Re: [patch, libgfortran] Fix PR27304 Data descriptors exhausted

[Brooks Moses <bmoses at stanford dot edu>]

Jerry DeLisle wrote:
> :ADDPATCH fortran:
>
> 2006-04-26  Jerry DeLisle  <jvdelisle@gcc.gnu.org>
>
> 	PR libgfortran/27304
> 	* io/transfer.c (formatted_transfer_scaler):  Generate error if data
> 	descriptors are exhausted.

That should be scalAr.  :)

> ! 	{
> ! 	  /* No data descriptors left.  */
> ! 	  if (n > 0)
> ! 	    generate_error (&dtp->common, NULL,
> ! 				"Data descriptors exhausted in format");
> ! 	  return;
> ! 	}

That error message isn't one that I'd necessarily understand correctly 
on first reading, despite the fact that I have a native-speaker 
understanding of the relevant secondary definition of "exhausted".

Also, it occurs to me that I don't understand it on second reading, 
either.  It sounds like the sort of error message that should occur when 
one has written something like

  write(*,'(2I6)'), 3, 4, 5

but that is perfectly acceptable, and simply puts the 5 in the next 
record (i.e., on the next line).

Thus, I'm a bit baffled: is it possible for this error to occur except 
in the case (from your test program) where there are no data descriptors 
in the format at all?  If not, perhaps an error message of "No data 
descriptors found in format" would be better.

I'd also like to see the testcase include a confirmation that the 
multi-line write statement above works correctly rather than producing 
the error.

- Brooks