autotools and ada

autotools and ada

[Ottavio Campana]

Hi,

I'm trying to develop a program using autotools to manage the makefiles.
I haven't found anything on the web, except for this

http://oopweb.com/Ada/Documents/AdaLinux/VolumeFrames.html?/Ada/Documents/AdaLinux/Volume/5.html

which says "It is possible to use autoconf and automake on Ada
Makefiles, but this topic is beyond the scope of this book" so I don't
know what to do.

I'm familiar with autotools and c/c++, but not with ada. If I try to set
up a project and to compile it, I obtain:

bott@pcimg12:~/Progetti/turbo$ make
Making all in src
make[1]: Entering directory `/home/bott/Progetti/turbo/src'
cc     -o subtypes
cc: no input files
make[1]: *** [subtypes] Error 1
make[1]: Leaving directory `/home/bott/Progetti/turbo/src'
make: *** [all-recursive] Error 1
bott@pcimg12:~/Progetti/turbo$

so I see that it calls the wrong compiler. do you have any hint?

Thank you for your help.

Re: autotools and ada

[Matthieu Moy]

Ottavio Campana <spam@campana.vi.it> writes:

> which says "It is possible to use autoconf and automake on Ada
> Makefiles, but this topic is beyond the scope of this book" so I don't
> know what to do.

I have no experience with the autotools with Ada, but I believe
automake is of no use if you use gnatmake or equivalent. autoconf
itself is still usefull, but it is anyway mostly language independant.

-- 
Matthieu

Re: autotools and ada

[Pascal Obry]

Matthieu Moy a �crit :
> I have no experience with the autotools with Ada, but I believe
> automake is of no use if you use gnatmake or equivalent. autoconf
> itself is still usefull, but it is anyway mostly language independant.

No automake is still useful for Ada. gnatmake only build the tool. But
you probably also want to install, setup the build, create tarball for
distribution... all this can be done in makefiles.

XML/Ada uses the autotools, maybe you should have a look. I have myself
avoided the autotools because this is a pain to use on Windows.

Pascal.

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|              http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595

Re: autotools and ada

[Georg Bauhaus]

Pascal Obry wrote:
> Matthieu Moy a écrit :
> 
>>I have no experience with the autotools with Ada, but I believe
>>automake is of no use if you use gnatmake or equivalent. autoconf
>>itself is still usefull, but it is anyway mostly language independant.
> 
> 
> No automake is still useful for Ada. gnatmake only build the tool. But
> you probably also want to install, setup the build, create tarball for
> distribution... all this can be done in makefiles.

If your programs are just Ada, chances are that you can do
software configuration more or less from within the language
using the Ada make tools provided with your toolset.

For example, use suitable package renamings.
Replace #ifdef hacks with one-spec many-bodies.
Place the different bodies in different directories,
etc etc etc.

autotools running on anything but well-autotested Unix platforms
tends to grow into a maintenance nightmare AFAICS.
The more I have to cope with auto*ed software on non-Unix
platforms, the more I think autotools is an oxymoron.

Re: autotools and ada

[Ottavio Campana]

Georg Bauhaus wrote:
> Pascal Obry wrote:
> 
>> Matthieu Moy a écrit :
>>
>>> I have no experience with the autotools with Ada, but I believe
>>> automake is of no use if you use gnatmake or equivalent. autoconf
>>> itself is still usefull, but it is anyway mostly language independant.
>>
>>
>>
>> No automake is still useful for Ada. gnatmake only build the tool. But
>> you probably also want to install, setup the build, create tarball for
>> distribution... all this can be done in makefiles.
> 
> 
> If your programs are just Ada, chances are that you can do
> software configuration more or less from within the language
> using the Ada make tools provided with your toolset.
> 
> For example, use suitable package renamings.
> Replace #ifdef hacks with one-spec many-bodies.
> Place the different bodies in different directories,
> etc etc etc.
> 
> autotools running on anything but well-autotested Unix platforms
> tends to grow into a maintenance nightmare AFAICS.
> The more I have to cope with auto*ed software on non-Unix
> platforms, the more I think autotools is an oxymoron.

yes, ok, but I only work on unix platforms. I don't care about windows.
I usually use autotools because they are useful to create makefiles and
to deal with tarballs.

