BUGs in egcs-1.1 on hppa1.1-hp-hpux10.20, delta, delta88

[Manfred Hollstein <manfred at s-direktnet dot de>]

These are the problems I've encountered while building egcs-1.1 release.

Sources:	- egcs -regcs_1_1_branch from CVS last updated at
		  1998/09/21 10:19:30 +0200
		- libg++-2.8.1.2

Patches:	Wed Sep 16 00:30:56 1998  Geoff Keating  <geoffk@ozemail.com.au>

		* gcse.c: New definition NEVER_SET for reg_first_set, reg_last_set,
		mem_first_set, mem_last_set; because 0 can be a CUID.
		(oprs_unchanged_p): Use new definition.
		(record_last_reg_set_info): Likewise.
		(record_last_mem_set_info): Likewise.
		(compute_hash_table): Likewise.

		1998-08-19  Mark Mitchell  <mark@markmitchell.com>

		* lib/old-dejagnu.exp (old-dejagnu): Remove sig6/sig11 handling
		code. 

		1998-08-09  Mark Mitchell  <mark@markmitchell.com>

		* lib/old-dejagnu.exp: Revise handling of `Internal compiler
		error'.

Environment:	 - GNU binutils 2.9.1
		 - dejagnu-971222 + look_for_dir_with_trigger patch
		 - gperf-2.7


Building  the   [class], [byval],  [call],  [convert],  and [call-var]
targets in ${target}/libg++/etc/benchmarks results in:

../../egcs-1.1/gcc/integrate.c:1733: Internal compiler error in function expand_inline_function

on   all  three   mentionened  platforms.    I've attached    the file
dhrystone.ii.gz from the hppa variant  for  [class]. This is what  gdb
shows:

$ gdb cc1plus 
GDB is free software and you are welcome to distribute copies of it
 under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.16 (hppa1.1-hp-hpux10.20), Copyright 1996 Free Software Foundation, Inc...
Breakpoint 1 at 0x2ba404
Breakpoint 2 at 0x50a04
Breakpoint 3 at 0xf5e6c: file ../../egcs-1.1/gcc/toplev.c, line 2114.
(gdb) cd ~/work/GNU/egcs-1.1-hppa1.1-hp-hpux10.20/hppa1.1-hp-hpux10.20/libg++/etc/benchmarks
Working directory /u/b60/manfred/work/GNU/egcs-1.1-hppa1.1-hp-hpux10.20/hppa1.1-hp-hpux10.20/libg++/etc/benchmarks
 (canonically /tmp_mnt/phys/b60/data62/manfred-work/GNU/egcs-1.1-hppa1.1-hp-hpux10.20/hppa1.1-hp-hpux10.20/libg++/etc/benchmarks).
(gdb) r dhrystone.ii -quiet -dumpbase dhrystone.cc -g -O9 -version -o dhrystone.s
Starting program: /tmp_mnt/phys/b60/data62/manfred-work/GNU/egcs-1.1-hppa1.1-hp-hpux10.20/gcc/cc1plus dhrystone.ii -quiet -dumpbase dhrystone.cc -g -O9 -version -o dhrystone.s
warning: Unable to find __d_pid symbol in object file.
warning: Suggest linking with /usr/lib/end.o.
warning: GDB will be unable to track shl_load/shl_unload calls
Breakpoint 1 at 0x7af496e8
Breakpoint 2 at 0x7af528fc
GNU C++ version egcs-2.91.57 19980901 (egcs-1.1 release) (hppa1.1-hp-hpux10.20) compiled by GNU C version egcs-2.91.57 19980901 (egcs-1.1 release).
../../egcs-1.1/gcc/integrate.c:1733: Internal compiler error in function expand_inline_function

