[Bug fortran/125314] ice in copy_reference_ops_from_ref, at tree-ssa-sccvn.cc:1147 with -O1.

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125314

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2026-05-15
                 CC|                            |rguenth at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |ice-on-valid-code,
                   |                            |wrong-code
     Ever confirmed|0                           |1

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
(gdb) p temp.opcode
$1 = FUNCTION_DECL


from

afun.i = 1;

(gdb) p debug_tree (lhs)
 <component_ref 0x7ffff722d180
    type <integer_type 0x7ffff76175e8 integer(kind=4) public SI
        size <integer_cst 0x7ffff761b018 constant 32>
        unit-size <integer_cst 0x7ffff761b030 constant 4>
        align:32 warn_if_not_align:0 symtab:0 alias-set 2 canonical-type
0x7ffff76175e8 precision:32 min <integer_cst 0x7ffff7602fc0 -2147483648> max
<integer_cst 0x7ffff7602fd8 2147483647>
        pointer_to_this <pointer_type 0x7ffff761fb28>>

    arg:0 <function_decl 0x7ffff7224600 afun
        type <function_type 0x7ffff722bdc8 type <record_type 0x7ffff72237e0
PDTty1_1_1>

so we are accessing a FUNCTION_DECL as record (it returns a record).

wrong code from the frontend.  I suppose intended GENERIC woudl be
sth like

 COMPONENT_REF <TARGET_EXPR <D.1234, CALL_EXPR (afunc)>, i>


I suppose I can make FRE not ICE but generate code to write to 'afun',
but I'm not sure RTL expansion does sth sensible to this.

struct S { int i; };
struct S foo();

void bar()
{
  ((struct S*)foo)->i = 1;
}

wraps the access in a MEM and that's handled fine.  I suppose IL verification
should catch the FUNCTION_DECL case, but for some reason it does not.

  MEM[(struct S *)foo].i = 1;