Fix argument pushes to unaligned stack slots

[Ian Lance Taylor]

"Joseph S. Myers" <joseph@codesourcery.com> writes:

> 2007-02-06  Joseph Myers  <joseph@codesourcery.com>
> 
> 	* expr.c (emit_push_insn): If STRICT_ALIGNMENT, copy to an
> 	unaligned stack slot via a suitably aligned slot.

This patch concerns me because the caller, in store_one_arg in
calls.c, will always pass PARM_BOUNDARY.  In the past this never
mattered since the alignment was effectively ignored for mode !=
BLKmode.  My concern is that your patch will start causing some
targets to unnecessarily stage large arguments through a stack
temporary.  Can you give me a reason why that will not happen?  Or
change the caller to pass in the correct alignment based on
arg->locate.boundary which is ultimately based on
FUNCTION_ARG_BOUNDARY?

Thanks.

Ian