f95 standard question

[Jerry DeLisle <jvdelisle at verizon dot net>]

(Third attempt, the message appears to be getting blocked)

While documenting and checking the count intrinsic I get the following:

program test_count
    integer, dimension(2,3) :: a, b
    a = reshape( (/ 1, 3, 5, 2, 4, 6 /), (/ 2, 3 /))
    b = reshape( (/ 0, 3, 5, 7, 4, 8 /), (/ 2, 3 /))
    !print *, shape(a)
    print '(3i6.1)', a
    print *
    !print *, shape(b)
    print '(3i6.1)', b
    print *
    !print *, shape(a.ne.b)
    print '(3l6)', a.ne.b
    print *
    print *, count(a.ne.b)
    print *
    print *, count(a.ne.b, DIM=1)
    print *
    print *, count(a.ne.b, DIM=2)
end program test_count

$ gfc count.f90
$ ./a.out
     1     3     5
     2     4     6

     0     3     5
     7     4     8

     T     F     F
     T     F     T

           3

           1           1           1

           1           2

I get the same result with ifort as gfc, however, the draft f95 standard
that I have to look at states that the case of DIM=1 should be:

	[2, 0, 1] and not [1, 1, 1]

which makes sense eyeballing it, but I have two compilers that agree to
diaagree with the standard.  Keeping in mind I have only a draft
standard (section 13.14.24) that could be wrong, (of course I could be
wrong too).

A second note, the program segfaults if I set DIM=3 which is illegal and
 ifort catches that at compile time.

Any opinions please?

Jerry