Re: [Plplot-devel] Re: Recent changes regarding octave/matwrap

[Alan W. Irwin <irwin@be...>]

On 2004-02-24 04:54-0000 Jo=E3o Cardoso wrote:

> On Monday 23 February 2004 08:42, Rafael Laboissiere wrote:
> |  I decided to completely revert my changes.  Looking more carefully to =
the
> |  dependencies, I noticed that plplot_octave.cc depends on plplot_octave=
=2Eh,
> |  which is generated by configure and depends on --with-double in its tu=
rn.
>
> Ah, yes, I forgot that issue.
> Some cvs versions ago this was not necessary, as our main config.h was
> included, removing the need for the substitution variable; but there was =
name
> clashes with Octave's own config.h, and I removed that include. Perhaps w=
e
> should try that approach again? It implies some changes to the matwrap
> options, but it was working fine, as far as I remember.
> But I don't have spare time.

Another approach might be to produce both single and double-precision
versions and use whatever is appropriate at make time depending on
with_double. That's the approach used for the swig-generated files for the
Python case, but I don't know whether that approach would be easy to do for
the matwrap case.

Alan

__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the PLplot scientific plotting software
package (plplot.org), the Yorick front-end to PLplot (yplot.sf.net), the
Loads of Linux Links project (loll.sf.net), and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________

[Plplot-devel] Recent changes regarding octave/matwrap

[Rafael Laboissiere <rlaboiss@us...>]

[Cc: to plplot-devel]

Joao,

I just committed some changes regarding the matwrap support.  We can always
revert them, if you dislike my ideas, but let us discuss here.

I have thought about the matwrap support in the Octave binding and concluded
that its situation is similar to that of Swig.  Therefore, I think that they
should be treated similarly.

Matwrap is only needed by developers who want to change the Octave binding
itself.  Regular users should never have need it.  In PLplot, matwrap
generates two architecture independent sources: plplot_stub.m and
plplot_octave.cc.  My changes to HEAD make these files to be generated by
make dist and included in the tarball.  Moreover, they will not be deleted
by make clean, only by make maintainer-clean.

Since matwrap is not needed for regular users anymore, the
bindings/octave/matwrap directory is not included in the tarball.  I did not
remove the matwrap files from CVS, though.  If you agree with my changes,
this removal is the next thing to do.

In sysloc.in, if the check for matwrap fails, the MATWRAP variable is set to:

    MATWRAP="-e 'print qq{matwrap not available in the system\n}' ; false "

and no warning message is output.   This means that the message:

    matwrap not available in the system

will be sent if no matwrap was detected at configure time and if there is an
attempt to use matwrap in bindings/octave/Makefile.

The advantages of my changes are:

   * Avoid distribution of third party stuff in the tarball
   
   * No need to build matwrap-generated files by regular users (faster
     build)

What do you think?

-- 
Rafael

[Plplot-devel] Re: Recent changes regarding octave/matwrap

[<jcard@fe...>]

On Sunday 22 February 2004 18:09, Rafael Laboissiere wrote:
|  [Cc: to plplot-devel]
|
|  Joao,
|
|  I just committed some changes regarding the matwrap support.  We can
| always revert them, if you dislike my ideas, but let us discuss here.
|
|  I have thought about the matwrap support in the Octave binding and
| concluded that its situation is similar to that of Swig.  Therefore, I
| think that they should be treated similarly.

Sure

|  Matwrap is only needed by developers who want to change the Octave binding
|  itself.  Regular users should never have need it.  In PLplot, matwrap
|  generates two architecture independent sources: plplot_stub.m and
|  plplot_octave.cc.  My changes to HEAD make these files to be generated by
|  make dist

And also by a regular make, I hope :). If we want to change something in 
plplot_octave.h do we need to reconfigure? Or a normal make will honour 
dependencies and will regenerate plplot_stub.m and plplot_octave.cc?

| and included in the tarball.  Moreover, they will not be deleted
|  by make clean, only by make maintainer-clean.
|
|  Since matwrap is not needed for regular users anymore, the
|  bindings/octave/matwrap directory is not included in the tarball.

OK

|  I did
| not remove the matwrap files from CVS, though.  If you agree with my
| changes, this removal is the next thing to do.

I don't understand this point. If you remove the matwrap directory, how will  
plplot_stub.m and  plplot_octave.cc be generate? Using a installed matwrap?
The built-in matwrap was slighly modified to work in-place, but if a installed 
matwrap works, that's fine.
However, I think that matwrap is not being maintained anymore, can't this be a 
problem in the future, if the distribution site disapears?

|  In sysloc.in, if the check for matwrap fails, the MATWRAP variable is set
| to:
|
|      MATWRAP="-e 'print qq{matwrap not available in the system\n}' ; false
| "
|
|  and no warning message is output.   This means that the message:
|
|      matwrap not available in the system
|
|  will be sent if no matwrap was detected at configure time and if there is
| an attempt to use matwrap in bindings/octave/Makefile.
|
|  The advantages of my changes are:
|
|     * Avoid distribution of third party stuff in the tarball

Agreed

|     * No need to build matwrap-generated files by regular users (faster
|       build)

Fine

|  What do you think?

That's fine, as long as a standard matwrap can be used and if a normal make 
honours dependencies.

