From: Simon Wright <simon.j.wright@mac.com>
Subject: Re: Only one Ada vendor?
Date: Sat, 20 Oct 2007 08:39:37 +0100
Date: 2007-10-20T08:39:37+01:00 [thread overview]
Message-ID: <m23aw66xqe.fsf@mac.com> (raw)
In-Reply-To: KrdSi.10507$oC3.1098@newsfe08.phx
"Nasser Abbasi" <nma@12000.org> writes:
> If this was say C or C++ code full of #ifdefs and each C compiler
> having its own set of a million different compiler switches I would
> understand, but Ada code? So I am just curious why would it be hard
> to "switch" Ada compilers?
Application code is very likely to depend on compiler-specific
extensions and support packages; eg GNAT's 'Unrestricted_Access and
GNAT.Sockets.
You might say 'so make your own socket binding', but our experience is
that that will support an imperfect subset of the underlying library;
for example, not providing TCP_NODELAY.
Another pressure is the need to avoid re-certification. We have strong
resistance to upgrading the version of GNAT we use in case it affects
application behaviour. And remember many of the projects that might
switch will be in long-term maintenance; if there's a complete set of
unit tests it would be less of a risk, but that can be a big IF.
--S
next prev parent reply other threads:[~2007-10-20 7:39 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-10-19 15:05 Only one Ada vendor? Sloan.Kohler
2007-10-20 0:49 ` Randy Brukardt
2007-10-20 2:08 ` Nasser Abbasi
2007-10-20 7:39 ` Simon Wright [this message]
2007-10-21 20:57 ` Florian Weimer
2007-10-21 17:23 ` Martin Krischik
2007-10-21 17:33 ` Gary Scott
2007-10-21 19:42 ` Maciej Sobczak
2007-10-21 23:57 ` Robert A Duff
2007-10-22 12:36 ` Georg Bauhaus
2007-10-22 20:55 ` Maciej Sobczak
2007-10-25 9:33 ` llothar
2007-10-25 9:58 ` Ludovic Brenta
2007-10-25 20:35 ` adaworks
2007-10-28 5:44 ` anon
2007-10-28 8:04 ` Pascal Obry
2007-10-28 11:13 ` Jerry van Dijk
2007-10-29 11:36 ` Georg Bauhaus
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox