[Announcement] New Project: ASIS for GNAT

[Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Hello,

I like to announce the beginning of a new Project.

The Project amis to provide an up to date implementations of ASIS = Ada
Semantic Interface Specification for GNAT based on gcc 3.5. The currently
available open source ASIS implementation is for GNAT on gcc 2.8.1.

The Project is located at sourceforge: 

http://sourceforge.net/projects/gnat-asis/

Current experinental versions are available via cvs.

With Regards

Martin 
-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Jeff C,]

"Martin Krischik" <krischik@users.sourceforge.net> wrote in message
news:1252369.os30E2mMWi@linux1.krischik.com...
> Hello,
>
> I like to announce the beginning of a new Project.
>
> The Project amis to provide an up to date implementations of ASIS = Ada
> Semantic Interface Specification for GNAT based on gcc 3.5. The currently
> available open source ASIS implementation is for GNAT on gcc 2.8.1.
>
> The Project is located at sourceforge:
>
> http://sourceforge.net/projects/gnat-asis/
>
> Current experinental versions are available via cvs.
>

Hmm..Might be nice to target gcc 3.4 since although it looks promising, I am
not sure that
GNAT will be released in gcc 3.5 (which is currently targeted for
tree-ssa)..

This is nice to have since ACT has not provided up to date public releases
since 3.15. Perhaps this will
entice them to release one at http://libre.act-europe.fr

I suspect part of the reason for the hold off is that the FSF releases
containing Ada have been generally more
broken than 3.15.

Gcc 3.4 was looking like it might actually be a pretty good release for
GNAT... I have not checked
back recently on progress.


Anyone run the test suite recently on any popular targets and want to report
results?

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Jeff C, wrote:

> 
> "Martin Krischik" <krischik@users.sourceforge.net> wrote in message
> news:1252369.os30E2mMWi@linux1.krischik.com...
>> Hello,
>>
>> I like to announce the beginning of a new Project.
>>
>> The Project amis to provide an up to date implementations of ASIS = Ada
>> Semantic Interface Specification for GNAT based on gcc 3.5. The currently
>> available open source ASIS implementation is for GNAT on gcc 2.8.1.
>>
>> The Project is located at sourceforge:
>>
>> http://sourceforge.net/projects/gnat-asis/
>>
>> Current experinental versions are available via cvs.
>>
> 
> Hmm..Might be nice to target gcc 3.4 since although it looks promising, I
> am not sure that
> GNAT will be released in gcc 3.5 (which is currently targeted for
> tree-ssa)..

You think they will remove gnat from gcc?

> This is nice to have since ACT has not provided up to date public releases
> since 3.15. Perhaps this will
> entice them to release one at http://libre.act-europe.fr

It is not to difficult. All you have to do is update gnat subdirectory with
a current version from the original gcc sources and switch -gnatwe off. 

Of cource I would like for a more automated approach.

> I suspect part of the reason for the hold off is that the FSF releases
> containing Ada have been generally more
> broken than 3.15.

True, but I think with gcc 3.5 we might be almost there. I have made good
experiences with my snapshot compile.

> Gcc 3.4 was looking like it might actually be a pretty good release for
> GNAT... I have not checked
> back recently on progress.

I have tried both gcc 3.4 and gcc 3.5. Strangely a compile from the gcc 3.4
branch does not create all the gnat file:

/work/gnu_3.4/x86_64-suse-linux-gnu  Linux  martin@linux2  Fr Mï¿œr 12
19:27:10  standart
>la gcc/*gnat*
-rwxrwxr-x+   1 martin   users         12M 2004-03-08 12:08 gcc/gnat1*
-rwxrwxr-x+   1 martin   users        1,9M 2004-03-08 12:08 gcc/gnatbind*
/work/gnu_3.4/x86_64-suse-linux-gnu  Linux  martin@linux2  Fr Mï¿œr 12
19:27:17  standart
>

Does not look very good, does it? The main branch, which should be
"instable" provides:

/work/gnu_3.5/x86_64-suse-linux-gnu  Linux  martin@linux2  Fr Mï¿œr 12
19:28:46  standart
>la gcc/*gnat*
-rwxrwxr-x+   1 martin   users        2,3M 2004-03-07 18:40 gcc/gnat*
-rwxrwxr-x+   1 martin   users         13M 2004-03-08 10:42 gcc/gnat1*
-rwxrwxr-x+   1 martin   users        1,9M 2004-03-08 10:42 gcc/gnatbind*
-rwxrwxr-x+   1 martin   users         58K 2004-03-08 10:46 gcc/gnatbl*
-rwxrwxr-x+   1 martin   users        843K 2004-03-07 18:40 gcc/gnatchop*
-rwxrwxr-x+   1 martin   users        2,3M 2004-03-07 18:41 gcc/gnatclean*
-rwxrwxr-x+   1 martin   users        1,3M 2004-03-07 18:40 gcc/gnatfind*
-rwxrwxr-x+   1 martin   users        386K 2004-03-07 18:40 gcc/gnatkr*
-rwxrwxr-x+   1 martin   users        928K 2004-03-08 10:46 gcc/gnatlink*
-rwxrwxr-x+   1 martin   users        2,1M 2004-03-07 18:40 gcc/gnatls*
-rwxrwxr-x+   1 martin   users        3,6M 2004-03-08 10:46 gcc/gnatmake*
-rwxrwxr-x+   1 martin   users        2,2M 2004-03-07 18:41 gcc/gnatname*
-rwxrwxr-x+   1 martin   users        1,6M 2004-03-07 18:40 gcc/gnatprep*
-rwxrwxr-x+   1 martin   users        1,3M 2004-03-07 18:40 gcc/gnatxref*
-rw-rw-r--+   1 martin   users           0 2004-03-08 10:46 gcc
stamp-gnatlib
-rw-rw-r--+   1 martin   users           0 2004-03-08 10:44 gcc
stamp-gnatlib1
-rw-rw-r--+   1 martin   users           0 2004-03-08 10:44 gcc
stamp-gnatlib2
/work/gnu_3.5/x86_64-suse-linux-gnu  Linux  martin@linux2  Fr Mï¿œr 12
19:28:49  standart
>

Better, isn't it.

> Anyone run the test suite recently on any popular targets and want to
> report results?

With Regards

Martin

--  
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Jeff C, <jcreem@yahoo.com> wrote:
: 
: Gcc 3.4 was looking like it might actually be a pretty good release for
: GNAT... I have not checked
: back recently on progress.

The test results I have (If I do still have them, that is) are
from last year. I have not managed to get a 3.4 or a 3.5 compiler
on an AMD Duron since last year, bootstrapping loops in stage1/gnat1
translating sem_prag.adb.  But I'm glad I still have compilers from
last year :-)



Georg

Re: [Announcement] New Project: ASIS for GNAT

[Jeff C,]

"Martin Krischik" <krischik@users.sourceforge.net> wrote in message
news:2143348.gl8qfD7ZKX@linux1.krischik.com...
> Jeff C, wrote:
>
> >
>>
> > Hmm..Might be nice to target gcc 3.4 since although it looks promising,
I
> > am not sure that
> > GNAT will be released in gcc 3.5 (which is currently targeted for
> > tree-ssa)..
>
> You think they will remove gnat from gcc?
>

Not forever but potentially 3.5...
Though there was talk about removing it for a release if it was not ready
(i.e. Having a functional
Ada is not considered part of the release criterea for GCC 3.5. There is
apparently some non trivial  work to make Ada work with the tree-ssa
backend. Looks like they are making progress...So perhaps this will not
happen).

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Jeff C, wrote:

> 
> "Martin Krischik" <krischik@users.sourceforge.net> wrote in message
> news:2143348.gl8qfD7ZKX@linux1.krischik.com...
>> Jeff C, wrote:
>>
>> >
>>>
>> > Hmm..Might be nice to target gcc 3.4 since although it looks promising,
> I
>> > am not sure that
>> > GNAT will be released in gcc 3.5 (which is currently targeted for
>> > tree-ssa)..
>>
>> You think they will remove gnat from gcc?
>>
> 
> Not forever but potentially 3.5...
> Though there was talk about removing it for a release if it was not ready
> (i.e. Having a functional
> Ada is not considered part of the release criterea for GCC 3.5. There is
> apparently some non trivial  work to make Ada work with the tree-ssa
> backend. Looks like they are making progress...So perhaps this will not
> happen).

tree-ssa backend?!

With Regards 

Marin


-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Martin Krischik <krischik@users.sourceforge.net> wrote:
: tree-ssa backend?!
http://gcc.gnu.org/projects/tree-ssa/


-- Georg

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

"Jeff C," <jcreem@yahoo.com> writes:

> "Martin Krischik" <krischik@users.sourceforge.net> wrote in message
[...]
> > http://sourceforge.net/projects/gnat-asis/
> >
> > Current experinental versions are available via cvs.
> 
> Hmm..Might be nice to target gcc 3.4 since although it looks
> promising, I am not sure that GNAT will be released in gcc 3.5
> (which is currently targeted for tree-ssa)..

AFAIK, GNAT will be part of all future releases of GCC, won't it?  The
tree-ssa branch you are referring to affects only the back-end unless
I am mistaken.  Of course, such a massive change to the back-end might
affect the GNAT font-end; this is a quality issue which is orthogonal
to the existence of the GNAT front-end.

> This is nice to have since ACT has not provided up to date public
> releases since 3.15. Perhaps this will entice them to release one at
> http://libre.act-europe.fr

They have told me privately that they do plan to make more "p"
releases in the future; these will have ASIS and GLADE.  I am not a
paying customer, so please do not take this as a promise.  Obviously,
they did not mention any release date; like Debian, it will be
released "when it's ready".

I certainly hope, though, that the ASIS project on SourceForge will
coordinate with ACT in order not to duplicate effort.  I'm sure ACT
will appreciate the help.  I do not know whether or not they would
agree to doing all further development of ASIS on SourceForge.  They
did move their development of GCC to gcc.gnu.org.

> I suspect part of the reason for the hold off is that the FSF
> releases containing Ada have been generally more broken than 3.15.

Yes, switching the back-end from GCC 2.8.1 to 3.1 was rather
disruptive.  The GNAT in GCC 3.1, 3.2 and 3.3 are essentially the
same; I would not trust them too much, and I second Jeff's comment
that it would be better to target GCC 3.4 for ASIS.

> Gcc 3.4 was looking like it might actually be a pretty good release
> for GNAT... I have not checked back recently on progress.

ACT have put in a lot of work in GCC 3.4.  It will be very different
from 3.3, and, I think, much better.  Although 3.4 has not been
released yet, the release branch has been created in January 2004, and
is in Stage 3.  This means that no new features are being added to it;
the only changes are regression fixes.  So, if Martin's project
targets it for ASIS, it won't suffer from the moving target syndrome.

ACT are still working actively on GCC, but only on the main branch
which will become GCC 3.5.

Since ACT still plan to make more "p" releases, I would suggest to
distinguish between "ASIS for GNAT", which refers to the "p" releases
made by ACT, and "ASIS for GCC", which refers to the the project
hosted on SourceForge.

Since GCC 3.4 has not been released yet, there is still a chance that
"ASIS for GCC 3.4" might be released at the same time as GCC 3.4.
That would be neat, but I may be day-dreaming :)
 
> Anyone run the test suite recently on any popular targets and want
> to report results?

That would definitely help.

For GNAT itself, the ACATS test suite is part of 3.4 and HEAD (which
will become 3.5), but not part of 3.3 or any earlier release.

Now, I am not aware of an ASIS test suite but creating one would
definitely help.  The existing examples from the source distribution
can serve as a starting point.  The ASIS project on SourceForge could
be the central point to publish reports.

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[David Starner]

On Sat, 13 Mar 2004 21:38:34 +0100, Ludovic Brenta wrote:

> AFAIK, GNAT will be part of all future releases of GCC, won't it?  The
> tree-ssa branch you are referring to affects only the back-end unless
> I am mistaken.  Of course, such a massive change to the back-end might
> affect the GNAT font-end; this is a quality issue which is orthogonal
> to the existence of the GNAT front-end.

No. tree-ssa is all about improving the connection between the front
end and the middle end so more intelligent optimization algorithms can
be used. So GNAT needs large changes to work properly with the tree-ssa.

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Ludovic Brenta wrote:

> "Jeff C," <jcreem@yahoo.com> writes:
 
> I certainly hope, though, that the ASIS project on SourceForge will
> coordinate with ACT in order not to duplicate effort.  I'm sure ACT
> will appreciate the help. 

I will contact them

> I do not know whether or not they would
> agree to doing all further development of ASIS on SourceForge.  They
> did move their development of GCC to gcc.gnu.org.
 
I think A.S.I.S. should ultimatly be moved right into the gcc cvs tree.

> Since GCC 3.4 has not been released yet, there is still a chance that
> "ASIS for GCC 3.4" might be released at the same time as GCC 3.4.
> That would be neat, but I may be day-dreaming :)

Creating a ASIS release for any working gnat is not that difficult. Just
fill the "gnat" directory with the sources from the release you like and
type "make".

Only one function has changed it's name and the version number is now stored
inside a C source.

Only I have not got a working gcc 3.4.

BTW: The project is looking for a Debian Release Technican ;-).

> Now, I am not aware of an ASIS test suite but creating one would
> definitely help.  The existing examples from the source distribution
> can serve as a starting point.  The ASIS project on SourceForge could
> be the central point to publish reports.

The A.S.I.S. from ACT has it's own test suite and some tools (gnatelim).
Also I have got adabrowse working.

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Rolf Ebert]

"Jeff C," <jcreem@yahoo.com> wrote in message news:<3zl4c.17228$mM.123491@attbi_s02>...
 
> Anyone run the test suite recently on any popular targets and want to report
> results?

I ran gcc 3.3.3 and 3.4 prereleases on the ACATS test suite at several
optimization levels on Linux and Windows during the last few weeks. 
As I cannot attach a pdf or Excel file here, I am willing to send the
summary sheet to anybody who is interested.

In short: the 3.4 prereleases are considerably better than the 3.3.3
release.  The number of failures dropped from 24 to 7 (Linux) or 5
(Windows) out of more than 2300 test files.  The remaining 7 errors on
Linux are all related to the complex calculation library and don't
show up on Windows.  The number of failures is stable at -O0 to -O2. 
Only further optimizations and inlining (-O3, -gnatN,
-funroll-all-loop, -finline-functions) trigger additional failures.

    Rolf

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

rolf.ebert_nospam_@gmx.net (Rolf Ebert) writes:

> I ran gcc 3.3.3 and 3.4 prereleases on the ACATS test suite at several
> optimization levels on Linux and Windows during the last few weeks. 
> As I cannot attach a pdf or Excel file here, I am willing to send the
> summary sheet to anybody who is interested.
> 
> In short: the 3.4 prereleases are considerably better than the 3.3.3
> release.  The number of failures dropped from 24 to 7 (Linux) or 5
> (Windows) out of more than 2300 test files.  The remaining 7 errors on
> Linux are all related to the complex calculation library and don't
> show up on Windows.  The number of failures is stable at -O0 to -O2. 
> Only further optimizations and inlining (-O3, -gnatN,
> -funroll-all-loop, -finline-functions) trigger additional failures.


Thank you for these data points.  It has been my experience with 3.15p
that -gnatVa tends to trigger ICE's quite often.  As I find the
additional checks valuable, I would suggest that running the ACATS
tests with -gnatVa (and possibly -gnatwa, too) might be worthwhile.

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Martin Krischik <krischik@users.sourceforge.net> writes:

> I think A.S.I.S. should ultimatly be moved right into the gcc cvs tree.

Yes, and glade too.

> Creating a ASIS release for any working gnat is not that difficult. Just
> fill the "gnat" directory with the sources from the release you like and
> type "make".
> 
> Only one function has changed it's name and the version number is now stored
> inside a C source.
> 
> Only I have not got a working gcc 3.4.
> 
> BTW: The project is looking for a Debian Release Technican ;-).

I see the hint, but I'm not going to jump at your project just now.
Let me explain why.

I am working on the next stable release of Debian, code-named Sarge,
which is due out "when it's ready".  The definition of "ready", by
Debian standards, is quite conservative, and stable really means
stable.  I am trying to make sure that Sarge is an attractive and
rock-solid platform for Ada development; the platform already contains
many tools and libraries.  The foundation of this platform currently
consists of GNAT 3.15p, Glade 3.15p, ASIS 3.15p and Florist 3.15p.
Each new release of GNAT is binary-incompatible with the previous,
therefore the entire platform must migrate at the same time.  I think
you're already getting the drift.

Before I can migrate the entire platform to a new version of GNAT, I
must also have all of ASIS, Glade and Florist available and
coordinated with it.  In addition, the new GNAT itself must prove to
be an improvement over 3.15p (GCC 3.3 is not an improvement; it does
not even support tasking on powerpc, and lacks the three complementary
packages).

So, basically, I will wait until all four of GNAT, ASIS, Glade and
Florist are upgraded.  When they are upgraded, then of course I will
package them for Debian, and then I will migrate all other Ada
packages to them and there will be a new stable Ada development
platform.

At this point, I do not know if this new platform will be based on GCC
3.4 or on a "p" release made by ACT.  There is also a timing issue.
It seems that Sarge will probably be released (i.e. declared "stable")
in the second quarter of 2004.  I do not know if the new platform will
be available by then.  But you can help make that happen :)

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Ludovic Brenta wrote:

> Martin Krischik <krischik@users.sourceforge.net> writes:
> 
>> I think A.S.I.S. should ultimatly be moved right into the gcc cvs tree.
> 
> Yes, and glade too.

I wonder how difficult that migh be. Asking the gcc team to be a maintainer, 
asking them to accept a new lib,creating the "configure" jobs, etc. pp.

>> BTW: The project is looking for a Debian Release Technican ;-).
> 
> I see the hint, but I'm not going to jump at your project just now.
> Let me explain why.

[snip]

I, of course, never expected you to change the SAGE release.

> be available by then.  But you can help make that happen :)

Im am toying with adapting glade as well. - I think that glade might be a
nice complement to AdaCL.CGI. But first I like to add "multipart/form-data"
support - which neiter AdaCGI nor GNAT.CGI has got.

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Martin Krischik <krischik@users.sourceforge.net> writes:

> Im am toying with adapting glade as well.

Excellent initiative.  I do hope that this, too, is coordinated with
ACT.  The last thing I want is a fork.

> - I think that glade might be a
> nice complement to AdaCL.CGI. But first I like to add "multipart/form-data"
> support - which neiter AdaCGI nor GNAT.CGI has got.

With Glade, you can do distributed applications provided that all
parts are written in Ada.

With CGI, you can extend a web server such as Apache in a standard way
to provide simple functionality.

And with AWS, you can build a standalone server that provides
standards-compliant web services, complete with session management.

Who says Ada lacks leverage?  Now I'd like to see applications built
with all these libraries.  I'd gladly package them for Debian :)

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Ludovic Brenta wrote:

> Martin Krischik <krischik@users.sourceforge.net> writes:
> 
>> Im am toying with adapting glade as well.
> 
> Excellent initiative.  I do hope that this, too, is coordinated with
> ACT.  The last thing I want is a fork.

ACT did not answer my e-Mail.

