invalid attribute slots with add_attribute/3

If the third argument of add_attribute/3 is an integer, it is used directly
as an index to the attribute slot. There is apparently no check if the index
is a valid slot, so I assume this can change random memory unsafely?

[eclipse 4]: add_attribute(V,a,999).

V = V
yes.
[eclipse 5]: add_attribute(V,a,-1).

V = V
yes.

although:

[eclipse 6]: add_attribute(V,a,0).
undefined variable attribute in add_attribute(V, a, 0)

Should this be considered a bug?