plplot-devel — Technical discussion about PLplot development

[Plplot-devel] Use $(VAR) instead of @VAR@ in Makefile.am

[Rafael L. <lab...@ps...>]

In doing my last changes, I noticed that in many Makefiles.am, constructs
like @MATWRAP@ and @MKOCTFILE@ are used.  This is bad practice.  Use
$(MATWRAP) and $(MKOCTFILE) instead, which is the standard, advised style.

The reason for this is simple: variables could be replaced in the invocation
of make, like this: 

    make MKOCTFILE=my-mkoctfile

Using the construct @VARIABLE@ prevents doing that.

This is not a particular problem in the PLplot project, since I guess that
people barely need the variable substitution feature.  However, try to
follow the good usage rule, please. 

I will fix those @VAR@ instances in the various Makefile.am, eventually.

-- 
Rafael

[Plplot-devel] Problem with Dynamic driver

[Pankaj L. <pa...@da...>]

Hi All,

I have,
0. built Plplot5.2.0 on Solaris with tcl/tk driver.
1. copied all the shared objects and "pkgIndex.tcl" to
/usr/ActiveTcl/lib/plplot5.2.0/ folder.
2. Active tcl 8.4 installed at /usr/ActiveTcl.
3. copied the "drivers" folder at /usr/ActiveTcl/lib/plplot5.2.0/drivers/
folder.
4. Taken out comments from "pkgIndex.tcl" to see what all drivers are searched
and corrected the file for "ver" variable bug as reported by me Y'day.

I gave the following instructions at wish prompt.
% package require Itcl
3.2
% package require Itk
3.2
% package require Pltk
looking for driver file: /usr/ActiveTcl/lib/plplot5.2.0/drivers/tkd_drv.so
looking for driver file: /usr/ActiveTcl/lib/plplot5.2.0/drivers/tk_drv.so
5.2.0
% plfram .f
Unable to load driver: xwin_drv.

*** PLPLOT ERROR ***
Unable to load driver
Program aborted

Does anybody has a clue what is going wrong here?

Thanks for any help.

-Pankaj.

[Plplot-devel] "It may be time to pull your projects from SourceForge"

[Joao C. <jc...@fe...>]

I crossed with this:


More recently, due to the increasing reliance on SourceForge to organize=20
FreeSoftware projects, SourceForge has growing power over what it may or =
may=20
not allow. Forking from SourceForge remains difficult, as things like the=
 bug=20
tracking database don't extract so easily. If SourceForge just shut down,=
 or=20
locked out non-paying users, many would be left stranded. See=20
Advogato:article/378.html. The RightToFork is technically hindered, as is=
 the=20
lesson by CodeAndOtherLawsOfCyberspace.

This is one of the reasons the Savannah project was started at=20
http://savannah.gnu.org/. With SourceForge moving to proprietary software=
, it=20
gets increasingly difficult to fork from SourceForge. In order to prevent=
=20
dependencies on a single company, certain parties such as the FSF Europe =
have=20
started to spread the word: It may be time to pull your projects from=20
SourceForge and move elsewhere. Savannah is based on SourceForge code and=
 the=20
maintainers are trying to package the Savannah software as a Debian=20
package... Thus, they are trying to build the RightToFork right into the=20
software.=20

http://www.usemod.com/cgi-bin/mb.pl?RightToFork

Joao

Re: [Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0

[Alan W. I. <ir...@be...>]

On Tue, 18 Feb 2003, Pankaj Laddha wrote:

> Did you require to make any changes in configure file, pkgIndex file to build
> the Plplotter on Solaris,  as I found "pkgIndex.tcl" still has "ver" variable as
> "5.1.0" which causing problem in loading the appropriate drivers.

Thanks, Pankaj, for pointing out this bug in 5.2.0 which I confirm.

Maurice, each of tcl/pkgIndex.tcl.in, tk/pkgIndex.tcl.in, and
tk-x-plat/pkgIndex.tcl.in had the string "5.1.0" in it.  I am sorry we all
missed that for 5.2.0, but to make sure it doesn't happen again, I have
CVS-committed a fix where these files are now configured with the correct
VERSION to produce pkgIndex.tcl in the 3 sub-directories which are then
concatanated to form pkgIndex.tcl.

I subsequently found that the package require business had been clobbered by
the new dynamic driver names in CVS so I have just committed a fix for that
problem as well.  Everything now seems to be fine if you use the recipes in
tcl/README.tcldemos and tk/README.tkdemos.

Maurice and Pankaj, please give these fixes a try for yourself.

Pankaj, I have included you in that invitation because I would value your
solaris/tcl/tk input in the next few weeks as we prepare for the release of
plplot-5.2.1.  Thus, if you are interested, the instructions for getting
anonymous read-only access to cvs are given in
http://sourceforge.net/cvs/?group_id=2915.  The CVS version of plplot
requires more tools than when you work with a plplot tarball, but explicit
instructions about the swig version and autotools versions that are required
have been given on this list recently.  Beyond those additional required
tools, the only difference with the tarball approach is you have to run

./bootstrap.sh -I $autotools-prefix/share/libtool/libltdl/

before configure; make; and make install.  ($autotools-prefix is the prefix
you use when you install the recommended autotools versions).  Feel free to
write me off list if you have any problem getting set up with autotools and
swig.  (BTW, autotools is always required for cvs users, but you don't have
to worry about swig if you are not going to build the java or python
interfaces to plplot.)

