Ada Mode and GNAT Project files

Ada Mode and GNAT Project files

[deadlyhead]

This has been discussed before, specifically by Ludovic Brenta in
2005. <http://groups.google.com/group/comp.lang.ada/browse_thread/
thread/8ca2d617a0d92ff/2dad47ea5cfa17ed?lnk=gst&q=ada+mode+gnat+project
+files#2dad47ea5cfa17ed> I'm wondering if anything ever came of plans
to replace Ada Mode Projects with GNAT Projects.  There are many
benefits to making this change, so I thought I'd ask if there was
anyone in the know.

I've sent a message to the emacs-ada-mode mailing list inquiring about
this and offering my help if Stephe is interested in incorporating
it.  Maybe Ludovic could supply the work that he's already done to get
things started?

I appreciate all replies.  I'll be starting my first large Ada project
soon, and this feature would be a huge help.  Ada + Emacs is already
one of the best development environments out there as I'm sure you all
know.  Tighter integration would make it even better.

--deadlyhead

Re: Ada Mode and GNAT Project files

[Georg Bauhaus]

deadlyhead schrieb:
> This has been discussed before, specifically by Ludovic Brenta in
> 2005. <http://groups.google.com/group/comp.lang.ada/browse_thread/
> thread/8ca2d617a0d92ff/2dad47ea5cfa17ed?lnk=gst&q=ada+mode+gnat+project
> +files#2dad47ea5cfa17ed> I'm wondering if anything ever came of plans
> to replace Ada Mode Projects with GNAT Projects.  There are many
> benefits to making this change, so I thought I'd ask if there was
> anyone in the know.

I'd vote against a complete replacement and in favor
of a simple abstraction:

 - where to find files and

 - where to find navigation information

(such as can be extracted from .ali files).  Then it is still
possible to steer the compiler with a -P switch and maybe
a Lisp variable for the "external" project settings.

One reason is that GNAT project files are, well, GNAT specific.
They are not standard, though open, but other Ada compilers
use different project descriptions, use different options,
and so on.
  However, all compilers need to know where to find sources,
and many if not all produce information that can be used
for navigation.

Re: Ada Mode and GNAT Project files

[deadlyhead]

On Aug 15, 1:56 am, Georg Bauhaus <rm.dash-bauh...@futureapps.de>
wrote:
> I'd vote against a complete replacement and in favor
> of a simple abstraction:
>
>  - where to find files and
>
>  - where to find navigation information
>
> (such as can be extracted from .ali files).  Then it is still
> possible to steer the compiler with a -P switch and maybe
> a Lisp variable for the "external" project settings.
>
> One reason is that GNAT project files are, well, GNAT specific.
> They are not standard, though open, but other Ada compilers
> use different project descriptions, use different options,
> and so on.
>   However, all compilers need to know where to find sources,
> and many if not all produce information that can be used
> for navigation.

Well, at the moment, Ada Mode for Emacs is GNAT specific.  This makes
perfect sense, though, as both GNAT and Emacs are part of the GNU
project.  There's nothing stopping proprietary compiler developers
from contributing their own generalizing patches to Ada Mode if they
like, but it's unreasonable, I think, to have a free software project
try to accommodate proprietary vendors' products without access to
their documentation in a free and open manner.

With that out of the way, a different approach would be to add a Lisp
variable to Ada Mode which indicates whether a GNAT *.gpr (or some
other vendors project file) is to be used, and having Ada Mode simply
use that rather than the current (or recently current) .adp files.  It
would take care of the general navigation information you're looking
for, but still provide the added power of whatever project facilities
are provided by the compilation system.

Re: Ada Mode and GNAT Project files

[Ludovic Brenta]

On Aug 15, 12:19 pm, deadlyhead <deadlyh...@gmail.com> wrote:
> On Aug 15, 1:56 am, Georg Bauhaus <rm.dash-bauh...@futureapps.de>
> wrote:
>
>
>
> > I'd vote against a complete replacement and in favor
> > of a simple abstraction:
>
> >  - where to find files and
>
> >  - where to find navigation information
>
> > (such as can be extracted from .ali files).  Then it is still
> > possible to steer the compiler with a -P switch and maybe
> > a Lisp variable for the "external" project settings.
>
> > One reason is that GNAT project files are, well, GNAT specific.
> > They are not standard, though open, but other Ada compilers
> > use different project descriptions, use different options,
> > and so on.
> >   However, all compilers need to know where to find sources,
> > and many if not all produce information that can be used
> > for navigation.
>
> Well, at the moment, Ada Mode for Emacs is GNAT specific.  This makes
> perfect sense, though, as both GNAT and Emacs are part of the GNU
> project.  There's nothing stopping proprietary compiler developers
> from contributing their own generalizing patches to Ada Mode if they
> like, but it's unreasonable, I think, to have a free software project
> try to accommodate proprietary vendors' products without access to
> their documentation in a free and open manner.
>
> With that out of the way, a different approach would be to add a Lisp
> variable to Ada Mode which indicates whether a GNAT *.gpr (or some
> other vendors project file) is to be used, and having Ada Mode simply
> use that rather than the current (or recently current) .adp files.  It
> would take care of the general navigation information you're looking
> for, but still provide the added power of whatever project facilities
> are provided by the compilation system.

