Soundly representing simultaneously mutable non-overlapping sub-slices of the same slice

[2e71828]

@gspr

Does that not mean that accessing any slice element using a reference to the slice invalidates any *mut to any element, including the one to the beginning of the slice (i.e. a *mut to the slice itself)?

Mostly, it does mean that. The only exception that I know of is if the &mut [T] was somehow derived from a particular *mut T or *mut [T], then that particular raw pointer isn't invalidated by using the reference. But messing with that parent pointer in certain ways (exactly which ones I'm a bit unclear on...) before the last use of the reference is also problematic.