https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86582
Bug ID: 86582
Summary: [debug] vla size reported as 0 at Og
Product: gcc
Version: 9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: debug
Assignee: unassigned at gcc dot gnu.org
Reporter: vries at gcc dot gnu.org
Target Milestone: ---
Consider this test-case, minimized from vla-1.c:
...
/* { dg-do run } */
/* { dg-options "-g" } */
int __attribute__((noinline, noclone))
f1 (int i)
{
char a[i + 1];
a[0] = 5;
return a[0]; /* { dg-final { gdb-test . "sizeof (a)" "6" } } */
}
int
main ()
{
volatile int j;
int i = 5;
asm volatile ("" : "=r" (i) : "0" (i));
j = f1 (i);
return 0;
}
...
This fails at Og:
...
FAIL: gcc.dg/guality/vla-1.c -Og -DPREVENT_OPTIMIZATION line . sizeof (a) ==
6
...
The size is reported as 0 by gdb:
...
$ gdb vla-1.exe -batch -ex "b 10" -ex "r" -ex "p sizeof(a)"
Breakpoint 1 at 0x40049b: file vla-1.c, line 10.
Breakpoint 1, f1 (i=<optimized out>) at vla-1.c:10
10 return a[0]; /* { dg-final { gdb-test . "sizeof (a)" "6" } }
*/
$1 = 0
...
We should be able to report the correct size, because the vla is not optimized
away.
AFAIU from debugging gdb, the evaluation of this location expression:
...
.value 0xe # Location expression size
.byte 0xf3 # DW_OP_GNU_entry_value
.uleb128 0x1
.byte 0x55 # DW_OP_reg5
.byte 0x23 # DW_OP_plus_uconst
.uleb128 0x1
.byte 0x8 # DW_OP_const1u
.byte 0x20
.byte 0x24 # DW_OP_shl
.byte 0x8 # DW_OP_const1u
.byte 0x20
.byte 0x26 # DW_OP_shra
.byte 0x31 # DW_OP_lit1
.byte 0x1c # DW_OP_minus
.byte 0x9f # DW_OP_stack_value
...
fails on the DW_OP_GNU_entry_value.
It fails due to this throw in dwarf_expr_reg_to_entry_parameter in gdb:
...
/* DW_TAG_call_site_parameter will be missing just if GCC could not
determine its value. */
throw_error (NO_ENTRY_VALUE_ERROR, _("Cannot find matching parameter "
"at DW_TAG_call_site %s at %s"),
paddress (gdbarch, caller_pc),
msym == NULL ? "???" : MSYMBOL_PRINT_NAME (msym));
...
I see in main though:
...
call f1
.LVL9:
...
and
...
.uleb128 0x6 # (DIE (0x6a) DW_TAG_call_site)
.quad .LVL9 # DW_AT_call_return_pc
.long 0x84 # DW_AT_call_origin
...
By modifying main to return i, I get in addition:
...
.uleb128 0x7 # (DIE (0x77) DW_TAG_call_site_parameter)
.uleb128 0x1 # DW_AT_location
.byte 0x55 # DW_OP_reg5
.uleb128 0x2 # DW_AT_call_value
.byte 0x73 # DW_OP_breg3
.sleb128 0
...
and then gdb prints sizeof(a) as expected.
This seems to be cause DW_TAG_call_site_parameter is generated based on
REG_CALL_ARG_LOCATION, and in the case without call_site_parameter we have:
...
(call_insn 15 39 44 2 (set (reg:SI 0 ax)
(call (mem:QI (symbol_ref:DI ("f1") [flags 0x3] <function_decl
0x7f7bc3fc2700 f1>) \
[0 f1 S1 A8])
(const_int 0 [0]))) "vla-1.c":19 722 {*call_value}
(expr_list:REG_CALL_ARG_LOCATION (nil)
(expr_list:REG_DEAD (reg:SI 5 di)
(expr_list:REG_EH_REGION (const_int 0 [0])
(nil))))
(expr_list:SI (use (reg:SI 5 di))
(nil)))
...
and in the case with call_site_parameter we have:
...
(call_insn 15 14 17 2 (set (reg:SI 0 ax)
(call (mem:QI (symbol_ref:DI ("f1") [flags 0x3] <function_decl
0x7f701d65a700 f1>) \
[0 f1 S1 A8])
(const_int 0 [0]))) "vla-1.c":19 722 {*call_value}
(expr_list:REG_CALL_ARG_LOCATION (expr_list:REG_DEP_TRUE (concat:SI
(reg:SI 5 di)
(reg:SI 3 bx [orig:89 i ] [89]))
(nil))
(expr_list:REG_DEAD (reg:SI 5 di)
(expr_list:REG_EH_REGION (const_int 0 [0])
(nil))))
(expr_list:SI (use (reg:SI 5 di))
(nil)))
...
So, why doesn't var-tracking generate the REG_CALL_ARG_LOCATION that we need?
