https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125396
Drea Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2026-05-20
Status|UNCONFIRMED |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |pinskia at gcc dot
gnu.org
--- Comment #1 from Drea Pinski <pinskia at gcc dot gnu.org> ---
Confirmed.
I think maybe_remove_forwarder_block is causing the issue so mine.
```
<bb 14> [local count: 206998870]:
# c4_41(ab) = PHI <c4_10(ab)(13), c4_29(15)>
# arr1$0_43(ab) = PHI <arr1$0_25(ab)(13), 9(15)>
# sj12_44(ab) = PHI <sj12_31(ab)(13), arr1$0_32(15)>
f8 ();
<bb 15> [local count: 182536112]:
# c4_29 = PHI <c4_9(ab)(11)>
# arr1$0_32 = PHI <arr1$0_36(ab)(11)>
g26.2_40 = g26;
goto <bb 14>; [100.00%]
```
Neither of the incoming edges into bb 14 are abnormal but the out going one is.
Let me think about this. Note I suspect this was a latent bug in this code just
not invoked until now since the merging of the phis part is invoked more often.
Maybe the check is the bb_has_abnormal_succ on the non-forwarder block when
there are phis on the forwarder branch.
