Improved test results on i586-pc-sco3.2v5.0.4

[Jeffrey A Law]

  In message <19980522222038.A16799@dgii.com>you write:
  > In light of the recent activity surrounding problems with x86 bootstrapping  > ,
  > I updated this morning and reran the tests.   There are now fewer 
  > "mysterious" failures, but in particular, the -fPIC for all languages
  > and g++ seem to have taken a hit in the shorts sometime in the last month.
  > I'll try to look into the g++ runtime failures in the next few days.
  > 
  > Here are the "obvious" code generation or test  case problems.  I picked
  > only the first occurrence of the ICE's.
  > 
  >  980329-1 fails becuase we're trying tin incLONG a byte register:
  > 	/usr/tmp/cca005z7.s:76:invalid register for instruction: %si in incl  
Ahhh.  Nevermind my earlier private question about 980329-1.c.  I fixed
this bug a little while ago (July 30).  It just took a little more context to
remind me.

jeff