Re: autotools and ada

[Stephen Leake]

Pascal Obry <pascal@obry.net> writes:

> XML/Ada uses the autotools, maybe you should have a look. I have myself
> avoided the autotools because this is a pain to use on Windows.

With the latest Cygwin distribution, and installing Cygwin at c:/
(that is, / is mounted as c:/, not c:/cygwin), automake works ok on
Windows.

I have an example project using automake and autoconf on Windows for
an Ada project, if anyone is interested. It may be broken with the
current tools; I've stopped using it.

I found the autotools useful but fragile (they are written in Perl and
m4). They are a reasonable solution to the problem of installing
projects on diverse systems.

One thing they do is encourage the Gnu standard directory structure,
which is a good idea.

To be fully useful with Ada, there needs to be a shared set of macros
written (in m4) for finding and executing GNAT tools (and/or other Ada
toolsets).

In addition, pkg-config now does some of what autoconf does, and is a
better solution for that subset of the problem, because pkg-config has
a mechanism for automatically including transitive dependencies.

-- 
-- Stephe

Re: autotools and ada

[Ottavio Campana]

Stephen Leake wrote:
> I have an example project using automake and autoconf on Windows for
> an Ada project, if anyone is interested. It may be broken with the
> current tools; I've stopped using it.

i'm keen on it, please.

Re: autotools and ada

[Ottavio Campana]

Ottavio Campana wrote:
> Stephen Leake wrote:
> 
>>I have an example project using automake and autoconf on Windows for
>>an Ada project, if anyone is interested. It may be broken with the
>>current tools; I've stopped using it.
> 
> 
> i'm keen on it, please.

ehm, sorry, if you want to mail it, just change my address, use ottavio
instead of spam

Re: autotools and ada

[Jerome Hugues]

In article <uwtizy8kf.fsf@acm.org>, Stephen Leake wrote:

> To be fully useful with Ada, there needs to be a shared set of macros
> written (in m4) for finding and executing GNAT tools (and/or other Ada
> toolsets).

You'll find some of them in Adasockets, PolyORB, XML/Ada (and possibly
others) depending on your needs ..

Chose AdaSockets or XML/Ada, their build structure is simpler than
PolyORB's one, so it should be easier to adapt it to your needs.

-- 
Jerome

Re: autotools and ada

[Samuel Tardieu]

>>>>> "Jerome" == Jerome Hugues <hugues@merlin.enst.fr> writes:

Jerome> You'll find some of them in Adasockets, PolyORB, XML/Ada (and
Jerome> possibly others) depending on your needs ..

Note that I never took the time to make the build process of
AdaSockets consistent with cross-builds. However, if the only thing
the original poster wants is native compilation, the autotools process
used in AdaSockets is indeed usable.

  Sam
-- 
Samuel Tardieu -- sam@rfc1149.net -- http://www.rfc1149.net/

Re: autotools and ada

[Jerome Hugues]

In article <87veyjmw2r.fsf@willow.rfc1149.net>, Samuel Tardieu wrote:
>>>>>> "Jerome" == Jerome Hugues <hugues@merlin.enst.fr> writes:
> 
>Jerome> You'll find some of them in Adasockets, PolyORB, XML/Ada (and
>Jerome> possibly others) depending on your needs ..
> 
> Note that I never took the time to make the build process of
> AdaSockets consistent with cross-builds. However, if the only thing
> the original poster wants is native compilation, the autotools process
> used in AdaSockets is indeed usable.

Note there is some support in PolyORB for cross compilation,
succesfully tested with ORK and (partially) MaRTE OS.

-- 
Jerome

Re: autotools and ada

[Stephen Leake]

Jerome Hugues <hugues@merlin.enst.fr> writes:

> In article <uwtizy8kf.fsf@acm.org>, Stephen Leake wrote:
>
>> To be fully useful with Ada, there needs to be a shared set of macros
>> written (in m4) for finding and executing GNAT tools (and/or other Ada
>> toolsets).
>
> You'll find some of them in Adasockets, PolyORB, XML/Ada (and possibly
> others) depending on your needs ..

