https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100073
Martin Sebor <msebor at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2021-04-14
Assignee|unassigned at gcc dot gnu.org |msebor at gcc dot
gnu.org
Status|UNCONFIRMED |ASSIGNED
Ever confirmed|0 |1
--- Comment #2 from Martin Sebor <msebor at gcc dot gnu.org> ---
Looks like powerpc64 doesn't warn for the code, and apparently neither does
aarch64 with -mabi=ilp32, due to what seems like a limitation in the uninit
pass which sees this IL:
<bb 3> [local count: 105119324]:
doloop.6_16 = (unsigned int) indx_9(D);
ivtmp.9_22 = (unsigned long) &MEM <int[10]> [(void *)&arr + -4B]; <<<
&arr[-1]
_27 = out_11(D) + 18446744073709551612;
ivtmp.11_26 = (unsigned long) _27;
<bb 4> [local count: 955630225]:
# doloop.6_20 = PHI <doloop.6_17(6), doloop.6_16(3)>
# ivtmp.9_7 = PHI <ivtmp.9_6(6), ivtmp.9_22(3)> <<< ivtmp.9_22(3)
points to arr[-1]
# ivtmp.11_24 = PHI <ivtmp.11_25(6), ivtmp.11_26(3)>
ivtmp.9_6 = ivtmp.9_7 + 4; <<< &arr[0]
_28 = (void *) ivtmp.9_6;
_1 = MEM[(int *)_28]; <<< uninitialized
read, no warning
