[Plplot-devel] Time to start extensive testing

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

I am glad to report that except for constraints in our tcl/tk configuration
system that make it impossible to test the tcl/tk interface to plplot on
netbsd, our netbsd port is fine, and all the nagging problems I discussed with
Rafael over the last few days are now solved.  For example, this morning on
the netbsd system, I used plplot-test.sh to generate all psc and png
non-interactive examples for the c, c++, f77, octave, and python front ends.
There were no obvious problems in building any of these ~300 plots.

Now that we have this quite comprehensive netbsd test completed as well as
Joao's test of plplot in a more limited OSF1 environment, I want to strongly
encourage expanding the testing activity to all accessible platforms.

I will make an announcement of the testing opportunity on plplot-general
early next week.  (I am holding off now because Rafael forgot to update the
tarball at http://people.debian.org/~rafael/plplot.html from Thursday's
version which had some problems, and he is out of e-mail contact this
weekend.)

Meanwhile, the developers here should be able to start testing this weekend
by generating their own tarballs.

There are 3 possible methods of building a tarball.

(1) The simplest procedure is still to install autotools (autoconf-2.57,
automake-1.7.2, and libtool-1.4.3; all these versions must be exact), put
the resulting executables on your path, and

cvs export -D now plplot
cd plplot
./bootstrap.sh
cd include
touch plConfig.h.in  plDevs.h.in
cd ../..
tar zcf plplot.tar.gz plplot

However, your tarball will be missing the documentation, the perl-generated
tcl/tk interface, and the swig-generated python and java interfaces.

(2) If you have autotools, perl, and swig-1.3.17 (only that exact version)
on your system, then the better procedure is

cvs export -D now plplot
cd plplot
scripts/make_tarball.sh
cd ..
tar zcf plplot.tar.gz plplot

scripts/make_tarball.sh takes care of the perl and swig generation of the
tcl/tk, python, and java interfaces, as well all the required touching.
However, the resulting tarball will be missing the documentation.

(3) If you have autotools, perl, and swig on your system, and you are also
able to build the documentation (see doc/docbook/README.developers for a
list of the required packages and other comments) then you can simply use
Rafael's convenient script plplot/scripts/make-cvs-tarball.sh to build the
tarball.  The script takes care of the cvs export details and finishes by
moving the created full tarball (including the built documentation) to the
current directory.

After the tarball is created from current CVS using any one of the above 3
methods, then simply copy it to the Unix or Linux system you want to test.
N.B. autotools and swig will not be required on the test system!  In fact,
all you have to do on that system is the normal PLplot build, i.e.,

* Possibly set some environment variables to help our configuration system
find installed headers and libraries on the test system.  See configure.ac
and sysloc.in for possible environment variables to set although these
generally wont be needed if your test system follows the LSB.  Note this
is an area for possible improvement in the next two weeks; ideally we would
have documented configuration parameters to give the same information to
our configuration system as is currently conveyed by environment variables.
For example, on the netbsd system I set the following environment variables:
export GDINCDIR=/usr/pkg/include
export TCLINCDIR=/usr/pkg/include
export TKINCDIR=/usr/pkg/include
export PYTHON_INC_DIR=/usr/pkg/include/python2.2
export FREETYPEINCDIR=/usr/pkg/include/freetype2
export GDLIBDIR=/usr/pkg/lib
export PNGLIBDIR=/usr/pkg/lib
export JPEGLIBDIR=/usr/pkg/lib
export TCLLIBDIR=/usr/pkg/lib
export TKLIBDIR=/usr/pkg/lib
export FREETYPELIBDIR=/usr/pkg/lib

but individual configure switches for all of these would be nice as well.

* Use ./configure --help to find out about useful configure options.  One of
the most important of these is --prefix, and the defaults for the rest of
the options are useful for most needs.  This is another area for improvement
in the next two weeks.  Some of the documented plplot options don't work,
and some of the plplot options are not completely documented (especially
what the default values are)!

* ./configure --prefix=???

* make

* make install

* Test the result (if no obvious problems with make or make install).

cd $prefix/lib/plplot$version/examples)
cd c; make; cd ../c++; make; cd ../f77; make; cd ../tk; make; cd ..
./plplot-test.sh --device=psc
./plplot-test.sh --device=png
... (use plplot-test.sh --help to find out list of possible devices).

Good luck with your testing!  The more you do, the better our release
candidate will be on on April 12 (two weeks from today!), and the better our
final version will be on April 19.

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
__________________________