Well, my point was that they need to be _shared_. By which I meant
"contributed to the autotools project, so _all_ Ada projects have
direct access to them".

I haven't looked, but it sounds like the Ada macros in the above
projects are all different from each other.

So the problem now is to get the various Ada projects that use
autotools to agree on a standard set of macros. Hmm, do you suppose
the ARG would accept that task :)?

-- 
-- Stephe

Re: autotools and ada

[Jerome Hugues]

In article <u3blmnvlg.fsf@acm.org>, Stephen Leake wrote:
> Jerome Hugues <hugues@merlin.enst.fr> writes:
> 
>> In article <uwtizy8kf.fsf@acm.org>, Stephen Leake wrote:
>>
>>> To be fully useful with Ada, there needs to be a shared set of macros
>>> written (in m4) for finding and executing GNAT tools (and/or other Ada
>>> toolsets).
>>
>> You'll find some of them in Adasockets, PolyORB, XML/Ada (and possibly
>> others) depending on your needs ..
> 
> Well, my point was that they need to be _shared_. By which I meant
> "contributed to the autotools project, so _all_ Ada projects have
> direct access to them".

That would be nice.

I could take some time to extract and publish them, as I'm familiar
with autotools with Ada in the context of PolyORB, but cannot provide
any deadline for the next few months :(

> I haven't looked, but it sounds like the Ada macros in the above
> projects are all different from each other.

They all have a common ancestor I suppose, but AFAICT, yes, they are
now different ..

> So the problem now is to get the various Ada projects that use
> autotools to agree on a standard set of macros. Hmm, do you suppose
> the ARG would accept that task :)?

What about a user-driven initiative ? I'm not convinced it is the role
of the ARG, and they have a more important task to finish ;)

-- 
Jerome

Re: autotools and ada

[Samuel Tardieu]

>>>>> "Jerome" == Jerome Hugues <hugues@nephilim.enst.fr> writes:

Jerome> They all have a common ancestor I suppose

Yes, to the best of my knowledge, this ancestor is a set of m4 macros
I wrote for the early versions of GLADE.

  Sam
-- 
Samuel Tardieu -- sam@rfc1149.net -- http://www.rfc1149.net/

Re: autotools and ada

[Martin Dowie]

Stephen Leake wrote:
> So the problem now is to get the various Ada projects that use
> autotools to agree on a standard set of macros. Hmm, do you suppose
> the ARG would accept that task :)?

Er, no...

...but perhaps you could persuade the WG's to collaberate on a common
format?

http://www.acm.org/sigada/wg/wg.html

Re: autotools and ada

[Jeffrey R. Carter]

Ottavio Campana wrote:
> 
> I'm trying to develop a program using autotools to manage the makefiles.

My experience, in over 21 yrs using Ada, is that make files are unnecessary. I 
have never used them.

-- 
Jeff Carter
"Mr. President, we must not allow a mine-shaft gap!"
Dr. Strangelove
33

Re: autotools and ada

[Brian May]

>>>>> "Jeffrey" == Jeffrey R Carter <spam@spam.com> writes:

    Jeffrey> Ottavio Campana wrote:
    >> I'm trying to develop a program using autotools to manage the
    >> makefiles.

    Jeffrey> My experience, in over 21 yrs using Ada, is that make files are
    Jeffrey> unnecessary. I have never used them.

... and complicated.

Most of the time maintaining Heimdal for Debian seems to be spent on
redoing the autotools stuff to do adhere to Debian requirements.  This
would be OK if upstream applied the same (or similar) changes,
frequently this doesn't happen. :-(

Argghhh!

This isn't that I dislike autotools, but I think it is important to
ensure the system remains, flexible, easy to understand, and easy to
debug.

I also have a suspicion if you tried to design something like
autotools from scratch without the requirement it work with old
standards (Makefiles, m4, etc) you would end up with something that
makes meeting these goals a lot simpler.

