plplot-devel — Technical discussion about PLplot development

Re: [Plplot-devel] PLplot release work status

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

Good news:  the psc and plmeta results are identical between the Debian and
RH 7.3 systems for the C, fortran, C++, python, tcl, and octave front ends.
png and jpeg results were completely different, but I attribute that
to the completely different libgd versions on the two systems.

I did no interactive testing on the RH7.3 system.

That completes my RH 7.3 testing.

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] Install relocatibility patch

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

I tried the latest autoconf-2.57, automake-1.7.2 (Maurice uses the earlier
1.7), and libtool-1.4.3, but the bug involving DESTDIR is still there.

Apparently, bug-libtool mailing list has had ongoing discussions about this
bug for the last year at least, but no satisfactory cross-platform solution
has been obtained yet.  However, the solution for Linux is straightforward,
and I copied the enclosed patch from the google cache:
http://216.239.33.100/search?q=cache:vjaWOUEnDsUC:mail.gnu.org/pipermail/bug-libtool/2002-February/003019.html+libtool+DESTDIR&hl=en&ie=UTF-8

(bug-libtool has somehow clobbered their mailing list archive so only google
cache is left, and will presumably disappear shortly).

This patch works for me on either Debian or RH 7.3, but I don't think it
should be part of our general distribution since it is Linux packager
specific.  Instead, it should be used by Linux packagers as a patch file
(Rafael or David will want to use something like this when they get around
to releasing debs for plplot-5.2.0) so that is why I have posted it here.

Using the attached patch (which is only relevant if you are using
libtool-1.4.3), and using a tarball built with the latest autotools, I have
now been able to "almost" build a complete binary rpm for RH-7.3.  Aside
from the tkwin driver (which had to be abandoned because tkInt.h is not
available on RH-7.3), the configure ; make; and make install went fine.
However, the rpm spec file still needs a bit of massaging of the file list
before a binary rpm can be produced.  But I am almost to the point where
I can start running plplot-test.sh on the RH7.3 machine.

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] PLplot release work status

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

Alan W. Irwin writes:
 > I have run up against an libtool bug.  It does not honor the DESTDIR variable
 > that is supposed to make packages install-relocatable (so you can install in
 > a temporary staging area).  This was reported as a bug last February so the
 > fix may be in later versions of libtool than the old version that is on
 > my Debian system.
 > 
 > Anyhow, I am going to install from scratch the combination of versions of
 > autotools that Maurice found would work, and see whether that solves
 > the problem.
 > 
 > Maurice, are there any gotcha's that you remember for the versions of
 > autotools packages that you installed?

No gotcha's that I can recall.  Versions of tools I'm working with: 
 
autoconf 2.57 
automake 1.7 
swig     1.3.17 
libtool  1.4.3 
tcl      8.3.2 
tk       8.3.2 
itcl     3.2 
blt      2.4u 

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

Re: [Plplot-devel] PLplot release work status

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

I have run up against an libtool bug.  It does not honor the DESTDIR variable
that is supposed to make packages install-relocatable (so you can install in
a temporary staging area).  This was reported as a bug last February so the
fix may be in later versions of libtool than the old version that is on
my Debian system.

Anyhow, I am going to install from scratch the combination of versions of
autotools that Maurice found would work, and see whether that solves
the problem.

Maurice, are there any gotcha's that you remember for the versions of
autotools packages that you installed?

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] prova.py example

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

Maurice LeBrun writes:
 > Alan W. Irwin writes:
 >  > sysloc.in sets this variable using
 >  > PYTHON_INSTDIR=`echo $PYTHON_DIR | sed s%$prefix/%%`
 >  > 
 >  > So I think you have to change the last line above.  Use
 >  > ...
 > 
 > OK, by only setting this in sysloc.in if it's not already set I can override
 > it as needed in my ~/config/cf_plplot.in.  After configuring it looks like
 > a winner so I've committed the change.

I finally was able to run this.. it's cool.. yay!

Meanwhile I have about 10 immediate plplot projects to work on, I just need to
figure out how to create a warp bubble around my office where time will run
more slowly and I can get it all done. :/

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

Re: [Plplot-devel] prova.py example

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

Alan W. Irwin writes:
 > sysloc.in sets this variable using
 > PYTHON_INSTDIR=`echo $PYTHON_DIR | sed s%$prefix/%%`
 > 
 > So I think you have to change the last line above.  Use
 > ...