I think that at this point you are more concerned with making tarbals, 
releases, etc, and making our users life easier, but please don't forget us, 
the developpers, developping (the modify, make, test cycle) must not be a 
pain  ;-)

Joao

[Plplot-devel] Re: Recent changes regarding octave/matwrap

[Rafael Laboissiere <rlaboiss@us...>]

* João Cardoso <jcard@...> [2004-02-23 04:54]:

> On Sunday 22 February 2004 18:09, Rafael Laboissiere wrote:
> |  Matwrap is only needed by developers who want to change the Octave binding
> |  itself.  Regular users should never have need it.  In PLplot, matwrap
> |  generates two architecture independent sources: plplot_stub.m and
> |  plplot_octave.cc.  My changes to HEAD make these files to be generated by
> |  make dist
> 
> And also by a regular make, I hope :). If we want to change something in 
> plplot_octave.h do we need to reconfigure? Or a normal make will honour 
> dependencies and will regenerate plplot_stub.m and plplot_octave.cc?

Yes, of course.

> | I did  not remove the matwrap files from CVS, though.  If you agree with
> | my changes, this removal is the next thing to do.
> 
> I don't understand this point. If you remove the matwrap directory, how
> will plplot_stub.m and plplot_octave.cc be generate? Using a installed
> matwrap? 

Yes, this would be the only way.

> The built-in matwrap was slighly modified to work in-place, but if a
> installed matwrap works, that's fine.

Like in the case of Swig.

> However, I think that matwrap is not being maintained anymore, can't this
> be a problem in the future, if the distribution site disapears?
>
> [...]
>
> I think that at this point you are more concerned with making tarbals, 
> releases, etc, and making our users life easier, but please don't forget us, 
> the developpers, developping (the modify, make, test cycle) must not be a 
> pain  ;-)

Ok, I did not know that matwrap is not being maintained.  In this case, let
us continue to distribute matwrap with PLplot.  I will allow the use of its
local version again.  However, I will include the generated files
plplot_stub.m and plplot_octave.cc in the tarball.

-- 
Rafael

Re: [Plplot-devel] Re: Recent changes regarding octave/matwrap

[Rafael Laboissiere <rlaboiss@us...>]

* Rafael Laboissiere <rlaboiss@...> [2004-02-23 07:31]:

> Ok, I did not know that matwrap is not being maintained.  In this case, let
> us continue to distribute matwrap with PLplot.  I will allow the use of its
> local version again.  However, I will include the generated files
> plplot_stub.m and plplot_octave.cc in the tarball.

I decided to completely revert my changes.  Looking more carefully to the
dependencies, I noticed that plplot_octave.cc depends on plplot_octave.h,
which is generated by configure and depends on --with-double in its turn.
This means that distributing plplot_octave.cc with the tarball is a bogus
procedure.
 
-- 
Rafael

Re: [Plplot-devel] Re: Recent changes regarding octave/matwrap

[<jcard@fe...>]

On Monday 23 February 2004 08:42, Rafael Laboissiere wrote:
|  * Rafael Laboissiere <rlaboiss@...> [2004-02-23 07:31]:
|  > Ok, I did not know that matwrap is not being maintained.

There has been no changes for several years now, I really don't know if it is 
not maintained. Maybe it's already perfect :-?

|  > In this case,
|  > let us continue to distribute matwrap with PLplot.  I will allow the use
|  > of its local version again.  However, I will include the generated files
|  > plplot_stub.m and plplot_octave.cc in the tarball.
|
|  I decided to completely revert my changes.  Looking more carefully to the
|  dependencies, I noticed that plplot_octave.cc depends on plplot_octave.h,
|  which is generated by configure and depends on --with-double in its turn.

Ah, yes, I forgot that issue.
Some cvs versions ago this was not necessary, as our main config.h was 
included, removing the need for the substitution variable; but there was name 
clashes with Octave's own config.h, and I removed that include. Perhaps we 
should try that approach again? It implies some changes to the matwrap 
options, but it was working fine, as far as I remember.
But I don't have spare time.

Joao

|  This means that distributing plplot_octave.cc with the tarball is a bogus
|  procedure.

Re: [Plplot-devel] Re: Recent changes regarding octave/matwrap

[Alan W. Irwin <irwin@be...>]

On 2004-02-24 04:54-0000 Jo=E3o Cardoso wrote:

> On Monday 23 February 2004 08:42, Rafael Laboissiere wrote:
> |  I decided to completely revert my changes.  Looking more carefully to =
the
> |  dependencies, I noticed that plplot_octave.cc depends on plplot_octave=
=2Eh,
> |  which is generated by configure and depends on --with-double in its tu=
rn.
>
> Ah, yes, I forgot that issue.
> Some cvs versions ago this was not necessary, as our main config.h was
> included, removing the need for the substitution variable; but there was =
name
> clashes with Octave's own config.h, and I removed that include. Perhaps w=
e
> should try that approach again? It implies some changes to the matwrap
> options, but it was working fine, as far as I remember.
> But I don't have spare time.

Another approach might be to produce both single and double-precision
versions and use whatever is appropriate at make time depending on
with_double. That's the approach used for the swig-generated files for the
Python case, but I don't know whether that approach would be easy to do for
the matwrap case.

Alan

__________________________
Alan W. Irwin
email: irwin@...
phone: 250-727-2902

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the PLplot scientific plotting software
package (plplot.org), the Yorick front-end to PLplot (yplot.sf.net), the
Loads of Linux Links project (loll.sf.net), and the Linux Brochure Project
(lbproject.sf.net).
__________________________

Linux-powered Science
__________________________