void_type_node change

[Richard Henderson]

On Wed, Mar 22, 2000 at 04:28:21PM -0500, Richard Kenner wrote:
> I'll think about it some more.  I agree it shouldn't be passed, but it
> definitely wants something of zero size so that it doesn't affect the
> parameter passing data.

It probably shouldn't be passed at all. 

The whole last arg thing appears to be a grotty hack that gets
abused by ppc and arm and probably others to control gen_call,
since the later doesn't get to look at the FUNCTION_DECL.

Anyway, I fixed up alpha's function_arg function this morning.


r~