[Bug middle-end/38505] [4.4 Regression] Revision 142061 caused ICE on __builtin_memcpy

["jakub at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Comment #5 from jakub at gcc dot gnu dot org  2008-12-12 20:19 -------
I'd say the bug is in may_propagate_address_into_dereference, called with addr:
 <addr_expr 0x7ffff7ffb780
    type <pointer_type 0x7ffff320f600
        type <integer_type 0x7ffff31f8480 short unsigned int sizes-gimplified
public unsigned HI
            size <integer_cst 0x7ffff31e7960 constant 16>
            unit size <integer_cst 0x7ffff31e7990 constant 2>
            align 16 symtab 0 alias set -1 canonical type 0x7ffff31f8480
precision 16 min <integer_cst 0x7ffff31e79c0 0> max <integer_cst 0x7ffff31e7930
65535>
            pointer_to_this <pointer_type 0x7ffff320f600>>
        unsigned DI
        size <integer_cst 0x7ffff31e7b70 constant 64>
        unit size <integer_cst 0x7ffff31e7ba0 constant 8>
        align 64 symtab 0 alias set -1 canonical type 0x7ffff320f600>
    constant
    arg 0 <var_decl 0x7ffff32bd960 d
        type <array_type 0x7ffff30f9c00 type <integer_type 0x7ffff31f8480 short
unsigned int>
            BLK
            align 16 symtab 0 alias set -1 canonical type 0x7ffff30f9c00
            pointer_to_this <pointer_type 0x7ffff31123c0>>
        addressable used public external common BLK file pr38505.c line 7 col
23
        align 16
        chain <function_decl 0x7ffff30f8f00 foo type <function_type
0x7ffff3210000>
            public static QI file pr38505.c line 10 col 1 align 8 initial
<block 0x7ffff310f2a0> result <result_decl 0x7ffff3110000 D.1597>
            (mem:QI (symbol_ref:DI ("foo") [flags 0x3] <function_decl
0x7ffff30f8f00 foo>) [0 S1 A8])
            saved-insns 0x7ffff32bda00>>>
and deref:
 <indirect_ref 0x7ffff7ffb540
    type <array_type 0x7ffff30f9a80
        type <integer_type 0x7ffff31f8480 short unsigned int sizes-gimplified
public unsigned HI
            size <integer_cst 0x7ffff31e7960 constant 16>
            unit size <integer_cst 0x7ffff31e7990 constant 2>
            align 16 symtab 0 alias set -1 canonical type 0x7ffff31f8480
precision 16 min <integer_cst 0x7ffff31e79c0 0> max <integer_cst 0x7ffff31e7930
65535>
            pointer_to_this <pointer_type 0x7ffff320f600>>
        sizes-gimplified BLK
        size <integer_cst 0x7ffff310d9c0 constant 320>
        unit size <integer_cst 0x7ffff310db70 constant 40>
        align 16 symtab 0 alias set 2 canonical type 0x7ffff30f9a80
        domain <integer_type 0x7ffff30f99c0 type <integer_type 0x7ffff31f8000
long unsigned int>
            sizes-gimplified DI
            size <integer_cst 0x7ffff31e7b70 constant 64>
            unit size <integer_cst 0x7ffff31e7ba0 constant 8>
            align 64 symtab 0 alias set -1 canonical type 0x7ffff30f99c0
precision 64 min <integer_cst 0x7ffff31e7720 0> max <integer_cst 0x7ffff310dab0
19>>
        pointer_to_this <pointer_type 0x7ffff3112240>>

    arg 0 <ssa_name 0x7ffff31e5d20
        type <pointer_type 0x7ffff3112240 type <array_type 0x7ffff30f9a80>
            static unsigned DI size <integer_cst 0x7ffff31e7b70 64> unit size
<integer_cst 0x7ffff31e7ba0 8>
            align 64 symtab 0 alias set -1 canonical type 0x7ffff3112240>
        visited var <var_decl 0x7ffff32bdbe0 d.0>def_stmt d.0_1 = (short
unsigned int[20] * {ref-all}) &d;

        version 1>>

It allows propagating of an incomplete type var, which it certainly shouldn't
allow.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38505