Description
Location
https://doc.rust-lang.org/std/vec/struct.Vec.html#method.from_raw_parts
Summary
Currently the documentation of Vec::from_raw_parts
claims that both:
- The size of
T
times thecapacity
(ie. the allocated size in bytes) needs to be the same size as the pointer was allocated with. capacity
needs to be the capacity that the pointer was allocated with.
This is somewhat confusing. The first condition arises from #95016 which explicitly removed the second condition in favor of the first. The second condition was then added back in in #99216 without any specific comment about it being necessary.
Together these conditions imply that T
needs to have the same size and alignment as what ptr
was allocated with. But the whole point of #95016 is that this isn't actually the case. (If this is actually the case, I think this could be reworded to be more clear)
Broadly what I am trying to understand is if the following 2 bits of code which convert between Vec<[T; N]>
and Vec<T>
are UB or not:
fn flatten<const N: usize>(vec: Vec<[T; N]>) -> Vec<T> {
let mut v = mem::ManuallyDrop::new(v);
let len = v.len() * N;
let cap = v.capacity() * N;
let p = v.as_mut_ptr() as *mut T;
unsafe {
Vec::from_raw_parts(ptr, new_len, new_cap)
}
}
fn recombine<const N: usize>(vec: Vec<T>) -> Vec<[T; N]> {
let mut v = mem::ManuallyDrop::new(v);
assert_eq!(v.len() % N == 0);
assert_eq!(v.capacity() % N == 0);
let len = v.len() / N;
let cap = v.capacity() / N;
let p = v.as_mut_ptr() as *mut [T; N];
unsafe {
Vec::from_raw_parts(ptr, new_len, new_cap)
}
}
As far as I can tell these should work (and seem to in the playground) but they rely on the second condition being overly strict and the first condition being the one which matters,