Re: seperate keyword and seperate compilation with Gnat?

[jsa@organon.com (Jon S Anthony)]

In article <dewar.836627714@schonberg> dewar@cs.nyu.edu (Robert Dewar) writes:

> Jon said
> 
> "You can't be serious.  Really.  An _INFINITELY_ clearer statement would
> have simply been:
> 
>   "A proper_body is not required in the compilation environment for the
>    compilation of the corresponding parent_body"."
> 
> Such a statement would  be entirely inappropriate in the main body of 
> the standard, which is in the business of telling you what *is* required
> not what is *not* required. From a formal point of view, the above 
> statement makes as much sense as saying:

Which only goes to show that the "formal point of view" can at times
simply be empty of content (this from a mathematician who specialized
in logics and foundations).  Also, slavishly trying to adhere to
stating things in positives is not always the best (clearest)
approach.  Sometimes a negative can go to the heart of an issue far
more succinctly and clearly and with no more logical
"inappropriateness" than the equivalent but far more turgid positive.
And, of course, there are times when the negative is all you _can_ say
(though for a "standard" this is probably an indication that something
has gone wrong).


> It would be legitimate to add your statement as a note, if there is
> agreement that this is something that confuses people. For me, the
> idea that it is even conceivable that proper bodies of stubs would
> have to be around to compile the parent is so obviously incorrect
> that the note would be redundant. If this were the case, you would
> have no separate compilation at all for stubs, and that would make

I was not saying that the statement _should_ have been in the
standard.  I can actually accept that the RM as it stands is clear
enough on the issue.  My only point is (and it _still_ is) that the
statement is FAR clearer than what the RM actually says.  I put it
forward only in response to your belief that what the RM says on the
issue could not possibly be clearer.  That's just plain not true.


> no sense at all, since the whole point of separate units is to be
> able to compile them separately.

Not completely.  You can make a maintenance case that they help from
the standpoint of localizing change and potential change.

/Jon
-- 
Jon Anthony
Organon Motives, Inc.
1 Williston Road, Suite 4
Belmont, MA 02178

617.484.3383
jsa@organon.com