Re: Why not gnat Ada in gcc?

Re: Why not gnat Ada in gcc?

[Robert Dewar]

Laurent Guerby asks

<<I don't know if ACT has debated the issue, but assuming the GNAT
sources end up in the GCC CVS repository what will be the updating
policy between the ACT tree and the public GCC one? One massive update
from ACT per public GNAT version? Minor updates when GCC backend
interface changes are made that break the compilation of the Ada
frontend? Or more frequent updates?
>>

Certainly new releases of GNAT always have major additions, which will
result in a fairly massive update to the tree (as has happened with
other GNU tools when major updates are made). We will also provide minor
updates and patches as we go along to correct significant problems or
deal with interface changes.

<<If some people want to develop non trivial patches to GNU Ada
(affecting multiple files), how should they proceed (I assume by first
contacting ACT, but then)?
>>

These patches should be submitted to someone with write access to the
tree in the usual manner. Ada Core Technologies will of course follow
any such patches, and incorporate them where appropriate (or correct
them where this is needed).

<<PS: I promised on this list I would package a test suite and a
performance test suite if the GNAT sources go in the GCC CVS tree.
>>

That would certainly be helpful, in particular, packaging the latest
ACVC test suite will be very useful.

Re: Why not gnat Ada in gcc?

[Jeffrey A Law]

  In message < 20001001153500.5FBCB34D84@nile.gnat.com >you write:
  > Laurent Guerby asks
  > 
  > <<I don't know if ACT has debated the issue, but assuming the GNAT
  > sources end up in the GCC CVS repository what will be the updating
  > policy between the ACT tree and the public GCC one? One massive update
  > from ACT per public GNAT version? Minor updates when GCC backend
  > interface changes are made that break the compilation of the Ada
  > frontend? Or more frequent updates?
  > >>
  > 
  > Certainly new releases of GNAT always have major additions, which will
  > result in a fairly massive update to the tree (as has happened with
  > other GNU tools when major updates are made). We will also provide minor
  > updates and patches as we go along to correct significant problems or
  > deal with interface changes.
I think that's the wrong model.  Dumping in massive changes like that makes
it nearly impossible for others outside ACT to be involved with development.  

I would much rather see the GCC CVS sources become the master sources and
GNU Ada work happen in that source tree on an incremental basis just like
the other parts of the GNU compiler suite.

Basically the whole GNAT development process is closed to developers outside
ACT.  That is terribly unfortunate.

jeff

Re: Why not gnat Ada in gcc?

[Laurent Guerby]

Disclaimer: I worked for ACT but I no longer do, I'm on the support client side now.

jeff wrote:
> I think that's the wrong model.  Dumping in massive changes like that makes
> it nearly impossible for others outside ACT to be involved with development.  
>
> I would much rather see the GCC CVS sources become the master sources and
> GNU Ada work happen in that source tree on an incremental basis just like
> the other parts of the GNU compiler suite.
>
> Basically the whole GNAT development process is closed to developers outside
> ACT.  That is terribly unfortunate.

IMHO, the problem we're trying to correct now is to get past the chicken-and-egg
bootstrap problem, no available GCC compatible source means no contribution
which means no incentive to provide public up to date sources...

If ACT provides sources compatible with the current GCC in CVS and the
GCC steering committee accepts it, that's a HUGE progress in the right
direction. So far, nothing has been decided (at least not publically
on this list) on the topic of the inclusion of the Ada sources in the
FSF GCC CVS repository.

The model proposed by Robert Dewar is the right one to start with, and
it will put the merging burden on ACT if some contributions are indeed
made by people having write access to the GNAT sources. If this
becomes significant, I assume the natural thing will happen (move of
the master source to the FSF CVS repository).

But I don't think we're here yet or anytime in less than a year, so I
don't think it is useful to argue on the source merging process right
now.

Up to now, GNAT source releases have been made something like every
year, so we have plenty of time ;-).

-- 
Laurent Guerby <guerby@acm.org>

Re: Why not gnat Ada in gcc?

[Jeffrey A Law]

  In message < 200010111937.VAA01637@ulmo >you write:
  > Disclaimer: I worked for ACT but I no longer do, I'm on the support client 
  > side now.
Understood.

  > IMHO, the problem we're trying to correct now is to get past the chicken-an
  > d-egg
  > bootstrap problem, no available GCC compatible source means no contribution
  > which means no incentive to provide public up to date sources...
We can get the compiler bootstrapped.  It's not the most pleasant thing
to do, but nor is it a terrible thing to do.  And it only has to be done
once and you can then use that compiler to bootstrap later versions.

