While stabby::result::Result's actual layout hasn't truly changed, this PR fixes a soundness hole which might still be triggered by exploiting a Result received from a binary with the previous representation.
stabby treats this release's Result as different from that of the previous.
Fixing an unsoundness in stabby::result::Result which infected all repr(stabby) enum
stabby::result::Result sometimes encodes determinants in structure padding. However, until this PR, rustc was free to treat these paddings as unused data, making reads in these regions UB.
Due to this, any repr(stabby) enum that used these determinants is unsound until this PR is merged.
This PR addresses this by ensure Rust treats enums as a blob of memory with the appropriate size and alignment.
A nice side effect of this PR is that Rust previously treated repr(stabby) enums as improper_ctypes due to the determinants being ZSTs, as highlighted by issue #68. This is no longer the case.
ABI BREAKING CHANGE
While
stabby::result::Result
's actual layout hasn't truly changed, this PR fixes a soundness hole which might still be triggered by exploiting aResult
received from a binary with the previous representation.stabby
treats this release'sResult
as different from that of the previous.Fixing an unsoundness in
stabby::result::Result
which infected allrepr(stabby)
enumstabby::result::Result
sometimes encodes determinants in structure padding. However, until this PR, rustc was free to treat these paddings as unused data, making reads in these regions UB.Due to this, any
repr(stabby)
enum that used these determinants is unsound until this PR is merged.This PR addresses this by ensure Rust treats enums as a blob of memory with the appropriate size and alignment.
A nice side effect of this PR is that Rust previously treated
repr(stabby)
enums asimproper_ctypes
due to the determinants being ZSTs, as highlighted by issue #68. This is no longer the case.