Regressions caused by copy_string patch.

[Steve Kargl]

On Fri, Jul 14, 2006 at 10:40:54PM +0200, Paul Thomas wrote:
> Are you all absolutely sure that FX's patch is the problem here?  No 
> matter what, FX's patch should produce an BUILT_IN_MEMMOVE someplace.  I 
> do not see such a call in the code produced by actual_array_substr_2.f90
> 
> Equally, however, I do not see any type mismatches in my patch.
> 

This reduced testcase still triggers the problem.

program pr28174
  implicit none
  character(len=12) :: a(2) = (/ "abc def ghij", "klm nop qrst" /)
  integer :: m = 7

  call foo (a(:)(m:m+5))

  contains
    subroutine foo (w)
      character(len=*), intent(in) :: w(:)
    end subroutine foo
end program pr28174

-- 
Steve