Compiler option to warn about argument declarations of x(1) instead of x(*)

[ivanpribec]

@feenberg

but not an expression.

That makes more sense now. I’ve asked myself the same question as @Beliavsky before, when I was fixing the issue in the thread:

Aliasing issue with zero-length array

I am refactoring a F77 code for solving a system of second order PDEs along with coupled ODEs (if any). While testing, I encountered an aliasing bug that goes unnoticed even when warning and instrumentation flags are turned on. Assume the global vector of unknowns is stored in an array U(NEQN) where NEQN = NPDE*NPTS + NV, and NPDE is the number of PDE variables NPTS is the number of grid points for the PDE semi-discretization NV is the number of ODE variables (a non-negative number) the eleme…

Nowadays expressions are allowed:

subroutine demo1(n,x)
    integer :: n
    real :: x(n*n/2)
    print *,x
end subroutine

subroutine demo2(x,ipar)
    integer :: ipar(7)
    real :: x((ipar(1)**2)/2)
    print *, x
end subroutine

and an array is allowed to have zero-length which clashes with the (non-standard) v(1) convention.