Failed to convert matlab function to dll

I was using matlab coder to convert my matlab function to dll. It converted to C code successfully but failed when compiling to dll. Target Build Log:

D:\Myfolder\codegen1\rt_nonfinite.c(21): ​error C2099: initializer is not a constant
D:\Myfolder\codegen1\rt_nonfinite.c(24): ​error C2099: initializer is not a constant

NMAKE: fatal error U1077: 'cl -c -nologo -GS -W4 -DWIN32 -D_MT -MT -D_CRT_SECURE_NO_WARNINGS /O2 /Oy- /source-charset:utf-8 -DMODEL=Recoil @Recoil_rtw_comp.rsp -Fo"rt_nonfinite.obj" "D:\Research\Achievement\RecoilCFD\FHZ30\Solve\matlabdll\codegen1\rt_nonfinite.c"' : return code '0x2'
Stop.
The make command returned an error of 2

I checked the rt_nonfinite.c and found that these lines:

#include "rt_nonfinite.h"
#include <math.h>

#if defined(__ICL) && __ICL == 1700
#pragma warning(disable : 264)
#endif

real_T rtNaN = (real_T)NAN;
real_T rtInf = (real_T)INFINITY;
real_T rtMinusInf = -(real_T)INFINITY;
real32_T rtNaNF = (real32_T)NAN;
real32_T rtInfF = (real32_T)INFINITY;
real32_T rtMinusInfF = -(real32_T)INFINITY;

NAN and INFINITY is not constant and they actually identified by compiler as function, which defined in corecrt_math.h

Anybody know how to solve this problem?

Thank you!

Answer 1

[Walter Roberson]

NAN and INFINITY is not constant and they actually identified by compiler as function, which defined in corecrt_math.h

In C, there is no automatic calling of functions. References to function names that do not include () to call the function, are semantically invalid unless proceeded by & .

But suppose naming the function without & or () were equivalent to taking the address of the function. Then your (real_T)NAN would be coercing the address of NAN to be real_T which is obvious nonsense.

At the very least you will need to provide function call syntax,

real_T rtNaN = (real_T)NAN();
real_T rtInf = (real_T)INFINITY();
real_T rtMinusInf = -(real_T)INFINITY();
real32_T rtNaNF = (real32_T)NAN();
real32_T rtInfF = (real32_T)INFINITY();
real32_T rtMinusInfF = -(real32_T)INFINITY();