DO construct rules not consistent with Fortran standards.

[tomedunn]

The current rules for do-constructs work perfectly well for standard modern Fortran syntax but there are two use cases missing from the older punchcard Fortran (Fortran 77) standards.

In most cases do-constructs are writen like

do i = 1, 10, 1
  ...
end do

However, one could also replace the enddo statement with a continue statement

do i = 1, 10, 1
  ...
continue

and still be correct. This change on it's own would be rather trivial to make in the rules however there is one last form of the do-construct that complicates things significantly.

The previous two examples are both forms of un-labeled do-constructs. A do-construct can also be written as a labeled do-construct, such as:

       do 99 i = 1, 10, 1
         ...
99     k = k + i

When written in this format the line terminating the loop can be almost any executable statement (including continue and end do). Furthermore multiple labeled do-constructs can terminate on the same line. For instance

       do 99 i = 1, 10, 1
         do 99 j = 2, 20, 2
           ...
99     k = k + i

where both the inner and outer loop terminate on line 99.

I've spent a little time toying around with fixed for this but I haven't come up with anything particularly satisfying that didn't involved abandoning the more structured rules approach (i.e., getting rid of the multi-line do-construct rules and replacing them with the old do/enddo keywords.