Status of GCC-GNAT versus GNAT 3.14-3.15

Status of GCC-GNAT versus GNAT 3.14-3.15

[Preben Randhol]

Hi

I'm just wondering about how the gcc-gnat is doing. Is it getting up to
speed with the gnat 3.14-3.15?

Preben

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[chris.danx]

Preben Randhol wrote:
> Hi
> 
> I'm just wondering about how the gcc-gnat is doing. Is it getting up to
> speed with the gnat 3.14-3.15?

I would also like to know.  Also are there regular updates to the source 
tree like other gcc front ends or are there big foldins after a few 
months?  Just wondering since there's little point in building the 3.3 
set regularly if there's no change in the Ada front end, assuming the 
other front ends are fine.

Anyone?

Chris
-- 
for personal replies change 'spamoff' to 'chris'

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[chris.danx]

Does no one know?  Are ACT keeping it pretty hush hush? ;)

-- 
for personal replies change 'spamoff' to 'chris'

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[nicolas]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1791 bytes --]

gnat in gcc 3.3 works much better than in gcc 3.1 and 3.2, but it is still
far from gnat 3.14p
There has been a big merge in original 3.2 version, but a 3.1 branch has
been renamed 3.2, and 3.2 has been renamed 3.3

From the tests we make regularly on our code (several times a week), gnat in
gcc 3.3 works more or less fine in -O0
higher optimisation levels lead often to internal error at compilation, or
crash at runtime

I think you can forget about using gnat in gcc 3.1 and gcc 3.2.
It is not reliable at all, even in -O0, and its main purpose is to be able
to build gcc 3.3, which seems ok.
(it is not a good idea to build gcc with a different compiler for C and for
Ada)

After the big merge (about 6 monthes ago if I remember correctly), there
have been almost no modification in 3.3 for Ada sources,
except small updates, consequences of modifications which don't concern Ada,
made by non-Ada maintainers

If you are willing to build gcc 3.3 on Windows, you'll have to do that with
a Mingw environment
You will need some patches for gcc sources.
I got them from Mingw maintainers.
Windows (especially Mingw) is not a very well considered platform for gcc,
and even less as far as Ada is concerned
Some elementary patches for Mingw have never been commited in official gcc
sources after more than a year, and specific source patches are distributed
with Mingw

Things could change if there were enough people asking for that, reporting
bugs, and helping Mingw maintainers to have their patches more easily
accepted and commited


"chris.danx" <spamoff.danx@ntlworld.com> a �crit dans le message de news:
g2Ej9.70$8d4.240@news13-win.server.ntlworld.com...
> Does no one know?  Are ACT keeping it pretty hush hush? ;)
>
> --
> for personal replies change 'spamoff' to 'chris'
>

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[Preben Randhol]

On Tue, 24 Sep 2002 10:05:43 +0200, nicolas wrote:
> 
> If you are willing to build gcc 3.3 on Windows, you'll have to do that

No I'm using Linux. I ask as I see that debian is now building gnat 3.2
in unstable. Gnat 3.14 is in the stable distribution so I'll keep using
this until a gnat 3.x comes that are as stable as gnat 3.14 (or more).

I guess the biggest problem is to get the compiler to work with the
newer gcc backend ?

Preben

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[nicolas]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1075 bytes --]

we build gcc 3.3 on linux (Redhat) with gnat 3.14p, but I guess you can do
it with gcc 3.1 or gcc 3.2 (with gnat executables included)
I think the most tested configuration is to build gcc 3.3 with gcc 3.2
there are no required source patches, linux is one of the primary platform
supported by gcc
the most important thing to avoid problems is to use the same gcc for C and
Ada when you build
We make more tests on Windows than on linux, because all of our customers
use Windows