Alan

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Alan W. I. <ir...@be...>]

On Tue, 18 Feb 2003, Rafael Laboissiere wrote:

I don't agree with your interpretation, but let's forget any more
discussion on this topic....EOT

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Rafael L. <lab...@ps...>]

* Alan W. Irwin <ir...@be...> [2003-02-18 10:29]:

> Yes it is, but it is also far from perfect so to encourage users to use
> uninstall is just plain wrong unless they don't mind if it screws up.
> 
> First, the automake documentation specifically spreads some doubt about
> uninstall
> 
> "Note that uninstall' is not meant as a replacement for a real packaging
> tool."

Your interpretation of this sentence in the manual is very biased.  The
authors are just stating that "make uninstall" does very simple things,
actually only deletion of files.  It is like "make install", that just
install files.  There could be a sentence in the manual like this:

"Note that install is not meant as a replacement for a real packaging tool."
           ^^^^^^^

Would you then say that "make install" is not trustful?	 
	   
> That doesn't sound like it is a high maintenance priority with the
> autotools developers. Also even if autotools uninstall is completely
> bug-free it is not clear that uninstall will actually work properly if you
> have gone outside the autotools paradigm.

Why are you suggesting that make uninstall is buggy?  Your statements sound
like FUD.

> Sorry, Rafael, but I like the KISS principle so I am going to stick with
> 
> rm -rf
> 
> and a special prefix for each of my handful of different tarball installs.
> Of course if a real packaging tool is used like debs or rpms I am happy to
> go with those since their package uninstalls are heavily tested by many
> users.

At any rate, I never pretended that "make uninstall" would work in any
possible case.  Neither that you should start using it.  I just pointed out
that "make uninstall" take care of removing every file installed through
"make install", including those installed elsewhere than $prefix.  Period.

-- 
Rafael

Re: [Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Alan W. I. <ir...@be...>]

On Tue, 18 Feb 2003, Rafael Laboissiere wrote:

> * Alan W. Irwin <ir...@be...> [2003-02-18 08:52]:
>
> > and once you have made such an install it is non-trivial to un-install.
> > (With my normal prefix of /usr/local/plplot_at,
> >
> > rm -rf /usr/local/plplot_at/*
>
> Wrong.  Just call "make uninstall" and all the files that have been
> installed (regardless if under $prefix or not) are removed.
>
> Automake is great.

Yes it is, but it is also far from perfect so to encourage users to use
uninstall is just plain wrong unless they don't mind if it screws up.

First, the automake documentation specifically spreads some doubt about
uninstall

"Note that uninstall' is not meant as a replacement for a real packaging
tool."

That doesn't sound like it is a high maintenance priority with the autotools
developers. Also even if autotools uninstall is completely bug-free it is
not clear that uninstall will actually work properly if you have gone
outside the autotools paradigm.

Sorry, Rafael, but I like the KISS principle so I am going to stick with

rm -rf

and a special prefix for each of my handful of different tarball installs.
Of course if a real packaging tool is used like debs or rpms I am happy to
go with those since their package uninstalls are heavily tested by many
users.

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Rafael L. <lab...@ps...>]

* Alan W. Irwin <ir...@be...> [2003-02-18 08:52]:

> and once you have made such an install it is non-trivial to un-install.
> (With my normal prefix of /usr/local/plplot_at,
> 
> rm -rf /usr/local/plplot_at/*

Wrong.  Just call "make uninstall" and all the files that have been
installed (regardless if under $prefix or not) are removed.  

Automake is great.

-- 
Rafael

Re: [Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Rafael L. <lab...@ps...>]

* Alan W. Irwin <ir...@be...> [2003-02-18 08:52]:

> OCTAVE_M_DIR = /usr/share/octave/site/m
> OCTAVE_OCT_DIR = /usr/lib/octave/2.0.16.92/oct/i386-pc-linux-gnu
> 
> Rafael, this is a bad result.  It means nobody can try plplot without having
> root privilege, and once you have made such an install it is non-trivial to
> un-install.  (With my normal prefix of /usr/local/plplot_at,
> 
> rm -rf /usr/local/plplot_at/*
> 
> is a wonderfully simple way to uninstall....;-)) Also, ignoring the install
> prefix seems to go against the design philosophy of autotools where every
> effort seems to be made to honor the installation prefix.
> 
> Please change to
> 
> $prefix/share/octave/site/m
> and
> $prefix/lib/octave/2.0.16.92/oct/i386-pc-linux-gnu

This problem has already been detected by Joao, and I worked in a fix since
yesterday evening.  During the day today, I ran some experiements to be sure
that things work.  At any rate, this is fixed and cvs committed.

At any rate, I have been aware of the problem from the very beginning, and I
do not need a long and tedious course on "How $prefix is useful and
wonderful" :-)

-- 
Rafael

[Plplot-devel] install now puts files in /usr/lib regardless of prefix

[Alan W. I. <ir...@be...>]

OCTAVE_M_DIR = /usr/share/octave/site/m
OCTAVE_OCT_DIR = /usr/lib/octave/2.0.16.92/oct/i386-pc-linux-gnu

Rafael, this is a bad result.  It means nobody can try plplot without having
root privilege, and once you have made such an install it is non-trivial to
un-install.  (With my normal prefix of /usr/local/plplot_at,

rm -rf /usr/local/plplot_at/*

is a wonderfully simple way to uninstall....;-)) Also, ignoring the install
prefix seems to go against the design philosophy of autotools where every
effort seems to be made to honor the installation prefix.

