Re: [PATCH] Fix FMT= with CHARACTER arrays (PR fortran/39865, take 2)

[Tobias Schlüter <tobias dot schlueter at physik dot uni-muenchen dot de>]

Hi,

Jerry DeLisle wrote:
> Hmm, g77 does compile the following without complaint.
>
>       subroutine foo (a)
>       complex :: a(1:5,1:10)
> c      complex :: b(:,:)
>       write (*, fmt=a) 1
> c      write (*, fmt=b(3,2)) 1
>       end subroutine
>
> In this case maybe give an error unless -std=legacy, since it goes 
> against all standards from f77 on, and warn otherwise. I would sure hope 
> nobody is actually using this.
>
> Jakub, if you are aware of s specific user issue with all this, let us 
> know,  If not, I would like to see a hard error and be done with this.

My take on this: we have to support Hollerith formatting for backwards 
compatibility.  But we don't have to support Hollerith formatting 
combined with Fortran 90 constructs, as we don't want new code written 
with this ugly extension.  The gaps in the arrays Jakub was talking 
about can only arise with Fortran 90 constructs (viz complex :: b(:,:) 
above).  Therefore,
   write(*, fmt=b(3,2)) 1
should give a hard error in my opinion "Error: legacy hollerith 
formatting is not allowed with assumed shape array" (modulo terminology, 
I always forget what is assumed-size, assumed-shape etc).

Cheers,
- Tobi