enable -Zrandomize-layout for miri tests
#229
Assignees
Comments
hawkw
added a commit
that referenced
this issue
Jun 19, 2022
`cordyceps` and `maitake` currently contain code that perform layout-dependent casts (in this case, casting a ptr to struct to a ptr to the struct's first subfield), which would be UB if those structs were not `#[repr(C)]`. the `-Zrandomize-layout` flag tells the Rust compiler to randomize the layout of all `#[repr(Rust)]` structs (see rust-lang/compiler-team#457 for details). if we ever perform a layout-dependent cast on a struct that is not `#[repr(C)]` (or `#[repr(transparent)]`), layout randomization would break that cast. enabling this flag while running the Miri tests will help to catch any bugs introduced by accidentally performing such a cast on a non-layout-dependent type. i also made some changes to the `bin/miri` script. this was primarily to add comments on the individual flags that are added to `$MIRIFLAGS`, so that we can remember what they're doing when we look back at the script. the actual behavior should be identical, but the values added to `$MIRIFLAGS` are now declared in an array so that each flag can have a comment. Closes #229 Signed-off-by: Eliza Weisman <eliza@buoyant.io>
hawkw
added a commit
that referenced
this issue
Jun 19, 2022
`cordyceps` and `maitake` currently contain code that perform layout-dependent casts (in this case, casting a ptr to struct to a ptr to the struct's first subfield), which would be UB if those structs were not `#[repr(C)]`. the `-Zrandomize-layout` flag tells the Rust compiler to randomize the layout of all `#[repr(Rust)]` structs (see rust-lang/compiler-team#457 for details). if we ever perform a layout-dependent cast on a struct that is not `#[repr(C)]` (or `#[repr(transparent)]`), layout randomization would break that cast. enabling this flag while running the Miri tests will help to catch any bugs introduced by accidentally performing such a cast on a non-layout-dependent type. i also made some changes to the `bin/miri` script. this was primarily to add comments on the individual flags that are added to `$MIRIFLAGS`, so that we can remember what they're doing when we look back at the script. the actual behavior should be identical, but the values added to `$MIRIFLAGS` are now declared in an array so that each flag can have a comment. Closes #229 Signed-off-by: Eliza Weisman <eliza@buoyant.io>
hawkw
added a commit
that referenced
this issue
Jun 21, 2022
`cordyceps` and `maitake` currently contain code that perform layout-dependent casts (in this case, casting a ptr to struct to a ptr to the struct's first subfield), which would be UB if those structs were not `#[repr(C)]`. the `-Zrandomize-layout` flag tells the Rust compiler to randomize the layout of all `#[repr(Rust)]` structs (see rust-lang/compiler-team#457 for details). if we ever perform a layout-dependent cast on a struct that is not `#[repr(C)]` (or `#[repr(transparent)]`), layout randomization would break that cast. enabling this flag while running the Miri tests will help to catch any bugs introduced by accidentally performing such a cast on a non-layout-dependent type. i also made some changes to the `bin/miri` script. this was primarily to add comments on the individual flags that are added to `$MIRIFLAGS`, so that we can remember what they're doing when we look back at the script. the actual behavior should be identical, but the values added to `$MIRIFLAGS` are now declared in an array so that each flag can have a comment. Closes #229 Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
we should be running Miri tests with
RUSTFLAGS="-Zrandomize-layout".cordycepsandmaitakecurrently contain code that perform layout-dependent casts (in this case, casting a ptr to struct to a ptr to the struct's first subfield), which would be UB if those structs were not#[repr(C)]. i believe that running the Miri tests with this flag would help catch any cases where a layout-dependent cast is inadvertently performed with a type that is not#[repr(C)](or#[repr(transparent)]?), so we should be enabling it in our Miri CI jobs to catch any potential bugs involving this...The text was updated successfully, but these errors were encountered: