doubt for using poingers as components of derived data types

[cepheid]

programtest

implicitnone

typedata
real::val
type(data),pointer::next
endtypedata
type(data)::a,b

!--BODY--
allocate(a%next)
a%next%val=1.

b=a
b%next%val=2.

write(*,*)a%next%val,b%next%val

endprogramtest



See the above program, what I mean is that "a" and "b" should be indepedant, the out result should be " 1. 2.".

While after compiling and running, the actual output is "2. 2.".

So Why "a" and "b" are linked as the same? How to keep "a" and "b" indepedant after "b=a"?

[mecej4]

The confusion here supports the view that pointers are troublesome and should be used, with care, only when necessary.

After the assignment b = a, the components a%next and b%next point to the same variable. Therefore, the next statement, b%next%val=2., results in changing a%next%val also (noting that a%next and b%next point to the same variable).

You may find it instructive to run this program under the debugger, watching the changes to the components of a and b.

[cepheid]

If I want to pass the a structure to b( the a%next has been allocated), and I don't want a and b linked together after b=a, So how to avoid the error?

[mecej4]

You cannot use pointers and ask that pointers not have the properties of pointers! Nor can you ask pointers to have certain properties only some of the time.

My recommendation: do not be seduced by the apparent convenience of pointers.