[Bug middle-end/49723] New: gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[Bug middle-end/49723] New: gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

           Summary: gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at
                    combine.c:707 at -O1 and above
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: danglin@gcc.gnu.org
              Host: hppa*-*-*
            Target: hppa*-*-*
             Build: hppa*-*-*


Executing on host: /test/gnu/gcc/objdir/gcc/xgcc -B/test/gnu/gcc/objdir/gcc/  
-O1  -w -c  -o pr46934.o
/test/gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/pr
46934.c    (timeout = 300)
/test/gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/pr46934.c: In function
'caller':/test/gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/pr46934.c:17:1:
internal c
ompiler error: in do_SUBST, at combine.c:707

[Bug middle-end/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|hppa*-*-hpux*               |hppa64-*-*
               Host|hppa*-*-hpux*               |hppa64-*-*
              Build|hppa*-*-hpux*               |hppa64-*-*

--- Comment #1 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-13 00:19:10 UTC ---
   0x4000000004d67bac <do_SUBST+172>:    b,l 0x4000000004d9a290 <.stub>,rp
   0x4000000004d67bb0 <do_SUBST+176>:    nop
   0x4000000004d67bb4 <do_SUBST+180>:    copy r4,dp
   0x4000000004d67bb8 <do_SUBST+184>:    cmpb,*=,n ret0,r6,0x4000000004d67bf0
<do_SUBST+240>

(gdb) p/x $r6
$6 = 0x80000000
(gdb) p/x $ret0
$7 = 0xffffffff80000000

newval ($r6) doesn't appear valid.

[Bug middle-end/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ebotcazou at gcc dot
                   |                            |gnu.org

--- Comment #2 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-13 01:09:39 UTC ---
(gdb) bt
#0  do_SUBST (into=0x417aa1e8, newval=0x4170cfb4) at
../../gcc/gcc/combine.c:706
#1  0x0029981a in subst (x=0x417aa1e0, from=0x417a52a0, to=0x4170cfb4,
in_dest=0, in_cond=0, unique_copy=0) at ../../gcc/gcc/combine.c:5227
#2  0x002996b7 in subst (x=0x41712ed0, from=0x417a52a0, to=0x4170cfb4,
in_dest=0, in_cond=0, unique_copy=0) at ../../gcc/gcc/combine.c:5189
#3  0x00292c99 in try_combine (i3=0x4176a84c, i2=0x4176a828, i1=0x0, i0=0x0,
new_direct_jump_p=0xbffff034, last_combined_insn=0x4176a84c) at
../../gcc/gcc/combine.c:3178
#4  0x0028cb6d in combine_instructions (f=0x417135c0, nregs=81) at
../../gcc/gcc/combine.c:1223
#5  0x002b1eb8 in rest_of_handle_combine () at ../../gcc/gcc/combine.c:13878
#6  0x0070bfda in execute_one_pass (pass=0xcf3360) at
../../gcc/gcc/passes.c:2063
#7  0x0070c477 in execute_pass_list (pass=0xcf3360) at
../../gcc/gcc/passes.c:2118
#8  0x0070c493 in execute_pass_list (pass=0xd44320) at
../../gcc/gcc/passes.c:2119
#9  0x0090a5ee in tree_rest_of_compilation (fndecl=0x41785300) at
../../gcc/gcc/tree-optimize.c:416
#10 0x00289022 in cgraph_expand_function (node=0x4178d000) at
../../gcc/gcc/cgraphunit.c:1804
#11 0x00289202 in cgraph_expand_all_functions () at
../../gcc/gcc/cgraphunit.c:1863
#12 0x00289a4a in cgraph_optimize () at ../../gcc/gcc/cgraphunit.c:2133
#13 0x002860d9 in cgraph_finalize_compilation_unit () at
../../gcc/gcc/cgraphunit.c:1304
#14 0x00038838 in c_write_global_declarations () at ../../gcc/gcc/c-decl.c:9844
#15 0x00830b46 in compile_file () at ../../gcc/gcc/toplev.c:564
#16 0x0083375f in do_compile () at ../../gcc/gcc/toplev.c:1886
#17 0x00833940 in toplev_main (argc=13, argv=0xbffff400) at
../../gcc/gcc/toplev.c:1958
#18 0x00149e0f in main (argc=13, argv=0xbffff400) at ../../gcc/gcc/main.c:36

