I have been testing building "large" program in gfortran. The project is one I routinely work on in ifort. It has thrown up one or two interesting standards issues but that is by the bye. I am getting this error tagged on a submodule statement. I have tried various things but it has me baffled. Any ideas anyone? The submodule does contain a BIND(C) routine among other things.
C:\test>gfortran -c bpatn_SUBS.f90
3 | submodule (patn) patn_SUBS
Error: BIND(C) attribute at (1) can only be used for variables or common blocks
The 1 is under the n in (patn).
Without seeing the code, I would say the error message is wrongly placed or wrongly formulated. Is the BIND(C) routine private to the submodule or is it public? And does the declaration in the main module have the BIND(C) attribute as well? Mostly thinking out loud here, I am afraid.
Thanks for the Reply Argen. The binc(c) is public and the declarations match. Those would both have thrown errors in Ifort. The thought about misplaced message is an interesting one because the error is on the first non-comment line of code. Does gfortran have history on that? I am a fortranner back to the early 80s but a gfortran newbie. My my next step is to take copies and remove things step by step to see if the error goes away at some step.
I just tried the small sample code with Intel Fortran 18.0 and gfortran 10.2 and both accept the code with no error messages whatsoever (for gfortran I had to create a dummy ifwin module, but that was all). I don't have a newer version of Intel Fortran installed here.
Andrew, for use with gFortran 9.3 (Cygwin Win10 x64), I created a fake ifwin.f90 with the content
MODULE IFWIN integer, parameter :: POINTER_LEN = 8 integer, parameter :: INT_PTR = POINTER_LEN integer, parameter :: HANDLE = 8 integer, parameter :: UINT = 4 integer, parameter :: UINT_PTR = POINTER_LEN integer, parameter :: LONG_PTR = POINTER_LEN integer, parameter :: fLPARAM = LONG_PTR integer, parameter :: fWPARAM = UINT_PTR END MODULE
and compiled with the command
gfortran -c mywinty.f90 bpatn_IF.f90 bpatn_SUBS.f90
The compilation did not cause any messages to be issued. HTH
Thanks for that. It would have been kinder if I had removed the Kinds. Thanks for that, it seems it is just a bug in the version I am running. I am not familiar with Cygwin that is a linux like environment for windows I think yes? would I be able to get a binary from somewhere for the compiler that I can run in a Windows command prompt? I had a look around but failed so far.
To elaborate a bit on Cygwin and MinGW (note: this is my understanding of the nature of these two environments):
- Cygwin is a Linux-like environment on Windows, which comes with all the commands from that platform. A program built under Cygwin with the Cygwin-version of the compilers runs in a Cygwin console, but not in an ordinary DOS box.
- MinGW can be regarded as a light version of Cygwin and the most important difference is that programs built with the MinGW-version of the compilers can indeed run in an ordinary DOS box, which makes distributing them a bit easier. (They do depend on some MinGW-specific DLLs, but that is not an uncommon phenomenon - Cygwin is more intrusive)
A little sad that the gfortran project is having some trouble. I used one of the binaries that mecej4 suggested and built in a CMD window under Windows 10. The main thrust was not to switch the gfortran but to make my code compiler "independent":
1) By using a different compiler a few small things got thrown up. One such was C_LOC of things where target attribute was not specified. Ifort does not complain about that even with /STAND. Another was use if C_LOC when C_FUNLOC should have been used....
2) removal of compiler specific code. Given I only want windows builds I made a few new wrappers for the windows SDK routines to get rid of some IFPORT routines.
3) use my own bindings using binc(c) for the subset of IFWIN that I actually use. easier said than done but nearly complete.
OK I will try a later gfortan binary I was using GNU Fortran (MinGW.org GCC Build-2) 9.2.0
I'm curious: can you try this simple case with your version of gfortran and see if it compiles without errors.
module m interface module subroutine sub() bind(C, name="sub") end subroutine end interface end module submodule(m) sm contains module subroutine sub() bind(C, name="sub") end subroutine end submodule
@andrew_4619 wrote: 8 | submodule(m) sm | 1 Error: BIND(C) attribute at (1) can only be used for variables or common blocks To the small reproducer.
That suggests to me the version of gfortran you are using is BUGGY.
I can't find anything in the standard that makes that small reproducer non-conforming.
In addition, a later version of gfortran as well as the latest Intel Fortran compiler processes the small reproducer code with no errors or warnings:
C:\Temp>type m.f90 module m interface module subroutine sub() bind(C, name="sub") end subroutine end interface end module submodule(m) sm contains module subroutine sub() bind(C, name="sub") end subroutine end submodule C:\Temp>gfortran --version -c -Wall -std=f2018 m.f90 GNU Fortran (GCC) 10.0.1 20200209 (experimental) Copyright (C) 2020 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. C:\Temp>ifort /c /standard-semantics /warn:all /stand:f18 m.f90 Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 126.96.36.1991 Build 20201010_000000 Copyright (C) 1985-2020 Intel Corporation. All rights reserved. C:\Temp>