enable-checking problems on PPC

[Jeffrey A Law]

  In message < 99101402250400.01713@ns1102.munich.netsurf.de >you write:
  > Now I understand the reason behind a long patch queue, just wait and it
  > will be superflous sometime :-).
:-) It's not planned that way, but it does happen.  Not as often as I might
like though.  [ When a patch is overtaken by events like this it usually
means we're cleaning up something that desperately needed a cleanup. ]

  > >I believe align_rtx will always be a CONST_INT in the case you're looking 
  > at;
  > >so accessing with INTVAL would be the right thing to do.
  > 
  > Yep, works. I used the following patch to catch possible glitches:
  > 
  > 	* rs6000.c (expand_block_move): Use INTVAL, not XINT to access
  > 	alignement
No need to abort -- emit_block_move is always going to pass in a constant
integer (it's actually defined as always being a CONST_INT in the docs too).

jeff