `-Cjump-tables=bool`]
  (rust-lang/rust#145974). The flag was
  previously called `-Zno-jump-tables`.
- [Promote `riscv64a23-unknown-linux-gnu` to Tier 2 (without host tools)]
  (rust-lang/rust#148435)

Platform Support
----------------

Refer to Rust's [platform support page][platform-support-doc]
for more information on Rust's tiered platform support.

[platform-support-doc]: https://doc.rust-lang.org/rustc/platform-support.html

Libraries
---------
- [Stop internally using `specialization` on the `Copy` trait as
  it is unsound in the presence of lifetime dependent `Copy`
  implementations. This may result in some performance regressions
  as some standard library APIs may now call `Clone::clone` instead
  of performing bitwise copies]
  (rust-lang/rust#135634)
- [Allow the global allocator to use thread-local storage and
  `std::thread::current()`]
  (rust-lang/rust#144465)
- [Make `BTree::append` not update existing keys when appending an
  entry which already exists]
  (rust-lang/rust#145628)
- [Don't require `T: RefUnwindSafe` for `vec::IntoIter<T>: UnwindSafe`]
  (rust-lang/rust#145665)

Stabilized APIs
---------------

- [`<MaybeUninit<T>>::assume_init_drop`]
  (https://doc.rust-lang.org/stable/core/mem/union.MaybeUninit.html#method.assume_init_drop)
- [`<MaybeUninit<T>>::assume_init_ref`]
  (https://doc.rust-lang.org/stable/core/mem/union.MaybeUninit.html#method.assume_init_ref)
- [`<MaybeUninit<T>>::assume_init_mut`]
  (https://doc.rust-lang.org/stable/core/mem/union.MaybeUninit.html#method.assume_init_mut)
- [`<[MaybeUninit<T>]>::write_copy_of_slice`]
  (https://doc.rust-lang.org/stable/std/primitive.slice.html#method.write_copy_of_slice)
- [`<[MaybeUninit<T>]>::write_clone_of_slice`]
  (https://doc.rust-lang.org/stable/std/primitive.slice.html#method.write_clone_of_slice)
- [`String::into_raw_parts`]
  (https://doc.rust-lang.org/stable/std/string/struct.String.html#method.into_raw_parts)
- [`Vec::into_raw_parts`]
  (https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.into_raw_parts)
- [`<iN>::unchecked_neg`]
  (https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_neg)
- [`<iN>::unchecked_shl`]
  (https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_shl)
- [`<iN>::unchecked_shr`]
  (https://doc.rust-lang.org/stable/std/primitive.isize.html#method.unchecked_shr)
- [`<uN>::unchecked_shl`]
  (https://doc.rust-lang.org/stable/std/primitive.usize.html#method.unchecked_shl)
- [`<uN>::unchecked_shr`]
  (https://doc.rust-lang.org/stable/std/primitive.usize.html#method.unchecked_shr)
- [`<[T]>::as_array`]
  (https://doc.rust-lang.org/stable/std/primitive.slice.html#method.as_array)
- [`<[T]>::as_array_mut`]
  (https://doc.rust-lang.org/stable/std/primitive.slice.html#method.as_mut_array)
- [`<*const [T]>::as_array`]
  (https://doc.rust-lang.org/stable/std/primitive.pointer.html#method.as_array)
- [`<*mut [T]>::as_array_mut`]
  (https://doc.rust-lang.org/stable/std/primitive.pointer.html#method.as_mut_array)
- [`VecDeque::pop_front_if`]
  (https://doc.rust-lang.org/stable/std/collections/struct.VecDeque.html#method.pop_front_if)
- [`VecDeque::pop_back_if`]
  (https://doc.rust-lang.org/stable/std/collections/struct.VecDeque.html#method.pop_back_if)
- [`Duration::from_nanos_u128`]
  (https://doc.rust-lang.org/stable/std/time/struct.Duration.html#method.from_nanos_u128)
- [`char::MAX_LEN_UTF8`]
  (https://doc.rust-lang.org/stable/std/primitive.char.html#associatedconstant.MAX_LEN_UTF8)
- [`char::MAX_LEN_UTF16`]
  (https://doc.rust-lang.org/stable/std/primitive.char.html#associatedconstant.MAX_LEN_UTF16)
- [`std::fmt::from_fn`]
  (https://doc.rust-lang.org/stable/std/fmt/fn.from_fn.html)
- [`std::fmt::FromFn`]
  (https://doc.rust-lang.org/stable/std/fmt/struct.FromFn.html)

Cargo
-----
- [Enable CARGO_CFG_DEBUG_ASSERTIONS in build scripts based on profile]
  (rust-lang/cargo#16160)
- [In `cargo tree`, support long forms for `--format` variables]
  (rust-lang/cargo#16204)
- [Add `--workspace` to `cargo clean`]
  (rust-lang/cargo#16263)

Rustdoc
-----
- [Remove `#![doc(document_private_items)]`](rust-lang/rust#146495)
- [Include attribute and derive macros in search filters for "macros"](rust-lang/rust#148176)
- [Include extern crates in search filters for `import`](rust-lang/rust#148301)
- [Validate usage of crate-level doc attributes](rust-lang/rust#149197).  This means if any of `html_favicon_url`, `html_logo_url`, `html_playground_url`, `issue_tracker_base_url`, or `html_no_source` either has a missing value, an unexpected value, or a value of the wrong type, rustdoc will emit the deny-by-default lint `rustdoc::invalid_doc_attributes`.

