scottmcm / rev_slice Goto Github PK

Without one of those attributes, my understanding is the transmute from &[T] to &RevSlice<T> is undefined behavior. The compiler is free to stick some additional data at the start of RevSlice<T> or give it a stricter alignment than [T]. For example some targets may require all struct types to have some minimum alignment that is not required for slices.

From nomicon's Alternative representations:

#[repr(C)] is also necessary to soundly do more elaborate tricks with data layout such as reinterpreting values as a different type.