Some Haifa scheduler bugs

[Jeffrey A Law]

  In message < Pine.SOL.3.90.970905134354.4356D-100000@tracy.informatik.rwth-aachen.de >you write:
  > Now I'm confused. Isn't the reload_address_base_reg_class stuff supposed to
  > handle that case?
I haven't read all the new code, but based on experiences of someone in
Cygnus porting to a register poor machine that passes arguments in registers
my impression was that the reload_address_base_reg_class stuff wasn't
complete or sufficient to prevent incorrect code from being generated.

Jeff