(gdb) p debug_rtx (i3)
(insn 9 8 10 2 (set (reg:SI 73)
        (minus:SI (reg:SI 26 %r26 [ reg_type ])
            (reg:SI 74)))
/test/gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/pr46934.c:3 168
{*pa.md:5239}
     (expr_list:REG_DEAD (reg:SI 26 %r26 [ reg_type ])
        (expr_list:REG_DEAD (reg:SI 74)
            (nil))))
$4 = void
(gdb) p debug_rtx (i2)
(insn 8 7 9 2 (set (reg:SI 74)
        (const_int 2147483648 [0x80000000]))
/test/gnu/gcc/gcc/gcc/testsuite/gcc.c-torture/compile/pr46934.c:3 71
{*pa.md:2130}
     (nil))
$5 = void

This is how insn 8 is generated:

(gdb) bt
#0  emit_move_sequence (operands=0xbfffe904, mode=SImode, scratch_reg=0x0) at
../../gcc/gcc/config/pa/pa.c:1541
#1  0x00171709 in gen_movsi (operand0=0x417a52a0, operand1=0x4170cfb4) at
pa.md:2048
#2  0x003a97e6 in emit_move_insn_1 (x=0x417a52a0, y=0x4170cfb4) at
../../gcc/gcc/expr.c:3280
#3  0x003a9c63 in emit_move_insn (x=0x417a52a0, y=0x4170cfb4) at
../../gcc/gcc/expr.c:3370
#4  0x00388537 in force_reg (mode=SImode, x=0x4170cfb4) at
../../gcc/gcc/explow.c:671
#5  0x00178945 in gen_casesi (operand0=0x41712e94, operand1=0x41712ea0,
operand2=0x4170c36c, operand3=0x4176a7e0, operand4=0x4176a72c) at pa.md:6918
#6  0x006fc425 in maybe_gen_insn (icode=CODE_FOR_casesi, nops=5,
ops=0xbfffebb0) at ../../gcc/gcc/optabs.c:7259
#7  0x006fc533 in maybe_expand_jump_insn (icode=CODE_FOR_casesi, nops=5,
ops=0xbfffebb0) at ../../gcc/gcc/optabs.c:7290
#8  0x006fc5cb in expand_jump_insn (icode=CODE_FOR_casesi, nops=5,
ops=0xbfffebb0) at ../../gcc/gcc/optabs.c:7316
#9  0x003dbe85 in try_casesi (index_type=0x417153c0, index_expr=0x411e8554,
minval=0x4172101c, range=0x4178ad3c, table_label=0x4176a7e0,
default_label=0x4176a72c, fallback_label=0x4176a750) at
../../gcc/gcc/expr.c:10152
#10 0x00814841 in expand_case (stmt=0x417a6000) at ../../gcc/gcc/stmt.c:2505
#11 0x0024c13d in expand_gimple_stmt_1 (stmt=0x417a6000) at
../../gcc/gcc/cfgexpand.c:1918
#12 0x0024ca1d in expand_gimple_stmt (stmt=0x417a6000) at
../../gcc/gcc/cfgexpand.c:2071
#13 0x00258312 in expand_gimple_basic_block (bb=0x417a7080) at
../../gcc/gcc/cfgexpand.c:3805
#14 0x0025b646 in gimple_expand_cfg () at ../../gcc/gcc/cfgexpand.c:4288
#15 0x0070bfda in execute_one_pass (pass=0xcf2f60) at
../../gcc/gcc/passes.c:2063
#16 0x0070c477 in execute_pass_list (pass=0xcf2f60) at
../../gcc/gcc/passes.c:2118
#17 0x0090a5ee in tree_rest_of_compilation (fndecl=0x41785300) at
../../gcc/gcc/tree-optimize.c:416
#18 0x00289022 in cgraph_expand_function (node=0x4178d000) at
../../gcc/gcc/cgraphunit.c:1804
#19 0x00289202 in cgraph_expand_all_functions () at
../../gcc/gcc/cgraphunit.c:1863
#20 0x00289a4a in cgraph_optimize () at ../../gcc/gcc/cgraphunit.c:2133
#21 0x002860d9 in cgraph_finalize_compilation_unit () at
../../gcc/gcc/cgraphunit.c:1304
#22 0x00038838 in c_write_global_declarations () at ../../gcc/gcc/c-decl.c:9844
#23 0x00830b46 in compile_file () at ../../gcc/gcc/toplev.c:564
#24 0x0083375f in do_compile () at ../../gcc/gcc/toplev.c:1886
#25 0x00833940 in toplev_main (argc=13, argv=0xbffff400) at
../../gcc/gcc/toplev.c:1958
#26 0x00149e0f in main (argc=13, argv=0xbffff400) at ../../gcc/gcc/main.c:36