OK, by only setting this in sysloc.in if it's not already set I can override
it as needed in my ~/config/cf_plplot.in.  After configuring it looks like
a winner so I've committed the change.

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

Re: [Plplot-devel] PLplot release work status

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

On Sat, 18 Jan 2003, Alan W. Irwin wrote:

The octave example problem is now fixed thanks to some quick work from Joao.

I have just completed a whole host of tests of various device drivers. The
png, jpeg, cgm, and plmeta devices give perfect results as far as I can see.
There are some 285 plot pages to look at for _each_ so bear with me if I
missed something minor.  There is a legend sizing problem for some of the
experimental "p" octave examples for png (and others), but that is nothing
new. I expect raw device coordinates are being used in those examples (which
will vary from device to device) rather than normalized device coordinates
or world coordinates.  I hope those "p" example legends get changed over to
some sort of normalized device coordinates so this will no longer be a
problem. However, Joao, if some sort of normalized device coordinates are
already being used, please let me know and we should follow up on this
further to see why we are getting device-dependent legend size results for
those examples.

The cgm results are particularly noteworthy in how fast they are produced,
and how nice they look afterward.  It is just a real shame that that
absolutely free (libcd is a public domain software library) format has never
gotten much mindshare in the Linux community.  I guess it just works too
easily, and people pay attention to other things that have constant
problems....;-)

Some of the poorly supported drivers look, well, poorly supported. I only
did a few spot checks for gnome, xfig, and pbm.  I noticed some rendering
problems (gnome, xfig), colour map problems (xfig), and extremely bad
resolution (pbm).  Again, however, I don't think anything new has cropped up
(i.e, the situation is not worse than 5.1.0) so I will continue.

The next step (as I said before, but I forgot how many plots I had to look
at before I could get to it), is making the rpm (which will test the install
relocatibility of the new configuration scheme).  I will start that early
tomorrow after I get some sleep.

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] PLplot release work status

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

On Sat, 18 Jan 2003, Alan W. Irwin wrote:

> The next step is to put together a script to build the tarball.  I am going
> with a script for the first time because the tarball build it is getting
> more complicated with execution of bootstrap.sh, building all the pre-built
> interfaces with swig plus the stuff I did before.

OK. that step is done, and the resulting tarball builds okay on my Debian
system, but plplot-test.sh revealed a problem with at least one octave
example (which I have informed Joao about).  Also, there was a problem with
one python example which I have now fixed.  So if I exclude octave, all
postscript results build fine now.

So I will move forward with the next step (making rpm's on RH7.3) and
hopefully Joao will get the octave bug fix in shortly.

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] prova.py example

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

On Sat, 18 Jan 2003, Maurice LeBrun wrote:

> Unfortunately this doesn't work.  Putting:
>
> # Fixes to work with python 1.5 under RH 7.3
>
> enable_python="yes"
> PY_VERSION=`python -c 'import sys ; print sys.version[0:3]'`
> export PYTHON_INC_DIR=/usr/include/python${PY_VERSION}/
> export PYTHON_MOD_DIR=/usr/lib/python${PY_VERSION}/
> export PYTHON_CFG_DIR=${PYTHON_MOD_DIR}/config
> export PYTHON_NUM_DIR=${PYTHON_INC_DIR}/Numeric/
> export PYTHON_MACH_DIR=${PYTHON_MOD_DIR}/site-packages
> export PYTHON_DIR=${PYTHON_MACH_DIR}
>
> in my ~/config/cf_plplot.in, gives the following line in
> bindings/python/Makefile:
>
> ourpythondir = ${prefix}//usr/lib/python1.5//site-packages
>
> so since I have set prefix to $HOME/tools, this results in absolute nonsense.
> The culprit is:
>
> ourpythondir = ${prefix}/@PYTHON_INSTDIR@
>
> in bindings/python/Makefile.am.  How to fix this?

Sorry, I gave you a specific example that is hard to generalize since it
mixes two different prefixes (the one for the python that is already
installed and the one for the plplot to be installed) which are both /usr in
this specific case of building rpm's.

