PLplot

Hello all,

I promised to look into the FORTRAN versions of the demo programs
a while back. Well, here is the overview:

Detailed comparison FORTRAN/C:
------------------------------
Demonstration programs in FORTRAN versus the C versions

Common differences:
- No handling of command-line arguments
- Call to plcol0 in C is call to plcol in FORTRAN
- Placement of the comments in FORTRAN is not one-to-one with the C
  version, making comparisons a bit awkward
- Sometimes the subroutines/function calls have a different order
- Sometimes constants (array-dimensions etc.) have different values

x01c/x01f:
- No subroutines plot2 and plot3 in x01f
- Subroutine plot1 uses different colours (arguments to plcol())
- Call to plwid(1) missing after plot2
- No fonts selected (based on user option)
- No "locate mode" (input from mouse/keyboard)
- No plot stored in file

x02c/x02f:
- Call to plwid(j) missing (do-loop 2)
- Order pladv/plcol

x03c/x03f:
- plcol/plcol0 - only difference

x04c/x04f:
- No subroutine plot1
- freql(i) calculated with 1.0 instead of -2.0
- Call to plwind with arguments 1.0, 6.0 instead of -2.0, 3.0
- Call to plptex (5.0 instead of 1.6)
- No switch on type (user option)
- No type=0 distinguished at the end

x05c/x05f:
- Constant 2048 in FORTRAN and 2047 in C
- Call to plhist uses this

x06c/x06f:
- Ordering comments/subroutine calls
- Difference in justification of number (10*i)

x07c/x07f:
- Call to plvpor with different numerical arguments
- Ordering of comments

x08c/x08f:
- Calls to plsurf3d missing
- C code distinguishes number of points in X and Y direction (different
  values too!)
- plinit() after initialisation of z array
- No distinction rosen/sombrero
- No call to plMinMax2dGrid() (or comments regarding this)

x09c/x09f:
- Little similarity between the two versions

x10c/x10f:
- No differences

x11c/x11f:
- Extremely little coherence between the two versions!!

x12c/x12f:
- FORTRAN version uses C convention for do-loop (1): do 1 i = 0,9

x13c/x13f:
- Bizarre diffences in use of reals and integers
- Different text strings used
- Possible bug in incrementing j (loop 2)

x14c:
- No FORTRAN equivalent

x15c:
- No FORTRAN equivalent

x16c/x16f:
- No likeness between the two versions

x17c-x21c:
- No FORTRAN equivalent


As you can see, some of the demos need a lot of work :(

Well, now the next step.

Regards,

Arjen

Hello all,

I have run into a bizarre problem under Windows that has an equivalent 
under Linux (but apparently not the same solution):

I need to set a field in the pls structure via a public routine that I
added 
to plcore.c - this is to get the interaction between my program and
PLplot 
correct (my program is supposed to manage the windows and any events, so 
under Linux I need to inform PLplot about the existing X connection and
under Windows I need to something similar).

On Linux:
I found that using plsc can be troublesome: there seems to be a double 
pointer plsc - in PLplot plsc points to a different structure than
within
the driver. I introduced a function "plP_setpxl2" and a function
"plP_setphy2"
to get around this.

On Windows:
I get the same type of problem again. The following fragment shows the
symptoms:

I added the following function (to plcore.c):
void
plsxwinpixmap(PLINT pixmap_id)
{
   plsc->pixmap_id = pixmap_id ;
}

This is called from my own program (not from within PLplot). At the end
of the routine:

pixmap_id = 117507625
plsc->pixmap_id = 0    !!!!
pls0.pixmap_id = 1175075625
plsc = 0x00a91c08 pls0
&pls0 = 0x00a91c08 pls0
(&pls0)->pixmap_id = 0 !!!!

Now can anyone shed a light here?

This goes beyond my comprehension, although I think it is a
quirk/bug/flaw
in the setup of the libraries (at least, that could be the case on
Linux):
I use PLplot as an archive/static library
the drivers are DLLs/shared libraries

The nett effect is: the pls structure does not appear updated once you 
get into the realm of the actual driver and it creates its own window
...

Regards,

Arjen

[ Moving this thread to plplot-devel. ]

* João Cardoso <jcard@...> [2003-10-06 16:41]:

> -The x21 c++ demos fails with "Floating point exception"; adding -mieee 
> to the gcc command line solves the problem. The -mieee gcc option is OS 
> dependent, and is set in sysloc.in (I think)

Yes, this is the relevant part of sysloc.in:

# ----------------------------------------------------------------------------
# CFLAGS to use ieee because of NaN issues in libcsironn and libcsirocsa
# ----------------------------------------------------------------------------

if test "$with_qhull" = "yes" -o "$with_csa" = "yes"; then
    case "$host_cpu" in
        i*86 )
            CFLAGS="$CFLAGS -mieee-fp"
        ;;
        alpha* )
            if test "$GCC" = yes; then
                CFLAGS="$CFLAGS -mieee"
            else
                CFLAGS="$CFLAGS -ieee"
            fi
        ;;
    esac

fi

I am puzzled, since you reported:

* João Cardoso <jcard@...> [2003-10-06 16:41]:

> Under a alphaev6-dec-osf4.0f, with gcc-3.0.3,
> 
> With a plain ./configure --prefix=...

From the code above, -mieee must be added to CFLAGS when $host_cpu starts
with alpha and gcc is in use.  What are we missing here?

-- 
Rafael

It looks like there is a problem with the plplot_libtool here. The C
examples are compiled with -mieee-fp defined, but the C++ examples
are not. I guess it must be because there is a different libtool rule
for C++ code and CFLAGS is not being passed through. I'm afraid I
don't know enough about libtool to know where the problem is though.
This occurs on my Debian machine as well, it's just that it doesn't
actually cause a crash there.

Andrew

On Wed, Oct 08, 2003 at 10:54:57AM +0200, Rafael Laboissiere wrote:
> [ Moving this thread to plplot-devel. ]
> 
> * Jo?o Cardoso <jcard@...> [2003-10-06 16:41]:
> 
> > -The x21 c++ demos fails with "Floating point exception"; adding -mieee 
> > to the gcc command line solves the problem. The -mieee gcc option is OS 
> > dependent, and is set in sysloc.in (I think)
> 
> Yes, this is the relevant part of sysloc.in:
> 
> # ----------------------------------------------------------------------------
> # CFLAGS to use ieee because of NaN issues in libcsironn and libcsirocsa
> # ----------------------------------------------------------------------------
> 
> if test "$with_qhull" = "yes" -o "$with_csa" = "yes"; then
>     case "$host_cpu" in
>         i*86 )
>             CFLAGS="$CFLAGS -mieee-fp"
>         ;;
>         alpha* )
>             if test "$GCC" = yes; then
>                 CFLAGS="$CFLAGS -mieee"
>             else
>                 CFLAGS="$CFLAGS -ieee"
>             fi
>         ;;
>     esac
> 
> fi
> 
> I am puzzled, since you reported:
> 
> * Jo?o Cardoso <jcard@...> [2003-10-06 16:41]:
> 
> > Under a alphaev6-dec-osf4.0f, with gcc-3.0.3,
> > 
> > With a plain ./configure --prefix=...
> 
> >From the code above, -mieee must be added to CFLAGS when $host_cpu starts
> with alpha and gcc is in use.  What are we missing here?
> 
> -- 
> Rafael
> 
> 
> -------------------------------------------------------
> This sf.net email is sponsored by:ThinkGeek
> Welcome to geek heaven.
> http://thinkgeek.com/sf
> _______________________________________________
> Plplot-devel mailing list
> Plplot-devel@...
> https://lists.sourceforge.net/lists/listinfo/plplot-devel

Arjen Markus writes:
 > Hello all,
 > 
 > I have run into a bizarre problem under Windows that has an equivalent 
 > under Linux (but apparently not the same solution):
 > 
 > I need to set a field in the pls structure via a public routine that I
 > added 
 > to plcore.c - this is to get the interaction between my program and
 > PLplot 
 > correct (my program is supposed to manage the windows and any events, so 
 > under Linux I need to inform PLplot about the existing X connection and
 > under Windows I need to something similar).
 > 
 > On Linux:
 > I found that using plsc can be troublesome: there seems to be a double 
 > pointer plsc - in PLplot plsc points to a different structure than
 > within
 > the driver. I introduced a function "plP_setpxl2" and a function
 > "plP_setphy2"
 > to get around this.

The drivers should never use "plsc", as "pls" is passed in directly.
I will fix this shortly.

-- 
Maurice LeBrun
Lightspeed Semiconductor Corp

On Mon, Oct 06, 2003 at 04:41:26PM +0100, Jo?o Cardoso wrote:
> On Sunday 05 October 2003 01:02, Rafael Laboissiere wrote:
> | A new CVS snapshot distribution tarball for PLplot is available at
> | the usual place:
> |
> |     http://people.debian.org/~rafael/plplot.html
> |
> | You will find below the changelog of the recent changes in CVS.
> |
> | Please, test and report.
> 
> 
> -After disabling it, a sucessefull make/make install (with only the C, 
> F77 and C++ bindings) make check (in the build tree) fails with
> 
>    x01.cc: In member function `void x01::plot1(int)':
>    x01.cc:276: `usleep' undeclared (first use this function)
> 
> -and also at
> 
>    x17.cc: In constructor `x17::x17(int, char**)':
>    x17.cc:148: `usleep' undeclared (first use this function)
> 
>   the corresponding C demos compile/work ok.
> 

This is a little strange. Can you see what the command line generated
by plplot_libtool is in both the C and C++ cases? Maybe this is the
same problem as the missing -mieee line because CFLAGS is not being
correctly used? It's only a guess.

Andrew

Andrew Ross writes:
 > On Mon, Oct 06, 2003 at 04:41:26PM +0100, Jo?o Cardoso wrote:
 > > On Sunday 05 October 2003 01:02, Rafael Laboissiere wrote:
 > > | A new CVS snapshot distribution tarball for PLplot is available at
 > > | the usual place:
 > > |
 > > |     http://people.debian.org/~rafael/plplot.html
 > > |
 > > | You will find below the changelog of the recent changes in CVS.
 > > |
 > > | Please, test and report.
 > > 
 > > 
 > > -After disabling it, a sucessefull make/make install (with only the C, 
 > > F77 and C++ bindings) make check (in the build tree) fails with
 > > 
 > >    x01.cc: In member function `void x01::plot1(int)':
 > >    x01.cc:276: `usleep' undeclared (first use this function)
 > > 
 > > -and also at
 > > 
 > >    x17.cc: In constructor `x17::x17(int, char**)':
 > >    x17.cc:148: `usleep' undeclared (first use this function)
 > > 
 > >   the corresponding C demos compile/work ok.
 > > 
 > 
 > This is a little strange. Can you see what the command line generated
 > by plplot_libtool is in both the C and C++ cases? Maybe this is the
 > same problem as the missing -mieee line because CFLAGS is not being
 > correctly used? It's only a guess.

I agree, looking at these I don't see what could be wrong -- they are
being handled just like the x01c.c and x17c.c demos.

-- 
Maurice LeBrun
Lightspeed Semiconductor Corp

On Wednesday 08 October 2003 09:54, Rafael Laboissiere wrote:
| [ Moving this thread to plplot-devel. ]
|
| * Jo=E3o Cardoso <jcard@...> [2003-10-06 16:41]:
| > -The x21 c++ demos fails with "Floating point exception"; adding
| > -mieee to the gcc command line solves the problem. The -mieee gcc
| > option is OS dependent, and is set in sysloc.in (I think)
|
| Yes, this is the relevant part of sysloc.in:
|
| #
| ---------------------------------------------------------------------
|------- # CFLAGS to use ieee because of NaN issues in libcsironn and
| libcsirocsa #
| ---------------------------------------------------------------------
|-------
|
| if test "$with_qhull" =3D "yes" -o "$with_csa" =3D "yes"; then
|     case "$host_cpu" in
|         i*86 )
|             CFLAGS=3D"$CFLAGS -mieee-fp"
|         ;;
|         alpha* )
|             if test "$GCC" =3D yes; then
|                 CFLAGS=3D"$CFLAGS -mieee"
|             else
|                 CFLAGS=3D"$CFLAGS -ieee"
|             fi
|         ;;
|     esac
|
| fi
|
| I am puzzled, since you reported:
|
| * Jo=E3o Cardoso <jcard@...> [2003-10-06 16:41]:
| > Under a alphaev6-dec-osf4.0f, with gcc-3.0.3,
| >
| > With a plain ./configure --prefix=3D...
|
| From the code above, -mieee must be added to CFLAGS when $host_cpu
| starts with alpha and gcc is in use.  What are we missing here?

Right now I can't check it, but perhaps it was the c++ demos, and=20
CXXFLAGS should be set. I will check it in a couple of hours.

Thanks,
Joao

On Wed, Oct 08, 2003 at 10:06:35AM +0100, Andrew Ross wrote:
> 
> It looks like there is a problem with the plplot_libtool here. The C
> examples are compiled with -mieee-fp defined, but the C++ examples
> are not. I guess it must be because there is a different libtool rule
> for C++ code and CFLAGS is not being passed through. I'm afraid I
> don't know enough about libtool to know where the problem is though.
> This occurs on my Debian machine as well, it's just that it doesn't
> actually cause a crash there.

OK. I've looked a little closer and I think we have two problems. It
actually appears to be a makefile and not a libtool issue.

1) When you run "make check" the makefile uses CXXFLAGS instead of
CFLAGS for a C++ compilation so we need to alter sysloc.in to add 
-mieee to CXXFLAGS (and possible in the future FFLAGS as well). 

2) When you install the examples and compile them using the
installed Makefile (Makefiles.example) then no flags are added at all
for any language. I guess we need to ensure that CFLAGS/CXXFLAGS are
imported into this makefile and used. Does anyone know how to go
about this?

Regards

Andrew

* Andrew Ross <andrewr@...> [2003-10-08 10:06]:

> It looks like there is a problem with the plplot_libtool here. The C
> examples are compiled with -mieee-fp defined, but the C++ examples
> are not. I guess it must be because there is a different libtool rule
> for C++ code and CFLAGS is not being passed through. 

Are you talking about examples/c and examples/c++?  AFAIK, plplot_libtool is
not responsible for passing the C and C++ flags when compiling objects, 
they must be explicitly given.  Indeed, I see here:

$ cd examples/c
$ make x01c
plplot_libtool --mode=link gcc x01c.c -I/usr/include/plplot -L/usr/lib
-lplplotd -o x01c
mkdir .libs
gcc x01c.c -I/usr/include/plplot -o x01c  -L/usr/lib /usr/lib/libplplotd.so
/usr/lib/libfreetype.so -lz /usr/lib/libcsirocsa.so /usr/lib/libcsironn.so
/usr/lib/libqhull.so -lm -ldl
$ cd ../c++
$ make x01cc
plplot_libtool --mode=link g++ x01cc.cc -I/usr/include/plplot -L/usr/lib
-lplplotcxxd -o x01cc
g++ x01cc.cc -I/usr/include/plplot -o x01cc  -L/usr/lib
/usr/lib/libplplotcxxd.so /usr/lib/libplplotd.so /usr/lib/libfreetype.so -lz
/usr/lib/libcsirocsa.so /usr/lib/libcsironn.so /usr/lib/libqhull.so -lm -ldl

There is no -mieee being passed, neither for compiling the C nor the C++
examples.  By the way, this flag is not necessary.  The only part of PLplot
where they are needed is for compiling the libcsironn and libcsirocsa
libraries, which only need a C compiler.  This is why the -mieee flag is not
added to the CXXFLAGS variable, but only to CFLAGS.

> I'm afraid I don't know enough about libtool to know where the problem is
> though. This occurs on my Debian machine as well, it's just that it doesn't
> actually cause a crash there.

Have you had crashes due to -mieee elsewhere than in Debian?

-- 
Rafael

Maurice LeBrun wrote:
> 
> Arjen Markus writes:
>  > Hello all,
>  >
>  > I have run into a bizarre problem under Windows that has an equivalent
>  > under Linux (but apparently not the same solution):
>  >
>  > I need to set a field in the pls structure via a public routine that I
>  > added
>  > to plcore.c - this is to get the interaction between my program and
>  > PLplot
>  > correct (my program is supposed to manage the windows and any events, so
>  > under Linux I need to inform PLplot about the existing X connection and
>  > under Windows I need to something similar).
>  >
>  > On Linux:
>  > I found that using plsc can be troublesome: there seems to be a double
>  > pointer plsc - in PLplot plsc points to a different structure than
>  > within
>  > the driver. I introduced a function "plP_setpxl2" and a function
>  > "plP_setphy2"
>  > to get around this.
> 
> The drivers should never use "plsc", as "pls" is passed in directly.
> I will fix this shortly.
> 

Note, the drivers use "plsc" via plP_setpxl() and the like. My version
takes pls as its argument and the rest is just like plP_setpxl().

Now, with a few modifications I have made progress (that is: drawing
happens in the window *I* have assigned, rather than in a new one). 
However, the scaling is awkward, as dev->xScale and dev->yScale 
are never set or set to odd values.

Looking in the code, I see that a bunch of routines are responsible
for setting them. (I first thought none of these were called, but 
that is not true). 

Okay, I will look into this and report back. 

Regards,

Arjen

On Wednesday 08 October 2003 12:45, Rafael Laboissiere wrote:
| * Andrew Ross <andrewr@...> [2003-10-08 10:06]:
| > It looks like there is a problem with the plplot_libtool here. The
| > C examples are compiled with -mieee-fp defined, but the C++
| > examples are not. I guess it must be because there is a different
| > libtool rule for C++ code and CFLAGS is not being passed through.
|
| Are you talking about examples/c and examples/c++?  AFAIK,
| plplot_libtool is not responsible for passing the C and C++ flags
| when compiling objects, they must be explicitly given.  Indeed, I see
| here:
|
| $ cd examples/c
| $ make x01c
| plplot_libtool --mode=link gcc x01c.c -I/usr/include/plplot
| -L/usr/lib -lplplotd -o x01c
| mkdir .libs
| gcc x01c.c -I/usr/include/plplot -o x01c  -L/usr/lib
| /usr/lib/libplplotd.so /usr/lib/libfreetype.so -lz
| /usr/lib/libcsirocsa.so /usr/lib/libcsironn.so /usr/lib/libqhull.so
| -lm -ldl
| $ cd ../c++
| $ make x01cc
| plplot_libtool --mode=link g++ x01cc.cc -I/usr/include/plplot
| -L/usr/lib -lplplotcxxd -o x01cc
| g++ x01cc.cc -I/usr/include/plplot -o x01cc  -L/usr/lib
| /usr/lib/libplplotcxxd.so /usr/lib/libplplotd.so
| /usr/lib/libfreetype.so -lz /usr/lib/libcsirocsa.so
| /usr/lib/libcsironn.so /usr/lib/libqhull.so -lm -ldl
|
| There is no -mieee being passed, neither for compiling the C nor the
| C++ examples.  By the way, this flag is not necessary.  The only part
| of PLplot where they are needed is for compiling the libcsironn and
| libcsirocsa libraries, which only need a C compiler.

Not quite, as x21c also uses nans, so it must be compiled with a nan 
aware compiler.

|  This is why the
| -mieee flag is not added to the CXXFLAGS variable, but only to
| CFLAGS.

But it should. I confirm that compiling x21.cc with -mieee solves the 
problem. In my sysloc.in I added CXXFLAGS, so I can commit it (latter) 
unless someone else does it first.

Joao

PS-I forgot to report that using the native C/F77 compilers under osf 
works OK. Even the C++ demos, compiled with g++, link/run against the 
native C compiled library (as it should, but nowadays we must not take 
anything for granted)

|
| > I'm afraid I don't know enough about libtool to know where the
| > problem is though. This occurs on my Debian machine as well, it's
| > just that it doesn't actually cause a crash there.
|
| Have you had crashes due to -mieee elsewhere than in Debian?

* João Cardoso <jcard@...> [2003-10-08 10:44]:

> Right now I can't check it, but perhaps it was the c++ demos, and 
> CXXFLAGS should be set. I will check it in a couple of hours.

After what you wrote, the bug only shows up in Alpha/GCC when compiling the
x21.cc example, which calls griddata. I apparently fixed the problem with my
last CVS commits.  The -mieee option is now added to CXXFLAGS when it is 
required for CFLAGS.  Besides that, the examples/{c,c++}/Makefile files set
now the CFLAGS and CXXFLAGS variables, which are used when invoking
plplot_libtool.

-- 
Rafael

On Wednesday 08 October 2003 10:12, Maurice LeBrun wrote:
| Andrew Ross writes:
|  > On Mon, Oct 06, 2003 at 04:41:26PM +0100, Jo?o Cardoso wrote:
|  > > On Sunday 05 October 2003 01:02, Rafael Laboissiere wrote:
|  > > | A new CVS snapshot distribution tarball for PLplot is
|  > > | available at the usual place:
|  > > |
|  > > |     http://people.debian.org/~rafael/plplot.html
|  > > |
|  > > | You will find below the changelog of the recent changes in
|  > > | CVS.
|  > > |
|  > > | Please, test and report.
|  > >
|  > > -After disabling it, a sucessefull make/make install (with only
|  > > the C, F77 and C++ bindings) make check (in the build tree)
|  > > fails with
|  > >
|  > >    x01.cc: In member function `void x01::plot1(int)':
|  > >    x01.cc:276: `usleep' undeclared (first use this function)
|  > >
|  > > -and also at
|  > >
|  > >    x17.cc: In constructor `x17::x17(int, char**)':
|  > >    x17.cc:148: `usleep' undeclared (first use this function)
|  > >
|  > >   the corresponding C demos compile/work ok.
|  >
|  > This is a little strange. Can you see what the command line
|  > generated by plplot_libtool is in both the C and C++ cases? Maybe
|  > this is the same problem as the missing -mieee line because CFLAGS
|  > is not being correctly used? It's only a guess.
|
| I agree, looking at these I don't see what could be wrong -- they are
| being handled just like the x01c.c and x17c.c demos.

There is something indeed strange.

plConfig.h defines HAVE_UNISTD_H, and unistd.h defines two flavors of 
usleep().

The fact is that if I add
    extern "C" int usleep(useconds_t useconds);
or
     extern "C" void usleep(useconds_t useconds);

to x01.cc it compiles and runs OK (with the -xor cmd line option).

The relevant unistd.h section:

#ifdef _XOPEN_SOURCE_EXTENDED
extern int usleep __((useconds_t ));    /* spec1170 requires a value to 
be returned */
#else

#ifndef _XOPEN_SOURCE
extern void usleep __((unsigned int));
#endif  /* _XOPEN_SOURCE */

#endif /* _XOPEN_SOURCE_EXTENDED */

If I compile with -E, no signal of usleep() appears.
If I compile with -D_XOPEN_SOURCE_EXTENDED, there is no problem.
What to do?

Joao

Hello,

I have a very crude and very temporary workaround for the problem with
xScale and yScale in plD_bop_win3() (win3.cpp). I simply set rect.bottom 
and rect.right to some value. 

Now at least the coordinates are in the right order.

But, I still do not get a picture. Is there any buffering going on that
any of you know about? Note: I circumvent the direct window handling 
in win3.cpp - maybe something is going on there that I do not get 
right yet.

Regards,

Arjen

A new CVS snapshot distribution tarball for PLplot is available at the usual
place:

    http://people.debian.org/~rafael/plplot.html

The tarball name is plplot-5.2.1.cvs.20031019.tar.gz.  It was generated in a
Debian unstable system with the following versions of the GNU autotools:

    Autoconf  2.57
    Automake  1.7.8
     Libtool  1.5.0a

You will find below the changelog of the recent changes in CVS (generated
with the command: cvs2cl -l "-d'2003-10-06<'").

Please, test and report.

-- 
Rafael


============== Changelog ==============

2003-10-17 15:16  rlaboiss

	* debian/changelog: Debian release 5.2.1-20

2003-10-17 14:16  rlaboiss

	* configure.ac: Put AC_CHECK_LIB(ltdl, ...) outside the
	  with_ltdlsystem conditional

	  For some mysterious reason, when the AC_CHECK_LIB was put inside
	  the conditional 'if test "$with_ltdlsystem" = no', configure was
	  failing with the error message: conditional "AMDEP" was never
	  defined.  This should be fixed now.

	  Please, test this change.

2003-10-16 13:22  rlaboiss

	* debian/: changelog, control, rules: Preparation for Debian
	  release 5.2.1-20

	  This will be uploaded to unstable only after version 5.2.1-19 had
	  hit the testing distribution.

	  From debian/changelog:

	      * Following a suggestion of Matthias Klose
	  <doko@...>, the
		libltdl library shipped with the upstream sources is not
	  compiled
		anymore as a convenience library:
		- Imported configure.ac and Makefile.am from upstream CVS
	  that
		  implement the confgiure option --with-ltdlsystem,
	  allowing the use
		  of the installed libltdl library.
		- debian/control: Build-depends on libltdl3-dev.
		- debian/rules: Call configure with option
	  --with-ltdlsystem.
	      * debian/control:
		- Fixed description of libplplot-dev package.
		- Bumped the build-dependency on debhelper to version
	  4.1.67, since we
		  are using dh_python (thanks again to Matthias Klose for
	  pointing
		  this out).

