Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.

catching error due to larger adjustable array at run-time

abhimodak
New Contributor I
304 Views

Hi

The test-case below captures the gist of an error; where is the cause that the adjustable array is bigger than what has been passed in. It cannot be caught by the compiler; I understand. I am wondering is there a way (as in a flag) to catch it at run-time. Compiling with ifort /check:all Test.f90 won't do.

As you can imagine this is extracted from rather deep inside a combination of legacy and new program. In one of the test-cases of that program, it revealed itself as "access violation" but that was at a different location. 

Abhi

---

      Module OM
         Implicit None
         Integer :: M, N
         Integer :: L
      Contains
      
         Subroutine Set()
            M = 3; N = 5
         End Subroutine Set
         
         Subroutine Tech(K)
            Integer :: K(N) ! Size is larger than what's passed; its incorrect.
            Integer :: L
            K = -1 ! any way to catch this error at run-time?
            L = K(N) - 1 ! or here?
            Write(*,"(2(A,X,I0,X))") "in Tech: K(N) =", K(N), " and L =", L
         End Subroutine Tech
         
      End Module OM
       
      Program Test
         Use OM
         Implicit None
         Integer, Allocatable :: I(:)
         
         Call Set()

         Allocate(I(M))
         I = 0
         Call Tech(I)
          Write(*,"(A,X,I0,X)") "after the call L =", L
         
      End Program Test      

 

0 Kudos
3 Replies
FortranFan
Honored Contributor II
304 Views

Did you try /check:bounds?

A question: can you not use assumed-shape array argument and if the length is inadequate, do some error handling instead?

subroutine Tech(K)
   
   ! Argument list
   integer :: K(:) !<-- assumed shape
   ..
   if ( size(K) < N ) then
    ! Size is shorter than desired, do the needful
     .. 

 

0 Kudos
abhimodak
New Contributor I
304 Views

Hi FortranFan

Thanks for your comment.

(a) As I mentioned, this doesn't get trapped with /check:all. So, no /check:bounds won't catch it.

(b) Yes, there are ways to rectify this. But since this is coming from some legacy things, I am wondering if it can't be caught. 

Abhi

0 Kudos
Steve_Lionel
Honored Contributor III
304 Views

Intel Fortran doesn't offer a check for this sort of thing. It would require passing extra information in a way that would break existing programs.

Better to use assumed-shape arrays.

0 Kudos
Reply