Issue concerning forward referencing classes

[Arjen_Markus]

Hello,

I am trying to forward-reference a class (aggregate in the code below) but the Intel Fortran compiler

complains that that class has not been defined yet. If I use the type(...) keyword instead of class(...),

the code is accepted.

This is with version 11.1, so it is quite possible this problem has already been taken care of.

Regards,

Arjen

----

module forward

implicit none

type component

class(aggregate), pointer :: holder ! <== Intel Fortran 11.1 does not accept this line

end type

type aggregate

class(component), pointer :: part

end type

end module forward

[abhimodak]

Looks like it does the same in composer XE update 1 as well.

Abhi

[Wendy_Doerner__Intel]

Abhi,

Yes I have reproduced this behavior andbelieve it agrees with the Fortran 2003 standard:

"21 C439 (R440) If the POINTER attribute is specified for a component, the declaration-type-spec in the

22 component-def-stmt shall be CLASS(*) or shall specify an intrinsic type or any accessible derived

23 type including the type being defined."

In this case, POINTER is specified so the only CLASS allowed is CLASS(*).

------

Wendy

Attaching or including files in a post

[Arjen_Markus]

Right, well, I am still learning the fine print of OOP in Fortran 2003 :). It seemed odd to me that there should be

such anasymmetry between type(...) and class(...), but a class(*) will help. (The code, by the way, is accepted

by gfortran 4.6, but that compiler does not support unlimited polymorphic variables yet)

Regards,

Arjen

[IanH]

I don't think the constraint quoted invalidates the code... the component has the pointer attribute, the type "aggregate" is accessible (it's declared in the same scope) - on with the show.

I don't think the fact that it is polymorphic matters. Compare the wording of the constraint quoted with the one immediately before (C438 in F2003) which applies to non-pointer components - if you applied the same logic/interpretation to it then you'd be saying that you couldn't have any allocatable polymorphic components that were not CLASS(*). That would be a bit of a bummer.

Besides which, F2008 relaxes things even further, so Arjen's just ahead of the game in terms of his understanding of the fine print...

[Steven_L_Intel1]

Yes - indeed F2008 does relax this rule.

[L__Richard_L_]

I want to use a "forward-reference" to a module.  Is there any way to do this; e.g., pre-compiled modules?  My desired code sequence would have the following pattern:

• module M1 source. This references subroutine S2.
• module M2 source. This contains subroutine S2.

[Steven_L_Intel1]

This is really a very different question from what this thread is about - in the future, please start a new topic with new questions.

In the scenario you describe, module M1 would USE M2. M2 could not use M1. The build system will build sources in the correct order - M2 would have to be compiled already before M1 is compiled.