Re: [Plplot-devel] Release planned for Saturday with tarball Rafael made today

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

IMPORTANT.  Andrew's recent changes are required to make DJGPP work so I
have decided to delay the release until when Rafael is in a position to make
a new tarball (according to the note he sent me Thursday, he plans to get
back into internet access Monday evening his time).

That will give Maurice and Geoffrey one last chance on Monday to check the
solaris situation as well.  I plan to do the file release as soon as Rafael
tags cvs and creates the tarball.  I assume that will be something like noon
on Monday. But I can wait until later Monday to make the release
announcement if there is some prospect of a solaris report from M&G in that
time frame.

Joao, apparently you are not available next week, but if you can work a bit
more on PLplot this weekend, this change in release date should give us a
window of opportunity to sort out the recently introduced colour problems.

I am still not quite sure the source of those problems on your systems.  This
morning I rebuilt from current CVS and redid the examples.  On my system the
resulting x08c.png.08, x08j.png.08, pythondemos.png.32, and tcldemos.png.32
are all identical to the nearest bit (!) according to cmp and similarly for the
appropriate x11c.png.08 comparison. I suggest you try adding and subtracting
a small relative change in plot3d.c where you assign values to fc_minz and
fc_maxz to see if that solves the problem on your system(s).

If that works, cvs commit it so that minor change will be in the release.

The rest of this e-mail concerns some additional colour clipping and
positional z clipping wish-list items I discovered this morning which should
(unless Joao is feeling really ambitious) be dealt with post-release.

I have just checked examples 8 and 11 with z range deliberately reduced in
plw3d (thus, mimicking the user who deliberately underestimates the z range
of her data to zoom in on an area). You get a usable zoomed plot but with a
whole bunch of

*** PLPLOT ERROR ***
plcol1: Invalid color map position: 1.094434, aborting operation

etc., sort of error messages.  Colour clipping without the
error messages would be better.

Note that the positional z clipping is done very well in example 8 thanks to
all of Gary Bishop's efforts upgrading zoomed versions of what is now
plsurf3d.  But the base contours are not properly clipped consistent with
the surface contours. Furthermore, there is more work to be done on the
positional z clipping for plot3d and plmeshc used in example 11.  The result
is not too bad when the z range is underestimated by the user, but I believe
properly clipped z values as in example 8 (and properly clipped base
contours as well) would be much more informative to the user for zoomed
plots.

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 affiliation with the PLplot scientific plotting software package
(plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Release planned for Saturday with tarball Rafael made today

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

On Sat, 19 Apr 2003, Joao Cardoso wrote:

> I forgot to set LD_LIBRARY_PATH. I always forgot this for java.

Some time ago I ran across a way to specify library path within java, but I
lost it again.  Anyhow, I know it is out there, and if I find it again I
will implement it in the examples so nobody will have to set LD_LIBRARY_PATH.

> Yes, I have 1.8.4 on suse-8.1.
>
> But those examples work OK under c and octave.
> And all the other plots with a big (256 colors) colormap are fine under
> tcl/python. Even the other plots from the same examples.
>
>  It's not a matter of colormap. But, using "display" to see than, I got
>    display: invalid colormap index (tcldemos.png.46).
>    display: invalid colormap index (tcldemos.png.32).
>
> This does not occurs with the C generated png files.
>
> I confirm the above under RH-8.0. It also has gd-1.8.4. The mentioned plots
> from x08 and x11 run OK under C but not under tcl.

Due to your recent C changes, zmin to zmax is mapped exactly (I believe from
your comments, although I haven't looked at the code) for examples 8 and 11
to the maximum possible cmap1 range from 0. to 1.  This may be a mistake
because of floating-point rounding errors. It might be better to map zmin to
zmax to the cmap1 range from epsilon to 1. - epsilon where epsilon is a
small fraction like 1.d-5.  This change would ensure that PLplot never runs
outside the correct cmap1 range due to floating point errors.  Note that
rounding errors would act slightly differently for different front ends so
you might have a colour range error for certain plots and not others or for
certain front ends and not others (just as you have found).  Note all my
quick tests today were on RC1 so I may also be able to confirm your results
tomorrow when I test the 5.2.1 tarball for the first time.  Anyhow, the
problem you have found is not a showstopper and only affects the looks of 4
plots, and I don't want to wait for Rafael to get back so he can make a new
tarball. Thus, let's investigate rounding error as the cause of this problem
(and all other hypotheses until we find the correct solution) after the
release.