2003-10-16 13:13  rlaboiss

	* configure.ac: Cosmetic changes

	  Removed an obsolete comment about AM_PROG_LIBTOOL and added
	  AC_HELP_STRING in MY_ARG_WITH(pthreads, ...).

2003-10-16 12:43  rlaboiss

	* Makefile.am, configure.ac: Better logic for use of system's
	  libltdl

	  The decision about when to use the libltdl library is improved.
	  Now, the default is to always compiled the shipped libtool's
	  libltdl as a convenience library, unless the user explicitly give
	  the option --with-ltdlsystem to configure.  When this option is
	  given, confgiure checks if the libltdl library is actually
	  installed.  If not, it falls back to the convenience compilation
	  case.

	  Alan, if you are hearing this, could you please comment?

2003-10-16 03:06  jcard

	* src/plcore.c: plInBuildTree() is always needed, move out of
	  conditional ENABLE_DYNDRIVERS

2003-10-16 02:05  rlaboiss

	* Makefile.am, configure.ac: Allow use of installed libltdl

	  A Debian fellow pointed out to me that there is no need to
	  compile the libltdl support when building the PLplot Debian
	  packages, since there is a libltdl3 package against which
	  libplplot can be linked.

	  In order to be able to disable configuration/compilation of the
	  libltdl support, I did the following changes:

	      * In configure.ac, AC_LIBTOOL_CONVENIENCE and
	  AC_LIBTOOL_INSTALLABLE are
		called conditionally on the presence of an installed
	  libltdl (using
		AC_CHECK_LIB).

	      * In order to avoid config/build of the libltdl directory,
		AC_CONFIG_SUBDIRS(libltdl) is called conditionally as well.
	   Also, in
		Makefile.am, the libltdl directory is included
	  conditionally on the
		(newly created) AM_CONDITIONAL variable
	  enable_ltdl_convenience.
		[N.B.: I changed the code for setting the SUBDIRS variable
	  in
		Makefile.am using the automake append (+=) operator,
	  instead of using
		the auxiliary variable drivers_src_dirs as before.]

	  If the user has libltdl installed in the system but wants to use
	  the version shipped with PLplot, she can override the configure
	  check by giving the --enable-ltdl-convenience option (this needs
	  further tests, though).

	  The advantage of using the system's libltdl are:

	      * Configuration time is reduced, since the libltdl dir does
	  not need to
		be configured.
	      * Build time is also reduced, for a similar reason.
	      * The size of the resulting libplplot library is around 10%
	  smaller (at
		least in Linux).

2003-10-15 23:36  rlaboiss

	* pkgcfg/Makefile.am: Fixed problem with sed option -i

	  Option -i of the sed command is quite useful, because it allows
	  files to be edited in-place.	It is too bad that GNU sed older
	  than version 4 do not support this feature.  Since this should be
	  the case for general sed's, the use of option -i was replaced by
	  an ugly (but working) code using a temporary file.

2003-10-15 23:31  rlaboiss

	* pkgcfg/gen-pc-files.sh: Added some output messages

	  In the for loop, messages like "Generated plplot(...).pc" are now
	  echoed.  This will help to diagnose problems in the future.

	  Also, some small reformatting of the code was done.

2003-10-15 22:53  rlaboiss

	* pkgcfg/gen-pc-files.sh: Added missing double quote in sed command

	  This fixes the 'command substitution: unexpected EOF while
	  looking for matching "' bug.	Quite interesting, this error
	  message is only triggered when /bin/sh is bash.  Other Bourne
	  shells like ash and dash accept happily this "syntax error".

2003-10-11 19:48  rlaboiss

	* pkgcfg/: Makefile.am, README: Fixed parsing of -L... entries in
	  dependency_libs

	  The -L... entries were incorrectly parsed in the
	  install-data-hook rule and disappeared from the Libs lines in the
	  *.pc files.  The documentation README is updated to reflect this
	  fix.

2003-10-11 15:40  rlaboiss

	* configure.ac, pkgcfg/Makefile.am, pkgcfg/README,
	  pkgcfg/gen-pc-files.sh, pkgcfg/plplot-float.pc.in,
	  pkgcfg/plplot-master-pc.in, pkgcfg/plplot.pc.in: Total redesign
	  of the pkg-config support for PLplot

	  Read the pkgcfg/README file for details.

2003-10-11 10:20  rlaboiss

	* pkgcfg/: Makefile.am, plplot-float.pc.in, plplot.pc.in:
	  pkg-config gives now complete list of dependency libraries

	  This is a first try to have "pkg-config plplot --libs" working
	  correctly in a cross-platform way.  For that, the complete list
	  of dependency libraries flags (-l*) must be included into
	  plplot.pc and plplot-float.pc.

	  This information is already available in the
	  src/.libs/libplplot*.lai files.  These files are parsed in the
	  install-data-hook rule of pkgcfg/Makefile.am in a similar way the
	  plplot_libtool script does at run time.  However, this is done at
	  install time.

	  This is now the result I am having in Linux (must be also tested
	  in other platform):

	      $ pkg-config plplot --libs
	      -lplplotd -lfreetype -lz -lcsirocsa -lcsironn -lqhull -lm
	  -ldl

2003-10-10 20:57  airwin

	* scripts/plplot-config.in: Update and partial redesign of
	  plplot-config, a simplified wrapper for plplot_libtool if you
	  don't want to use the power of that script directly.

	  * fixed bugs where it failed for libtool-1.5 (a linker must be
	  specified), and also failed for --with-c++ (wrongly named
	  library).

	  * introduce --with-f77 and --with-tcl options

	  * non-cumulative library modes.  The idea here is you ordinarily
	  only tell plplot_libtool about libplplot, and it figures out the
	  correct link options for the local platform based on that.
	  However, there are some additional mutually exclusive options a
	  user can specify:

	  --with-c++, --with-f77, or --with-tcl.

	  If you specify any of those additional switches you tell
	  plplot_libtool only about the appropriate library (libplplotcxx
	  for --with-c++, libplplotf77 for --with-f77, and libplplottcltk
	  for --with-tcl), and it determines the correct link options for
	  your local platform.	The above switches should be mutually
	  exclusive, but my scripting skills aren't up to checking for that
	  (help, please!), but for now I settled for an increasing order of
	  precedence being in the above order (so anything lower on the
	  list means anything above is ignored).

2003-10-10 20:26  jcard

	* examples/: tcl/README.tcldemos, tk/README.tkdemos: Adjust to show
	  how to run the demos in the build tree.

2003-10-10 20:26  jcard

	* bindings/: tcl/pkgIndex.tcl.in, tk/pkgIndex.tcl.in,
	  tk-x-plat/pkgIndex.tcl.in: Fix to enable the demos to be run in
	  the build tree.

2003-10-10 15:17  rlaboiss

	* pkgcfg/: plplot-float.pc.in, plplot.pc.in: Fixed Libs and Cflags
	  entries

	  Added "/plplot" to the includedir variable, such that there is no
	  need to explicitly #include <plplot/...> in user programs (this
	  is the case for the C and C++ examples shiped with PLplot).

	  Also, removed -DPL_DOUBLE from Cflags (this was producing
	  compilation warnings, since PL_DOUBLE is already defined in
	  plConfig.h) and removed @FREETYPELIBCMD@ from Libs.  All the
	  dependecy libraries of libplplot are guessed by the default
	  linker of Linux.  I am not sure this will work like this in other
	  Unices.  More work is needed here.

	  At any rate, compilation of the C examples work correctly now
	  with:

	      gcc `pkg-config plplot --cflags` x01c.c -o x01c `pkg-config
	  plplot --libs`

2003-10-10 10:45  rlaboiss

	* include/plConfig.h.in: Removed #undef *_DIR

	  The *_DIR #defines are only needed in plConfg.h, which is
	  included in user programs.  These #defines are only needed at
	  build time by the core functions of the PLplot library.  Since
	  they are already included in config.h (thanks to the
	  AC_DEFINE_UNQUOTED calls in configure.ac) and this later file is
	  included by plConfig.h at build time, then they can be deleted
	  from plConfig.h.in.

2003-10-09 23:08  airwin

	* bindings/java/Makefile.am, bindings/java/README.javaAPI,
	  examples/java/Makefile.am, examples/java/README.javademos,
	  test/test_java.sh: Change java install procedure so that
	  CLASSPATH is not required.  This solves the configuration problem
	  we were discussing.