Please change to

$prefix/share/octave/site/m
and
$prefix/lib/octave/2.0.16.92/oct/i386-pc-linux-gnu

at least. (I am being over-specific about the version here which I presume
is fully configured in your present scheme.)

Something similar is done for the python packages where the python
plplot-related modules and extension modules are installed in

$prefix/lib/python2.1/site-packages/

(I am being over-specific about the version here which is fully configured.)
Also, there is a configurable 3-line module,
examples/python/plplot_python_start.py(.in), that keeps track of the install
location if different from standard.

With the python scheme, $prefix=/usr works well for packagers, and also
everything works well for those who want to install to their own personal
prefix without having to be root.  Please adopt the same idea for octave.

Alan

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0

[Pankaj L. <pa...@da...>]

Thanks for your reply Maurice.

Did you require to make any changes in configure file, pkgIndex file to build
the Plplotter on Solaris,  as I found "pkgIndex.tcl" still has "ver" variable as
"5.1.0" which causing problem in loading the appropriate drivers.

Also when Build the Plplotter package, I am getting three packages like "Pltcl",
"Pltk", "Plplotter". Which of these packages is backward compatible with
Plplotter 5.1.0?

Thanks again for your time.
-Pankaj.

----- Original Message -----
From: Maurice LeBrun <mj...@ga...>
To: Alan W. Irwin <ir...@be...>
Cc: Pankaj Laddha <pa...@da...>; PLplot development list
<Plp...@li...>
Sent: Tuesday, February 18, 2003 12:10 PM
Subject: Re: [Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0


> Alan W. Irwin writes:
>  > On Thu, 13 Feb 2003, Pankaj Laddha wrote:
>  >
>  > > Hi All,
>  > >
>  > > I built plplotter 5.2.0 on Solaris (Active tcl version 8.4). I have
disabled
>  > > dynamic drivers and enable tcl/tk driver. I am attaching log of configure
to
>  > > this mail for refer of various options I have used for building the
shared
>  > > libraries.
>  > >
>  > > I tried the sequence given at end of this post at wish prompt. I observed
two
>  > > things.
>  > > 1. Why do Plplotter search for tk_drv.so or tkd_drv.so, even when I have
built
>  > > the shared library without any dynamic driver?
>  > > 2. Why am I getting the XCreatPixmap errors when I try to creat second
instance
>  > > of plframe ?
>  > >
>  > > Any idea what is going wrong? Please help me.
>  > >
>  > > Thanks
>  > > Pankaj.
>  >
>  > This is just a guess because I only have a short time to answer this, but I
>  > think we have a built-in assumption in the package require logic to look
for
>  > most stuff in the shared objects created when you specify dynamic drivers.
>  > (Any volunteers to fix this assumption for 5.2.1?)
>
> Sorry for the late reply.  A few comments:
>
> 1. To get 'package require' to work, you must have dynamically-loadable
> modules.  So try with dynamic drivers.
>
> 2. I have never tried PLplot with tcl/tk 8.4.  I believe one of our developers
> (Geoff Furnish) did try some time back and had problems.  So maybe try to
> stick with 8.3.x for now.  We will look at validating PLplot with tcl/tk 8.4
> asap.
>
> 3. I only got the basic package built & a few simple tests.  I don't recall if
> I tested the X driver but if I did, it was using a RH7.3 box as client.
>
> --
> Maurice LeBrun    mj...@ga...
> Research Organization for Information Science and Technology of Japan (RIST)

Re: [Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0

[Maurice L. <mj...@ga...>]

Alan W. Irwin writes:
 > On Thu, 13 Feb 2003, Pankaj Laddha wrote:
 > 
 > > Hi All,
 > >
 > > I built plplotter 5.2.0 on Solaris (Active tcl version 8.4). I have disabled
 > > dynamic drivers and enable tcl/tk driver. I am attaching log of configure to
 > > this mail for refer of various options I have used for building the shared
 > > libraries.
 > >
 > > I tried the sequence given at end of this post at wish prompt. I observed two
 > > things.
 > > 1. Why do Plplotter search for tk_drv.so or tkd_drv.so, even when I have built
 > > the shared library without any dynamic driver?
 > > 2. Why am I getting the XCreatPixmap errors when I try to creat second instance
 > > of plframe ?
 > >
 > > Any idea what is going wrong? Please help me.
 > >
 > > Thanks
 > > Pankaj.
 > 
 > This is just a guess because I only have a short time to answer this, but I
 > think we have a built-in assumption in the package require logic to look for
 > most stuff in the shared objects created when you specify dynamic drivers.
 > (Any volunteers to fix this assumption for 5.2.1?)

Sorry for the late reply.  A few comments:

1. To get 'package require' to work, you must have dynamically-loadable
modules.  So try with dynamic drivers.

2. I have never tried PLplot with tcl/tk 8.4.  I believe one of our developers
(Geoff Furnish) did try some time back and had problems.  So maybe try to
stick with 8.3.x for now.  We will look at validating PLplot with tcl/tk 8.4
asap.

