[Bug fortran/67933] [4.9/5/6 Regression] ICE for array of a derived type with allocatable class in derived type object

["pault at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67933

Paul Thomas <pault at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pault at gcc dot gnu.org
           Assignee|unassigned at gcc dot gnu.org      |pault at gcc dot gnu.org

--- Comment #3 from Paul Thomas <pault at gcc dot gnu.org> ---
(In reply to Mikael Morin from comment #2)
> The generated finalization wrapper has code doing
>   deallocate(ptr2%classes(:)%class_var)
> where ptr2 is a pointer of type list_t to be finalized.
> This is missing scalarization, a correct code would do instead:
>   do i=1,size(ptr2)
>     ptr3 => ptr2%classes(i)
>     deallocate (ptr3%class_var)
>   end do

Hi Mikael,

This does not seem to be the problem. It is actually the
              gfc_reset_vptr (&se.pre, al->expr);
at trans-stmt.c:6048.

Once gfc_reset_vptr is modified to quit if it doesn't find a vptr, compilation
succeeds and the finalizer has:

S.7 = 0;
while (1)
  {
    if (S.7 > 9) goto L.11;
    if (ptr2->classes[S.7].class_var._data != 0B &&
                           ptr2->classes[S.7].class_var._vptr->_final != 0B)
      {
        desc.8.dtype = 296;
        desc.8.data = (void * restrict) ptr2->classes[S.7].class_var._data;
        ptr2->classes[S.7].class_var._vptr->_final (&desc.8, (integer(kind=8)) 
        ptr2->classes[S.7].class_var._vptr->_size, 1);
      }
    if (ptr2->classes[S.7].class_var._data != 0B)
      {
         __builtin_free ((void *) ptr2->classes[S.7].class_var._data);
      }
    ptr2->classes[S.7].class_var._data = 0B;
    S.7 = S.7 + 1;
  }
L.11:;

I'll post the patch as soon as I understand where the vptr's go to.

Cheers

Paul