AUTOMATIC keyword

[Tim Prince]

On 1/6/2014 3:48 PM, Janus Weil wrote:
> Hi Mark,
>
>> I have been working on adding the 'automatic' keyword as a declaration
>> opposite to 'save'. This would obviously have no effect except when using
>> --fno-automatic and wouldn't be allowed with the 'save' keyword.
>>
>> Unfortunately --fno-automatic is at the moment implemented using
>> max-stack-var-size=0, so this would have to be re-implemented to allow
>> variables to go back on the stack.
>>
>> Does this sound like a good idea? Is there any support for adding the
>> 'automatic' keyword as a feature of gfortran?
> a few questions to evaluate the usefulness of your proposal:
>
> 1) Is the AUTOMATIC keyword part of any official Fortran standard?
> (Note that gfortran people usually less fond of implementing
> non-standard stuff. The main focus of the current development efforts
> is certainly to implement all aspects of the F03 and F08 standards,
> which is hard enough as it is.)
>
> 2) Is it implemented in any other compilers? (A quick web search seems
> to suggest that it is supported at least by Intel, Sun/Oracle and f2c
> ...)
>
> Cheers,
> Janus
automatic storage keyword extension is confusing vs. automatic objects 
(dynamic allocation) which is in the Fortran standard.  I think ifort 
implemented that keyword only for compatibility with some customer code, 
not because it does anything different from what is done without that 
keyword, provided that RECURSIVE attribute is given to the procedure, as 
is recommended when that effect is required.
gfortran already gave very careful consideration to g77 extensions which 
included everything in f2c.

-- 
Tim Prince