>> - I think that glade might be a
>> nice complement to AdaCL.CGI. But first I like to add
>> "multipart/form-data" support - which neiter AdaCGI nor GNAT.CGI has got.
> 
> With Glade, you can do distributed applications provided that all
> parts are written in Ada.

I know. And I was thinking that Glade might be a nice alternative to the
very C'ish FastCGI.

> With CGI, you can extend a web server such as Apache in a standard way
> to provide simple functionality.

Currently I am outfitting AdaCL with a more complete GCI implentation for
more complex functionality. For example: Allready all CGI-data is stored in
tagged objects which can be streamed.

> And with AWS, you can build a standalone server that provides
> standards-compliant web services, complete with session management.

Session management - interesting. 

> Who says Ada lacks leverage?  Now I'd like to see applications built
> with all these libraries.  I'd gladly package them for Debian :)

Would you package AdaCL into Debian? If so I would gladly add you to the
project developers so we could do any needed changes together.

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Martin Krischik <krischik@users.sourceforge.net> writes:

> Would you package AdaCL into Debian? If so I would gladly add you to the
> project developers so we could do any needed changes together.

Sure.  I have already looked at it very briefly, and I'd have a few
questions first, like what is exactly the relationship with the Booch
Components.  But in principle, yes I would package it.

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Arnaud Charlet]

> ACT did not answer my e-Mail.

FWIW, I haven't seen your email.

Note that having a reliable ASIS working with a particular GNAT version
is a lot of work, and certainly much more than just putting some sources
together as you described, since there's a very close relationship
between the GNAT and ASIS data structures. I'm afraid your resulting source 
package will suffer from the same trouble gcc 3.3 did: frozen Ada sources
with a moving (incompatible) GCC back-end that compiles but is pretty
unreliable.

Same comment for glade btw: the glade run time is tightly coupled with the
corresponding GNAT run time.

Arno

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Hi Arnaud, it's nice to see you on c.l.a.

Arnaud Charlet <charlet@gnat.com> writes:

> Note that having a reliable ASIS working with a particular GNAT version
> is a lot of work, and certainly much more than just putting some sources
> together as you described, since there's a very close relationship
> between the GNAT and ASIS data structures. I'm afraid your resulting source 
> package will suffer from the same trouble gcc 3.3 did: frozen Ada sources
> with a moving (incompatible) GCC back-end that compiles but is pretty
> unreliable.

How would you recommend to solve this problem?

I would tend to carefully select a "known good" combination of the
GNAT front-end and GCC back-end, port ASIS to it, and stick with it
until a new "known good" combination comes out.  The problem is, I
don't know if a "known good" combination even exists today, apart from
trusty old GNAT 3.15p and GNAT Pro.  Since you seem to think so little
about GCC 3.3, would you recommend 3.4 instead?  I suppose HEAD (the
future 3.5) is too prone to the moving target syndrome.

And what would be the best way for Martin to help in a way ACT can
use?

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Arnaud Charlet wrote:

>> ACT did not answer my e-Mail.
 
> FWIW, I haven't seen your email.

It went to libre - where I got the original sources from.

> Note that having a reliable ASIS working with a particular GNAT version
> is a lot of work, and certainly much more than just putting some sources
> together as you described, since there's a very close relationship
> between the GNAT and ASIS data structures.

True, but the ASIS and GNAT parts are seperated inside two directrorys:
"./asis/" and "./gnat/". They comunicate via the specs - as it should be.

It is not that diffcult to replace the "./gnat/" sources with a more current
one. The resulting ASIS uses then the correct data structures and works
with adabrowse which is the most complex asis application I know.

> I'm afraid your resulting
> source package will suffer from the same trouble gcc 3.3 did: frozen Ada
> sources with a moving (incompatible) GCC back-end that compiles but is
> pretty unreliable.

As I said on the project web side: any resulting ASIS will only work
reliable with the GNAT from which it got its "./gnat/" sources.

> Same comment for glade btw: the glade run time is tightly coupled with the
> corresponding GNAT run time.

Which proves Ludovic and my poing: both sources should be moved into gcc
tree from where they can draw the current gnat sources all the time.

We don't know why ACT did not do so. We can only guess it was either because
of restricted resources or for comertial interest. Both of which are fair
enough - ACT is a comertial company after all.

The only thing I found sad was that libre.act-europe.fr:/anoncvs did not
have a asis or glade directory.

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Ludovic Brenta wrote:

> Martin Krischik <krischik@users.sourceforge.net> writes:
> 
>> Would you package AdaCL into Debian? If so I would gladly add you to the
>> project developers so we could do any needed changes together.
> 
> Sure.  I have already looked at it very briefly, and I'd have a few
> questions first, like what is exactly the relationship with the Booch
> Components.

Well, AdaCL contains an extension to the booch components. The extension
provides comonents for indefinite elements. You know, for storing String or
'Class.

AdaCL does not replace the booch components. You will need the booch
components to use AdaCL.

I would, of course, prefer to move the extension into the booch components
themself, but Simon does not use a public cvs server :-( so it would be
rather tricky.

