https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122197
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Blocks| |99098
Keywords| |missed-optimization
--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
One thing I noticed is we have:
```
MEM[(struct _Sp_counted_deleter *)&storage].D.29810._vptr._Sp_counted_base =
&MEM <int (*) ()[7]> [(void
*)&_ZTVSt19_Sp_counted_deleterIP1XSt14default_deleteIS0_ESaIvELN9__gnu_cxx12_Lock_policyE2EE
+ 16B];
MEM[(struct _Impl *)&storage + 16B]._M_ptr = _9;
__libc_single_threaded.9_40 = __libc_single_threaded;
...
<bb 7> [local count: 115887544]:
MEM <vector(2) int> [(int *)&storage + 8B] = { 0, 0 };
_87 = MEM[(struct _Sp_counted_base *)&storage]._vptr._Sp_counted_base;
_88 = MEM[(int (*) () *)_87 + 16B];
PROF_91 = [obj_type_ref] OBJ_TYPE_REF(_88;(struct
_Sp_counted_base)&storage->2B);
if (PROF_91 == _M_dispose)
goto <bb 12>; [24.00%]
else
goto <bb 8>; [76.00%]
<bb 8> [local count: 88074534]:
PROF_93 = [obj_type_ref] OBJ_TYPE_REF(_88;(struct
_Sp_counted_base)&storage->2B);
if (PROF_93 == _M_dispose)
goto <bb 9>; [52.63%]
else
goto <bb 11>; [47.37%]
```
in optimized but we have 2 cehcks for different _M_dispose here one seems to be
other virtual function that is almost definition not called.
Now on the subject middle-end warnings, maybe the inlined devirtualization
regions we should mark those such not warn for these middle-end warnings
because most likely the code is never reachable.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99098
[Bug 99098] [meta-bug] invalid/missing -Wfree-nonheap-object warnings