I think almost all of these variables are essentially unused for the new
configuration system which does the same thing an entirely different way.
For example, see how the python version is set with the new configuration
scheme. (See info automake and search there for PYTHON_VERSION.) For now, I
am not going to hack out all this stuff from sysloc.in so you will still
have to set all those variables (although they will be mostly meaningless
except to bypass the bad sysloc.in logic when you have a split python
system).  To my knowledge the only variable that matters now is
PYTHON_INSTDIR which is used in bindings/python/Makefile.am and which
therefore is used to create ourpythondir.

sysloc.in sets this variable using
PYTHON_INSTDIR=`echo $PYTHON_DIR | sed s%$prefix/%%`

So I think you have to change the last line above.  Use

export PYTHON_DIR=$HOME/tools/lib/python${PY_VERSION}/site-packages

(or is that ${HOME}?) instead.

That should turn the outpythondir definition into

ourpythondir = ${prefix}/lib/python1.5//site-packages

which is exactly what you want if the version associated with the python
command (not python2!) is 1.5.

Anyhow, please change the last export line and let me know whether that works.

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] PLplot release work status

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

So far so good:

(1) I have rebuilt the documentation, spot-checked it, and installed it on
the live website.

(2) I have tagged the current CVS as v5_2_0_rc1 which gives me a nice option
to ignore or include any commits done by other developers today.

The next step is to put together a script to build the tarball.  I am going
with a script for the first time because the tarball build it is getting
more complicated with execution of bootstrap.sh, building all the pre-built
interfaces with swig plus the stuff I did before.

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] prova.py example

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

Trying to clean out my inbox and finally getting back to this.

Alan W. Irwin writes:
 > On Wed, 25 Dec 2002, Maurice LeBrun wrote:
 > 
 > > OK, I found it on my RH7.3 system in the python1.5 install.  But
 > > unfortunately, not under python2.2.  Probably the biggest flaw of RH7.3 in my
 > > book is this split python distribution.
 > 
 > I agree 100 per cent.  When making the RH 7.3 rpm, I got around the RH split
 > python problem by forcing a number of variables for plplot-5.1.0.  I haven't
 > tried this for CVS HEAD, but I believe it should work since I copied most of
 > plplot/cf/sysloc.in to plplot/sysloc.in where these variables should
 > override the python search that ordinarily fails completely on RH 7.3
 > because of their split system.
 > 
 > PY_VERSION=python -c 'import sys ; print sys.version[0:3]'
 > export
 > PYTHON_INC_DIR=/usr/include/python${PY_VERSION}/
 > export PYTHON_MOD_DIR=/usr/lib/python${PY_VERSION}/
 > export PYTHON_CFG_DIR=${PYTHON_MOD_DIR}/config
 > export PYTHON_NUM_DIR=${PYTHON_INC_DIR}/Numeric/
 > export PYTHON_MACH_DIR=${PYTHON_MOD_DIR}/site-packages
 > export PYTHON_DIR=${PYTHON_MACH_DIR}
 > ./configure --prefix=/usr --with-double --enable-dyndrivers --enable-gnome
 > --ena
 > ble-ntk --disable-linuxvga

Unfortunately this doesn't work.  Putting:

# Fixes to work with python 1.5 under RH 7.3

enable_python="yes"
PY_VERSION=`python -c 'import sys ; print sys.version[0:3]'`
export PYTHON_INC_DIR=/usr/include/python${PY_VERSION}/ 
export PYTHON_MOD_DIR=/usr/lib/python${PY_VERSION}/ 
export PYTHON_CFG_DIR=${PYTHON_MOD_DIR}/config 
export PYTHON_NUM_DIR=${PYTHON_INC_DIR}/Numeric/ 
export PYTHON_MACH_DIR=${PYTHON_MOD_DIR}/site-packages 
export PYTHON_DIR=${PYTHON_MACH_DIR} 

in my ~/config/cf_plplot.in, gives the following line in
bindings/python/Makefile: 

ourpythondir = ${prefix}//usr/lib/python1.5//site-packages

so since I have set prefix to $HOME/tools, this results in absolute nonsense.
The culprit is:

ourpythondir = ${prefix}/@PYTHON_INSTDIR@

in bindings/python/Makefile.am.  How to fix this?