  > If ACT provides sources compatible with the current GCC in CVS and the
  > GCC steering committee accepts it, that's a HUGE progress in the right
  > direction. So far, nothing has been decided (at least not publically
  > on this list) on the topic of the inclusion of the Ada sources in the
  > FSF GCC CVS repository.
I'd very much like to see it in the repo.  That's the first step, move
development into the open where others can contribute in whatever way
is most appropriate for them.

  > 
  > The model proposed by Robert Dewar is the right one to start with, and
  > it will put the merging burden on ACT if some contributions are indeed
  > made by people having write access to the GNAT sources. If this
  > becomes significant, I assume the natural thing will happen (move of
  > the master source to the FSF CVS repository).
But that's there the burden belongs -- on ACT.  Cygnus has dealt with
this burden for 10 years -- it's the price (IMHO) for wanting to be in
the business of supporting and custom development for tools you do not
own/control.  I'm more than happy to give ACT some pointers on how to
manage the process.


  > But I don't think we're here yet or anytime in less than a year, so I
  > don't think it is useful to argue on the source merging process right
  > now.
I would claim the opposite.  We need this to happen sooner, not later.
The longer we wait, the more difficult our work becomes and the longer
we have to deal with the problem of developers making changes, but 
being unable to test that they haven't broken Ada.

Jeff

Re: Why not gnat Ada in gcc?

[Richard Stallman]

I agree with your arguments that the Ada sources should be included in
the main GCC repository, and I would like to see this done soon.

But it is unreasonable to ask ACT to put its repository onto a machine
run by Cygnus, and associated by the public with Cygnus.  If we want
to say to ACT that using the GCC repository is the only right thing to
do, we need to move the GCC repository to a neutral GNU site first.

    Also note that I've got the financial committement from Red Hat I need
    to co-locate the existing box at an AT&T site which should provide a huge
    increase in network bandwidth.

That is nice, but the problems we should solve by moving the
repository are not a matter of to network bandwidth.  Increasing
bandwidth to the repository, while not a bad thing, won't help the GNU
Project in any of the larger ways.

repository [was: Re: Why not gnat Ada in gcc?]

[Geoff Keating]

Hi Richard,

Richard Stallman <rms@gnu.org> writes:

> But it is unreasonable to ask ACT to put its repository onto a machine
> run by Cygnus, and associated by the public with Cygnus.  If we want
> to say to ACT that using the GCC repository is the only right thing to
> do, we need to move the GCC repository to a neutral GNU site first.

* It wasn't well publicised, but Cygnus Solutions merged with Red Hat,
Inc., on July 1 of 2000.  The name of the merged company is Red Hat,
Inc., and so it is no longer appropriate to speak of 'Cygnus', as
it no longer exists as a separate entity.

* It is also not correct to say that the machine `sourceware.cygnus.com'
(aka sources.redhat.com and gcc.gnu.org) is `run by Red Hat' (or
Cygnus).  It's currently administered by a group of people, who are
all Red Hat employees, but do not generally administer it in their
capacity as a Red Hat employee; as an example, the on-site
administrator is neither a member of the main Red Hat I/S organization
nor the Red Hat Engineering Services I/S organization, but is
currently a staff engineer.  Many of the other admins are doing it in
their capacity as FSF project maintainers or steering committee
members, which is a personal capacity not part of their employment.

It (the machine and the network connection) is _funded_ by Red Hat,
but this is of course not the same thing.

* I question as to whether any "neutral GNU site"s exist with
the capacity to provide such a service.  I don't believe the FSF has
the resources or competence by itself; the other obvious candidate
would be sourceforge, but they are "associated by the public" with VA
Linux, and according to their web site the admins' salaries are paid
by VA Linux: "VA Linux Systems now pays our salary to offer you these
services."

* Finally, I'm not sure you're asking the right question.  You said
"put its repository...".  I would never ask this of anyone, as I would
expect that ACT has information in its repository, for instance in
checkin messages, which is confidential, or private, or was not
intended to be released.  Instead, the request is that ACT take their
current tree, as they would ship it to a nonspecific customer, and
check it in, so that the only information released is that encoded in
the GPLed sources.

-- 
- Geoffrey Keating <geoffk@cygnus.com>

Re: repository [was: Re: Why not gnat Ada in gcc?]

[Chris Faylor]

On Thu, Oct 12, 2000 at 06:44:45PM -0700, Geoff Keating wrote:
>* It is also not correct to say that the machine `sourceware.cygnus.com'
>(aka sources.redhat.com and gcc.gnu.org) is `run by Red Hat' (or
>Cygnus).  It's currently administered by a group of people, who are
>all Red Hat employees,

This is not strictly true.  The primary administrators may be from Red
Hat, but there are a couple of non Red Hat employees who also assist in
administration duties.

cgf

Re: repository [was: Re: Why not gnat Ada in gcc?]

[Richard Stallman]

    * I question as to whether any "neutral GNU site"s exist with
    the capacity to provide such a service.  I don't believe the FSF has
    the resources or competence by itself;

I could try to defend the people who run our CVS server, but there is
a more important thing to be said here: you're making a hostile
supposition about another part of the GNU Project, and that is not
the right way to treat other GNU contributors.

GCC development is part of the GNU Project, and all parts of the GNU
Project are supposed to cooperate with each other.  So if you want to
participate, please treat other parts of the GNU Project decently, and
give them the benefit of the doubt.  If you observe a problem that
calls for criticism, offer constructive criticism, not an attack.

I would ask people to give you the same consideration as a GNU
contributor.

Re: repository [was: Re: Why not gnat Ada in gcc?]

[Geoff Keating]

> Date: Fri, 13 Oct 2000 16:02:39 -0600 (MDT)
> From: Richard Stallman <rms@gnu.org>
> CC: law@cygnus.com, guerby@acm.org, dewar@gnat.com, gcc@gcc.gnu.org
> cc: rms@gnu.org
> Reply-to: rms@gnu.org
> 
>     * I question as to whether any "neutral GNU site"s exist with
>     the capacity to provide such a service.  I don't believe the FSF has
>     the resources or competence by itself;
> 
> I could try to defend the people who run our CVS server, but there is
> a more important thing to be said here: you're making a hostile
> supposition about another part of the GNU Project, and that is not
> the right way to treat other GNU contributors.

Oops!

I'm sorry.  I had completely forgotten about subversions.  I certainly
didn't mean to impute anything about its admins, and apologize to
anyone I might have offended.

-- 
- Geoffrey Keating <geoffk@cygnus.com>

Re: repository [was: Re: Why not gnat Ada in gcc?]

[Richard Stallman]

    I'm sorry.  I had completely forgotten about subversions.  I certainly
    didn't mean to impute anything about its admins, and apologize to
    anyone I might have offended.

Thank you.  Everything is cool now.

Re: Why not gnat Ada in gcc?

[Jeffrey A Law]

  In message < 200010122240.QAA09837@wijiji.santafe.edu >you write:
  > I agree with your arguments that the Ada sources should be included in
  > the main GCC repository, and I would like to see this done soon.
Good.


  > But it is unreasonable to ask ACT to put its repository onto a machine
  > run by Cygnus, and associated by the public with Cygnus.  If we want
  > to say to ACT that using the GCC repository is the only right thing to
  > do, we need to move the GCC repository to a neutral GNU site first.
This is awful funny in a sick sort of way.

Red Hat (and formerly Cygnus) has bent over backwards to open up the GCC
development environment and try to put all developers on a level playing
field.    Yet, we have to do even more before you'll encourage a company
(ACT) that has de-facto stranglehold on GNU Ada development to open up
development of GNU Ada and put all developers on a level playing field.

Isn't that somewhat of a double standard?

Yes, the machine is physically at Red Hat's Sunnyvale site (Cygnus as a
corporate entity no longer exists), but everyone is encouraged to use
gcc.gnu.org and with a few exceptions gcc.gnu.org is logically out of
the redhat.com/cygnus.com domains.

Yes, there has been some content on sources.redhat.com which could be
mis-leading in terms of implying control over GCC, GDB or binutils.  But
as I've always maintained if there is something on that site that is 
objectionable and gives the wrong impression that we'll fix it.

jeff

Re: Why not gnat Ada in gcc?

[Hartmut Schirmer]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 698 bytes --]

On Wed, 11 Oct 2000, Laurent Guerby wrote:
>IMHO, the problem we're trying to correct now is to get past the chicken-and-egg
>bootstrap problem, no available GCC compatible source means no contribution
>which means no incentive to provide public up to date sources...

Configure GNAT to produce java byte code and let it compile itself:
JGNAT.

This (and any binary produced by JGNAT) can be executed in a JVM or
directly on the host using GCC´s java front end:

         JGNAT               gcj
Sources -------> byte code -------> hosted GNAT (stage0)


Creating JGNAT once (or every few years) should be sufficient.

This scheme may work for other self implemented languages as well.

Hartmut

Re: Why not gnat Ada in gcc?

[Laurent Guerby]

I wrote:
> IMHO, the problem we're trying to correct now is to get past the chicken-and-egg
> bootstrap problem, no available GCC compatible source means no contribution
> which means no incentive to provide public up to date sources...

The use of "bootstrap" here was unfortunate, but well...

Harmut wrote:
> Configure GNAT to produce java byte code and let it compile itself:

I'm not sure it would work since JGNAT might not support at this point
some Ada features used by the GNAT code itself (parts of it are fairly
low level and might not work on JVM IIRC). But definitely an
interesting thing to look at ;-).

-- 
Laurent Guerby <guerby@acm.org>