Re: Next public release of GNAT(?)

[dewar@gnat.com (Robert Dewar)]

Stephen Leake <stephen.a.leake.1@gsfc.nasa.gov> wrote in message news:<uadvn1ouh.fsf@gsfc.nasa.gov>...

> Actually, that will be closer to GNAT 3.15; a lot of work 
> has been done since 3.14, and that work is reflected in 
> the gcc source tree.

Closer, but not the same at all. There are major differences between
the source bases for 3.15 and 5.00.
 
> In addition, any release based on the gcc source tree 
> will not have the same quality aura as an ACT public 
> release, since ACT has much more strict quality controls > than gcc in general.

In particular, we have access to the ACT internal test suite and the
Compaq (formerly DEC) test suite, which
together provide a very extensive regression testing
facility. Of course bugs we find and fix will eventually
find their way into the GCC tree, but changes are constantly being
made to GCC that break things, so if
you take the snapshots on a particular day, there are
no guarantees.

> ACT seems to be getting into the habit of releasing the 
> "previous" version as public, keeping the "current" 
> version for supported customers.

Actually the two schedules are completely unrelated. The
3.14 public release has been delayed because we have been
putting those resources into the GCC 3 based source release for
gnu.org instead.

> I'm not objecting; any public release is fine with me. 
> On that basis, the answer to the question above is "the 
> day after the ACT customer release of 3.15" :).

For sure that is false, the period coming up to a major
release of GNAT Pro and following it for a couple of months
are a time when our release facilities are fully engaged.
We are trying to get the 3.14p release out, at least for
GNU/Linux in the near future, but if we don't make it,
then it will be delayed until well after the 3.15 releases.

Actually we have some internal debate here about the value
of the public releases given the development of the public
FSF GCC version. The latter is perfectly suitable for
research and student use, which is the primary target of
the public versions. Moreover, for general Free Software
being developed in Ada, we think it much more important
that it compiles and runs on the GCC 3 based FSF version
than on an obsolete 2.8 based technology. The requirements
for production use are of course radically different.

Robert Dewar
Ada Core Technologies