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

Re: [Plplot-devel] Zombied file cleanup completed for repository

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

On Sat, 18 Jan 2003, Maurice LeBrun wrote:

> Alan W. Irwin writes:
>  > Also, Maurice, when you have
>  > a chance could you look in the Attic of the top-level directory and the
>  > Attic of the doc directory of the cvs backup (as soon as you get one dated
>  > later than my massive deletes of zombied files) to see whether you spot
>  > anything that is still wrong with the status flags.
>
> I did a find..grep through every Attic file and they all have "state dead;"
> associated with their latest revision.  Nor did I see anything else out of
> whack when comparing specific files under doc/Attic/ to say, those under
> src/Attic/.  So everything looks ok AFAIK; I can't comment on the viewcvs
> problem.

Thanks, Maurice, for doing this additional check.  That eases my mind
considerably!

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] Zombied file cleanup completed for repository

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

Alan W. Irwin writes:
 > Well, the viewcvs problems have not gone away.  If you click on the doc
 > directory or the "don't hide" button on the Attic in
 > http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/plplot/plplot/ there are
 > still python exceptions that occur within viewcvs that don't usually occur
 > anywhere else in my experience.  So it may be a coincidence that the Attic
 > and doc (two areas which were loaded with zombied files) still show
 > problems, but I urge the next person that merges one branch with another to
 > be extra careful that they are not ressurecting a bunch of old deleted files
 > due to some still existing repository problem.  Also, Maurice, when you have
 > a chance could you look in the Attic of the top-level directory and the
 > Attic of the doc directory of the cvs backup (as soon as you get one dated
 > later than my massive deletes of zombied files) to see whether you spot
 > anything that is still wrong with the status flags.

They're all dead, Jim.

I did a find..grep through every Attic file and they all have "state dead;"
associated with their latest revision.  Nor did I see anything else out of
whack when comparing specific files under doc/Attic/ to say, those under
src/Attic/.  So everything looks ok AFAIK; I can't comment on the viewcvs
problem.

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

Re: [Plplot-devel] Current plan is to release PLplot-5.2.0 on 18th January

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

On Sat, 18 Jan 2003, Maurice LeBrun wrote:

> Sorry it took so long to get back to plplot work.. I've been pretty low on my
> free time budget recently.  If the release goes out w/o me having solved my
> portability problems, so be it.  I rarely use the release version anyway. :/

Sorry your timing constraints have not allowed you to make any further
contributions to this release, but that is the way it goes sometimes.  What
I am going to do is get out the release anyhow and mention in particular
that it is fine for Linux, but there are some cross-platform Unix issues
which we are still sorting out.  Once you do have a substantial improvement
in the configuration for even just one more platform, I would be willing to
put out one or more "patch" releases, 5.2.x, with your configuration (or
other) bug fixes.

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] Current plan is to release PLplot-5.2.0 on 18th January

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

Sorry it took so long to get back to plplot work.. I've been pretty low on my
free time budget recently.  If the release goes out w/o me having solved my
portability problems, so be it.  I rarely use the release version anyway. :/

Alan W. Irwin writes:
 > Maurice, do you find a similar xwin problem on your OSF1 platform?

Probably.  I noticed a problem at link time so I didn't even bother trying an
example program.  But errors are easy to miss given the large amount of
output.

...
 /bin/bash ../../libtool --mode=install /usr2/mjl/local/bin/install -c  libplplottcltk.la /usr2/mjl/tools/lib/libplplottcltk.la
libtool: install: warning: relinking `libplplottcltk.la'
(cd /usr2/mjl/dev/plplot/latest/bindings/tcl; /bin/bash ../../libtool --mode=relink gcc -g -O2 -o libplplottcltk.la -rpath /usr2/mjl/tools/lib -version-info 7:0:2 -rpath /usr2/mjl/tools/lib -no-undefined -L/usr2/mjl/tools/lib -ltcl8.3 -L/usr2/mjl/tools/lib -litcl3.2 -L/usr2/mjl/tools/lib -litk3.2 -L/usr2/mjl/tools/lib -ltk8.3 -L../../src -lplplot -L. -ltclmatrix tclAPI.lo tclMain.lo Pltk_Init.lo plframe.lo plr.lo tcpip.lo tkMain.lo)
gcc -shared  tclAPI.o tclMain.o Pltk_Init.o plframe.o plr.o tcpip.o tkMain.o  -Wl,-rpath -Wl,/usr2/mjl/tools/lib  -L/usr2/mjl/tools/lib -ltcl8.3 -litcl3.2 -litk3.2 -ltk8.3 -L/usr2/mjl/dev/plplot/latest/src -lplplot -L/usr2/mjl/dev/plplot/latest/bindings/tcl -ltclmatrix  -Wl,-msym -Wl,-soname -Wl,libplplottcltk.so.5 `test -n "7.2.0:5.0:6.0:7.0" && echo -Wl,-set_version -Wl,7.2.0:5.0:6.0:7.0` -Wl,-update_registry -Wl,.libs/so_locations -o .libs/libplplottcltk.so.7.2.0
/usr/ucb/ld:
Warning: Unresolved:
cos
sin
XLookupString
XWarpPointer
XQueryPointer
XDrawLines
XClearWindow
XFlush
XParseColor
XSync