There are two ways to provide support for GNAT project files in ada-
mode: my first approach was to completely replace the .adp project
file with the .gpr file, parse it and extract the source path. I never
went so far as to also find and parse "with"ed project files, though.
The more complete and portable approach, implemented by Stephe Leake,
was to call "gnat ls -P", pass it the project file, and parse the
result. The advantage is that it simplifies parsing and also supports
"with"ed project files.

Stephe's approach is better, of course. That's why I never merged my
fork into Stephe's ada-mode. But if you are still interested in my
version, I put it on Ada-France's monotone server [1,2,3]. [2] is my
fork and [3] is Stephe's upstream version, probably outdated by now.

[1] http://www.ada-france.org/article131.html
[2] http://www.ada-france.org:8081/branch/changes/org.ludovic-brenta.ada-mode
[3] http://www.ada-france.org:8081/branch/changes/org.gnu.emacs.ada-mode

--
Ludovic Brenta.

Re: Ada Mode and GNAT Project files

[Georg Bauhaus]

deadlyhead schrieb:

> Well, at the moment, Ada Mode for Emacs is GNAT specific.

A major part of Ada mode is consistent with Emacs in that
it is not at all GNAT specific but addresses just Ada.
The 5_000 lines of Lisp code in ada-mode.el are Ada specific,
not GNAT specific.  So the statement that Ada Mode for Emacs
is GNAT specific is overly general.

OTOH, replacing GNAT specific parts with other GNAT specific
parts does not do any harm---except if this kind of approach
affects the entire structure and approach of Ada Mode for Emacs.

A form of abstraction inversion.


>  This makes
> perfect sense, though, as both GNAT and Emacs are part of the GNU
> project.

I'd contend that Free Software may be, and is written to
support systems that (unfortunately) are not currently
targetted by the GNU system.

Also, and in particular, Emacs has tended to provide for
"the system compiler", which may or may not be GNAT.
A related example, it has been fairly easy to hook a scanner for
error messages printed by the free Aonix compiler and relatives.

If you have FSF GNAT and the compiler does not currently
support Ada feature X, you can try a different compiler.
If you are stuck with GNAT (or maybe just
the other way around) *because*  Ada Mode for Emacs has
made supporting other compilers difficult, well...

What if you want to work productively using beloved Emacs
on a platform that FSF GNAT does not support?


>  There's nothing stopping proprietary compiler developers
> from contributing their own generalizing patches to Ada Mode if they
> like, but it's unreasonable, I think, to have a free software project
> try to accommodate proprietary vendors' products without access to
> their documentation in a free and open manner.

On the same narrow basis, one might argue that we should not
develop for most of PC computer hardware.  Or write control
software for flying machine parts, using portable Ada yet a
compiler that happens to produce code for a non-GNU target,
still using Emacs.
Because Emacs is a free software project and jet engines are
not free, we should not be using Emacs when writing such
Ada software?

A ban on everything else also narrows sources of inspiration,
and comparison.


> With that out of the way, a different approach would be to add a Lisp
> variable to Ada Mode which indicates whether a GNAT *.gpr (or some
> other vendors project file) is to be used, and having Ada Mode simply
> use that rather than the current (or recently current) .adp files.  It
> would take care of the general navigation information you're looking
> for, but still provide the added power of whatever project facilities
> are provided by the compilation system.


IMHO this view reflects a Microsoft style approach to software:
We do our own non-standard thing; we are guided by this
non-standard thing.  If others write free contributions
matching our special APIs, fine.  If not, that's the problem of
the Emacs user who needs to use Emacs with a non-GNAT Ada
compiler.

Does this put the long time Emacs user, and maybe contributor,
in the center?

Re: Ada Mode and GNAT Project files

[Stephen Leake]

Ludovic Brenta <ludovic@ludovic-brenta.org> writes:

> There are two ways to provide support for GNAT project files in ada-
> mode: my first approach was to completely replace the .adp project
> file with the .gpr file, parse it and extract the source path. I never
> went so far as to also find and parse "with"ed project files, though.
> The more complete and portable approach, implemented by Stephe Leake,
> was to call "gnat ls -P", pass it the project file, and parse the
> result. The advantage is that it simplifies parsing and also supports
> "with"ed project files.

For the record, the current Emacs Ada mode is at:

http://stephe-leake.org/emacs/ada-mode/emacs-ada-mode.html

It supports using a GNAT project file directly as an Emacs Ada mode
project file, or including a GNAT project file as part of a larger
Emacs Ada mode project file.

The version in Emacs 22.2 does not support GNAT project files; future
Emacs releases will include this support.

-- 
-- Stephe

Re: Ada Mode and GNAT Project files

[Stephen Leake]

Georg Bauhaus <rm.dash-bauhaus@futureapps.de> writes:

> What if you want to work productively using beloved Emacs
> on a platform that FSF GNAT does not support?

Patches are always welcome.

-- 
-- Stephe