Hello,
We are currently using the latest official build of IVF (2011 Update 11) and are still experiencing quite a few problems (corrupt data, seg faults) when using allocatable class elements in derived types. Unfortunately, so far we couldn't create simple reproducers for some of our problems. However, we succeeded in creating one simple reproducer for at least one case of data corruption when using allocatable class elements in a derived type.
In the following sample code, setting ClassMain%AllocType%c corrupts the value of ClassMain%AllocType%i (Note, this also happens with a pointer element instead of allocatable).
The problem seems to happen regardless of build options.
regards,
Thomas
PS: The debugger shows different values for ClassMain%AllocType%i than printed. We also have experienced such behavior quite often.
[fortran]
PROGRAM Test1
IMPLICIT NONE
TYPE :: tAlloc
REAL(8) :: a
INTEGER(4) :: i
REAL(8) :: b
REAL(8) :: c
END TYPE
TYPE :: tClassMain
LOGICAL :: f ! note, these unused fields are required to recreate the problem.
REAL(8) :: d ! removing any of these, makes the problem go away (or at least invisible)
CLASS(tAlloc), ALLOCATABLE :: AllocType ! The problem occurs only with class here. Type works fine.
END TYPE
TYPE(tClassMain) :: ClassMain
ALLOCATE(tAlloc::ClassMain%AllocType)
ClassMain%AllocType%i = 1
ClassMain%AllocType%a = 0.6789D0
WRITE(*,*) ClassMain%AllocType%i ! Here, i is 1 as expected.
ClassMain%AllocType%c = 0.12345D0
WRITE(*,*) ClassMain%AllocType%i ! Here, after writing c, i is changes as well and now reads 1353773692
END PROGRAM Test1
[/fortran]
