execute/950503-1.c failing at all optimization levels on x86

[Jeffrey A Law]

  In message < 199908020452.VAA25544@zack.bitmover.com >you write:
  > 
  > execute/950503-1.c is a test for proper handling of compile time
  > shifts of "long long" quantities.  It's failing at all optimization
  > levels with current CVS.  1.1.2 had no trouble.
  > 
  > The appended assembly dump is cc1 output at -O0 for this test case.
  > Note the anonymous constants at the top of the file:
[ ... ]
This seems to be working OK now.
jeff