3. I only got the basic package built & a few simple tests.  I don't recall if
I tested the X driver but if I did, it was using a RH7.3 box as client.

-- 
Maurice LeBrun    mj...@ga...
Research Organization for Information Science and Technology of Japan (RIST)

[Plplot-devel] Re: [Plplot-cvs] CVS update notice (plplot/bindings/octave/PLplot)

[Rafael L. <lab...@ps...>]

[Moving this discussion to plplot-devel.]

* Joao Cardoso <jc...@fe...> [2003-02-17 00:52]:

> I don't disagree, but I don't think that "Without this change, or something 
> functionally equivalent, the Octave bindings are unusable."  is accurate, as 
> you said in another e-mail. Indeed, plplot_octave_path is only used by 
> use_plplot, which is buggy and don't work.

I agree with you, as I already wrote.  I was overstating the problem, just
to make more impact.

> Also you might want to know that octave-plplot does not depends by itself
> on octave-2.1, as Alan makes its testing with octave-2.0.16 and promptly
> reports incompatibilities he founds and that I'm happy to fix. Does
> Debian-stable use the octave-2.1 series? I think that RH-8.0 uses 
> octave-2.1.3?. Suse-8.1 don't has octave! (but suse-8.0 had 2.0.16).

There are both octave2.0 and octave2.1 packages in Debian.  Since the 2.1 is
far more superior, is fully stable and supported, and is a dependency to the
great octave-forge package, I do not see the point in making octave-plplot
depend on octave2.0.

> As you have expertise that I lack, could you comment on the following setup
> to avoid rebuilding the .o file whenever a "make install" or plain make
> command is issued?
> I am planning to look at the This works for both octave-2.1.36 and 2.0.16. Remember that mkoctfile-2.0.16 
> does not supports the -c option.
> 
> plplot_octave.oct: plplot_octave.o
> 
> # this compiles (slow) and links, generating .o and .oct,
> # but links with the wrong LD_RUN_PATH (doesn't matter)
> plplot_octave.o: plplot_octave.cc
> 	@MKOCTFILE@ -v -I. \
> 	-L../../src/.libs plplot_octave.cc $(octave_libs)
> 
> # relinks unconditionally, fast
> all-local: plplot_octave.o
> 	LD_RUN_PATH=`pwd`/../../src/.libs/ @MKOCTFILE@ -v -I. \
> 	plplot_octave.o -L../../src/.libs $(octave_libs)
> 
> # relinks  unconditionally, fast
> install-exec-local: plplot_octave.o
> 	LD_RUN_PATH=$(libdir)  @MKOCTFILE@ -v -I. \
> 	plplot_octave.o -L$(libdir) $(octave_libs)

I am planning to look in detail at the PLplot_Octave configuration scheme
soon.  I will keep you informed.

-- 
Rafael

Re: [Plplot-devel] Recent octave configuration changes

[Alan W. I. <ir...@be...>]

On Mon, 17 Feb 2003, Rafael Laboissiere wrote:

> Of course, everything will work fine if the correct path is set
> independently by the user....

That explains it then.  That is what is done by my non-interactive test
scripts so that is why I was under the mistaken impression that octave was
working well for 5.2.0.

There is a lesson to be learned here; it should be emphasized that most of
my release tests are non-interactive in nature (there is just no time for
more) so in order to have a decent release, **the developers here must back
up my test efforts with a variety of tests of your own**. For example, I
leave all interactive octave tests before a release to Joao (and now Rafael)
since I have never had time to learn octave.  Also, the major testing
responsibility for tcl/tk falls on the shoulders of Maurice, Geoffrey, and
Vince although I do a modest amount of interactive testing of the GUI's in
that case. Probably the non-interactive tests are good enough for the C,
C++, f77, and python front ends (so long as they are tried by everybody on
this list with access to a variety of platforms), but tcl/tk and octave are
special cases which need extra testing attention by the experts in those
languages.

Rafael, I am glad you are applying such testing attention to the octave
case.  Also, I want to take this opportunity to say how pleased I am by your
recent successful Debian packaging efforts which are now accessible from our
website.  I suspect those (and the official Debian packages once the 5.2.1
tarball release is completed) are going to bring in a lot more octave and
other front end users and testers for PLplot.

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Recent octave configuration changes

[Rafael L. <lab...@ps...>]

* Alan W. Irwin <ir...@be...> [2003-02-16 15:08]:

> However, I assure you everything worked fine for 5.2.0 so your statement:
> "Without this change, or something functionally equivalent, the Octave
> bindings are unusable." indicates to me you haven't looked clearly at the
> previous working functionality in 5.2.0.

I checked this with a "make install" from the official 5.2.0 tarball with
Octave enabled.  I am sorry to tell you, but the Octave installation is
broken, as I suspected.  In file octave_plplot_path.m the string
PLPLOT_OCTAVE_PATH does not get substituted as it was in the pre-AT era.

The reason for this misbehavior is simple, as I already explained in the cvs
commit log: the substitution of PLPLOT_OCTAVE_PATH was done in
cf/pkg_octave.in and this code is lost now, since it was part of a Makefile.

