Intrensic error

[Astrodude]

Hello everyone

I'm using gfortran on SUSE to compile my code and receive the following error:

       yr_step = sign(1,delta_yr)                             
                       1
Error: 'b' argument of 'sign' intrinsic at (1) must be the same type and kind as 'a'

Here's my code:

        subroutine solar_loc(yr,t1,mean_long,declination)
        implicit none
        

        integer*2 yr,yr_step,delta_yr,i,iyr
        integer*4 t1
        integer*4 index
        real*4 mean_long,declination
        real*8 d,lambda,g,eps,L
        real*8 L0(10)/279.642,279.403,279.165,278.926,279.673,279.434,
     1                279.196,278.957,279.704,279.465/
        real*8 DL/0.985647/
        real*8 G0(10)/356.892984,356.637087,356.381191,356.125295,
     1    356.854999,356.599102,356.343206,356.087308,356.817011,356.56/
        real*8 DG/0.98560028/
        real*8 EPS0(10)/23.440722,23.440592,23.440462,23.440332,
     1    23.440202,23.440072,23.439942,23.439811,23.439679,23.439548/
        real*8 DE/-0.00000036/
        real sind,asind
        save

        d = 0

        if (yr .lt. 1900) then
           index = yr - 88
        else
           index = yr - 1988
        endif

        if (index .le. 0) then   
           delta_yr = index - 1  
        else if (index .gt. 10) then
           delta_yr = index - 10
        else 
           delta_yr = 0
        endif

        if (index .le. 0) index = 1
        if (index .gt. 10) index = 10

        yr_step = sign(1,delta_yr)
        delta_yr = abs(delta_yr)  


        do i = 1,delta_yr
          if (yr_step .gt. 0) then
            iyr = 98 + i - 1
          else 
            iyr = 89 - i
          end if
          if (mod(iyr,4).eq.0)then
            d = d + 366*yr_step   
          else
            d = d + 365*yr_step
          end if
        end do  

        d = d + t1/86400.

        L = L0(index) + DL*d
        g = G0(index) + DG*d
        do while (L.lt.0)   
          L = L + 360.      
        end do
        do while (g.lt.0)
          g = g + 360.   
        end do

        L = mod(L,360.0d0)
        g = mod(g,360.0d0)

        lambda = L + 1.915*sind(g) + 0.020*sind(2*g)
        eps = EPS0(index) + DE*d

        declination = asind(sind(eps)*sind(lambda))
        mean_long = L
        return
        end

Any suggestions? Thanks in advance!

 

[xwb]

sign(a, b) returns a with the sign of b. Both must be of the same type. In your case, you have a being INTEGER and b being INTEGER*2. Try

yr_step = sign(1_2,delta_yr)

Alternatively

integer*2 yrpositive
...
yrpositive = 1
...
yr_step = sign(yrpositive, delta_yr)

I prefer the first method but people who don't know the notation will moan about it in code reviews.

 

[Astrodude]

Good thinking! Thanks.

 

[Astrodude]

Both of the methods work well, but I'm sticking with the first one.

Now however I receive the error:

Function 'sind' has no IMPLICIT type

My guess is to fix this with:

real sind

Agree?

 

[xwb]

Agreed.

sind is one of those routines that may be intrinsic on some compilers (eg IVF). On others you have to define it.

 

[Astrodude]

Right.

So I added

real sind

But now the compiler complains

undefined reference to sind

Why is that?

 

[xwb]

sind is an intrinsic on some compilers but not on others. You can define it as

real function sind(angle_deg)
real angle_deg
sind = sin(angle_deg * 3.1415926535 / 180.0)
end function sind

 

[Astrodude]

Thanks! That does the trick.

But what about...

 lambda = L + 1.915*sind(g) + 0.020*sind(2*g)             
1
Error: Unclassifiable statement at (1)

 

[xwb]

At a guess, it doesn't like mixing the types. Why not just make everything real*8