	  Also drop all mention of CLASSPATH in the documentation except
	  where relevant; i.e., only at the time when you execute the
	  installed compiled java examples.

2003-10-09 23:02  jcard

	* configure.ac, include/plConfig.h.in: Create and set BUILD_DIR.

2003-10-09 23:00  jcard

	* src/plcore.c, src/plctrl.c, bindings/tcl/tclAPI.c: If the current
	  directory is in the build tree, change files search order so as
	  to load them locally.

	  The modifications guarantee that files/drivers/scripts are loaded
	  from the build tree if the current directory is somewhere in the
	  build tree.

	  The objective is that one should be able to change/test/evaluate
	  PLplot without the necessity of installing it; recently this was
	  needed (and for some front ends it still is), requiring the user
	  to have a separate directory to test/change/evaluate PLplot,
	  which means two steps of configure/make, one with the test
	  --prefix and the other with the final one, which is confusing and
	  error prone. Also, for developpers, the make/make install step
	  turns the development cycle sloww and boring.

	  All the compiled front-ends, C, C++ and F77, was tested and work
	  OK with all drivers.	The Tcl and Tk front-ends also work OK, but
	  the tk loadable extensions, Pltk and Plplotter, as well as the
	  tcl extension Pltcl, still need adjustments in the pkgIndex.tcl
	  script.

	  The Octave front-end also works OK.

	  Python and Java was not yet tested.

	  This is not yet finish, as some problems might yet appear;
	  however, care has been taken to guarantee that the old behavior
	  still works OK.

	  tclAPI.c:    PLbasicInit(): search the init script in the build
	  tree	  pls_auto_path(): add bindings/tk to auto_path

	  plcore.c:    add plInBuildTree() that check if the current
	  directory in within the build source tree    plGetDriverDir(): if
	  in the build tree, load the drivers locally

	  plctrl.c:	plFindCommand(): if in the build tree, search
	  locally     plLibOpenPdfstrm(): if in the build tree, search
	  locally

2003-10-08 19:19  rlaboiss

	* sysloc.in: Reintroduced deleted AC_MSG_RESULT in Nan awareness
	  test

	  Also, the logic of the test is fixed now.  Strangely, the last
	  commit made AC_RUN_IFELSE have only two arguments, instead of
	  three.  Besides that, the test must return 1 in case of failure
	  and not 0 as was the case.  I did minimal tests, but evrything
	  seems to work now.

2003-10-08 18:37  jcard

	* sysloc.in: Reformating of "Check for NaN awareness" test, to
	  avoid using another variable.

2003-10-08 17:22  rlaboiss

	* sysloc.in: Surrounded NaN awareness test with
	  AC_MSG_CHECKING/RESULT

	  Configure shows now: "checking for NaN awareness in C compiler...
	  yes".  Please change the message text if it is not appropriate.

2003-10-08 16:51  jcard

	* sysloc.in: Make test on simple NaN operations

	  libscirocsa/nn can only be compiled if the OS/C compiler is NaN
	  aware.  If this test fails they are disabled.  For this test to
	  suceed, the correct CFLAGS must be passed to the C compiler.

2003-10-08 16:15  jcard

	* examples/c/x21c.c: The #, as in #define, should be in the first
	  column, as Compaq C complains.

2003-10-08 15:51  rlaboiss

	* examples/: c/Makefile.examples.in, c++/Makefile.examples.in:
	  Added CFLAGS and CXXFLAGS

	  The CFLAGS and CXXFLAGS variables have been added to the calls of
	  plplot_libtool in examples/c/Makefile and examples/c++/Makefile,
	  respectively.  They get their values from the AC_SUBST variables
	  of same name.

2003-10-08 15:48  rlaboiss

	* sysloc.in: Added -mieee to CXXFLAGS when neeeded.

	  Setting the CXXFLAGS variable seems to be necessary for compiling
	  the C++ examples on the Alpha architecture, even when the -mieee
	  option is only needed for compiling the libcsirocsa and
	  libcsironn libraries.

2003-10-08 11:17  mlebrun

	* drivers/: tk.c, tkwin.c, xwin.c: Removed references to plsc --
	  drivers should use pls, which is passed in.  Also minor
	  formatting fixes.

2003-10-07 18:23  rlaboiss

	* configure.ac: Avoid abortion of configure when python is not
	  found

	  Redefine AC_MSG_ERROR as AC_MSG_WARN around the call to
	  AM_PATH_PYTHON, such that if python is not found, configure will
	  not stop.

A new CVS snapshot distribution tarball for PLplot is available at the usual
place:

    http://people.debian.org/~rafael/plplot.html

The tarball name is plplot-5.2.1.cvs.20031027.tar.gz.  The changes in
relation to the last CVS tarball (released one week ago) are quite minimal.

The tarball was generated in a Debian unstable system with the following
versions of the GNU autotools:

    Autoconf  2.57
    Automake  1.7.8
     Libtool  1.5.0a

You will find below the changelog of the recent changes in CVS (generated
with the command: cvs2cl -l "-d'2003-10-17<'").

Please, test and report.

Notice that this tarball will be used in the forthcoming release of the
Debian packages.  This means that I will abandon the Debian series 5.2.1-*
in favor of the 5.2.1.cvs.*-* series.  In concrete terms, this means that
the Debian packages in the testing distribution will be actually the
"release candidates" for the next version of PLplot (probably 5.2.2).  The
presence of the CVS snapshot version inDebian will help us in the testing
cycle of the next release of PLplot.

-- 
Rafael


============== Changelog ==============

2003-10-27 08:48  rlaboiss

	* debian/rules: Fixed "--with-pthread" -> "--with-pthreads" in call
	  to configure.

2003-10-27 08:46  rlaboiss

	* scripts/make-cvs-tarball.sh: Made make-cvs-tarball script more
	  robust to premature exit.

	  Added a cleanup function that is called for the trapped signals
	  0, HUP, QUIT, INT, TERM, and PIPE.  This insures that the
	  temporary directory $CVSTMPDIR is removed at exit.  Also, a
	  caveat note is added at the comments on the top of the file.

2003-10-26 16:07  rlaboiss

	* bootstrap.sh: Call aclocal in libltdl directory.

	  This insures that the aclocal.m4 files in both the topdir and the
	  libltdl directory are generated with by same version of Automake
	  and no errors regarding incompatible calls of AM_AUTOMAKE_VERSION
	  will happen.

2003-10-26 05:37  airwin

	* examples/f77/x05f.fm4: AWI for Arjen Markus plus significant
	  changes of my own as well.  Update fifth standard fortran example
	  so it produces results consistent with results for other language
	  interfaces.  diff of postscript results showed no difference with
	  c result for example.

2003-10-26 05:23  airwin

	* examples/f77/x03f.fm4: AWI for Arjen Markus.	Minore update of
	  third standard fortran example so it avoids deprecated plcol API
	  and replaces by plcol0 instead.  More work required, however,
	  before it will produce results consistent with results for other
	  language interfaces.

2003-10-26 05:17  airwin

	* examples/f77/x02f.fm4: AWI for Arjen Markus.	Update second
	  standard fortran example so it produces results consistent with
	  results for other language interfaces.  diff of postscript
	  results showed no difference with c result for example.

2003-10-26 03:55  airwin

	* examples/f77/: Makefile.am, x17f.fm4, x18f.fm4, x19f.fm4: Initial
	  commit of contributed standard examples from Arjen Markus with
	  some changes by AWI to allow them to compile.  x17f and x19f
	  don't link (so must use, e.g, make -k check for now to compile
	  rest of examples) because of missing fortran API. x18f actually
	  links and gives almost the correct standard results for example
	  18, but "almost" means there is probably an API problem to deal
	  with there.

2003-10-26 02:49  airwin

	* examples/f77/: Makefile.am, fmacs.m4, x01f.fm4, x02f.fm4,
	  x03f.fm4, x04f.fm4, x05f.fm4, x06f.fm4, x07f.fm4, x08f.fm4,
	  x09f.fm4, x10f.fm4, x11f.fm4, x12f.fm4, x13f.fm4, x16f.fm4: Drop
	  single-precision support for examples for now.  I intend to
	  re-instate it later with a sed script.  The *.fm4 are now a
	  misnomer (probably the name should be changed to *.fdouble or
	  something, but I have to think about it) since they are no longer
	  processed by m4.  Instead, these files are filled by straight
	  double-precision fortran (with no embedded m4 commands) and
	  simply copied to *.f.

