Reputation: 75
I have this Fortran code.
subroutine DLL_TEST_PROJECT_001(dielconst, tandelta, kappa)
! Expose subroutine DLL_TEST_PROJECT_001 to users of this DLL
!
!DEC$ ATTRIBUTES DLLEXPORT::DLL_TEST_PROJECT_001
! Variables
implicit none
real*8::dielconst
real*8::tandelta
complex*16::kappa
! Body of DLL_TEST_PROJECT_001
kappa = dielconst * dcmplx(1.d0, tandelta)
end subroutine DLL_TEST_PROJECT_001
The return value is a complex (complex in C++).
Here's the C++ code.
typedef complex<double> (*forTest)(double, double);
library.load("C:\\forTest");
// The library loads ok.
forTest ft = (forTest)library.resolve("DLL_TEST_PROJECT_001");
// The function resolves ok and we have an address in ft.
// Now if I call the function...
complex<double> d = ft(1.0, 1.0);
// or just...
ft(1.0, 1.0);
// The app crashes with a segmentation fault.
I'm guessing that the crash has something to do with the return value of the Fortran function.
Any help?
Upvotes: 2
Views: 434
Reputation: 60088
Fortran subroutine
subroutine DLL_TEST_PROJECT_001(dielconst, tandelta, kappa)
real*8 :: dielconst
real*8 :: tandelta
complex*16 :: kappa
is equivalent to the C++ void function
void c_name(double *dielconst, double *tandelta, complex <double> *kappa)
(provided double
is equivalent to real*8
, which is common)
The wrong signature used in the original code caused a mismatch and the runtime error.
Upvotes: 3