egcs-{core|g77}-970901 on m68k-next-nextstep3

[Jeffrey A Law]

  In message < 9709031803.AA29073@moene.indiv.nluug.nl >you write:
  > Hi All,
  > 
  > I got egcs-core-970901.tar.gz and egcs-g77-970901.tar.gz from  
  > ftp.cygnus.com, unpacked and configure'd:
Cool.  Even more so since you managed to build core + g77 :-)

  > No problems, except for a corrupted specs.h (12 zero characters in  
  > it - probably another NEXT utility giving me problems)
One way to check this is to do a "gcc -dumpspecs" since that's how
the specs file gets created.

If the stage1 gcc produces a corrupted specs file, then either there's
a bug in the code to create specs files or "cc" mis-compiled the code
to create specs files.

  > make compare succeeded too.
Good.  It's worth noting that I've got the hp400 set up with netbsd, so
I can do bootstraps of the 68k port now, so if we run into problems again
it shouldn't take so long to get them resolved.

  > Then I proceeded to build HIRLAM with the Fortran compiler (-O2  
  > -fforce-addr) without much trouble.  There was again one routine  
  > drawing a
  > 
  > LOWPASS.F:131: Non-optional registers need a spill register
  > (insn:HI 860 2140 2141 (set (mem:SI (plus:SI (reg:SI 14 a6)
  >                 (const_int -96976)))
  >         (plus:SI (mem:SI (plus:SI (reg:SI 14 a6)
  >                     (const_int -96976)))
  >             (reg/v/u:SI 28))) 122 {*addsi3_internal} (insn_list 859  
  > (nil))
  >     (expr_list:REG_DEAD (reg:SI 462)
  >         (nil)))
  > gcc: Internal compiler error: program f771 got fatal signal 6
Any chance you could forward the .f file to us?  This shouldn't happen,
even if you use -fforce-addr.

  > because that's the frame-pointer.  Note that an address like  
  > -96976(a6) is invalid on an m68k, because the offset must remain  
  > between -32768 and 32767 (actually, on an m680x0 with x >= 2 this is  
  > not true - therefore normally I won't encounter this problem,  
  > because I hack m68k.h to accept these offsets).
Can you send me that hack?

  > Performance figures tomorrow.
Looking forward to them.

jeff