Compatibility Notes
-------------------
- [Introduce `pin_v2` into the builtin attributes namespace]
  (rust-lang/rust#139751)
- [Update bundled musl to 1.2.5]
  (rust-lang/rust#142682)
- [On Emscripten, the unwinding ABI used when compiling with
  `panic=unwind` was changed from the JS exception handling ABI to
  the wasm exception handling ABI.]
  (rust-lang/rust#147224) If linking C/C++
  object files with Rust objects, `-fwasm-exceptions` must be passed
  to the linker now. On nightly Rust, it is possible to get the
  old behavior with `-Zwasm-emscripten-eh=false -Zbuild-std`, but
  it will be removed in a future release.
- The `#[test]` attribute, used to define tests, was previously
  ignored in various places where it had no meaning (e.g on trait
  methods or types). Putting the `#[test]` attribute in these places
  is no longer ignored, and will now result in an error; this may
  also result in errors when generating rustdoc. [Error when `test`
  attribute is applied to structs]
  (rust-lang/rust#147841)
- Cargo now sets the `CARGO_CFG_DEBUG_ASSERTIONS` environment
  variable in more situations. This will cause crates depending on
  `static-init` versions 1.0.1 to 1.0.3 to fail compilation with
  "failed to resolve: use of unresolved module or unlinked crate
  `parking_lot`". See [the linked issue]
  (rust-lang/rust#150646 (comment))
  for details.
- [User written types in the `offset_of!` macro are now checked to
  be well formed.]
  (rust-lang/rust#150465)
- `cargo publish` no longer emits `.crate` files as a final artifact
  for user access when the `build.build-dir` config is unset

…adding, r=JonathanBrouwer,traviscross

const-eval: always do mem-to-mem copies if there might be padding involved

This is the final piece of the puzzle for rust-lang#148470: when copying data of a type that has padding, always do a mem-to-mem copy, so that we always preserve the source padding exactly. That prevents rustc implementation choices from leaking into user-visible behavior.

This is technically a breaking change: the example at the top of rust-lang#148470 no longer compiles with this. However, it seems very unlikely that anyone would have depended on this. My main concern is not backwards compatibility, it is performance.

Fixes rust-lang#148470

---

> Actually that seems to be entirely fine, it even helps with some benchmarks! I guess the mem-to-mem codepath is actually faster than the scalar pair codepath for the copy itself. It can slow things down later since now we have to do everything bytewise, but that doesn't show up in our benchmarks and might not be very relevant after all (in particular, it only affects types with padding, so the rather common wide pointers still always use the efficient scalar representation).
>
> So that would be my proposal to for resolving this issue then: to make const-eval behavior consistent, we always copy the padding from the source to the target. IOW, potentially pre-existing provenance in the target always gets overwritten (that part is already in rust-lang#148259), and potentially existing provenance in padding in the source always gets carried over (that's rust-lang#148967). If there's provenance elsewhere in the source our existing handling is fine:
> - If it's in an integer, that's UB during const-eval so we can do whatever.
> - If it's in a pointer, the the fragments must combine back together to a pointer or else we have UB.
> - If it's in a union we just carry it over unchanged.
>
> @traviscross we should check that this special const-eval-only UB is properly reflected in the reference. Currently we have [this](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.const-transmute-ptr2int) but that only talks about int2ptr, not about invalid pointer fragments at pointer type. I also wonder if this shouldn't rather be part of ["invalid values"](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.validity) to make it clear that this applies recursively inside fields as well.
> EDIT: Reference PR is up at rust-lang/reference#2091.

 _Originally posted by @RalfJung in [rust-lang#148470](rust-lang#148470 (comment)

> Worth noting that this does not resolve the concerns @theemathas had about `-Zextra-const-ub-checks` sometimes causing *more* code to compile. Specifically, with that flag, the behavior changes to "potentially existing provenance in padding in the source never gets carried over". However, it's a nightly-only flag (used by Miri) so while the behavior is odd, I don't think this is a problem.

 _Originally posted by @RalfJung in [rust-lang#148470](rust-lang#148470 (comment)

---

Related:

- rust-lang#148470
- rust-lang/reference#2091

…adding, r=JonathanBrouwer,traviscross

const-eval: always do mem-to-mem copies if there might be padding involved

This is the final piece of the puzzle for rust-lang#148470: when copying data of a type that has padding, always do a mem-to-mem copy, so that we always preserve the source padding exactly. That prevents rustc implementation choices from leaking into user-visible behavior.

