Re: Ada Generic vs. C++ Templates

["Marin David Condic" <marin.condic.auntie.spam@pacemicro.com>]

Well, I think it is true that *technically* speaking, an Ada compiler could
read any Ada program and generate code that produced the exception
Program_Error (or some other appropriate exception).(1) All you have to do
is posit that the target implementation is too small to contain *any* legal
Ada program of *any* size. This would be perfectly legal and presumably
ought to pass the validation suite. (This is why I have argued in the past
against the "perfectly legal" defense when a compiler doesn't do what I want
it to do. Its "Perfectly Legal" and "Totally Useless" :-)

I agree that it is less important to me to have a validated compiler than it
is to have one that produces correct, efficient and usable results. It does
give one a warm fuzzy feeling though if the compiler *is* validated because,
if nothing else, it implies that the validation suite has been used as a
test case and the bulk of it must have worked. Its significantly more than a
smoke test of the software. Also, it makes you feel comfortable that you can
use any of the language constructs and (with the possible exception of some
corner cases) get the compiler to recognize what you're trying to do &
hopefully do it.

(1) Hey! Anybody want to gang up and implement the worlds smallest/fastest
Ada compiler & get it validated? :-)

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"DuckE" <nospam_steved94@home.com> wrote in message
news:WMRy6.676244$U46.20968423@news1.sttls1.wa.home.com...
> I believe GNAT was the first (and may still be the only, I'm not sure) to
be
> "validated" for Ada95 including all annexes.  This does not mean the
> compiler is without bugs, nor does it mean that all features described by
> the Ada95 standard are implemented.  It merely means that the validation
> test suite was passed.
>
> Don't get me wrong, I think validation is a good thing.  But you must be
> careful about what it really means.
>
> BTW: I don't really care whether the compiler I use is validated, but it
is
> reassuring.
>