Re: [Fortran] Improved dependency analysis of pointer variables

[FX Coudert <fxcoudert at gmail dot com>]

> Finally, the third change is that a pointer variable "p" and another
> variable "v" can not reference the same object, if the base type of
> "p" is not the same as the base type of "v".  To demonstrate this
> consider the example code below extracted from the polyhedron benchmark
> capacita.f90:
>
>
>     integer, private, pointer, dimension(:,:) :: Grid
>     real, allocatable, dimension(:,:) :: Y
>
>     where (Grid == 0)
>       Y = 0
>     end where

Sorry for the possible very naive question, but doesn't that interfere 
with equivalences, such as:

$ cat a.f90
  integer :: i(5)
  real(4) :: x(5)
  equivalence(x,i)

  i = (/ 1, 0, 3, 5, 0 /)
  where (i == 0)
    x = -42.
  end where
  print *, i
  end
$ gfortran a.f90 && ./a.out
           1 -1037565952           3           5 -1037565952

(this is with the current mainline gfortran). I know this example is 
not valid from the standard point of view, but it's a construct that is 
used a lot (although usually in the other way around, setting fp values 
by equivalencing them to integers), including our testcase and tons of 
already available and widely used code.

FX