2003-10-24 20:31  jcard

	* configure.ac: Disable dynamic drivers if enable_shared != yes

	  If the user don't want (or the system is not capable of) to build
	  shared libraries, then disable dynamic drivers, as they are
	  dynamically loaded objects.

	  This must to be done after AM_PROG_LIBTOOL.  There is a test
	  if test "$enable_dyndrivers" = "yes" above that should, if
	  possible, be put after this test.

2003-10-23 11:48  rlaboiss

	* debian/: changelog, rules: Debian release 5.2.1-21

	  From the debian/changelog file:

	    * sysloc.in: Applied patch from upstream CVS that assigns an
	  unversioned
	      path to OCTAVE_OCT_DIR.
	    * debian/rules:
	      - Loosen the restriction on the Octave version dependency,
	  now that
		OCTAVE_OCT_DIR is unversioned.
	      - Configure with --with-pthreads, such that the xwin driver
	  refreshes
		the plot when the window is moved or re-exposed.

2003-10-23 09:17  rlaboiss

	* sysloc.in: Unversioned path for Octave oct dir

	  The OCTAVE_OCT_DIR variable is not versioned anymore (i.e. it
	  does not contain the Octave version 2.*.* in it).  This parallels
	  what is done for the OCT_M_DIR variable.  OCTAVE_OCT_DIR takes it
	  value from the octave_config_info structure, being either
	  localoctfiledir (for Octave version 2.1.*) or the first component
	  of localoctfilepath (version 2.0.*).

	  The advantage of this change is that the Octave bindings are not
	  tied to a specific version of Octave, and will continue to be
	  visible if the user upgrades Octave.	Remember though that there
	  is no guarantee of backward compatibility of the *.oct API within
	  the unstable series of Octave (2.1.*).

2003-10-21 12:22  rlaboiss

	* drivers/gnome.c: Use HAVE_PTHREAD instead of USE_THREADS

	  This means that pthreads support will be activated in the gnome
	  driver whenever the --with-pthreads option is given to configure.

2003-10-17 15:16  rlaboiss

	* debian/changelog: Debian release 5.2.1-20

2003-10-17 14:16  rlaboiss

	* configure.ac: Put AC_CHECK_LIB(ltdl, ...) outside the
	  with_ltdlsystem conditional

	  For some mysterious reason, when the AC_CHECK_LIB was put inside
	  the conditional 'if test "$with_ltdlsystem" = no', configure was
	  failing with the error message: conditional "AMDEP" was never
	  defined.  This should be fixed now.

	  Please, test this change.

Rafael Laboissiere wrote:
> 
> A new CVS snapshot distribution tarball for PLplot is available at the usual
> place:
> 
>     http://people.debian.org/~rafael/plplot.html
> 
> The tarball name is plplot-5.2.1.cvs.20031027.tar.gz.  The changes in
> relation to the last CVS tarball (released one week ago) are quite minimal.
> 

I took this snapshot and created a whole new installation. But when I
run the
C and FORTRAN examples with xwin as the driver, I get the message:

X Error of failed request:  BadMatch (invalid parameter attributes)
  Major opcode of failed request:  1 (X_CreateWindow)
  Serial number of failed request:  10
  Current serial number in output stream:  14

Any idea why? (The jpeg driver works without problems and the xterm
driver
too)

Regards,

Arjen

* Arjen Markus <arjen.markus@...> [2003-10-27 14:06]:

> I took this snapshot and created a whole new installation. But when I
> run the
> C and FORTRAN examples with xwin as the driver, I get the message:
> 
> X Error of failed request:  BadMatch (invalid parameter attributes)
>   Major opcode of failed request:  1 (X_CreateWindow)
>   Serial number of failed request:  10
>   Current serial number in output stream:  14

Very strange, indeed.  Have any previous tarball worked for you? Could you
give more informations on your system and also on your configuration
options?

-- 
Rafael

Rafael Laboissiere wrote:
> 
> * Arjen Markus <arjen.markus@...> [2003-10-27 14:06]:
> 
> > I took this snapshot and created a whole new installation. But when I
> > run the
> > C and FORTRAN examples with xwin as the driver, I get the message:
> >
> > X Error of failed request:  BadMatch (invalid parameter attributes)
> >   Major opcode of failed request:  1 (X_CreateWindow)
> >   Serial number of failed request:  10
> >   Current serial number in output stream:  14
> 
> Very strange, indeed.  Have any previous tarball worked for you? Could you
> give more informations on your system and also on your configuration
> options?
> 
> --
> Rafael

I do not quite remember - I took the official release of PLplot 5.2.1 
when I started with this and I am not sure whether the examples worked
correctly with the X Window driver (I messed around with the
installation
so I can not check this anymore).

What I do know is that in order to make it work in my own application
I had to fiddle a bit with the X Window driver. It turned out that at 
some deep level the plsc stream was doubly defined - it took me several
hours to find this out and it still frightens me at night :D

The changes I had to make were:
plP_setpxl() and plP_setphy() take an argument pls (instead implicitly
using plsc)

I will try to see if this is a solution now too.

Well, here is the summary:

Configure results:

command:        ./configure
--prefix=/p/delft3d/users/markus/plplot-new/plplot-5.2.1
system:         i686-pc-linux-gnu
have_x:         yes
prefix:         /p/delft3d/users/markus/plplot-new/plplot-5.2.1
CC:             gcc
CXX:            g++
F77:            g77
LIB_TAG:        d
devices:         dg300 png jpeg hp7470 hp7580 lj_hpgl imp ljii ljiip mem
ntk null pbm plmeta ps psc pstex xterm tek4010 tek4107 mskermit
versaterm vlt conex tek4010f tek4107f tk xfig xwin

Available device drivers:
static:
dynamic:         dg300.la gd.la hpgl.la impress.la ljii.la ljiip.la
mem.la ntk.la null.la pbm.la plmeta.la ps.la pstex.la tek.la tk.la
xfig.la xwin.la

Compilation options:
with_debug:     no              with_opt:       yes
with_warn:      no              with_profile:   no

Library options:
enable_shared:  yes             enable_static:  yes
with_rpath:     yes             with_double:    yes

Optional libraries:
with_qhull:     no              with_csa:       yes
with_freetype:  yes             with_pthreads:  no


Language Bindings:
enable_tcl:     yes             enable_itcl:    yes
enable_cxx:     yes             enable_f77:     yes
enable_java:    no              enable_python:  no
enable_octave:  yes


Here is the output from uname -a:
Linux hydrax 2.4.20 #10 SMP Wed May 14 17:26:48 CEST 2003 i686 unknown

glibc:
/lib/libc-2.2.5.so

gcc:
gcc version 2.96 20000731 (Red Hat Linux 7.3 2.96-113)


Regards,

Arjen

* Arjen Markus <arjen.markus@...> [2003-10-28 08:54]:

> What I do know is that in order to make it work in my own application
> I had to fiddle a bit with the X Window driver. It turned out that at 
> some deep level the plsc stream was doubly defined - it took me several
> hours to find this out and it still frightens me at night :D
> 
> The changes I had to make were:
> plP_setpxl() and plP_setphy() take an argument pls (instead implicitly
> using plsc)
> 
> I will try to see if this is a solution now too.

If this is a real bug, please try to find the fix again and contribute it.

Thanks,

-- 
Rafael

Rafael Laboissiere wrote:
> 
> * Arjen Markus <arjen.markus@...> [2003-10-28 08:54]:
> 
> > What I do know is that in order to make it work in my own application
> > I had to fiddle a bit with the X Window driver. It turned out that at
> > some deep level the plsc stream was doubly defined - it took me several
> > hours to find this out and it still frightens me at night :D
> >
> > The changes I had to make were:
> > plP_setpxl() and plP_setphy() take an argument pls (instead implicitly
> > using plsc)
> >
> > I will try to see if this is a solution now too.
> 
> If this is a real bug, please try to find the fix again and contribute it.
> 
> Thanks,
> 
> --
> Rafael

Hm, I tried to implement my original changes - no effect. :( It happens
way before that.

After some debugging sessions, I think the problem is that the depth
for the main window is not correct. 

In the debugger I see a depth of 8 (in xwd) whereas XGetGeometry returns
a depth of 24! 

Oh, one other thing: header in InitMain is a string of 80 characters,
right now the full name of my program is used and it turns out to be 72
characters ... With one more directory level the buffer would overflow, 
giving even more mysterious errors. 