Might be an interesting project in fact, how about

(autoconf+automake+libtool functionality)+(Ada syntax)+(strict Ada checking)
  --> new make tool for any language and portable to any environment?

Obviously I haven't given this much consideration. Yet.
-- 
Brian May <bam@snoopy.apana.org.au>

Re: autotools and ada

[Samuel Tardieu]

>>>>> "Jeffrey" == Jeffrey R Carter <spam@spam.com> writes:

Jeffrey> My experience, in over 21 yrs using Ada, is that make files
Jeffrey> are unnecessary. I have never used them.

Have your programs be installed on totally different target systems?
At various location? Can they easily be cross-compiled? Can compiler
flags be changed easily? Can you switch compilers easily?

All this requires work and this work can be alleviated using
autotools. Makefiles are not for compilation only.

  Sam
-- 
Samuel Tardieu -- sam@rfc1149.net -- http://www.rfc1149.net/

Re: autotools and ada

[Georg Bauhaus]

On Fri, 2005-11-25 at 15:31 +0100, Samuel Tardieu wrote:
> >>>>> "Jeffrey" == Jeffrey R Carter <spam@spam.com> writes:
> 
> Jeffrey> My experience, in over 21 yrs using Ada, is that make files
> Jeffrey> are unnecessary. I have never used them.
> 
> Have your programs be installed on totally different target systems?
> At various location? Can they easily be cross-compiled? Can compiler
> flags be changed easily? Can you switch compilers easily?
> 
> All this requires work and this work can be alleviated using
> autotools. Makefiles are not for compilation only.

This work can be alleviated in Unix(-like) environments only.

$ variable=`program1 | program2 'parameters'`

requires a specific flavor of process control, piping, availability
of program1 and program2. If the output of such constructs is
just some string identifying the systems, I can only think of this
as a kludge. autotools are that way throughout.

My suspicion is that people will hardly ever have a chance
of learning how to use their language to good effect to reduce
the need for configuration. Reason: autotools is popular and has
the well known seduction trick built in: Users think they don't
need to think. 

-- Georg 

Re: autotools and ada

[Pascal Obry]

Georg,

> This work can be alleviated in Unix(-like) environments only.
> 
> $ variable=`program1 | program2 'parameters'`

No this can't be :)

One interresting feature of Makefile is that it "documents" the build
processes: how to configure the project, how to build the differents
modules (some can be C, Java, Ada...), how to build a distrib tarball,
how to install the project, how to run non regression tests... and lot more.

You'll probably agree that it is far better to do:

	$ make build test

Than to look at the documentation or remmember a set of commands to do
the same... and this also avoid procedure mistakes... you know that you
are running the tests the very same way each time!

Using the makefiles to build automatically projects using cron-script or
running the tests is then quite easy...

Frankly I do not see what is wrong with makefile and Ada. It is quite
easy to use and brings a lot to the project. Lot of mess that C/C++
projects needs for files dependancies are not needed here, neither the
right CC, CFLAGS or whatever. Use the make tools from your Ada vendor
(gnatmake for example).

Now you can use makefile without the autotools which I find quite hard
to grasp ;)

Pascal.

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|              http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595

Re: autotools and ada

[Georg Bauhaus]

On Fri, 2005-11-25 at 19:53 +0100, Pascal Obry wrote:

> You'll probably agree that it is far better to do:
> 
> 	$ make build test

I do agree and this is where GNU make is the most useful tool.

I always have *small* makefiles with the first target being

all: test doc

And they have a dist: target that asks the version control
system to make an archive  e.g.

$ darcs dist
 or
$ prcs --execute | cpio -oc > ../dist-$(date ...).cpio

BUT, all this can easily be done without running autotools, as
you have mentioned at the end of your post.
GNU make is comparatively powerful as a make tool! 
vpath, $(function ...) etc. come to mind. They are underused.

>  Use the make tools from your Ada vendor
> (gnatmake for example).