>  But in principle, yes I would package it.

Thanks,

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Martin Krischik]

Ludovic Brenta wrote:

> 
> Hi Arnaud, it's nice to see you on c.l.a.
> 
> Arnaud Charlet <charlet@gnat.com> writes:
> 
>> Note that having a reliable ASIS working with a particular GNAT version
>> is a lot of work, and certainly much more than just putting some sources
>> together as you described, since there's a very close relationship
>> between the GNAT and ASIS data structures. I'm afraid your resulting
>> source package will suffer from the same trouble gcc 3.3 did: frozen Ada
>> sources with a moving (incompatible) GCC back-end that compiles but is
>> pretty unreliable.
 
> How would you recommend to solve this problem?
 
> I would tend to carefully select a "known good" combination of the
> GNAT front-end and GCC back-end, port ASIS to it, and stick with it
> until a new "known good" combination comes out.  The problem is, I
> don't know if a "known good" combination even exists today, apart from
> trusty old GNAT 3.15p and GNAT Pro.  Since you seem to think so little
> about GCC 3.3, would you recommend 3.4 instead?  I suppose HEAD (the
> future 3.5) is too prone to the moving target syndrome.

Somebody suggested the GNAT is going to skip one gcc release. And I frear
3.4 is just that version. While the 3.5 brach version at least compiles ok
the 3.4 bracne has not produced gnat executables for the last 2 weeks.

That is also the reason why there is no 3.4 release of ASIS.

I do updates on both on a regular basis. Once you have the source the update
does not take that long.

With Regards

Martin

-- 
mailto://krischik@users.sourceforge.net
http://www.ada.krischik.com

Re: [Announcement] New Project: ASIS for GNAT

[Rolf Ebert]

Martin Krischik <krischik@users.sourceforge.net> wrote in message news:<1728933.b9v3IVdcgI@linux1.krischik.com>...
> Somebody suggested the GNAT is going to skip one gcc release. And I frear
> 3.4 is just that version. While the 3.5 brach version at least compiles ok
> the 3.4 bracne has not produced gnat executables for the last 2 weeks.

AFAIK it will be accepted that Ada will be broken after the tree-ssa
merge.  The merge will take place on the main trunk that will
eventually lead to gcc-3.5.  If an upcoming version of gcc will not
have Ada, it will be gcc-3.5.

I regularily build gcc from the 3.3 and 3.4 branches on Linux and
Windows (about once per week) and don't have major problems.  (see my
other posting about ACATS results).

> That is also the reason why there is no 3.4 release of ASIS.

I hope that will change.

     Rolf

Booch Components and AdaCL in Debian (was Re: [Announcement] New Project: ASIS for GNAT)

[Ludovic Brenta]

Martin Krischik <krischik@users.sourceforge.net> writes:

> Well, AdaCL contains an extension to the booch components. The
> extension provides comonents for indefinite elements. You know, for
> storing String or 'Class.
> 
> AdaCL does not replace the booch components. You will need the booch
> components to use AdaCL.

OK, see Debian bug #234551.

> I would, of course, prefer to move the extension into the booch
> components themself, but Simon does not use a public cvs server :-(
> so it would be rather tricky.

Is it not possible for you to send him the patches so he can apply
them?  Alternately, would Simon consider hosting the Booch Components
on SourceForge, Berlios, Alioth or some such?  Or even as a directory
inside the AdaCL?  Merging the two projects would simplify packaging
for distributions like Debian.

I have looked at the AdaCL project on SourceForge, and I have to say I
am quite impressed with the activity that takes place there.  However,
the pace of new releases is also a concern to me; I already maintain
17 source packages and it would be difficult for me to package a new
version of AdaCL every couple of weeks.  I just won't have enough
time.  In addition, if the binary interface to the library changes too
often, people will be reluctant to use the library in their programs.
The binary interface normally changes at every major release; this is
reflected in the soname of the shared library.  For the current
version of AdaCL, we would have in /usr/lib:

libadacl.so.4.0.2                    the actual shared library
libadacl.so.4 -> libadacl.so.4.0.2   the link used at runtime: soname
libadacl.so -> libadacl.so.4.0.2     the link used at link time

I can see two solutions to this problem of too frequent releases:

- let somebody else maintain AdaCL in Debian; this includes not only
  packaging each new version, but also tracking bugs, being the
  front-line of support for Debian users, and interacting with the
  upstream authors on SourceForge.  I could provide assistance to get
  going.  In fact, generally speaking, I would really like it if
  someone else stepped up to package more Ada software in Debian.

- slow down the pace of new releases to, say, one every 2 months.
  This implies no bug-fix releases (unless absolutely required, of
  course).  This implies an automated test suite to ensure top quality
  of each release (hint: AUnit is already in Debian).

I don't think it is desirable to skip some versions of AdaCL in
Debian, because newer versions seem to provide valuable bug fixes.

What do you think?

