Re: [gfortran,documentation patch] Improve the documentation of GNU extensions

[Brooks Moses <bmoses at stanford dot edu>]

FX Coudert wrote:
> +@node Integer-logical conversions
> +@section Integer-logical conversions
> +@cindex Integer-logical conversions
> +
> +@command{gfortran} allows integer-to-logical and logical-to-integer
> +conversions.

Given the recent comp.lang.fortran thread, this should probably describe 
the conversions -- i.e., what range of integers are converted to which 
logical, and what integers result when a logical TRUE or FALSE are 
converted.

Also, do these conversions work with all sizes of integers and logicals?

Does this apply (without generating an error) even when -std=legacy is 
not declared?  (I'm not immediately sure whether these are in a section 
that implies that all of these require -std=legacy or not, but I note 
that the goto-and-label-in-different-blocks section mentions the option 
explicitly.)  If there is not at least a warning when this occurs under 
-std=gnu, I should like to complain, but that's not a matter for your 
patch!  :)

> +@node Non scalar character formats
> +@section Non scalar character formats
> +@cindex Non scalar character formats
> +
> +@command{gfortran} allows a character array to be used as in a
> +@code{FORMAT} tag; the format is then understood as the concatenation of
> +all the character strings of the array. The following example program
> +@smallexample
> +  character*4,parameter :: str(2) = (/ "(I5,", "A4) " /)
> +  write (*,str) 12345, "6789"
> +  end
> +@end smallexample

This interpretation of character arrays in FORMAT statements is standard 
F95, not an extension.  See 10.1.2, p. 162, lines 3-6 of the standard.

> +outputs "123456789". Assigned Hollerith constants can also be used as
> +@code{FORMAT} variables, when code is compiled with the
> +@code{-std=legacy} option.

This, however, presumably is not standard F95, and should be described 
separately.

- Brooks