[gnu.gcc.bug] Bugs in GCC 1.31 distribution

jdiaz@hqsun1.oracle.uucp (Jean Marie Diaz) (11/23/88)

My distribution is version 1.30, patched to 1.31 (patches obtained from
osu-cis).  So far, I have run into two minor problems:

The routine set_preference in global-alloc.c, appears to be missing three
right braces ('}').  I include a context diff that shows where I added
braces.

On a Sun3 running SunOS4.0, gnulib builds okay, but every other file causes
ccom to core dump.  My Sun4 running 4.0 is building 1.31 happily (we're
halfway through stage2).  Using GCC 1.30 as the bootstrap doesn't seem to
help the Sun3.

			AMBAR

jdiaz@hqsun1.oracle.uucp (Jean Marie Diaz) (11/23/88)

(mind is clearly not in gear today --)
Also, using GCC 1.30 to bootstrap GCC 1.31 on a Sun3 running SunOS4.0 works
fine.  Please ignore my previous assertion to the contrary.

				   AMBAR
ambar%oracle.uucp@hplabs.hp.com			 {hplabs,pyramid}!oracle!ambar

*** global-alloc.c~	Tue Nov 22 15:36:45 1988
--- global-alloc.c	Tue Nov 22 15:36:20 1988
***************
*** 944,950 ****
  		      reg_allocno[src_regno], dest_regno);
  	  SET_HARD_REG_BIT (regs_someone_prefers, dest_regno);
  	}
! 
    if (src_regno < FIRST_PSEUDO_REGISTER && dest_regno >= FIRST_PSEUDO_REGISTER
        && reg_allocno[dest_regno] >= 0)
      {
--- 944,950 ----
  		      reg_allocno[src_regno], dest_regno);
  	  SET_HARD_REG_BIT (regs_someone_prefers, dest_regno);
  	}
!     }
    if (src_regno < FIRST_PSEUDO_REGISTER && dest_regno >= FIRST_PSEUDO_REGISTER
        && reg_allocno[dest_regno] >= 0)
      {
***************
*** 957,962 ****
--- 957,966 ----
  		      reg_allocno[dest_regno], src_regno);
  	  SET_HARD_REG_BIT (regs_someone_prefers, src_regno);
  	}
+     }
+ }
+   
+ 
  
  /* Print debugging trace information if -greg switch is given,
     showing the information on which the allocation decisions are based.  */