README.md

Compile using either of:

./compile_gnu

and you should get the result:

           3 T F
true -> false
           3 F T
false -> true
Done.

Now compile using Intel:

./compile_intel

and you get the result:

           3 T F
true -> true
           3 F T
false -> true
Done.

Which is incorrect. Enable -fpscomp logical in compile_intel, recompile, then you get:

           3 T F
true -> false
           3 F T
false -> true
Done.

a.f90

module a
use iso_c_binding, only: c_int, c_bool
implicit none
private
public foo

contains

subroutine foo(n, relat, relat_out) bind(c, name="foo")
integer(c_int), intent(in), value :: n
logical(c_bool), intent(in), value :: relat
logical(c_bool), intent(out) :: relat_out
relat_out = .not. relat
print *, n, relat, relat_out
end subroutine

end module

compile_cray

# you can only find Cray compilers on a Cray machine...
module load PrgEnv-cray
module swap PrgEnv-intel PrgEnv-cray
module swap PrgEnv-gnu PrgEnv-cray
module swap PrgEnv-pgi PrgEnv-cray
cc -V
cc -o m.o -c m.c
ftn -o a.o -c a.f90
cc m.o a.o -o cray

compile_gnu

gcc -o m.o -c m.c
gfortran -o a.o -c a.f90
gcc m.o a.o -lgfortran -o gnu

compile_ibm

xlc -o m.o -c m.c
xlf2003 -o a.o -c a.f90
xlc m.o a.o -lxlf -o ibm

compile_intel

icc -o m.o -c m.c
#ifort -fpscomp nologicals -o a.o -c a.f90
ifort -o a.o -c a.f90
icc m.o a.o -lifcore -o intel

compile_pgi

pgcc -o m.o -c m.c
pgf90 -o a.o -c a.f90
pgcc m.o a.o -pgf90libs -o pgi

m.c

#include <stdio.h>
#include <stdbool.h>

void foo(int n, bool relat, bool *relat_out);

int main()
{
    bool relat_out;
    foo(3, true, &relat_out);
    printf("true -> %s\n", relat_out ? "true" : "false");
    foo(3, false, &relat_out);
    printf("false -> %s\n", relat_out ? "true" : "false");
    printf("Done.\n");
    return 0;
}

Compiling with Cray:

jhammond@edison02:~/ELEMENTAL/f2c-logical> cat compile_cray 
# you can only find Cray compilers on a Cray machine...
module load PrgEnv-cray || module swap PrgEnv-intel PrgEnv-cray || module swap PrgEnv-gnu PrgEnv-cray || module swap PrgEnv-pgi PrgEnv-cray
cc -o m.o -c m.c
ftn -o a.o -c a.f90
cc m.o a.o -o cray

Results with Cray:

jhammond@edison02:~/ELEMENTAL/f2c-logical> ./cray 
           3 T F
true -> true
           3 F T
false -> true
Done.

Thanks! So Cray fails as well. What about IBM?

With the latest XLF AIX 14.1 compiler and the XLC AIX 12.1 compiler, without any additional option required, it produces

3 T F
true -> false
3 F T
false -> true
Done.

Not sure how Jeff got those results. Here is what I get with the current (8.2.5) Cray compiler:

> ftn -c modu.f90
> cc main.c modu.o
> ./a.out
3,  T,  F
true -> false
3,  F,  T
false -> true
Done.

Additionally, I tried a simpler code:

use,intrinsic :: iso_c_binding 
logical(c_bool) :: x
x = .true.
print *, c_bool
write (*,"(g10.8,L5)") transfer (x, 0_1), x
x = .not. x
write (*,"(g10.8,L5)") transfer (x, 0_1), x

end

which results in 

> ftn test.f90
> ./a.out
1
        1    T
        0    F

illustrating that the Cray compiler does use 1 and 0 for true and false, so does have correct interoperability.

Here's the BGQ release of XL on the POWER7 Linux logins...

[jhammond@vestalac1 f2c-logical]$ xlc -qversion=verbose
IBM XL C/C++ for Blue Gene, V12.1
Version: 12.01.0000.0006
Driver Version: 12.01(C/C++) Level: 131009
C Front End Version: 12.01(C/C++) Level: 131009
C++ Front End Version: 12.01(C/C++) Level: 131023
High-Level Optimizer Version: 12.01(C/C++) and 14.01(Fortran) Level: 131022
Low-Level Optimizer Version: 12.01(C/C++) and 14.01(Fortran) Level: 131009

[jhammond@vestalac1 f2c-logical]$ xlf -qversion=verbose
IBM XL Fortran for Blue Gene, V14.1
Version: 14.01.0000.0006
Driver Version: 14.01(Fortran) Level: 131009
Fortran Front End and Run Time Version: 14.01(Fortran) Level: 131009
Fortran Transformer Version: 14.01(Fortran) Level: 131009
High-Level Optimizer Version: 12.01(C/C++) and 14.01(Fortran) Level: 131022
Low-Level Optimizer Version: 12.01(C/C++) and 14.01(Fortran) Level: 131009

[jhammond@vestalac1 f2c-logical]$ ./ibm 
 3 T F
true -> false
 3 F T
false -> true
Done.

[jhammond@vestalac1 f2c-logical]$ cat compile_ibm 
xlc -o m.o -c m.c
xlf2003 -o a.o -c a.f90
xlf m.o a.o -o ibm

As I noted through another channel, Bill is correct and my attempts to test Cray's behavior were at first flawed.

Thanks everybody for testing this! Conclusion: gfortran, Cray, IBM work. Intel and PGI need an extra option to work.

I've documented this behavior here:

http://www.fortran90.org/src/gotchas.html#c-fortran-interoperability-of-logical