Here is the error message in Octave:

    $ octave
    octave> use_plplot
    error: PLPLOT_OCTAVE_DIR: No such file or directory
    error: called from ùse_plplot'in file `/usr/share/octave/site/m/PLplot/use_plplot.m'

Of course, everything will work fine if the correct path is set
independently by the user, but this is not documented.  Joao's documentation
tells the user to use the use_plplot script, which is broken in 5.2.0.  I
repeat: the Octave bindings _are_ broken in the 5.2.0 tarball.

I will work together with Joao to improve the Octave
configuration/installation schemes.

-- 
Rafael

[Plplot-devel] Recent octave configuration changes

[Alan W. I. <ir...@be...>]

Rafael, I will let Joao comment on whether you have made improvements in the
octave configuration design.  I am more interested in whether it works or
not but won't be able to test it for a while.

However, I assure you everything worked fine for 5.2.0 so your statement:
"Without this change, or something functionally equivalent, the Octave
bindings are unusable." indicates to me you haven't looked clearly at the
previous working functionality in 5.2.0.

Or were you referring to some post-5.2.0 breakage which you have now fixed
again?

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] trying to synchronize PyDVT with plplot repository

[Alan W. I. <ir...@be...>]

On Fri, 14 Feb 2003, Alessandro Mirone wrote:

> I would like to try now to get PyDVT  working with 5.2.0 ( or eventually
> CVS snapshots ).

Excellent.  5.2.0 has some important bug fixes and some very neat new
contouring API for 3D plots.  For a recent screenshot look at
http://plplot.sourceforge.net/examples/demo08/index.php.

>
> But by now  I fail compiling plplot.
>
>   __ defining CPPFLAGS or LDFLAGS as environment variable breaks the
> compilation of libltdl
>           ( CPPFLAGS=-I.... is given as argument to libltdl configure
> script )
>
>   --  Double precision breaks the compilation of drivers : xwind.c is
> required but only xwin
>       is there

Is this for an unmodified plplot-5.2.0?  What are the details of your
platform?  I am not sure of that first issue, but the second one (xwind.c)
sounds bizarre to me.  We do not have a precision tag for the names of
our source code.

>
>   -- I tried to modify makefile.am but then configure.in is missing.
>

automake, autoconf, and libtool all must be the appropriate latest versions
(autoconf-2.57, automake-1.7.2, and libtool-1.4.3).  We use configure.ac and
not configure.in.  If your autotools are demanding configure.in, it means
they are dated.  autotools are in a state of flux at the moment so you must
be quite precise in the versions that you adopt.  The above versions are
what we use for preparing the CVS version of plplot for the configure step
using the ./bootstrap.sh command.  bootstrap.sh has already been executed
for the plplot-5.2.0 tarball so no autotools are needed (and should not be
used) in the tarball case at all.

>
> Is it possible to disactivate  libltdl( see the static library issue
> below )?

That library is used to dynamically load our drivers as needed. Specify
--disable-dyndrivers if you don't want this feature.  With dynamic drivers
disabled, the drivers are absorbed into libplplot.  However, there is a
well-known bug in 5.2.0 such that --disable-dyndrivers and --with-double
cannot be specified at the same time.  Use our current CVS (which is being
prepared for release of 5.2.1 in April) if you want the combination of
--disable-dyndrivers and --with-double.  If you use CVS, you will need the
above autotools versions as well as swig version 1.3.17. (Like the autotools
case, swig is not required if you just use the tarball, but it sounds
like you will want to try the CVS version.)

>
>
> *** LIST of modifications
>
> -- very small modifications, a part this one :
>
>    two python modules are created :
>         ** pyqt_module.so
>         ** plplotmodule.so
>
>     that's because pyqt_module is used for displaying while plplotmodule
>     for printing, and such two different operations need different
> settings( while plplot, if
>       I remember well has a global pointer to a static structure
>        that cannot be duplicated to have different settings for
>          different instances of the structure )

Before answering that I need to give you some background.  our python
interface is swig-generated now, and the results are organized differently
than before.  There is now a hand-crafted plplot_widgetmodule.so extension
module that contains a small number of functions (mostly donated by you long
ago) that are useful for plplot widgets.  There is a user-friendly plplot.py
module that wraps plplotc.py which in turn wraps the extension module
_plplotcmodule.so.  swig (with help from autotools) automatically generates
both plplotc.py and _plplotcmodule.so.  Although setup.py.in is still in
CVS bindings/python it is only there for historical reference because we
no longer use the setup.py method on the Unix side.

With that background out of the way, I think what you should do if you need
two versions of the results configured in different ways is configure,
build, and install the first version and move the installed results in, for
example, $prefix/lib/python2.1/site-packages/ out of the way to a special
directory.  Then configure, build, and install your second version. When you
want to use the first version specify the appropriate directory in a wrapper
and similarly for the second version for a differently named wrapper.  The
wrappers should keep the namespaces from the two separate versions from
colliding with each other.  By the way, if you don't know how to specify a
special directory for a special python module you should be able to easily
generalize what is done in examples/python/plplot_python_start.py(.in).)
Also, look at plplot.py to see how easy it is to wrap one namespace with
another.