But this was my first time when I forgot to put --enable-dyndrivers, so this
points to a problem in the static driver configuration.  Note there's no
-lX11 in the above link line. :)

Unfortunately, --enable-dyndrivers doesn't help -- it just moves the creation
of the error to build-time.  It gives the same error as before:

gcc -shared  tclAPI.o tclMain.o Pltk_Init.o plframe.o plr.o tcpip.o tkMain.o  -Wl,-rpath -Wl,/usr2/mjl/dev/plplot/latest/src/.libs:/usr2/mjl/dev/plplot/latest/bindings/tcl/.libs:/usr2/mjl/tools/lib  -L/usr2/mjl/tools/lib -ltcl8.3 -litcl3.2 -litk3.2 -ltk8.3 -L/usr2/mjl/dev/plplot/latest/src /usr2/mjl/dev/plplot/latest/src/.libs/libplplot.so -L/usr2/mjl/dev/plplot/latest/bindings/tcl /usr2/mjl/dev/plplot/latest/bindings/tcl/.libs/libtclmatrix.so    -Wl,-msym -Wl,-soname -Wl,libplplottcltk.so.5 `test -n "7.2.0:5.0:6.0:7.0" && echo -Wl,-set_version -Wl,7.2.0:5.0:6.0:7.0` -Wl,-update_registry -Wl,.libs/so_locations -o .libs/libplplottcltk.so.7.2.0
/usr/ucb/ld:
Warning: Unresolved:
cos
sin
XLookupString
XWarpPointer
XQueryPointer
XDrawLines
XClearWindow
XFlush
XParseColor
XSync

As well as a new one:

gcc -shared  tk_drv_la-tk.o  -Wl,-rpath -Wl,/usr2/mjl/dev/plplot/latest/bindings/tcl/.libs:/usr2/mjl/tools/lib  -L/usr2/mjl/dev/plplot/latest/bindings/tcl/.libs -L/usr2/mjl/dev/plplot/latest/src/.libs -L/usr2/mjl/dev/plplot/latest/src -L/usr2/mjl/tools/lib -L/usr2/mjl/dev/plplot/latest/bindings/tcl /usr2/mjl/dev/plplot/latest/bindings/tcl/.libs/libplplottcltk.so -ltcl8.3 -ltk8.3   -Wl,-msym -Wl,-soname -Wl,libtk_drv.so `test -n "" && echo -Wl,-set_version -Wl,` -Wl,-update_registry -Wl,.libs/so_locations -o .libs/libtk_drv.so
/usr/ucb/ld:
Warning: Unresolved:
plsc
c_pltext
c_plgra
pdf_wr_header
plexit
plP_setpxl
plP_setphy
pdf_wr_1byte
pdf_wr_2bytes
pdf_wr_2nbytes
pdf_wr_ieeef
plGinInit
plabort
pdf_close
plstrdup
plFindCommand
pdf_bopen
plTranslateCursor

Note I'm using a really old version of gcc on this machine but it looks like a
config error rather than being gcc's fault:

zion$ gcc -v
Reading specs from /usr/local/lib/gcc-lib/alphaev56-dec-osf4.0b/2.8.1/specs
gcc version 2.8.1


Aside:  I really hate how many hundreds of lines of output there is.  Makes it
much harder to find errors.  Unfortunately you can't just redirect stdout b/c
you lose context of the errors that do occur.  I suppose you could do:

zion$ make install >foo.log
libtool: install: warning: relinking `libplplottcltk.la'
/usr/ucb/ld:
Warning: Unresolved:
cos
sin
XLookupString
XWarpPointer
XQueryPointer
XDrawLines
XClearWindow
XFlush
XParseColor
XSync

and then to see the context of that do:

zion$ make install >foo.log 2>&1

What we need is a program that does for stderr what "tee" does for stdout,
that way we could send output+errors to a file but only errors to the console.
Should actually pretty easy to write one.