Regards,

Arjen

Arjen Markus wrote:
> 
> Rafael Laboissiere wrote:
> >
> > * Arjen Markus <arjen.markus@...> [2003-10-28 08:54]:
> >
> > > What I do know is that in order to make it work in my own application
> > > I had to fiddle a bit with the X Window driver. It turned out that at
> > > some deep level the plsc stream was doubly defined - it took me several
> > > hours to find this out and it still frightens me at night :D
> > >
> > > The changes I had to make were:
> > > plP_setpxl() and plP_setphy() take an argument pls (instead implicitly
> > > using plsc)
> > >
> > > I will try to see if this is a solution now too.
> >
> > If this is a real bug, please try to find the fix again and contribute it.
> >
> > Thanks,
> >
> > --
> > Rafael
> 
> Hm, I tried to implement my original changes - no effect. :( It happens
> way before that.
> 
> After some debugging sessions, I think the problem is that the depth
> for the main window is not correct.
> 
> In the debugger I see a depth of 8 (in xwd) whereas XGetGeometry returns
> a depth of 24!
> 

Yes, that was it: I used the option -drvopt defvis and all worked well!

Hm, I think it would be wise to turn that option on by default!

Regards,

Arjen

A new CVS snapshot distribution tarball for PLplot is available at the usual
place:

    http://people.debian.org/~rafael/plplot.html

The tarball name is plplot-5.2.1.cvs.20031030.tar.gz.  The main change in
this tarball in respect to the previous one (20031027) is the SOVERSION in
configure.ac.  The libraries will have (in Linux) the soversion 9 now,
instead of 5 as previously.  The Debian packages will be renamed as a
consequence of this.

The tarball was generated in a Debian unstable system with the following
versions of the GNU autotools:

    Autoconf  2.57
    Automake  1.7.8
     Libtool  1.5.0a

You will find below the changelog of the recent changes in CVS (generated
with the command: cvs2cl -l "-d'2003-10-27<'").

Please, test and report.  We are heading towards the next release of PLplot,
and this tarball can be considered as a "pre-candidate-release" one.

-- 
Rafael


============== Changelog ==============

2003-10-29 22:30  rlaboiss

	* debian/: changelog, control, libplplot9.README.Debian,
	  plplot-gd.files, plplot-gnome.files, plplot-xwin.files,
	  plplot9-driver-gd.files, plplot9-driver-gnome.files,
	  plplot9-driver-xwin.files, rules: Renaming of the driver module
	  Debian packages.

	  The naming changes are:

	      plplot-gd    -> plplot9-driver-gd
	      plplot-xwin  -> plplot9-driver-xwin
	      plplot-gnome -> plplot9-driver-gnome

2003-10-29 20:40  jcard

	* examples/tk/xtk01.c, include/plplotP.h, src/pdfutils.c,
	  src/plbox.c, src/plcore.c, src/plfreetype.c, src/plot3d.c,
	  utils/pltcl.c, bindings/tk/Pltk_Init.c, bindings/tk/plframe.c,
	  bindings/tk/plserver.c, bindings/tk-x-plat/Plplotter_Init.c,
	  bindings/tk-x-plat/plplotter.c, drivers/gd.c,
	  drivers/get-drv-info.c, drivers/tk.c, drivers/tkwin.c,
	  drivers/xwin.c, examples/tk/xtk02.c, examples/tk/xtk04.c,
	  bindings/tcl/matrixInit.c, bindings/tcl/tclAPI.c,
	  bindings/tcl/tclMain.c, bindings/tcl/tclMatrix.c,
	  bindings/tk/tcpip.c, bindings/tk/tkMain.c, bindings/tk/tkshell.c:
	  Mostly cosmetic changes that enable plplot to compiled with
	  (almost) no warnings, even with gcc -Wall.

	  Most changes are just casts, and most of them are tcl/tk related.
	  For tcl/tk-8.4, no warnings occurs.  Also tested with tcl/tk-8.3,
	  where some warnings remain.  There are no
	  java/f77/cxx/python/octave changes.

2003-10-29 20:19  rlaboiss

	* debian/: changelog, control, libplplot-dev.files,
	  libplplot9.README.Debian, libplplot9.files, octave-plplot.files,
	  plplot-doc.doc-base, plplot-doc.files, plplot-tcl-dev.files,
	  python-plplot.files, rules: Preparation for the next Debian
	  release.

	  From the debian/changelog file:

	    * NOT YET RELEASED!
	    * Preparation for the new upstream version, which changed the
	  soversion
	      of the PLplot library.  Changed all the references to
	  libplplot5 to
	      libplplot9.  Affected files:
	      - control
	      - libplplot-dev.files
	      - octave-plplot.files
	      - plplot-doc.doc-base
	      - plplot-doc.files
	      - plplot-tcl-dev.files
	      - python-plplot.files
	      - rules
	      - libplplot9.README.Debian (new)
	      - libplplot9.files (new)

2003-10-29 18:24  airwin

	* examples/f77/Makefile.examples.in: Add examples 15, 17, 18, and
	  19.

	  Note 17 and 19 don't link properly yet (missing fortran API) so
	  the installed examples will have to be built using make -k.

2003-10-29 17:14  airwin

	* examples/f77/: x01f.fm4, x04f.fm4, x06f.fm4, x07f.fm4, x08f.fm4,
	  x09f.fm4, x11f.fm4, x12f.fm4, x13f.fm4, x16f.fm4: AWI for Arjen
	  Markus <arjen.markus@...>

	  White space changes.

2003-10-29 17:14  airwin

	* examples/f77/Makefile.am: Add new example 15.

2003-10-29 17:13  airwin

	* examples/f77/x15f.fm4: AWI for Arjen Markus
	  <arjen.markus@...>.

	  Initial commit.  This compiles and executes, but there is a
	  problem with the colours which needs debugging.

2003-10-29 15:13  rlaboiss

	* debian/: changelog, rules: Debian release
	  plplot-5.2.1.cvs.20031027-3.

	  From the debian/changelog file:

	    * debian/rules:
	      - Added symlinks /usr/lib/lib{nn,csa}.so.0, such that this
	  release of
		the libplplot5 package does not break binary compatiblity
	  with previous
		released versions of PLplot (closes: #217895).	This is an
	  interim
		solution.  Probably, the upstream authors will have to
	  change the
		soversion or revert the changing of names of the
	  libscsiro{nn,csa}
		libraries in order to get the problem cleanly fixed.
	      - Call configure with --disable-java, since enable_java has
	  been made
		the default in the upstream configure.ac.

2003-10-29 14:46  rlaboiss

	* debian/TODO: Marked all tasks as "DONE".

	  I just realized that this list of ToDo tasks is finished now.  I
	  am keeping this file here out of historical interest.

2003-10-29 14:38  rlaboiss

	* configure.ac: Upgraded SOVERSION to 9:0:0.

	  This is necessary since the recent changes in the naming of the
	  libraries libnn -> libcsironn and libcsa -> libcsirocsa have
	  broke backward binary compatibility for the PLplot library.

2003-10-29 14:10  rlaboiss

	* examples/c/lena.pgm: Dummy commit to get the sticky -kb option
	  (hopefully).

	  No changes have been made to this file.

2003-10-29 13:47  rlaboiss

	* debian/control: Changed the build-dependency for GD library.

	  Instead of using the "transitional" package libgd2-dev, use the
	  explicit dependencies to the real packages (libgd2-noxpm-dev |
	  libgd2-xpm-dev).

2003-10-28 17:16  vincentdarley

	* sys/win-tk/: makePlplotStarkit.tcl, makefile.vc, testPlplot.tcl:
	  win-tk build now works again

2003-10-27 11:54  rlaboiss

	* debian/: changelog, libplplot-dev.files,
	  libplplot5.README.Debian: Debian release 5.2.1.cvs.20031027-2.

	  From the debian/changelog file:

	    * debian/libplplot-dev.files: Added README.pkg-config file.
	    * debian/libplplot5.README.Debian: Fixed typos.

2003-10-27 11:35  rlaboiss

	* debian/: changelog, control, libplplot5.README.Debian, rules:
	  Debian release 5.2.1.cvs.20031027-1.

	  From the debian/changelog file:

	    * New CVS snapshot upstream release.  This release is actually
	  quite
	      close to the latest Debian 5.2.1-21, since I have been
	  tracking CVS
	      developments since the upstream release 5.2.1.  One of
	  biggest
	      advantages of this move is the enourmous reduction in the
	  size of the
	      diff.gz file:

		du -b plplot*diff*
		2686036 ../plplot_5.2.1-21.diff.gz
		10198	../plplot_5.2.1.cvs.20031027-1.diff.gz

	      This huge difference is due to the patched documentation file
	  and the
	      necessity of an uuencoded file for the Debian packages.