So to summarize this, I think it should be entirely straightforward to have
several differently configured plplot modules that are wrapped under
different names.  But before you get to that, you should first make sure you
can build and use the unmodified CVS version of plplot on at least a Linux
machine. For example, I have maintained prova.py (original version donated
by you some time ago) in the installed examples
$prefix/lib/plplot5.2.0/examples/python/.  It still works for me, and should
also work for you for the unmodified CVS version of PLplot.

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

[Plplot-devel] trying to synchronize PyDVT with plplot repository

[Alessandro M. <mi...@es...>]

Hi , 

  as you probably know some time ago Alexandre Gobbo  developed
a visualisation toolkit called PyDVT that implements plplot.

The plplot version that we are using is 5.1.0 with small modifications.

I would like to try now to get PyDVT  working with 5.2.0 ( or eventually
CVS snapshots ).

Several ( in principle small ) problems should be solved. See the list
*** at the end.

But by now  I fail compiling plplot. 

  __ defining CPPFLAGS or LDFLAGS as environment variable breaks the
compilation of libltdl
          ( CPPFLAGS=-I.... is given as argument to libltdl configure
script )

  --  Double precision breaks the compilation of drivers : xwind.c is
required but only xwin
      is there

  -- I tried to modify makefile.am but then configure.in is missing. 


Is it possible to disactivate  libltdl( see the static library issue
below )?  


*** LIST of modifications

-- very small modifications, a part this one :

   two python modules are created : 
        ** pyqt_module.so 
        ** plplotmodule.so 

    that's because pyqt_module is used for displaying while plplotmodule 
    for printing, and such two different operations need different
settings( while plplot, if
      I remember well has a global pointer to a static structure 
       that cannot be duplicated to have different settings for
         different instances of the structure )  
 
   Such modules contains monolitically plplot lib ( we know how to trick
out configure script 
    to skip the test on shared library )


We don't want to force you to adopt this scheme, or switch to a more
object oriented 
architecture. By now for us would be enough to be able, given a
snapshot, to
create with minor efforts, pyqt_module.so and plplotmodule.so

Thank you very much for any help

      Alessandro Mirone

Re: [Plplot-devel] tcl/tk driver build results on Solaris : NeedsHelp

[Alan W. I. <ir...@be...>]

On Thu, 13 Feb 2003, Vince Darley wrote:

> Looks like you are getting closer -- all this static/dynamic driver stuff
> is very unix-specific and not something I know anything about, except that
> it seems extraordinarily complex.

Yep, the old paradigm of throwing the fortran, c++, java, and tcl/tk
front-ends, and *all* drivers (and the kitchen sink and the bathtub) into
libplplot was much easier to get right since there were few linking issues
to worry about.  But we started abandoning that paradigm in 5.1.0 with the
introduction of dynamic drivers, and we completely separated the front-ends
from the libplot library for 5.2.0.  I believe the lean/mean results of the
new paradigm (for example being able to dynamically load the tcl/tk plplot
interface from wish and tclsh using "package require") are worth the
additional trouble we have had to take with the linking. I would never want
to go back to the 5.0.x style of linking.

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0

[Alan W. I. <ir...@be...>]

On Thu, 13 Feb 2003, Pankaj Laddha wrote:

> Hi All,
>
> I built plplotter 5.2.0 on Solaris (Active tcl version 8.4). I have disabled
> dynamic drivers and enable tcl/tk driver. I am attaching log of configure to
> this mail for refer of various options I have used for building the shared
> libraries.
>
> I tried the sequence given at end of this post at wish prompt. I observed two
> things.
> 1. Why do Plplotter search for tk_drv.so or tkd_drv.so, even when I have built
> the shared library without any dynamic driver?
> 2. Why am I getting the XCreatPixmap errors when I try to creat second instance
> of plframe ?
>
> Any idea what is going wrong? Please help me.
>
> Thanks
> Pankaj.

This is just a guess because I only have a short time to answer this, but I
think we have a built-in assumption in the package require logic to look for
most stuff in the shared objects created when you specify dynamic drivers.
(Any volunteers to fix this assumption for 5.2.1?)

I cannot recall the rest of this thread, but did we advise you to use
static drivers for some reason?  Although those are the traditional solution,
they were not as well-tested as dynamic drivers for 5.2.0, and in fact there
is a well known bug if you use the combination of static drivers and double.

So please try dynamic drivers.  They certainly worked for me on one limited
solaris test without tcl/tk, and I think they are necessary for our current
tcl/tk package require logic.

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

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

Programming affiliations with the Canadian Centre for Climate Modelling and
Analysis (www.cccma.bc.ec.gc.ca) and the PLplot scientific plotting software
package (plplot.org).

__________________________

Linux-powered Science
__________________________

[Plplot-devel] Problem with Solaris - tcl/tk Plplot 5.2.0

[Pankaj L. <pa...@da...>]

Hi All,

I built plplotter 5.2.0 on Solaris (Active tcl version 8.4). I have disabled
dynamic drivers and enable tcl/tk driver. I am attaching log of configure to
this mail for refer of various options I have used for building the shared
libraries.

I tried the sequence given at end of this post at wish prompt. I observed two
things.
1. Why do Plplotter search for tk_drv.so or tkd_drv.so, even when I have built
the shared library without any dynamic driver?
2. Why am I getting the XCreatPixmap errors when I try to creat second instance
of plframe ?

