Re: Ada -vs- GNAT

[dewar@gnat.com (Robert Dewar)]

"Steve Doiel" <nospam_steved94@attbi.com> wrote in message news:<sXCH8.103053$Po6.225647@rwcrnsc52.ops.asp.att.net>...

>   pragma Warnings (Off, entitiy)
> 
> could be replaced by
> 
>   pragma Unreferenced (entity)
> 
> because the latter is "neater, reads better, and is clearer why it is
> there."
> 
> In the GCC context this type of change requires that GCC 3.2 be built with
> GCC 3.1 and no earlier version.

We have *always* had an absolute rule that any version of GNAT must be
built
with the previous version. In the GCC FSF context, some have argued
for relaxing this requirement, and indeed have tried violating it,
sometimes
with the result of running into very strange and difficult problems.

So I *strongly* advise anyone building GNAT to do so using the
previous
version, whatever you might read elsewhere. No point in buying
problems
you don't need.

Why do we recommend this?

a) because that's the way we test things, why pioneer untested
territory

b) because we dislike keeping bootstrap path kludges in the sources
any longer
than necessary. The particular example above is a good one. Pragma
Warnings (Off) is a blunt tool which may suppress other relevant
warnings (as you
probably know GNAT sources require no warnings ever to be generated).
The
new Unreferenced pragma is extremely useful for keeping the valuable
new
warning on unreferenced parameters whlie allowing it to be selectively
turned off where necessary.

I actually am arguing very strongly for 3.2 to require 3.1 in the FSF
environment for all sorts of reasons (it is interesting to note that
those
arguing against it are people who are typically outside the Ada
community
who are used to being able to compile C with any old compiler :-)

Robert Dewar
Ada Core Technologies