Re: SGI GNAT Question? (Long)

[dewar@gnat.com]

In article <876789s6or.fsf@mihalis.ix.netcom.com>,
  Chris Morgan <mihalis@ix.netcom.com> wrote:

> In case you're wondering what the point of this
> explanation is, I should add that when I was a supported
> user, the Ada user group in our
> company suggested it would be prudent to simply stick to
> the public releases on the grounds that we wanted the
> most stable versions.

Just to clarify things here. The most recent commercial
release is almost always the same source base as the
public release. For example, the current commercial release
is 3.11b2, which is the commercial release to which 3.11p
corresponds.

Generally we expect our customers to be using 3.11b2, since
this is the latest stable commercial release. We advise
them to use 3.11b2 rather than 3.11p for several reasons:

1. THe 3.11b2 comes from us. When you get 3.11p from some
   public site, you really don't know what you are getting,
   and neither do we, we obviously can't guarantee what is
   out there is the same as what we put out. It probably is
   but we have no control and no assurance of this.

2. The commercial version is the one which we support and
   for which we provide a guarantee. For example we will
   provide formal Y2K certification for 3.11b2, but we
   never make any such guarantees for the public system.

Now of course a number of our customers are using later
versions. We don't make these automatically available
(Chris in fact did NOT always have his hands on the latest
version!) We only make them available if in the judgment
of both ACT and the customer, it makes sense to move to
what we call a "wavefront version", identified by a w in
the version number, as in 3.12w, to fix a specific problem.

Where possible, both ACT and most customers prefer to work
around problems than to get new compiler versions, but the
development of GNAT is very rapid, and already the 3.12w
wavefront contains not only a large number of fixes, but
also some very important new functionality, developed for,
and in some cases funded by, customers, and of course these
customers get access to this wavefront version to test out
these new capabilities (an example is the -gnatR feature
that I highlighted the other day).

The features file for 3.12w already contains about 130
lines listing some 40-50 new features and fixes.

We are working towards getting this new level of the
technology releasable as fast as possible. First we will
make a 3.12a release for our customers, and then if there
are no glitches, make a corresponding 3.12p. if there are
minor glitches, we will fix them first. Note that although
some customers move to wavefronts rapidly, many have a
(very reasonable) policy of only looking at official
releases, so when we announce 3.12a, we get a lot of
people looking at it in a short time. If we do find
glitches we will make a 3.12a1, 3.12a2 etc and iterate
(that's why we arrived at 3.11b2) before we make the
public release.

Finally I note that Chris refers to EGCS and Linux. IN fact
the development situation for these projects is much the
same as in the GNAT case, with regard to major
developments.

Cygnus is doing major work on gcc on its internal tree.
You won't know the details unless you are a Cygnus customer
under an appropriate non-disclosure agreement. They will
synchronize these changes with EGCS at an appropriate
point. Similarly within redhat and the other Linux
companies all sorts of major internal development is
taking place that has not seen the light of day yet.

What is missing in the GNAT case is more active playing and
contributing with/to the public versions. Note that Marcus
Kuhn and his band of Linux enthusiasts are trying to do
something about this, and we are working with them to
figure out how to make this work smoothly.

We are also working out how to integrate GNAT into the
EGCS release, and perhaps this will also encourage that
kind of activity.

I don't mean to say there is NO useful work going on, not
at all, Jerry's announcement this morning for example is
a great case of important volunteer contributions, and
there are many others. Everyone is in favor of seeing a
more vital activity there!

Robert Dewar
Ada Core Technologies

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own