https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92586
Paul Thomas <pault at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |pault at gcc dot gnu.org
--- Comment #11 from Paul Thomas <pault at gcc dot gnu.org> ---
(In reply to G. Steinmetz from comment #5)
> A simplification might be :
>
>
> $ cat z1.f90
> program p
> type t
> integer, allocatable :: a
> end type
> type t2
> type(t) :: b
> end type
> type t3
> type(t2) :: c
> end type
> contains
> function g(x)
> class(t2) :: x
> type(t) :: g(2)
> end
> function f(x)
> class(t3) :: x
> type(t) :: f(2)
> f = g(x%c)
> end
> end
>
>
> $ gfortran-10-20191215 -c z1.f90
> gimplification failed:
> &parm.14 <addr_expr 0x7f89646c6e40
The tree dump contains the obviously wrong:
....snip....
struct array01_t parm.15;
....snip....
if ((&parm.15).a != 0B)
{
__builtin_free ((void *) (&parm.15).a);
(&parm.15).a = 0B;
}
which is why changing allocatable to pointer, for example, works.
Now to find out where that is happening.
Paul