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
: Martin Krischik <krischik@users.sourceforge.net> wrote in message news:<1728933.b9v3IVdcgI@linux1.krischik.com>...
: 
: I regularily build gcc from the 3.3 and 3.4 branches on Linux and
: Windows (about once per week) and don't have major problems.  (see my
: other posting about ACATS results).

Interesting. Do you have any experience of whether or not --enable-checking
turns out to be relevant?



-- Georg

Re: [Announcement] New Project: ASIS for GNAT

[Jeff C,]

"Martin Krischik" <krischik@users.sourceforge.net> wrote in message
news:1728933.b9v3IVdcgI@linux1.krischik.com...
> Ludovic Brenta wrote:
> Somebody suggested the GNAT is going to skip one gcc release. And I frear
> 3.4 is just that version. While the 3.5 brach version at least compiles ok
> the 3.4 bracne has not produced gnat executables for the last 2 weeks.
>


The someone was me and I did not say it would skip a release. I just said it
seemed
likely....But the release that is likely to skip is 3.5 not 3.4.....and
hopefully this wont happen

Re: [Announcement] New Project: ASIS for GNAT

[Rolf Ebert]

Georg Bauhaus <sb463ba@l1-hrz.uni-duisburg.de> wrote in message news:<c3bvd3$kcj$2@a1-hrz.uni-duisburg.de>...
> Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
> : I regularily build gcc from the 3.3 and 3.4 branches on Linux and
> : Windows (about once per week) and don't have major problems.  (see my
> : other posting about ACATS results).
> 
> Interesting. Do you have any experience of whether or not --enable-checking
> turns out to be relevant?
> 

No, I have no experience with --enable-checking. I had to look up the
doc:

<< 
--enable-checking 
--enable-checking=list 
When you specify this option, the compiler is built to perform
checking of tree node types when referencing fields of that node, and
some other internal consistency checks. This does not change the
generated code, but adds error checking within the compiler. This will
slow down the compiler and may only work properly if you are building
the compiler with GCC. This is on by default when building from CVS or
snapshots, but off for releases. [...]
>>

I build gcc from CVS sources, therefor the tested gcc always had
checking enabled.  I don't see how that has any influence on the ACATS
results, though, as it does not change the generated code.

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
: 
: I build gcc from CVS sources, therefor the tested gcc always had
: checking enabled.  I don't see how that has any influence on the ACATS
: results, though, as it does not change the generated code.

I haven't got this far :-) No successful build since last year
on my machine in question (Duron, GNU/Linux). (It seems that
there is a loop in a stage1/gnat1 analysing sem_prag.adb)

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Georg Bauhaus <sb463ba@l1-hrz.uni-duisburg.de> writes:

> Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
> : 
> : I build gcc from CVS sources, therefor the tested gcc always had
> : checking enabled.  I don't see how that has any influence on the ACATS
> : results, though, as it does not change the generated code.
> 
> I haven't got this far :-) No successful build since last year
> on my machine in question (Duron, GNU/Linux). (It seems that
> there is a loop in a stage1/gnat1 analysing sem_prag.adb)

I can build 3.3.3 without a problem; if you try again, you need to
delete your entire source tree before doing CVS update; see GCC PR
10996 for details.

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
: 
: In short: the 3.4 prereleases are considerably better than the 3.3.3
: release.  The number of failures dropped from 24 to 7 (Linux)

news:
GNAT 3.15p -> GCC 3.3.3 -> GCC 3.4 (20040303):
.../configure --enable-checking --enable-languages=c,ada[,f77]

                === acats Summary ===
# of expected passes            2322
# of unexpected failures        0


The fortran part doesn't report any FAIL.
The C part has 17 unexpected failures.
(one in va-arg-25.c, others in visibility-[1-8].c (scan-assembler...))

~/build-3.4> uname -a
Linux strudel 2.4.18 #1 Fri May 2 17:22:29 CEST 2003 i686 unknown
~/build-3.4> cat /etc/debian_version
3.0
CPU: AMD Athlon(tm) Processor stepping 02

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Ludovic Brenta <ludovic.brenta@insalien.org> wrote:
 
: that -gnatVa tends to trigger ICE's quite often.  As I find the
: additional checks valuable, I would suggest that running the ACATS
: tests with -gnatVa (and possibly -gnatwa, too) might be worthwhile.

GCC 3.4 as mentioned in my previous post, ACATS with -gnatVa

# of expected passes            2312
# of unexpected failures        10
*** FAILURES: c32111a c380004 c460010 c46014a cdd2a02 cxa5a03 cxa5a04 \
	cxaca02 cxacb02 cxh1001 

Re: [Announcement] New Project: ASIS for GNAT

[Florian Weimer]

Georg Bauhaus <sb463ba@l1-hrz.uni-duisburg.de> writes:

> Rolf Ebert <rolf.ebert_nospam_@gmx.net> wrote:
> : 
> : In short: the 3.4 prereleases are considerably better than the 3.3.3
> : release.  The number of failures dropped from 24 to 7 (Linux)
> 
> news:
> GNAT 3.15p -> GCC 3.3.3 -> GCC 3.4 (20040303):
> .../configure --enable-checking --enable-languages=c,ada[,f77]
> 
>                 === acats Summary ===
> # of expected passes            2322
> # of unexpected failures        0

With or without NPTL/futexes?

