g77 testsuite failures under hpux 10.20

[Jeffrey A Law]

  In message < 200006292104.RAA15633@hiauly1.hia.nrc.ca >you write:
  > >   > Executing on host: /xxx/gnu/gcc-2.96/objdir/gcc/g77 -B/xxx/gnu/gcc-2.
  > 96/obj
  > >   > dir/gcc/ /xxx/gnu/gcc-2.96/gcc/testsuite/g77.f-torture/compile/199902
  > 18-0.f
  > >   >   -w  -O0   -L/xxx/gnu/gcc-2.96/objdir/hppa1.1-hp-hpux10.20/threads/l
  > ibf2c 
  > >   > -c  -threads -o /xxx/gnu/gcc-2.96/objdir/gcc/testsuite/19990218-0.o  
  >   (tim
  > >   > eout = 300)
  > >   > /var/tmp/cc8IPMWB.s: Assembler messages:
  > >   > /var/tmp/cc8IPMWB.s:179: Error: Invalid Nullification: (%)
  > >   > /var/tmp/cc8IPMWB.s:179: Error: Invalid operands
  > 
  > > A breakpoint in output_call would be where you'd want to start debugging 
  > this.
  > > jeff
  > 
  > The problem occurs much before this.  Here is the rtl for the problematic
  > call from 19990218-0.f.00.rtl:
  > 
  > (call_insn 17 16 20 (parallel[
  >             (set (reg:SI 28 %r28)
  >                 (call (mem:SI (symbol_ref/v:SI ("")) 0)
  >                     (const_int 16 [0x10])))
  > 	    (clobber (reg:SI 2 %r2))
  >             (use (const_int 0 [0x0]))
  > 	] ) -1 (nil)
  >     (nil)
  >     (expr_list (use (reg:SI 26 %r26))
  > 	(nil)))
OK.  Wow.

You might put a conditional breakpoint in make_call_insn_raw on
cur_insn_uid == 17, then start walking backwards from there through
the call expanders.


Jeff