This is the mail archive of the
mailing list for the GCC project.
Re: Adding files for Ada
- To: Laurent Guerby <guerby at acm dot org>
- Subject: Re: Adding files for Ada
- From: Geert Bosch <bosch at gnat dot com>
- Date: Wed, 23 May 2001 16:04:41 -0400 (EDT)
- Cc: kenner at vlsi1 dot ultra dot nyu dot edu, dewar at gnat dot com, jsm28 at cam dot ac dot uk,gcc at gcc dot gnu dot org
On Wed, 23 May 2001, Laurent Guerby wrote:
There will probably some choices to make, should we discuss them on
firstname.lastname@example.org, or is it possible/better to set up a
email@example.com list or something? I don't expect a flood of
messages, but may be it's a good idea to avoid adding quite language
specific traffic when GCC 3.0 launch is scheduled (but on the other
hand we'll need some GNATS and DejaGNU expertise out of GCC long
I think it is best to keep everything on firstname.lastname@example.org for now.
As you say, during initial integration most traffic will *not* be
about Ada specific details. Later on, if we find that a large amount
of Ada specific traffic is generated, we can always decide to have
a new mailing list.
Robert Dewar sent me a few of those piece by piece by email. It would
be more practical for me to get the whole thing in a tarball on some
ftp server (or my ACT customer account).
The biggest issue is finding out, per test, whether it can be submitted
or not. Since obviously we cannot permit any mistakes here, don't count on
a tarball anytime soon.
Apart from that, the Ada validation suite contains some 2300 executable
tests and 1300 compile-only tests. Since this comprehensive test suite
is freely distributable, I think we should focus on using that. It is our
experience that once the ACVC tests compile and run without problems,
the compiler is in pretty good shape.
BTW, what is the state of GNAT + GCC 3 vs the ACT testing system?
(At least on GNU/Linux or some other major platform.)
On GNU/Linux, out of about 6000 regression tests there are two crashes
remaining and about 40 other failures, most of which are script issues in the
regression tests (due to linking with dynamic libgcc and things like that).
This was using last nights sources from the mainline GCC. SPARC/Solaris
and DEC/Alpha are OK too, but I don't have recent test results for those.