One problem in D that has no clear answer and continuously rears it's head
is the void[] type being used for buffering.
The main concern is that the runtime cautiously may remove the NO_SCAN bit
from a void[] array. This means false pointers.
But can we approach this from a different angle? What if you *couldn't*
create void arrays? Oh, you could cast arrays to void[] but you could
never create them as void. In addition, you couldn't append to void[],
concatenate void[]'s or dup them.
What problems would this cause? Essentially, if you wanted a void[] array
that contains pointers, you'd have to change:
auto x = new void[...];
to
void[] x = new void*[...];
And if you wanted one that doesn't contain pointers, you could do:
void[] x = new ubyte[...];
Any other issues? Who does frequent void[] appending or concatenation?
-Steve
