[patch, testsuite] PR32057 random failures in secnds.f

[Jerry DeLisle <jvdelisle at verizon dot net>]

I have committed the attached revised version of the test case.  I looped it 
10000 times with no failures.  It has to be better than what we had.

Thanks to Dominique for his investigations and ideas on this.

Regards,

Jerry

2007-06-05  Jerry DeLisle  <jvdelisle@gcc.gnu.org>

	PR testsuite/32057
	* gfortran.dg/secnds.f: Update test to avoid roundoff errors and
	midnight issues.

C { dg-do run }
C { dg-options "-O0 -ffloat-store" }
C Tests fix for PR14994 - SECNDS intrinsic not supported.
C
C Contributed by Paul Thomas  <pault@gcc.gnu.org>
C
      character*20 dum1, dum2, dum3
      real t1, t1a, t2, t2a
      real*4 dat1, dat2
      integer i, j, values(8), k
      t1 = secnds (0.0)
      call date_and_time (dum1, dum2, dum3, values)
      t1a = secnds (0.0)
      dat1 = 0.001 * real(values(8)) + real(values(7)) +
     &        60.0 * real(values(6)) + 3600.0 * real(values(5))
      ! handle midnight shift
      if ((t1a - t1) < -12.0*3600.0 ) t1 = t1 - 24.0*3600.0
      if ((t1a - dat1) < -12.0*3600.0 ) dat1 = dat1 - 24.0*3600.0
      if ((dat1 < nearest(t1, -1.)) .or. (dat1  > nearest(t1a, 1.)))
     &    call abort ()
      do j=1,10000
        do i=1,10000
        end do
      end do
      t2a = secnds (t1a)
      call date_and_time (dum1, dum2, dum3, values)
      t2 = secnds (t1)
      dat2 = 0.001 * real(values(8)) + real(values(7)) +
     &        60.0 * real(values(6)) + 3600.0 * real(values(5))
      ! handle midnight shift
      if ((dat2 - dat1) < -12.0*3600.0 ) dat1 = dat1 - 24.0*3600.0
      if (((dat2 - dat1) < t2a - 0.008) .or.
     &    ((dat2 - dat1) > t2 + 0.008)) call abort ()
      end