weird storage behaviotr leading to segfault

[svar]

I need to pass a set of arrays to a program that does the work (e.g. an ODe solver). Now the way this is written is it allows a parameter array RPAR which is a 1-d double precision array
However, the parameters that must be passed are a number of multidimensional arrays, and in once case a complex *16 array. As far as I know 2D arrays in Fortran are always stored by columns and a complex *16 value is two contiguous double precision reals. The minimal example showing the problem is this:

implicit double precision(a-h,o-z)​

complex*16, DIMENSION ,, ALLOCATABLE :: evecs​

double precision, dimension ), allocatable :: evals,rpar​

n=3​

c allocate space

allocate (evecs(N,n ), stat=iaLLOCATEstatus)​

if (IALLOCATEstatUS /= 0) then​

write(6,*)'ERROR trying to allocate evecs '​

stop 1​

end if​

allocate (evals(N ), stat=iaLLOCATEstatus)​

if (IALLOCATEstatUS /= 0) then​

write(6,*)'ERROR trying to allocate evals '​

stop 1​

end if​

allocate (rpar(N*(2*N+1) ), stat=iaLLOCATEstatus)​

if (IALLOCATEstatUS /= 0) then​

write(6,*)'ERROR trying to allocate rpar '​

stop 1​

end if​

c fill the data structures we will work with

do 1 i=1,n​

evals(i)=1.d0*i​

write(6,*)'evals(',i,')=',evals(i)​

do 2 j=1,n​

evecs(j,i)=dcmplx(i*1.d0,-j*1.d0)​

write(6,*)'evecs(',j,i,')=',evecs(j,i)​

2

continue​

1

continue​

c because the processing routine wants all parameters in a single double precision array rpar, do that

DO 11 I=1,N​

RPAR(I)=EVALS(I)​

11

CONTINUE​

indx=n​

DO 12 I=1,N​

DO 13 J=1,N​

INDX=INDX+1​

RPAR(INDX)=DREAL(EVECS(J,I))​

INDX=INDX+1​

RPAR(INDX)=DIMAG(EVECS(J,I))​

13

CONTINUE​

12

CONTINUE​

call bigroutine(n,rpar)​

deallocate (rpar, stat=IALLOCATEstatus)​

deallocate (evals, stat=IALLOCATEstatus)​

deallocate (evecs, stat=IALLOCATEstatus)​

end​

subroutine bigroutine(n,rpar)​

double precision rpar(*)​

c here the main work is none. But it needs to work with evals ane evecs

call process(n,rpar)​

return​

end​

subroutine process(n,evals,evecs)​

complex*16 evecs(n,n)​

double precision evals​

do 1 i=1,n​

write(6,*)'evals(',i,')=',evals(i)​

1

continue​

indx=n​

do 2 i=1,n*n​

c row:

icol=1+(i-irow)/n​

irow=mod(i,n)​

if(irow.eq.0)then​

irow=n​

icol=i/n​

endif​

write(6,*)'evecs(',irow,icol,')=',evecs(irow,icol)​

2

continue​

return​

end​