调用返回数组的函数时出现Fortran Error 256问题求助

Hey there, let's tackle that Error 256 you're hitting when calling swap, scale, and pivot in your Fortran program. In FTN95, this error almost always boils down to argument mismatches between your subroutine calls and the subroutine definitions—let's walk through the most likely fixes, since you need these functions to modify your input matrix.

Common Causes & Fixes #

1. Incorrect Argument Intent #

If your subroutines are supposed to modify the input matrix, their array parameters must be declared with intent(inout) (explicit intent is better practice than omitting it). If you accidentally used intent(in), FTN95 will throw Error 256 because you're trying to write to a read-only argument.

Correct Example for a Swap Subroutine:

subroutine swap(mat, row1, row2)
    implicit none
    real, intent(inout) :: mat(:,:)  ! Allows reading and modifying the matrix
    integer, intent(in) :: row1, row2
    real :: temp(size(mat,2))  ! Temp array to hold a full row
    
    ! Add bounds checking to avoid invalid indices
    if (row1 < 1 .or. row1 > size(mat,1) .or. row2 < 1 .or. row2 > size(mat,1)) then
        print *, "Error: Row indices out of bounds"
        return
    end if
    
    ! Perform the row swap
    temp = mat(row1,:)
    mat(row1,:) = mat(row2,:)
    mat(row2,:) = temp
end subroutine swap

2. Array Dimension Mismatch #

Your input is a 3x4 matrix (3 equations, each with 3 coefficients + 1 solution). If your subroutines use fixed-size arrays (e.g., real :: mat(3,4)) but you're passing a dynamic array, or vice versa, FTN95 will flag a mismatch.

Fix: Use assumed-shape arrays (mat(:,:)) in your subroutines. This makes them flexible for any size of 2D matrix and ensures compatibility with your input array, as long as you pass a 2D array when calling the subroutine.

3. Missing or Incorrect Subroutine Interfaces #

If your swap, scale, and pivot subroutines are not in a module (or you haven't included an explicit interface block), FTN95 uses implicit interface checking, which can miss subtle mismatches (like integer vs. real arguments, or wrong array rank).

Fix:

• Wrap your subroutines in a module and use that module in your main program, or
• Add an explicit interface block in your main program for each subroutine. This tells the compiler exactly what arguments to expect.

Example interface block for swap:

interface
    subroutine swap(mat, row1, row2)
        real, intent(inout) :: mat(:,:)
        integer, intent(in) :: row1, row2
    end subroutine swap
end interface

4. FTN95 Compiler Settings in Plato #

Plato's default FTN95 settings might have strict type checking enabled. Double-check that:

• You're passing arguments of the correct type (e.g., integer row indices, not real numbers)
• Your input matrix is indeed a 2D array (not 1D, which would cause a rank mismatch)

To get more details about the error, enable Verbose Error Messages in Plato:

1. Go to Project > Project Settings
2. Under Compiler Options, check the box for Verbose
3. Recompile—this will show exactly which argument is causing the mismatch.

Final Check #

Make sure your main program calls the subroutines with the right arguments. For example, if your matrix is declared as real :: eqn_matrix(3,4), your call should look like:

call swap(eqn_matrix, 1, 2)  ! Swap rows 1 and 2
call scale(eqn_matrix, 2, 0.5)  ! Scale row 2 by 0.5
call pivot(eqn_matrix, 1, 1)  ! Pivot on element (1,1)

If you still hit issues, sharing the exact code for your subroutines and main program calls would help narrow things down further!

内容的提问来源于stack exchange，提问作者Matt Johnson