(Over here, GNAT 3.15p only runs with LD_ASSUME_KERNEL=2.4.1 or a
similar environment variable setting.)

-- 
Current mail filters: many dial-up/DSL/cable modem hosts, and the
following domains: postino.it, tiscali.co.uk, tiscali.cz, tiscali.it,
voila.fr.

Re: [Announcement] New Project: ASIS for GNAT

[Georg Bauhaus]

Florian Weimer <fw@deneb.enyo.de> wrote:
:> 
:>                 === acats Summary ===
:> # of expected passes            2322
:> # of unexpected failures        0
: 
: With or without NPTL/futexes?

Uhm, I hadn't know futexes until reading your letter :-).
(And there is NGPT too...)

: (Over here, GNAT 3.15p only runs with LD_ASSUME_KERNEL=2.4.1 or a
: similar environment variable setting.)

The compiler is as unpacked from NYU. The kernel is a 2.4.18.
Nothing special (except SGI XFS support).
The sources have been updated by the Debian team but I haven't
found the time yet to reconfigure/recompile, so it's a bit old.

If it helps I can do that (time permitting as I have to find
a new job.)

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Georg Bauhaus <sb463ba@l1-hrz.uni-duisburg.de> writes:

> Florian Weimer <fw@deneb.enyo.de> wrote:
> :> 
> :>                 === acats Summary ===
> :> # of expected passes            2322
> :> # of unexpected failures        0
> : 
> : With or without NPTL/futexes?
> 
> Uhm, I hadn't know futexes until reading your letter :-).
> (And there is NGPT too...)
> 
> : (Over here, GNAT 3.15p only runs with LD_ASSUME_KERNEL=2.4.1 or a
> : similar environment variable setting.)

NPTL has been integrated into Linux 2.6 and glibc 2.3.  Apparently,
there are issues with GNAT 3.15p tasking when using these two.

I have not personally tried Linux 2.6 yet.  I would appreciate it if
someone running Debian could test tasking with Linux 2.6 and GNAT
3.15p.  If there are problems, I'll add release notes in
README.Debian.

> The compiler is as unpacked from NYU. The kernel is a 2.4.18.
> Nothing special (except SGI XFS support).

In this case, you are not using NPTL.

> The sources have been updated by the Debian team but I haven't
> found the time yet to reconfigure/recompile, so it's a bit old.
> 
> If it helps I can do that (time permitting as I have to find
> a new job.)

Would you consider making additional Ada packages for Debian?

-- 
Ludovic Brenta.

Re: [Announcement] New Project: ASIS for GNAT

[Adrian Knoth]

Ludovic Brenta <ludovic.brenta@insalien.org> wrote:

>> : With or without NPTL/futexes?
>> : (Over here, GNAT 3.15p only runs with LD_ASSUME_KERNEL=2.4.1 or a

> I have not personally tried Linux 2.6 yet.  I would appreciate it if
> someone running Debian could test tasking with Linux 2.6 and GNAT
> 3.15p.  If there are problems, I'll add release notes in
> README.Debian.

I'm running 2.6.4 and Debian unstable. Tasking fails without the
LD_ASSUME_KERNEL-statement above.

adi@drcomp:~/ada/work/test/neu/task/ab$ ./demo
ababababa
Counting:  4
Usermods:  0
adi@drcomp:~/ada/work/test/neu/task/ab$ unset LD_ASSUME_KERNEL
adi@drcomp:~/ada/work/test/neu/task/ab$ ./demo
abaaaaaaa



-- 
mail: adi@thur.de  	http://adi.thur.de	PGP: v2-key via keyserver

Wer die H�nde in den Scho� legt, mu� noch lange nicht unt�tig sein

Re: [Announcement] New Project: ASIS for GNAT

[Ludovic Brenta]

Adrian Knoth <adi@thur.de> writes:

> Ludovic Brenta <ludovic.brenta@insalien.org> wrote:
> 
> >> : With or without NPTL/futexes?
> >> : (Over here, GNAT 3.15p only runs with LD_ASSUME_KERNEL=2.4.1 or a
> 
> > I have not personally tried Linux 2.6 yet.  I would appreciate it if
> > someone running Debian could test tasking with Linux 2.6 and GNAT
> > 3.15p.  If there are problems, I'll add release notes in
> > README.Debian.
> 
> I'm running 2.6.4 and Debian unstable. Tasking fails without the
> LD_ASSUME_KERNEL-statement above.

Thanks.  I'll make gnat 3.15p-8 when I have time, just to update the
release notes.  I'm a perfectionist :)

-- 
Ludovic Brenta.