Breakpoint 2, 0x7af528fc in ___exit ()
(gdb) bt
#0  0x7af528fc in ___exit ()
#1  0x1a2118 in expand_inline_function (fndecl=0x4010a180, parms=0x7b03c4f0, 
    target=0x4016f060, ignore=0, type=0x400cd378, structure_value_addr=0x0)
    at ../../egcs-1.1/gcc/integrate.c:1733
#2  0x14c7a0 in expand_call (exp=0x40163500, target=0x4016f060, ignore=0)
    at ../../egcs-1.1/gcc/calls.c:735
#3  0x66fa8 in cplus_expand_expr (exp=0x7b03c4e8, target=0x4016f060, tmode=1075196456, 
    modifier=1075196576) at ../../../egcs-1.1/gcc/cp/expr.c:130
#4  0x143b74 in expand_expr (exp=0x401632d8, target=0x0, tmode=QImode, 
    modifier=EXPAND_NORMAL) at ../../egcs-1.1/gcc/expr.c:7774
#5  0x13a4a0 in store_expr (exp=0x401632d8, target=0x4016f060, want_value=0)
    at ../../egcs-1.1/gcc/expr.c:3414
#6  0x143070 in expand_expr (exp=0x401632f8, target=0x4016f060, tmode=VOIDmode, 
    modifier=EXPAND_CONST_ADDRESS) at ../../egcs-1.1/gcc/expr.c:7452
#7  0x1436dc in expand_expr (exp=0x40163420, target=0x0, tmode=VOIDmode, 
    modifier=EXPAND_SUM) at ../../egcs-1.1/gcc/expr.c:7580
#8  0x140f00 in expand_expr (exp=0x40163450, target=0x0, tmode=SImode, modifier=EXPAND_SUM)
    at ../../egcs-1.1/gcc/expr.c:6458
#9  0x1a18a4 in expand_inline_function (fndecl=0x400d1de8, parms=0x0, target=0x0, ignore=0, 
    type=0x400303f0, structure_value_addr=0x0) at ../../egcs-1.1/gcc/integrate.c:1437
#10 0x14c7a0 in expand_call (exp=0x40163498, target=0x0, ignore=0)
    at ../../egcs-1.1/gcc/calls.c:735
#11 0x140dbc in expand_expr (exp=0x40163498, target=0x0, tmode=SImode, 
    modifier=EXPAND_NORMAL) at ../../egcs-1.1/gcc/expr.c:6421
#12 0x140f4c in expand_expr (exp=0x401634b8, target=0x4016f040, tmode=SImode, 
    modifier=EXPAND_NORMAL) at ../../egcs-1.1/gcc/expr.c:6471
#13 0x128308 in expand_return (retval=0x4016f040) at ../../egcs-1.1/gcc/stmt.c:2655
#14 0xbff84 in c_expand_return (retval=0x401634b8) at ../../../egcs-1.1/gcc/cp/typeck.c:7384
#15 0xe13b0 in finish_return_stmt (expr=0x40163498)
    at ../../../egcs-1.1/gcc/cp/semantics.c:317
Cannot access memory at address 0x0.
(gdb) up
#1  0x1a2118 in expand_inline_function (fndecl=0x4010a180, parms=0x7b03c4f0, 
    target=0x4016f060, ignore=0, type=0x400cd378, structure_value_addr=0x0)
    at ../../egcs-1.1/gcc/integrate.c:1733
1733            abort ();
(gdb) l
1728        /* There is no return value to worry about.  */
1729        ;
1730      else if (GET_CODE (loc) == MEM)
1731        {
1732          if (! structure_value_addr || ! aggregate_value_p (DECL_RESULT (fndecl)))
1733            abort ();
1734      
1735          /* Pass the function the address in which to return a structure value.
1736             Note that a constructor can cause someone to call us with
1737             STRUCTURE_VALUE_ADDR, but the initialization takes place
(gdb) p structure_value_addr
$1 = (struct rtx_def *) 0x0
(gdb) p lineno
$2 = 414

Line 414/415 of dhrystone.ii contain:

inline  void  Int::operator += (const int b)
{ rep += b;  ; }

dhrystone.ii.gz

Contact me if you need more information or if you can't reproduce this
behaviour yourself.

For Jeff: Geoff Keating's GCSE patch doesn't fix the problem I've reported
as `BUG with libio/dbz/rdbz test case on m88k-motorola-sysv3'. I'm still
looking...

manfred