This is technically a breaking change: the example at the top of rust-lang#148470 no longer compiles with this. However, it seems very unlikely that anyone would have depended on this. My main concern is not backwards compatibility, it is performance.

Fixes rust-lang#148470

---

> Actually that seems to be entirely fine, it even helps with some benchmarks! I guess the mem-to-mem codepath is actually faster than the scalar pair codepath for the copy itself. It can slow things down later since now we have to do everything bytewise, but that doesn't show up in our benchmarks and might not be very relevant after all (in particular, it only affects types with padding, so the rather common wide pointers still always use the efficient scalar representation).
>
> So that would be my proposal to for resolving this issue then: to make const-eval behavior consistent, we always copy the padding from the source to the target. IOW, potentially pre-existing provenance in the target always gets overwritten (that part is already in rust-lang#148259), and potentially existing provenance in padding in the source always gets carried over (that's rust-lang#148967). If there's provenance elsewhere in the source our existing handling is fine:
> - If it's in an integer, that's UB during const-eval so we can do whatever.
> - If it's in a pointer, the the fragments must combine back together to a pointer or else we have UB.
> - If it's in a union we just carry it over unchanged.
>
> @traviscross we should check that this special const-eval-only UB is properly reflected in the reference. Currently we have [this](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.const-transmute-ptr2int) but that only talks about int2ptr, not about invalid pointer fragments at pointer type. I also wonder if this shouldn't rather be part of ["invalid values"](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.validity) to make it clear that this applies recursively inside fields as well.
> EDIT: Reference PR is up at rust-lang/reference#2091.

 _Originally posted by @RalfJung in [rust-lang#148470](rust-lang#148470 (comment)

> Worth noting that this does not resolve the concerns @theemathas had about `-Zextra-const-ub-checks` sometimes causing *more* code to compile. Specifically, with that flag, the behavior changes to "potentially existing provenance in padding in the source never gets carried over". However, it's a nightly-only flag (used by Miri) so while the behavior is odd, I don't think this is a problem.

 _Originally posted by @RalfJung in [rust-lang#148470](rust-lang#148470 (comment)

---

Related:

- rust-lang#148470
- rust-lang/reference#2091

Rollup merge of #148967 - RalfJung:const-eval-preserve-src-padding, r=JonathanBrouwer,traviscross

const-eval: always do mem-to-mem copies if there might be padding involved

This is the final piece of the puzzle for #148470: when copying data of a type that has padding, always do a mem-to-mem copy, so that we always preserve the source padding exactly. That prevents rustc implementation choices from leaking into user-visible behavior.

This is technically a breaking change: the example at the top of #148470 no longer compiles with this. However, it seems very unlikely that anyone would have depended on this. My main concern is not backwards compatibility, it is performance.

Fixes #148470

---

> Actually that seems to be entirely fine, it even helps with some benchmarks! I guess the mem-to-mem codepath is actually faster than the scalar pair codepath for the copy itself. It can slow things down later since now we have to do everything bytewise, but that doesn't show up in our benchmarks and might not be very relevant after all (in particular, it only affects types with padding, so the rather common wide pointers still always use the efficient scalar representation).
>
> So that would be my proposal to for resolving this issue then: to make const-eval behavior consistent, we always copy the padding from the source to the target. IOW, potentially pre-existing provenance in the target always gets overwritten (that part is already in #148259), and potentially existing provenance in padding in the source always gets carried over (that's #148967). If there's provenance elsewhere in the source our existing handling is fine:
> - If it's in an integer, that's UB during const-eval so we can do whatever.
> - If it's in a pointer, the the fragments must combine back together to a pointer or else we have UB.
> - If it's in a union we just carry it over unchanged.
>
> @traviscross we should check that this special const-eval-only UB is properly reflected in the reference. Currently we have [this](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.const-transmute-ptr2int) but that only talks about int2ptr, not about invalid pointer fragments at pointer type. I also wonder if this shouldn't rather be part of ["invalid values"](https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html#r-undefined.validity) to make it clear that this applies recursively inside fields as well.
> EDIT: Reference PR is up at rust-lang/reference#2091.

 _Originally posted by @RalfJung in [#148470](#148470 (comment)

> Worth noting that this does not resolve the concerns @theemathas had about `-Zextra-const-ub-checks` sometimes causing *more* code to compile. Specifically, with that flag, the behavior changes to "potentially existing provenance in padding in the source never gets carried over". However, it's a nightly-only flag (used by Miri) so while the behavior is odd, I don't think this is a problem.

 _Originally posted by @RalfJung in [#148470](#148470 (comment)

---

Related:

- #148470
- rust-lang/reference#2091