Fortran 200[3,8,X] permits a primary in a specification expression to be a dummy argument that is not OPTIONAL. I don't think ifort is enforcing this, as written in the standard.
MODULE m
IMPLICIT NONE
CONTAINS
FUNCTION optional_fun(arg)
INTEGER, INTENT(IN), OPTIONAL :: arg
INTEGER :: optional_fun(pure_fun(arg)) !<-- Notional badness
optional_fun = 1
END FUNCTION optional_fun
PURE FUNCTION pure_fun(arg)
INTEGER, INTENT(IN), OPTIONAL :: arg
INTEGER :: pure_fun
IF (PRESENT(arg)) THEN
pure_fun = arg
ELSE
pure_fun = 0
END IF
END FUNCTION pure_fun
END MODULE m
PROGRAM p
USE m
IMPLICIT NONE
PRINT *, SIZE(optional_fun(1))
PRINT *, SIZE(optional_fun())
END PROGRAM p
>ifort /check:all /warn:all /standard-semantics /stand:f2003 "2016-09-03 optional-fun.f90"
Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 17.0 Beta Build 20160517
Copyright (C) 1985-2016 Intel Corporation. All rights reserved.
ifort: NOTE: The Beta evaluation period for this product ends on 7-oct-2016 UTC.
Microsoft (R) Incremental Linker Version 14.00.24213.1
Copyright (C) Microsoft Corporation. All rights reserved.
"-out:2016-09-03 optional-fun.exe"
-subsystem:console
"2016-09-03 optional-fun.obj"
>"2016-09-03 optional-fun.exe"
1
0
The effective restriction on using an optional argument as an actual argument to a pure function that has a corresponding optional dummy argument in a specification expression seems a little artificial to me, given that F2008 explicitly (F08/0095) permits references to the PRESENT intrinsic in specification expressions.
I can't think of any way of providing default values for not-present optional arguments in specification expressions :(