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

Re: [Plplot-devel] Zombied file cleanup completed for repository

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

On Wed, 15 Jan 2003, Alan W. Irwin wrote:

> [...] Now, after a lot of removes and commits there are no differences left
> between plplot1 and plplot3.  Therefore, I believe I have cleaned up all the
> state problems that existed in our CVS.  This should mean that cvs merges
> and the cvs -D option should work properly (for the first time in at least 7
> years), and the viewcvs errors that occurred for directories (e.g. doc) that
> had zombied files should be eliminated.  I don't plan to check the merge
> point any time soon, but for those of you who merge in the feature, please
> be careful to make sure you are not reintroducing these zombied files (lots
> of them in the top-level directory and the doc directory). I do plan to
> check the viewcvs result as soon as that server is up again at sourceforge.

Well, the viewcvs problems have not gone away.  If you click on the doc
directory or the "don't hide" button on the Attic in
http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/plplot/plplot/ there are
still python exceptions that occur within viewcvs that don't usually occur
anywhere else in my experience.  So it may be a coincidence that the Attic
and doc (two areas which were loaded with zombied files) still show
problems, but I urge the next person that merges one branch with another to
be extra careful that they are not ressurecting a bunch of old deleted files
due to some still existing repository problem.  Also, Maurice, when you have
a chance could you look in the Attic of the top-level directory and the
Attic of the doc directory of the cvs backup (as soon as you get one dated
later than my massive deletes of zombied files) to see whether you spot
anything that is still wrong with the status flags.

In any case we do know my massive deletes did fix the cvs -D option.

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] support for Windowses

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

On Thu, 16 Jan 2003, Joachim Wuttke wrote:

> [...]I propose the following solution
> for the forthcoming release:
>
> ../sys/win32/README            # new, text below
> ../sys/win32/msdev             # unchanged
> ../sys/win32/dll/README        # new, text below
> ../sys/win32/dll/plconsole.cpp # take from my patch of jan 15th
> ../sys/win32/dll/plstub.cpp    # take from my patch of jan 15th
> ../sys/win32/dll/win3.cpp      # take from my patch of jan 15th
>
> The big advantage of this solution is that I (or any future co-
> developer) could continue to work on the dll directory without
> caring for backward compatibility with the static library and
> old Windowses.

Joachim, thanks for clarifying your solution which seems like a good one to
me. But since msdev is unaffected, and the rest is experimental I think we
should keep it out of 5.2.0 as you originally proposed (but which I didn't
understand until now) just to simplify my life at this critical time.  Olof,
if you agree with what Joachim is proposing would you please evaluate and
assemble the patch for me say on a time scale of a week or two from now?

Thanks in advance.

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] support for Windowses

[Joachim W. <wu...@cr...>]

Alan: I fully agree that we must not withdraw existing drivers
for Win95/98/ME. My proposal is rather to put them into a
maintenance-only mode, and to concentrate further development
on the NT line - for the simple reason that it is better to
do something for some users than to do nothing for all users.

Olof: as you know, I modified the Win drivers in order to build
a dynamic library. I have absolutely no idea whether the modified
sources still allow to build a static library (a) under NT, and
(b) under W95/... It would be great if you could find out.

Alan: in case Olof has no time to assemble a unique directory for
static and dynamic Win drivers, I propose the following solution
for the forthcoming release:

../sys/win32/README            # new, text below
../sys/win32/msdev             # unchanged
../sys/win32/dll/README        # new, text below
../sys/win32/dll/plconsole.cpp # take from my patch of jan 15th
../sys/win32/dll/plstub.cpp    # take from my patch of jan 15th
../sys/win32/dll/win3.cpp      # take from my patch of jan 15th

The big advantage of this solution is that I (or any future co-
developer) could continue to work on the dll directory without
caring for backward compatibility with the static library and
old Windowses.

Here comes some text for the READMEs:

../sys/win32/README:
--------------------
The directory ../sys/win32/msdev contains everything you need
for building a static library under Win32 (Win95/98/ME/NT/2k/XP).=20

The directory ../sys/win32/dll contains additional experimental
code for building a shared library (DLL). Development is done
for WinNT/2k/XP only; tests under Win95/98/ME would be welcome.

