chorman0773
commented
1 month ago It does not permit that, no, as we can relax validity invariants.
On Thu, Oct 10, 2024, 21:20 Joshua Liebow-Feeser @.***> wrote:
@.**** commented on this pull request.
In src/type-layout.md https://github.com/rust-lang/reference/pull/1630#discussion_r1796293210:
+Padding bytes have no bit validity requirement: it is always well-defined +to write any byte value - including an uninitialized byte - to any byte of +padding.
IMO bit validity (at least as I've used it) is a tad stronger than that: it also permits unsafe code to assume that it won't see certain values produced by safe Rust. That implies that it also constrains what low-level representations can be produced by the "encoding" operation you described.
I don't see any reason why explicitly documenting encoding and decoding separately is a problem. I'd just want to make sure we can still derive the stronger notion of bit validity that I described from whatever guarantees are made regarding encoding and decoding.
— Reply to this email directly, view it on GitHub https://github.com/rust-lang/reference/pull/1630#discussion_r1796293210, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABGLD22HV3WP7ABLDQLOSETZ24RVTAVCNFSM6AAAAABOZMPY6WVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDGNRRG4ZTIOJZGM . You are receiving this because you commented.Message ID: @.***>
traviscross
This intentionally omits discussions of enums and unions. Enum bit validity is well-defined, but it is difficult to articulate concisely. Union bit validity is still not nailed down.