Any idea what is going wrong? Please help me.

Thanks
Pankaj.


% package require Itcl
3.2
% package require Itk
3.2
% package require Plplotter
5.2.0
% plframe .p
.p
% pack .p
% .p cmd plvsta
% .p cmd plwind 10 100 10 100
% .p cmd plbox b 0 0 b 0 0
% package require Pltk
looking for driver file:
/usr/ActiveTcl/lib/plplot5.2.0/drivers/tkd_drv.so
looking for driver file:
/usr/ActiveTcl/lib/plplot5.2.0/drivers/tk_drv.so  looking for driver file:
/usr/ActiveTcl/lib/plplot5.2.0/./libplplotd.so.5.2.0
looking for driver file:
/usr/ActiveTcl/lib/plplot5.2.0/./libplplot.so.5.2.0
file
"/usr/ActiveTcl/lib/plplot5.2.0/./libplplot.so.5.2.0"
is already loaded for package "Plplotter"
% matrix m f 2
m
% m 0 = 10
% m 1 = 30
% matrix l f 2
l
% l 0 = 20
% l 1 = 40
% .p cmd plline 2 m l
% plframe .p1
Error in XCreatePixmap: BadWindow (invalid Window
parameter).
Error in XCreatePixmap: BadDrawable (invalid Pixmap or
Window parameter).
Error in XCreatePixmap: BadDrawable (invalid Pixmap or
Window parameter).
X Error of failed request:  BadWindow (invalid Window
parameter)
  Major opcode of failed request:  2
(X_ChangeWindowAttributes)
  Resource id in failed request:  0x5000009
  Serial number of failed request:  63
  Current serial number in output stream:  66
/usr/ActiveTcl/lib/plplot5.2.0 >

Re: [Plplot-devel] tcl/tk driver build results on Solaris : NeedsHelp

[Vince D. <vi...@sa...>]

Looks like you are getting closer -- all this static/dynamic driver stuff
is very unix-specific and not something I know anything about, except that
it seems extraordinarily complex.

-- Vince

p.s. I'm on this mailing list, so no need to email me direct.

On Thu, 13 Feb 2003, Pankaj Laddha wrote:

> Hi Maurice,
>
> Did you try building tcl/tk driver on Solaris? Can you please share you
> results/changes with me, as I am trying to build plplot 5.2.0 on Solaris. I am
> getting following errors .
>
> 1. I was getting multiple defination error of "TclSetStartupScriptFileName,
> TclGetStartupScrip" symbols. I fixed this by commenting these function in
> tclMain.c. I have also defined -D USE_TCL_STUBS -D USE_TK_STUBS flags using
> CFLAGS. Are these flags required for building Plplotter with tcl/tk driver. I
> could build the shared libraries. I have enabled dynamic drivers.
>
> 2. After loading the Plplotter package in wish8.4, I am getting following error
> when try to creat a frame.
>
> % plfram .f
> Unable to load driver: tkwin_drv.
>
> *** PLPLOT ERROR ***
> Unable to load driver
> Program aborted
>
> 3. When I build the library with only static drivers, I try following I am
> getting some more errors.
> % package require Plplotter
> 5.2.0
> % Plplotwin .p
> .p
> % Plplotwin .p2
> Error in XCreatePixmap: BadWindow (invalid Windowparameter).
> Error in XCreatePixmap: BadDrawable (invalid Pixmap orWindow parameter).
> Error in XCreatePixmap: BadDrawable (invalid Pixmap orWindow parameter).
> X Error of failed request: BadWindow (invalid Windowparameter)
> Major opcode of failed request: 2
> (X_ChangeWindowAttributes)
> Resource id in failed request: 0x440001a
> Serial number of failed request: 38
> Current serial number in output stream: 41
>
> Can you please help me to sort out above problems?
>
> Thanks for you time and help.
>
> Regards,
> Pankaj.
>
>
> ----- Original Message -----
> From: Alan W. Irwin <ir...@be...>
> To: Pankaj Laddha <pa...@da...>
> Cc: PLplot development list <Plp...@li...>
> Sent: Wednesday, February 12, 2003 10:14 PM
> Subject: Re: [Plplot-devel] tcl/tk driver build results on Solaris : NeedsHelp
>
>
> > I don't have access to the combination of tcl/tk and solaris, but I believe
> > Maurice got tcl/tk to work on solaris for plplot-5.2.0.  I don't know the
> > details of what he had to do to get it to work.  My understanding is Maurice
> > is completely under new job pressure at the moment, but eventually I hope he
> > will be able to answer you.
> >
> > Alan
>
>
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Plplot-devel mailing list
> Plp...@li...
> https://lists.sourceforge.net/lists/listinfo/plplot-devel
>

Re: [Plplot-devel] tcl/tk driver build results on Solaris : NeedsHelp

[Pankaj L. <pa...@da...>]

Hi Maurice,

Did you try building tcl/tk driver on Solaris? Can you please share you
results/changes with me, as I am trying to build plplot 5.2.0 on Solaris. I am
getting following errors .