> Go ahead with the release!

I agree unless somebody else finds additional problems before tomorrow.

>
> Good work,

Thanks very much for your good work as well.

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 affiliation with the PLplot scientific plotting software package
(plplot.org).

__________________________

Linux-powered Science
__________________________

Re: [Plplot-devel] Release planned for Saturday with tarball Rafael made today

[Joao Cardoso <jcard@fe...>]

On Friday 18 April 2003 22:22, Alan W. Irwin wrote:
|  On Fri, 18 Apr 2003, Joao Cardoso wrote:
|  > |  http://people.debian.org/~rafael/plplot.html
|  > |
|  > |  md5sum is edb09c27ae0b977a4994efbe538bbe78
|  >
|  > On suse-8.1 configure/make/install runs OK.
|  >
|  > All tests run OK, but java. The instructions in
|  > examples/java/README.java are not accurate
|
|  Ugh.  You are right about the last part of that file.
|
|  > And the java examples directory is not in the usual
|  > <prefix>/lib/plplot<version>/examples directory. Why?
|
|  That is a leftover from Geoffrey's original organization of the interface
|  code and examples in the same installed directory tree.  This should
|  probably be looked at again post-release.
|
|  Here is what I do to get java to work.  Adjust for your own locations (and
|  from tcsh syntax to bash if appropriate).
|
|  setenv JAVA_HOME /home/software/java/IBMJava2-14/
|  setenv PATH $PATH":$JAVA_HOME/bin:/usr/local/plplot_at/bin"
|  setenv CLASSPATH /usr/local/plplot_at/lib/java
|
|  ./configure --prefix=/usr/local/plplot_at --enable-java (and other
| options). make
|  make install
|
|  cd /usr/local/plplot_at/lib/plplot5.2.1/examples
|
|  Compile the installed java examples.
|
|  javac -d $CLASSPATH $CLASSPATH/plplot/examples/x??.java

OK, it now works with java2-1.3.1 on suse-8.1. I couldn't test on RH-8.0 
because it hasn't the development package installed.

I forgot to set LD_LIBRARY_PATH. I always forgot this for java.

|
|  Compile the rest of the examples.
|
|  cd c; make; cd c++; make; cd f77; make; cd tk; make; cd ..
|
|  Build both psc and png non-interactive (file) example results.
|  ./plplot-test.sh --device=psc
|  ./plplot-test.sh --device=png
|
|  > The python/tcl generated  files *.png.34/46 are not OK.
|
|  For RC1, I have just stared at the second plot of x09c.ps (which
| corresponds to 34)

Sorry, I mean 32. I will send them to you.

| and the last plot of x11c.ps (which corresponds to 46)
| and I can see no problems or differences with the 34th and 46th page from
| pythondemos.ps, tcldemos.ps or the equivalent pythondemos.png.* and
| tcldemos.png.* files. Of course, my eyes may have just missed something
| subtle that you spotted. Could you please be really specific?
|
|  The differences you see between the postscript and png results may be a
|  result of the limited (paletted) colour support for older libgd versions.
| I am using libgd version 2.0.1 here, with truecolour support.
|
|  > The above small problems are minor. The java problem can be the result
|  > of some misunderstanding from me.
|
|  Thanks for your tests.  I hope you are able to do java now following the
|  above instructions.  Furthermore, if on both your platforms (I have read
|  your second e-mail as well) you have older libgd than 2.0.1,

Yes, I have 1.8.4 on suse-8.1.

But those examples work OK under c and octave.
And all the other plots with a big (256 colors) colormap are fine under 
tcl/python. Even the other plots from the same examples.

 It's not a matter of colormap. But, using "display" to see than, I got
   display: invalid colormap index (tcldemos.png.46).
   display: invalid colormap index (tcldemos.png.32).

This does not occurs with the C generated png files.

I confirm the above under RH-8.0. It also has gd-1.8.4. The mentioned plots 
from x08 and x11 run OK under C but not under tcl. 

| I think we
|  should ignore that problem for the release.  If you have 2.0.1 or later,
|  then I think we better follow it up (and probably delay the release until
|  Rafael gets back) although I cannot seem to reproduce it here.  Let me
| know what you think the best course is to take.

Go ahead with the release!

|  BTW, I am making good progress today writing up the formal release
|  announcement. This is the first formal announcement since 5.1.0 so there
| is a lot to cover.  At release time, I will be putting the result up on the
| website.

Good work,
Joao

|
|  Alan