[Bug debug/42861] [4.4/4.5 Regression] Spill slots not tracked during var-tracking

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

------- Comment #3 from jakub at gcc dot gnu dot org  2010-01-26 12:14 -------
Note the http://gcc.gnu.org/ml/gcc-patches/2010-01/msg01282.html doesn't
bootstrap on the redhat/gcc-4_4-branch on ppc, while it bootstraps on the trunk
on x86_64-linux and i686-linux and on the branch on those arches too.
I will try to bootstrap trunk with that patch on ppc soon to see.
Anyway, the bootstrap problem can be simplified to:
typedef struct { char **b; int c[]; } T;
int a;
extern int d[], e;
extern char f[30];

static void
foo (unsigned *x, unsigned *y, int z)
{
  int i;
  for (i = 0; i < z; i++)
    x[i] = y[i / 2];
}

void
bar (T *y)
{
  unsigned x[4];
  foo (x, (unsigned *) d, e);
  __builtin_memcpy (y->b [y->c[0]], f, a);
}

at -O2 -g -m64 this doesn't assemble, as .debug_loc contains:
.LLST3:
        .8byte  .LVL1-.Ltext0    # Location list begin address (*.LLST3)
        .8byte  .LVL7-.Ltext0    # Location list end address (*.LLST3)
        .2byte  0x1      # Location expression size
        .byte   0x59     # DW_OP_reg9
        .8byte  .LVL7-.Ltext0    # Location list begin address (*.LLST3)
        .8byte  .LVL9-1-.Ltext0  # Location list end address (*.LLST3)
        .2byte  0xc      # Location expression size
        .byte   0x72     # DW_OP_breg2
        .byte   0x0      # sleb128 0
        .byte   0x3      # DW_OP_addr
        .8byte  .LC2@toc
        .byte   0x22     # DW_OP_plus
        .8byte  0x0      # Location list terminator begin (*.LLST3)
        .8byte  0x0      # Location list terminator end (*.LLST3)
and @toc isn't allowed in .8byte.

There is:
(insn 41 43 40 2 random.i:18 (set (reg/f:DI 9 9 [orig:148 d.0 ] [148])
        (mem/u/c:DI (plus:DI (reg:DI 2 2)
                (const:DI (unspec:DI [
                            (symbol_ref/u:DI ("*.LC2") [flags 0x2])
                        ] 49))) [7 S8 A8])) 350 {*movdi_internal64}
(expr_list:REG_DEAD (reg:DI 2 2)
        (expr_list:REG_EQUIV (symbol_ref:DI ("d") [flags 0xc0]  <var_decl
0x7fdb7e28de60 d>)
            (nil))))
...
(debug_insn 47 46 48 2 random.i:18 (var_location:DI y (reg/f:DI 9 9 [orig:148
d.0 ] [148])) -1 (nil))

before var-tracking and apparently var-tracking doesn't consider that
REG_EQUIV, which leads to:
(note 142 75 89 5 (var_location y (expr_list:REG_DEP_TRUE (mem/u/c:DI (plus:DI
(reg:DI 2 2)
            (const:DI (unspec:DI [
                        (symbol_ref/u:DI ("*.LC2") [flags 0x2])
                    ] 49))) [7 S8 A8])
    (const_int 0 [0x0]))) NOTE_INSN_VAR_LOCATION)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=42861