-fno-automatic option (PR 20592)

[Richard E Maine]

On Aug 31, 2005, at 1:42 PM, François-Xavier Coudert wrote:

>> IMO recursive functions and variables with the ALLOCATABLE attribute 
>> should silently ignore -fno-automatic.
>
> There I disagree. I think the most easy thing is to say 
> "-fno-automatic means blanket SAVE statement in each programming 
> unit". If people don't want to have the nasty effects, then they don't 
> use it.

I think that ill-advised. The current equivalents of -fno-automatic are 
"safe" in that they sometimes make non-standard code work, but they 
never break standard-conforming code. As such, it is common in some 
environments for people to routinely use the switches. Whether that is 
ill-advised or not, it is quite common. People assume that it can't 
hurt.

I think that it is a very different matter to break standard-conforming 
code. I understand that it is simple to describe (and presumably to 
implement) that way, but I don't think it wise.

I would advocate for having the switch ignore recursive procedures (not 
just functions) and automatic variables.

-- 
Richard Maine                |  Good judgment comes from experience;
Richard.Maine@nasa.gov       |  experience comes from bad judgment.
                             |        -- Mark Twain