Re: GNAT GPL 2005 Edition is now available

[Ludovic Brenta <ludovic@ludovic-brenta.org>]

Florian Weimer <fw@deneb.enyo.de> writes:

> * Ludovic Brenta:
>
>> OK, here are the votes so far.  There are four contenders for the
>> title of "next Ada compiler for Debian": GNAT GPL 2005 Edition
>> (gnat-gpl below), gnat-3.4, gnat-4.0 and gnat-3.4+patches backported
>> from gnat-gpl.
>
> How does the new GNAT compiler behave from a technical point of
> view?  Is it better than the GCC 3.4 one?  Comparable to GNAT 3.15p?
> Will there be regular releases from now on, incorporating bug fixes?
>
> I don't think I'll have answers to those questions by next
> Tuesday. 8-(

Good point.  I have a more new data items from Matthias Klose, the
main GCC maintainer in Debian.

* The target release date for Debian Etch is 18 months after Sarge,
  i.e. December 2006.

* It is likely that Debian Etch will switch again from GCC 4.0 to GCC
  4.1, after 4.1 is released, and if 4.1 proves good enough.  The C++
  ABI will not (should not) change, so the transition should be much
  smoother than the current transition from 3.4 to 4.0 (which is already
  quite smooth).

* If Debian Etch moves to GCC 4.1, then GCC 4.0 will be dropped.  GCC
  3.4 will be kept, because the upstream maintainers of Fortran
  recommend shipping g77 (a Fortran 77 front-end and run-time library),
  which has been replaced in 4.0 by gfortran (Fortran 95), which is
  new and has incompatibilities.

* Whichever compiler is chosen for Etch (be it 4.0 or 4.1) will be
  frozen one month before the rest of Debian.  The plan is to freeze
  Debian for a couple of months before release.  This means that the
  compiler (and binutils and glibc) will probably be frozen in
  September 2006, and the rest of Debian in October 2006.

Nothing is 100% certain at this point, except that there is no hurry
to decide on a definitive Ada compiler for Debian Etch.  However, I
sense that some people are in a hurry to move away from gnat 3.15p.
Am I correct in this?

I encourage people to investigate the relative technical merits of the
contenders.  Most votes so far have ignored this issue altogether,
which is bad (IMHO).  If some people need more time, then I won't
close the vote on tuesday, but OTOH I'd like to reach some kind of
decision and then stick to it.

BTW, how many people here use gnat-3.4 or gnat-4.0 regularly?  Any
success or horror stories you might want to share?

And here is more food for thought:

If gnat-3.4 is chosen, its future availability in Etch is guaranteed,
so I can start a transition of the existing Debian packages soon
(e.g. October-November 2005).  No matter what, this transition will
take time and effort on my part.

If gnat-4.0 is chosen, then I'll probably wait until Matthias and the
Debian GCC maintainers decide on whether or not to move to 4.1, and
only then will I start the transition.  This places the transition in
the May-June 2006 timeframe.

None of the above affects Sarge users; only users of Etch or Sid, the
unstable, bleeding-edge branch of Debian.

-- 
Ludovic Brenta.