[Bug tree-optimization/115991] [15 Regression] ICE on linux-6.10 in ix86_print_operand_address_as() since r15-1945-g9d20529d94b232

[slyfox at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115991

--- Comment #11 from Sergei Trofimovich <slyfox at gcc dot gnu.org> ---
Looking at indunction variable candidates:

<Invariant Expressions>:
inv_expr 1:     (signed int) offset_5 + 12
inv_expr 2:     (signed int) offset_5 + 16
inv_expr 3:     (signed int) offset_5 - (signed int) (unsigned long) ((u32 *)
&hdr + 4)
inv_expr 4:     (signed int) offset_5 - (signed int) (unsigned long) &hdr
inv_expr 5:     (signed int) offset_5 - (signed int) (u32 *) &hdr

I think we select `inv_expr 5:     (signed int) offset_5 - (signed int) (u32 *)
&hdr` on `x86_64`. Is it a valid expression? Or should it be `(signed int)
offset_5 - (signed int) (unsigned long) (u32 *) &hdr`?