../sys/win32/dll/README:
------------------------
Building a shared (dynamic link) library plplot.dll is possible
in principle. However, development is still in an experimental
stage. A makefile is not yet available; users will have to use
Visual Studio to build their own plplot project. Input files
should be taken from the ../msdev directory, except for the
modified files contained in the present directory.

In one of the next releases of plplot we would like to distribute
a self-contained win32/dll directory, with all sources and scripts
needed for automatically building plplot.dll; contributions to
this project will be welcome.

[Plplot-devel] More details about the release

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

I have just reorganized and updated the PROBLEMS file to mark the ending of
the zombie file problems and the completion of my part of the example
consistency work.

The next PLplot issue on my agenda was lots of documentation work, but today
I have already used up all the time I can spare (and then some!) from my
paid job before Saturday morning so I am going to have to let the
documentation updates slide until the next release.

On Saturday morning, my plan is to tag a release candidate, rebuild the
current documentation without changing it, and make an rpm on a RH 7.3
machine that is accessible to me.  The only tests I intend to do that day
are to make sure my Debian box and the RH box produce consistent file
results for a number of drivers.  Once I am happy with those tests, I will
tag the release final, upload the final tarball to SF, and make it
accessible in our file release area to mark the official release.  I plan to
make the actual official announcement of the release by mid-week or so.  The
announcement venue that gets the most hits (by far) is LT, but they tend to
accumulate such announcements during the week and finally post them on the
weekend so there is no real rush to get the announcement out early in the
week.

If you have any CVS plans between now and Saturday morning, please let me
know so I will know what to expect.

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] Example consistency

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

As you are probably aware I have just finished the changes to the python
examples to make them consistent with the C examples and Java examples. I
hope to do the same thing with the tcl examples before the release Saturday
morning as a test that the tcl API (including the new API that Joao has
recently created for contoured 3D shade and mesh plots) is implemented
properly, and also as a teaching tool for showing how to do *exactly* the
same PLplot plots in different languages.  I won't have time for doing the
same thing for octave, fortran, and c++, but I am still hoping somebody else
will take responsibility for making the examples from those languages
consistent with the C (and therefore java, python, and tcl) examples.

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] Zombied file cleanup completed for repository

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

Alan W. Irwin writes:
 > ...
 > Now, after a lot of removes and commits there are no differences left
 > between plplot1 and plplot3.  Therefore, I believe I have cleaned up all the
 > state problems that existed in our CVS.  This should mean that cvs merges
 > and the cvs -D option should work properly (for the first time in at least 7
 > years), and the viewcvs errors that occurred for directories (e.g. doc) that
 > had zombied files should be eliminated.  ...

This is really good to see.  Thanks Alan!

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

Re: [Plplot-devel] Zombied file cleanup completed for repository

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

On Wed, 15 Jan 2003, Alan W. Irwin wrote:

> I don't plan to check the merge
> point any time soon, but for those of you who merge in the feature, please
> be careful...

TTA (terrible typo alert)  feature --> future.

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] modified drivers allow building a DLL under Windows NT / W2000

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

On Wed, 15 Jan 2003, Joachim Wuttke wrote:

> I uploaded modified versions of Win32 drivers.
> Please find them on the sourceforge patch page
> (https://sourceforge.net/tracker/?group_id=2915&atid=302915).
> They contain essentially the modifications I described
> in several mailings in December.
>
> I presume they are not compatible with archaic (non-NT)
> Windows idiolects; I have not tested.

Let's discuss whether it is important to continue to support win98.
According to http://www.google.com/press/zeitgeist/zeitgeist-dec02.html
(which I feel is a fairly reliable indication of OS share among the general
public desktop users) we have the following results:

w95  3 per cent
w98 38 per cent
wNT  5 per cent
w2K 21 per cent
wXP 25 per cent

To summarize this, there are a lot of legacy windows users in the general
public that use google! (In comparison, Mac share is 4 per cent and Linux
share is 1 per cent which I think is roughly correct for their respective
desktop share amongst the general public although I think Linux *business*
desktop use is somewhat higher.)