It's slightly confusing to call gnatmake a make tool when discussing
the traditional make program, I think. :)

There is a nice separation here. I can, to some extent, use the same
Makefile for producing docs and running some tests, and leave the
Ada parts largely to the language and to the build tools provided by
the Ada toolmakers. That's why I don't see a pressing need for
autotools for Ada programs. (And for Eiffel programs, etc.)

After all they were originally invented to cope with messy,
non-standardised C installations on Unix, on a text macro
replacement Unix toolbox basis. I hope that Ada
installations won't follow the same path. KISS.
See e.g. the origins of GCC PR 23400, which is in the
way of running $ make check when you have built a 4 series GCC.
The failure has some obscure reason originating in, huh!,
fixincludes.

Re: autotools and ada

[Pascal Obry]

Ok, so we agree 100% on this :)

Pascal.

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|              http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595

Re: autotools and ada

[Björn Persson]

Pascal Obry wrote:
> One interresting feature of Makefile is that it "documents" the build
> processes

The word is "automates", not "documents". The auto-generated makefiles 
I've seen have been practically unreadable. They don't document anything.

-- 
Bj�rn Persson                              PGP key A88682FD
                    omb jor ers @sv ge.
                    r o.b n.p son eri nu

Re: autotools and ada

[Pascal Obry]

Bj�rn Persson a �crit :
> Pascal Obry wrote:
> 
>> One interresting feature of Makefile is that it "documents" the build
>> processes
> 
> 
> The word is "automates", not "documents". The auto-generated makefiles
> I've seen have been practically unreadable. They don't document anything.

And we are not speaking about *generated* makefiles... so I really meant
"document".

Pascal.

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|              http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595

Re: autotools and ada

[Björn Persson]

Pascal Obry wrote:
> And we are not speaking about *generated* makefiles... so I really meant
> "document".

I thought this whole thread was about Autotools. But I did think your 
standpoint was somewhat unclear, so I suppose I misunderstood you somewhere.

-- 
Bj�rn Persson                              PGP key A88682FD
                    omb jor ers @sv ge.
                    r o.b n.p son eri nu

Re: autotools and ada

[Pascal Obry]

Bj�rn Persson a �crit :

> I thought this whole thread was about Autotools. But I did think your
> standpoint was somewhat unclear, so I suppose I misunderstood you
> somewhere.

No problem, as you know a thread is quite convoluted after some time :)

Pascal.

-- 

--|------------------------------------------------------
--| Pascal Obry                           Team-Ada Member
--| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
--|------------------------------------------------------
--|              http://www.obry.net
--| "The best way to travel is by means of imagination"
--|
--| gpg --keyserver wwwkeys.pgp.net --recv-key C1082595

Re: autotools and ada

[Stephen Leake]

"Jeffrey R. Carter" <spam@spam.com> writes:

> Ottavio Campana wrote:
>> I'm trying to develop a program using autotools to manage the
>> makefiles.
>
> My experience, in over 21 yrs using Ada, is that make files are
> unnecessary. I have never used them.

Amazing :).

Here are some of the things I use Makefiles for, that the various
Ada-specific tools don't do:

running unit tests automatically

creating zip files for archiving or installation

running compilers for languages other than Ada (LaTeX, VHDL). These
tools are essential to the core Ada project.

I gather you work on simpler (Ada-only) projects, and have a different
mechanism for running unit tests?

Hmm. I guess you could use shell scripts. But I find make's ability
to perform only the minimally required operations enhances
productivity enormously compared to shell scripts.

-- 
-- Stephe

Re: autotools and ada

[Jeffrey R. Carter]

Stephen Leake wrote:

> I gather you work on simpler (Ada-only) projects, and have a different
> mechanism for running unit tests?

I work on a lot of things, including complex, Ada-only embedded systems. Unit 
tests, where used*, are run by the developer as part of the development process 
and not as part of the build process.

*Note that the clean room process, the process used by the space shuttle people, 
who produced the 1 error in 400 KLOC SW, does not allow developers to execute 
their SW, and has no unit testing.

-- 
Jeff Carter
"Alms for an ex-leper!"
Monty Python's Life of Brian
75