1. I was getting multiple defination error of "TclSetStartupScriptFileName,
TclGetStartupScrip" symbols. I fixed this by commenting these function in
tclMain.c. I have also defined -D USE_TCL_STUBS -D USE_TK_STUBS flags using
CFLAGS. Are these flags required for building Plplotter with tcl/tk driver. I
could build the shared libraries. I have enabled dynamic drivers.

2. After loading the Plplotter package in wish8.4, I am getting following error
when try to creat a frame.

% plfram .f
Unable to load driver: tkwin_drv.

*** PLPLOT ERROR ***
Unable to load driver
Program aborted

3. When I build the library with only static drivers, I try following I am
getting some more errors.
% package require Plplotter
5.2.0
% Plplotwin .p
.p
% Plplotwin .p2
Error in XCreatePixmap: BadWindow (invalid Windowparameter).
Error in XCreatePixmap: BadDrawable (invalid Pixmap orWindow parameter).
Error in XCreatePixmap: BadDrawable (invalid Pixmap orWindow parameter).
X Error of failed request: BadWindow (invalid Windowparameter)
Major opcode of failed request: 2
(X_ChangeWindowAttributes)
Resource id in failed request: 0x440001a
Serial number of failed request: 38
Current serial number in output stream: 41

Can you please help me to sort out above problems?

Thanks for you time and help.

Regards,
Pankaj.


----- Original Message -----
From: Alan W. Irwin <ir...@be...>
To: Pankaj Laddha <pa...@da...>
Cc: PLplot development list <Plp...@li...>
Sent: Wednesday, February 12, 2003 10:14 PM
Subject: Re: [Plplot-devel] tcl/tk driver build results on Solaris : NeedsHelp


> I don't have access to the combination of tcl/tk and solaris, but I believe
> Maurice got tcl/tk to work on solaris for plplot-5.2.0.  I don't know the
> details of what he had to do to get it to work.  My understanding is Maurice
> is completely under new job pressure at the moment, but eventually I hope he
> will be able to answer you.
>
> Alan

Re: [Plplot-devel] Re: debian plplot & octave-plplot-0.3

[David S. <ds...@sc...>]

On Wed, Feb 12, 2003 at 08:59:17PM +0100, Rafael Laboissiere wrote:
> I am doing a big overhaul in the Debian packages and they will be available
> when the next version of PLplot will be released (5.2.1 or 5.3.0).  They
> will be much improved.  All the tk drivers will be included oin plplot-tcl.

If you're going to do this, do you need to wait for a new release?

I implemented many of these changes in December, but it tended to
break a lot of stuff, and I didn't have the time to complete and
upload it.  If you want to continue to maintain the Debian packages,
that's great, but I need to know that I'm _not_ (de facto)
maintaining them. :)  You haven't done an upload in almost 2 years.

By the way, I haven't been doing uploads recently because I
don't want to make buggy changes which might hold up python-2.2
transition.  Doesn't look very likely in the near future, but
we can hope.



dave...

Re: [Plplot-devel] Re: debian plplot & octave-plplot-0.3

[Joao C. <jc...@fe...>]

On Wednesday 12 February 2003 18:49, Christophe TROESTLER wrote:
> On Wed, 12 Feb 2003, Jo=E3o Cardoso <jc...@fe...> wrote:
=2E..
> BTW, there are some problems when compiling plplot-5.2.0 with the
> option --with-double, make ends with the message:
>
> =09...
> =09make[1]: *** No rule to make target `dg300d.lo', needed by
> `libplplotdrv.la'.  Stop. make[1]: Leaving directory
> `/tmp/PLPLOT/plplot-5.2.0/drivers'
> =09make: *** [all-recursive] Error 1

It turns out that this question has already been answered:

It turns out the fatal combination is static drivers and --with-double on
*all* platforms.  (We gave the mistaken impression before this problem on=
ly
occurred on some platforms.) Thus, for any platform a workaround is to
specify --enable-dyndrivers.  An alternative workaround is to avoid
specifying --with-double.  Its a Boolean OR of those two workarounds. Thu=
s,
the combination of --enable-dyndrivers and --with-double builds fine sinc=
e
the first workaround condition is fulfilled.  Similarly, default options
(nothing specified about the kind of drivers or precision) build fine sin=
ce
the second condition is fulfilled.

>
>
> Finally, if I issue =AB xhost +local: =BB, the Tk driver crashes with:
>
> =09X server insecure (must use xauth-style authorization); command igno=
red
> =09    while executing
> =09"send $client "set server_name [list $server_name]""
> =09    (procedure "plserver_link_init" line 25)
> =09    invoked from within
> =09"plserver_link_init"
> =09    (procedure "plserver_init" line 85)
> =09    invoked from within
> =09"plserver_init"
>
> Why?  (There is no problem with the xwin driver.)

This question was also already answered (not a plplot issue): you have to=
 use=20
xauth in order to securely run Tk (as tk has a mechanism that would allow=
=20
others to execute commands on your computer -- no need to be a hacker).
The easier way to setup xauth is to use a xdm login manager, like xdm, kd=
m,=20
etc. Otherwise you must use the xauth program and *do not* enable xhosts.=
=20
Other possibility is to rebuild tk without security.

Thanks,
Joao

>
> Cheers,
> ChriS
>
>
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Plplot-devel mailing list
> Plp...@li...
> https://lists.sourceforge.net/lists/listinfo/plplot-devel