(gdb) p debug_rtx (operands[1])
(const_int 2147483648 [0x80000000])
$15 = void

The problem would be avoid if the const_int was sign extended to
HOST_WIDE_INT, but who's supposed to do it?

[Bug middle-end/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #3 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-07-13 08:24:40 UTC ---
Well, the expander - or rather the constant should be properly extended in
the first place (if it is signed).

[Bug middle-end/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.07.13 10:34:54
     Ever Confirmed|0                           |1

--- Comment #4 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-07-13 10:34:54 UTC ---
Indeed, (const_int 2147483648 [0x80000000]) isn't valid for SImode, it should
be (const_int -2147483648 [0x80000000]) as RTL constants are always
sign-extended for their mode.  Usually a GEN_INT used in lieu of gen_int_mode.

[Bug middle-end/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #5 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2011-07-13 10:49:16 UTC ---
> Indeed, (const_int 2147483648 [0x80000000]) isn't valid for SImode, it should
> be (const_int -2147483648 [0x80000000]) as RTL constants are always
> sign-extended for their mode.

Actually (const_int -2147483648 [0xFFFFFFFF80000000]) on a 64-bit host.

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |target

--- Comment #6 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-13 15:24:34 UTC ---
Looks like a target bug in pa.md.

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #7 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-15 18:11:23 UTC ---
Author: danglin
Date: Fri Jul 15 18:11:18 2011
New Revision: 176319

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176319
Log:
    PR target/49723
    * config/pa/pa.md (casesi): Use gen_int_mode instead of GEN_INT.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/pa/pa.md

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #8 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-15 18:13:16 UTC ---
Author: danglin
Date: Fri Jul 15 18:13:13 2011
New Revision: 176320

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176320
Log:
    PR target/49723
    * config/pa/pa.md (casesi): Use gen_int_mode instead of GEN_INT.


Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/config/pa/pa.md

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #9 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-15 18:15:19 UTC ---
Author: danglin
Date: Fri Jul 15 18:15:16 2011
New Revision: 176321

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176321
Log:
    PR target/49723
    * config/pa/pa.md (casesi): Use gen_int_mode instead of GEN_INT.


Modified:
    branches/gcc-4_5-branch/gcc/ChangeLog
    branches/gcc-4_5-branch/gcc/config/pa/pa.md

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

--- Comment #10 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-16 18:33:46 UTC ---
Author: danglin
Date: Sat Jul 16 18:33:43 2011
New Revision: 176359

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=176359
Log:
    PR target/49723
    * config/pa/pa.md (casesi): Use gen_int_mode instead of GEN_INT.


Modified:
    branches/gcc-4_4-branch/gcc/ChangeLog
    branches/gcc-4_4-branch/gcc/config/pa/pa.md

[Bug target/49723] gcc.c-torture/compile/pr46934.c: ICE in do_SUBST, at combine.c:707 at -O1 and above

[danglin at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49723

John David Anglin <danglin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #11 from John David Anglin <danglin at gcc dot gnu.org> 2011-07-16 18:40:23 UTC ---
Fixed.