Now our users (mostly scientists, I presume) are not ordinary general public
users, and presumably their Linux and newer windows platform shares are much
higher than in this survey. Nevertheless, I think it still may be important
to support w98.  Olof, if you agree will you please test these changes for
that platform and any newer windows platform that you have access to?  On
the other hand if you decide the cutoff should be wNT as the oldest windows
platform to support that is fine with me as well.  But the important thing
is to test on all windows platforms that you want to support.

Once you are happy with the changes to your driver, please send me a patch
in the form we are most familiar with and I will make the commit (hopefully
before the Saturday release).

>
> Alan: I am sorry for not submitting this patch earlier;
> now it is certainly too late for including it in 5.2.

No problem.  I will include it if Olof has a chance to evaluate it in time
and approves of it.

>
> Olof: since none of us has time to support splitted
> versions for pre-NT and NT, I would propose to
> formally freeze generic Windows support after 5.2,
> and to concentrate on NT in further releases.

I will let you guys decide that between you, but please remember that the
above numbers are about the only solid ones we have on market share and any
discussions of how much lower the w98 share is for scientific users is
pretty much speculation.

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] Zombied file cleanup completed for repository

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

On Tue, 16 Jul 2002, Maurice LeBrun wrote:

> Alan W. Irwin writes:
>  > I am still quite interested in why
>  >
>  > cvs checkout -D '15 days ago' plplot
>  >
>  > doesn't give me a clean snapshot for the date specified since it is likely
>  > I would like to get date-specified snapshots from CVS in the future.
>
> I'm not really sure, but I've seen something like it before.
> [...]When I looked at the files in the repository, they
> showed 'state Exp' instead of "state dead" like they should have.  After
> changing the most recent state entry to "dead" like they should have been,
> the -D option worked fine.
>
> We really need direct access to the repository. :(

Just found out how to get rid of these problem with cvs commands.

cvs checkout -d plplot1 plplot
cvs checkout -D '1 sec ago' -d plplot2 plplot

diff -Naur --exclude CVS plplot1 plplot2 |less

shows all the (hundreds of) zombied files (those with bad state
information).  Virtually all had dates prior to 1995 (presumably due to some
cvs client or server problem in that era).  The few with newer dates
were due to an attempt to reuse a removed file (which is a no-no and results
in a file that cannot be checked out [unless you use the -D option]).

The fix is as follows:

cd plplot2
cvs update -A
cvs remove -f filename1 filename2 ... filenamezillion
cvs commit

Periodically, I checked on progress by

rm -rf plplot3
cvs checkout -D '1 sec ago' -d plplot3 plplot
diff -Naur --exclude CVS plplot1 plplot3 |less

Now, after a lot of removes and commits there are no differences left
between plplot1 and plplot3.  Therefore, I believe I have cleaned up all the
state problems that existed in our CVS.  This should mean that cvs merges
and the cvs -D option should work properly (for the first time in at least 7
years), and the viewcvs errors that occurred for directories (e.g. doc) that
had zombied files should be eliminated.  I don't plan to check the merge
point any time soon, but for those of you who merge in the feature, please
be careful to make sure you are not reintroducing these zombied files (lots
of them in the top-level directory and the doc directory). I do plan to
check the viewcvs result as soon as that server is up again at sourceforge.
(It was deliberately taken down yesterday by SF staff, and they hope to get
it back today.)

At some point we should weed the Attic so that the hundreds of file names
that have been removed will be available again for CVS.  For now, though,
our core developers with cvs access should be careful to not reuse old
deleted file names when creating a new file.

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] modified drivers allow building a DLL under Windows NT / W2000

[Joachim W. <wu...@cr...>]

I uploaded modified versions of Win32 drivers.
Please find them on the sourceforge patch page
(https://sourceforge.net/tracker/?group_id=3D2915&atid=3D302915).
They contain essentially the modifications I described
in several mailings in December.

I presume they are not compatible with archaic (non-NT)
Windows idiolects; I have not tested.

Alan: I am sorry for not submitting this patch earlier;
now it is certainly too late for including it in 5.2.
Better regard this as a first contribution for 5.3.

Olof: since none of us has time to support splitted
versions for pre-NT and NT, I would propose to
formally freeze generic Windows support after 5.2,
and to concentrate on NT in further releases.

Juan: don=B4t hesitate to post further questions directly
to the mailing list - even the most obscure technical
discussion might turn out helpful for someone else.

- Joachim