"Preben Randhol" <randhol+news@pvv.org> a �crit dans le message de news:
slrnap0eng.1q8.randhol+news@kiuk0156.chembio.ntnu.no...
> On Tue, 24 Sep 2002 10:05:43 +0200, nicolas wrote:
> >
> > If you are willing to build gcc 3.3 on Windows, you'll have to do that
>
> No I'm using Linux. I ask as I see that debian is now building gnat 3.2
> in unstable. Gnat 3.14 is in the stable distribution so I'll keep using
> this until a gnat 3.x comes that are as stable as gnat 3.14 (or more).
>
> I guess the biggest problem is to get the compiler to work with the
> newer gcc backend ?
>
> Preben

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[chris.danx]

Taken from gcc mailing list
(link to thread: http://gcc.gnu.org/ml/gcc/2002-09/msg00887.html)

--

On Monday, Sep 23, 2002, Andreas Jaeger wrote:
So, what's the status?  What are the plans that the Ada maintainers
have?

Part of the reason no more merges have been performed incorporating
changed made by Ada Core is that we have baselined on GCC 3.1 for
a while in order to have a stable back end while working on remaining
regressions using the GCC 3.1 back end compared to GCC 2.8.1.

One lesson learned from the last big merge I did, is that it only makes
sense to prepare such a merge after we have merged in all changes
made in the GCC tree. Last week Joseph assigned the GNATS PR about
dropped patches to me, and I will look at that PR and see what remains
to be fixed.

I'd like to encourage the Ada maintainers to set up at least a web
page documenting the Ada support in GCC and the plans they have...

The next major development we plan is converting the Ada front end
to function-at-a-time mode. We hope to be able to contribute this
before the end of the year.

   -Geert

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[chris.danx]

appologies, quoting got messed up in copy.  This is correct...

--

On Monday, Sep 23, 2002, Andreas Jaeger wrote:
 > So, what's the status?  What are the plans that the Ada maintainers
 > have?

Part of the reason no more merges have been performed incorporating
changed made by Ada Core is that we have baselined on GCC 3.1 for
a while in order to have a stable back end while working on remaining
regressions using the GCC 3.1 back end compared to GCC 2.8.1.

One lesson learned from the last big merge I did, is that it only makes
sense to prepare such a merge after we have merged in all changes
made in the GCC tree. Last week Joseph assigned the GNATS PR about
dropped patches to me, and I will look at that PR and see what remains
to be fixed.

 > I'd like to encourage the Ada maintainers to set up at least a web
 > page documenting the Ada support in GCC and the plans they have...

The next major development we plan is converting the Ada front end
to function-at-a-time mode. We hope to be able to contribute this
before the end of the year.

   -Geert

--

Again appologies for the misquoting!

Link to thread: http://gcc.gnu.org/ml/gcc/2002-09/msg00887.html

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[Preben Randhol]

On Tue, 24 Sep 2002 21:25:05 +0100, chris.danx wrote:
> 
> Link to thread: http://gcc.gnu.org/ml/gcc/2002-09/msg00887.html

I think this is also a problem:

   http://gcc.gnu.org/ml/gcc/2002-09/msg00906.html

Preben

Re: Status of GCC-GNAT versus GNAT 3.14-3.15

[Jerry van Dijk]

"chris.danx" <spamoff.danx@ntlworld.com> writes:

> > I'm just wondering about how the gcc-gnat is doing. Is it getting up to
> > speed with the gnat 3.14-3.15?
> 
> I would also like to know.  Also are there regular updates to the source tree
> like other gcc front ends or are there big foldins after a few months?  Just
> wondering since there's little point in building the 3.3 set regularly if
> there's no change in the Ada front end, assuming the other front ends are
> fine.

There has been a big merge in march, which was too late for the 3.2 release.

However it is in the current CVS HEAD, which will become 3.3 at some later
date. For Windows -which is what I am working on- the improvement from 3.2
to 3.3 is huge. However there are still a number of backend problems and 
about 100 ACVC tests fail.

As for the update policy, you should ask the maintainers, which currently are 
Robert Dewar and Geert Bosch.

gr.
Jerry.

-- 
--  Jerry van Dijk   | email: jvandyk@attglobal.net
--  Leiden, Holland  | web:   users.ncrvnet.nl/gmvdijk