Re: Someone has caused regressions in gfortran

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Tue, Sep 04, 2007 at 08:37:15PM -0700, Steve Kargl wrote:
> On Tue, Sep 04, 2007 at 07:48:08PM -0700, Steve Kargl wrote:
> > > My fortran-fu is however not at level to figure out what precisely is
> > > going wrong in those two testcases.
> > 
> > I'll try to reduce the do_3.F90 code to a minimum testcase.  Unfortunately,
> > my middle/back-end knowledge is probably much worse than your Fortran-fu.
> > 
> 
> Honza,
> 
> I've reduce do_3.F90 to the following code:
> 
> program test
>   integer :: count

You can delete the above declaration.

>   integer(kind=1) :: i1
>   integer(kind=1),  parameter :: h = huge(h)
> 
>   count = 0
>   do i1 = -huge(i1)-1_1, huge(i1), 1_1
>      count = count + 1
>   end do

You can also delete the above 4 lines.

>   if (test_i1(-h-1_1, h, 1_1, h+1_1) /= int(h) * 2 + 2) call abort
> 
> contains
> 
>   function test_i1 (from, too, step, final) result(res)
>     integer(kind=1), intent(in) :: from, too, step, final
>     integer(kind=1) :: i
>     integer :: res
>     res = 0
>     do i = from, too, step
>       res = res + 1
>     end do
> 
>     if (i /= final) call abort
> 
>   end function test_i1
> end program test
> 
> A couple comments:
> 
> 1) AFAIK, gfortran will only inline CONTAIN'd function (see the contains statement above).
>    There are essentially static functions private to the test program.
> 
> 2) If I comment out either IF statement, then the test will not abort.  In particular,
>    in the function test_i1, i == final to the 'call abort' is never executed, but 
>    apparently inlining doesn't like the IF statement.
> 
> 3) integer(kind=1) is equivalent to int8_t.  I'll see if I can translate the Fortran
>    into a failing C program.
> 
> -- 
> Steve

-- 
Steve