There seems to be a bug in ifort Version 13.0.1.117 regarding the finalization of allocated variables of extended type.
I don't think it has been reported in the forum.
Consider the following code:
[fortran]
MODULE modu
TYPE :: a_t
CONTAINS
FINAL :: d_a
END TYPE a_t
TYPE, EXTENDS (a_t) :: b_t
CONTAINS
FINAL :: d_b
END TYPE b_t
CONTAINS
SUBROUTINE d_a(this)
TYPE (a_t) :: this
PRINT *, "finalize a"
END SUBROUTINE d_a
SUBROUTINE d_b(this)
TYPE (b_t) :: this
PRINT *, "finalize b"
END SUBROUTINE d_b
END MODULE modu
PROGRAM test
USE modu
PRINT *, " sub1 "
CALL sub1()
PRINT *, " sub2 "
CALL sub2()
PRINT *, " sub3 "
CALL sub3()
CONTAINS
SUBROUTINE sub1()
TYPE (b_t) :: c
END SUBROUTINE sub1
SUBROUTINE sub2()
CLASS (a_t), ALLOCATABLE :: c
ALLOCATE(b_t :: c)
END SUBROUTINE sub2
SUBROUTINE sub3()
CLASS (a_t), ALLOCATABLE :: c
ALLOCATE(b_t :: c)
DEALLOCATE(c)
PRINT *, "deallocated"
END SUBROUTINE sub3
END PROGRAM test
[/fortran]
The finalization process for the variables in each of the subroutines sub1, sub2, sub3 is different (although it shouldn't be):
The variable in sub1 is finalized correctly (the final routine for the extended type is called, which, in turn, calls the final routine of the base type).
The variable in sub2 is finalized by calling the final routine of the base type (the declared type) only.
The variable in sub3 is finalized when deallocated explicitly by calling twice (!) the final routine of the base type. It is finalized again (although it shouldn't) at the end of the subroutine, by calling once the final routine of the base type.
