https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92893
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |diagnostic
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-12-10
Blocks| |88443
Ever confirmed|0 |1
--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
The warning sees the IL below. It calls compute_objsize() to determine the
size of the destination (what _34 points to) which, after failing to determine
the size of the object the pointer points to, falls back on its type, which it
determines from the assignment to _30 to be 1.
A solution that I think would avoid the warning in this case is to work a
harder and before falling back on using the type consider where the pointer
comes from. Here, it's ADDR_EXPR (MEM_REF (SSA_NAME (get()), 1)), so by
drilling into the operands we arrive at offset 1 from Buf*. Determining that
offset 1 from Buf* corresponds to a trailing array would let us avoid the
warning. I think all the bits to implement this are in place (particularly
field_at_offset in gimple-ssa-sprintf.c), so they just need to be made use of
here.
#pragma GCC warning is unfortunately known to work poorly for middle-end
warnings. It's also on my list of things to do for GCC 10 but I haven't gotten
around to it yet.
f (const char * p)
{
char * D.2771;
struct Cat c;
struct View D.2558;
struct Cat D.2568;
struct Buf * _6;
long unsigned int _8;
char * _30;
char * _34;
<bb 2> [local count: 1073741824]:
_8 = __builtin_strlen (p_2(D));
_6 = get ();
MEM[(char * {ref-all})_6] = 0;
_30 = &MEM <char[1]> [(void *)_6 + 1B]; <<< char[1]
__builtin_memcpy (_30, p_2(D), _8);
_34 = _30 + _8;
MEM <unsigned int> [(char * {ref-all})_34] = 7895160; <<<
-Wstringop-overflow
c ={v} {CLOBBER};
D.2568 ={v} {CLOBBER};
D.2558 ={v} {CLOBBER};
return;
}
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88443
[Bug 88443] [meta-bug] bogus/missing -Wstringop-overflow warnings
