Strange error message under VS2010 - function reference does not mathc function definition

[Arjen_Markus]

Hello,

I have a rather curious problem running Intel Fortran 12.1 under VS2010. My code looks essentially like this:

program xx
    implicit none

    logical, external :: modelinitialize

    logical           :: success

    success = ModelInitialize()
end program xx

When I compile the actual source under VS2010 (as part of a much larger project/solution), I get the error message:

error #7977: The type of the function reference does not match the type of the function definition.   [MODELINITIALIZE]
        success = ModelInitialize()
^

(as the routine needs to be callable from Java, it is not contained in a module)

If I compile the source outside of VS2010 with the same compile options, this problem does not occur. None of my colleagues working on the same project see this problem. Even creating a new project with just this file does not give any compil problem.

What can be the matter?

[mecej4]

The names differ as to case (upper/lower/combination). Do you have directives that affect the case of external symbols? Perhaps to make the function work with Java, which is a language that honors case?

[Arjen_Markus]

No, just the standard ones - nothing that I can think of. Besides if case sensitivity were enabled, then wouldn't the two occurrences be seen as two entirely different routines/functions?

[Anthony_Richards]

Shouldn't you show us where ModelInitialize is defined so we can see what the compiler is expecting its interface to be?

[Arjen_Markus]

Well, that is in the top part of the file: logical, external :: ModelInitialize It is a more modern version of the FORTRAN 77 way of dealing with external functions (modern, because of the combination into one statement)

[Steven_L_Intel1]

In Visual Studio you get generated interface checking. The compiler is then noticing that the actual function is declared as returning a type other than LOGICAL. You need to show us the called routine.

[Arjen_Markus]

Steve, you are absolutely right! The source file containing the function ModelInitialize defines it as an integer function (because of the Java connection). Stupid blind spot - I had all but forgotten these automatically generated interfaces. Well, easy to fix. Thanks very much, Arjen