Re: [FORTRAN PATCH]: Improved integer SIGN intrinsic (& bug-fix)

[Brooks Moses <brooks dot moses at codesourcery dot com>]

Roger Sayle wrote:
[...]
> where we clearly call the function ia(), the X argument, multiple times.
> When this function has side-effects, we can end up generating incorrect
> results.  In the middle-end, we'd normally use a SAVE_EXPR to handle this,
> but the gfortran scalarizer conveniently provides a gfc_evaluate_now
> function for this purpose.  If I could write FORTRAN, I might even have
> attempted a new testcase for the testsuite.

Here, try this:

! { dg-do run }
! At one point, SIGN() evaluated its first argument twice.
program sign1
  integer :: i = 1
  if (sign(foo(i), 1) /= 1) call abort
  if (sign(foo(i), -1) /= -1) call abort
contains
  integer function foo(i)
    integer :: i
    foo = i
    i = i + 1
  end function
end

This currently fails, but it shouldn't (and commenting out the "i=i+1" 
line causes it to stop failing, as expected from the bug you found).

- Brooks