Re: m68k stuff

[Jeffrey A Law <law at cygnus dot com>]

  In message <199808141613.MAA22303@ahab.rutgers.edu>you write:
  > > Looking over the gcc test results for m68k-sun-sunos4.1, I see one
  > > gcc regression.  compile/950612-1.c.
  > > 
  > > I built a cross compiler for m68k-sun-sunos4.1 and tried the test
  > > and it worked fine.
  > > 
  > > Do you still have the build dir for your m68k 19980727 directory
  > > lying around?    Any chance I could get you to send me the gcc.log
  > > file.  Or better yet, can you look in the log file and find out
  > > why dejagnu thought compile/950612-1.c failed at -O2 optimization
  > > levels?
  > 
  > I no longer have the original log file, but I can reproduce the
  > problem:
  > 
  > skink% gcc 950612-1.c  -w  -O2   -c --save-temps
  > 950612-1.s: Assembler messages:
  > 950612-1.s:182: Error: operands mismatch -- statement `addl a6@(-160),a6@(-132)' ignored
  > 950612-1.s:195: Error: operands mismatch -- statement `addl a6@(-172),a6@(-136)' ignored
Thanks.  That was enough for me to track it down (my assembly code had
the same bogus instruction).

The constraints on a couple of instructions in m68k.md were blatlently
wrong.  Amazing we haven't tripped this one before.  Anyway, it should
be fixed now.

jeff