https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88775
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
So after the patch we have
__x.5_4 = (long unsigned int) "hello";
__y.6_5 = (long unsigned int) _3;
if (__x.5_4 != __y.6_5)
goto <bb 3>; [75.00%]
else
goto <bb 4>; [25.00%]
<bb 3> [local count: 805306369]:
__builtin_memcpy (_3, "hello", 5);
goto <bb 14>; [100.00%]
<bb 4> [local count: 268435456]:
if (_3 >= &MEM[(void *)"hello" + 5B])
goto <bb 5>; [50.00%]
else
goto <bb 6>; [50.00%]
where we fail to elide this condition. Well, in theory "hello"
could be followed by "hello", thus in .str "hello\0hello\0" and
thus the condition could be true. Somehow w/o the libstdc++ patch
this condition either doesn't appear or is elided.