	    * debian/rules: Removed all manipulations (targets "build" and
	      "clean") related to the plplot-5.2.1-new-doc.tar.gz.uu file,
	  which is
	      not included in the source package anymore.  Also, added flag
	  -k to
	      make distclean.

	    * debian/control: Removed build-dependency on sharutils.

	    * debian/libplplot5.README.Debian: Added note about pkg-config
	  support.

2003-10-27 11:18  rlaboiss

	* bindings/tk/Makefile.am: Added tclIndex to the CLEANFILES list.

2003-10-27 11:17  rlaboiss

	* pkgcfg/Makefile.am: Install README file in
	  $(docdir)/README.pkg-config.

A new CVS snapshot distribution tarball for PLplot is available at the usual
place:

    http://people.debian.org/~rafael/plplot.html

The tarball name is plplot-5.2.1.cvs.20031226.tar.gz. It was generated in a
Debian unstable system with the following versions of the GNU autotools:

    Autoconf  2.58
    Automake  1.7.9
     Libtool  1.5.0a

Several changes and bug fixes since the last cvs distribution tarball
(tagged 20031030) are included, but I think that the two most important
improvements are the extension of the Fortran bindings API (done by Alan),
and the inclusion of demos for the Perl/PDL bindings.

You will find below the changelog of the recent changes in CVS, generated
with the cvsreport utility.

Please, test and report.  As with the previous ones, this tarball can be
considered as a "pre-candidate-release" for the next official version of
PLplot.  I am looking forward to strengthening our occupation in different
niches like Fortran and Perl/PDL bindings (I hope to catch up with the
upcoming PDL 2.4.1 release), as well as among users of Mac OS X.

-- 
Rafael


============== ChangeLog ==============

Commit from rlaboiss (2003-12-26 17:17 CET)
--------------------

Preparation for Debian release 5.2.1.cvs.20031226-1

  plplot  debian/changelog  1.55


Commit from rlaboiss (2003-12-26 17:15 CET)
--------------------

Fixed typo in regular expression for parsing AC_INIT command.

  plplot  debian/get-upstream-version.pl  1.3


Commit from rlaboiss (2003-12-26 17:08 CET)
--------------------

Disabled inclusion of drivers/*.rc in the distribution tarball

Revert the previous changes regarding includion of the drivers/*.rc files
into the tarball, because a .rc file depend on the .la file, which gets
built with "make" by the user.  The result is that the *.rc and get-drv-info
are built anyway.

I will think later on a better way to include the *.rc files in the tarball.

  plplot  drivers/Makefile.am          1.33
  plplot  scripts/make-cvs-tarball.sh  1.10


Commit from rlaboiss (2003-12-26 16:06 CET)
--------------------

Modernized some Autotools constructs

The following old-style calls in configure.ac:

    AC_INIT(src/plcore.c)
    AM_INIT_AUTOMAKE(plplot, 5.2.1)

have been modernized to:

    AC_INIT(plplot, 5.2.1, plplot-devel@...)
    AC_CONFIG_SRCDIR(src/plcore.c)
    AM_INIT_AUTOMAKE

(see "info autoconf" and "info automake" for details).

The three scripts that parsed configure.ac and relied on the old-style call
of AM_INIT_AUTOMAKE have been changed accordingly.

  plplot  configure.ac                    1.118
  plplot  bootstrap.sh                    1.26
  plplot  debian/get-upstream-version.pl  1.2
  plplot  scripts/check-tarball-cvs.sh    1.3


Commit from rlaboiss (2003-12-26 15:31 CET)
--------------------

Disabled generation of reconfig script by configure

As per our discussion in plplot-devel in September 2003 (and since Maurice
agreed), the section in configure.ac responsible for the generation of the
reconfig script is removed.  The command "./config.status --recheck" has the
same effect (i.e. calling configure with the previously specified options).

In order to keep the functionality for the fingers too used to type
"./reconfig", a one-line shell script called reconfig has been added to the
repository.

Also, added reconfig to the EXTRA_DIST variable in the top-dir Makefile.am,
such that the script gets included into the distribution tarball.

  plplot  configure.ac  1.117
+ plplot  reconfig      1.1
  plplot  Makefile.am   1.30


Commit from rlaboiss (2003-12-26 15:14 CET)
--------------------

Include most drivers/*.rc files in the distribution tarball

The goal of this change is to avoid the building of the drivers/*.rc files
by the users by including them into the distribution tarball.  This also
avoids the building of the get-drv-info utility when running "make" from the
tarball.

The price paid for this improvement on the user's side is that "make" is
needed before "make dist" (see scripts/make-cvs-tarball.sh). This happens
because get-drv-info needs libplplot, which needs libcsirocsa and libcsironn
in turn.  Also, libltdl needs to be built.

Since this change only affects the tarball construction and since the price
paid is too small compared with the gain obtained, I decided to implement
the new behavior. If anybody objects, please let us discuss in plplot-devel.

  plplot  scripts/make-cvs-tarball.sh  1.9
  plplot  drivers/Makefile.am          1.32


Commit from rlaboiss (2003-12-26 12:58 CET)
--------------------

Fixed debug and error messages.

In function plOpenFile, the message "Opened <filename>" is sent now through
pldebug.  This means that it will not spuriously appear on stderr anymore,
but will be issued when option -debug is specified.  Also, the message
"Can't open <filename>" is sent now to stderr instead of stdout.

In function plLibOpenPdfstr, two calls to pldebug are fixed now (called with
two argments instead of three).

  plplot  src/plctrl.c  1.74


Commit from airwin (2003-12-25 01:14 CET)
------------------

* Change so that file device result is visually identical with corresponding
C example result. diff shows that file device results still sometimes
disagree in last digit with corresponding C example (presumbably due to
16-bit rendering issues).

* Style changes: comments, do loops, indentation, and plcol ==> plcol0.

  plplot  examples/f77/x07f.fm4  1.8


Commit from airwin (2003-12-25 00:48 CET)
------------------

* Style changes: comments, do loops, indentation, and plcol ==> plcol0.
File device result still disagree with corresponding C example.  java
results also disagree (both with c and f77) for this example.  I believe
all these differences would disappear for higher resolution 32-bit rendering
as opposed to the current 16-bit rendering we now use for all devices.

  plplot  examples/f77/x06f.fm4  1.8


Commit from airwin (2003-12-24 23:25 CET)
------------------

* Style changes: comments, do loops, and indentation.  File device result
still agrees with corresponding C example.

  plplot  examples/f77/x05f.fm4  1.9


Commit from airwin (2003-12-24 21:09 CET)
------------------

* Change so that gives identical file device result to corresponding C example.

* Stylistic changes: comments, do loops, indentation, plcol ==> plcol0.

  plplot  examples/f77/x04f.fm4  1.8


Commit from airwin (2003-12-24 20:33 CET)
------------------

* Change so that gives identical file device result to C example.

* Stylistic changes: comments, do loops, and indentation.

  plplot  examples/f77/x03f.fm4  1.9


Commit from airwin (2003-12-24 02:16 CET)
------------------

Style improvements: enddo, indentation ==> still produces identical file
result to C example.

  plplot  examples/f77/x02f.fm4  1.7


Commit from airwin (2003-12-24 01:44 CET)
------------------

* Change code so that file results agree with C example file results.

* float ==> dble

* Stylistic changes: do ... enddo

  plplot  examples/f77/x01f.fm4  1.9


Commit from airwin (2003-12-23 07:29 CET)
------------------

Add the following API wrappers to the fortran interface:

PL_SETCONTLABELFORMAT
PL_SETCONTLABELPARAM
PLCALC_WORLD
PLCLEAR
PLCPSTRM
PLFILL3
PLGCOL0
PLGCOLBG
PLGCOMPRESSION
PLGDIDEV
PLGDIORI
PLGDIPLT
PLGRIDDATA
PLGVPD
PLGVPW
PLLIGHTSOURCE
PLMKSTRM
PLREPLOT
PLSCMAP1N
PLSCOMPRESSION
PLSDIDEV
PLSDIMAP
PLSDIORI
PLSDIPLT
PLSDIPLZ
PLSMEM
PLSTRIPA
PLSTRIPD
PLXORMOD

All of these new wrappers build without problems, but there are no examples
to run-time test them yet.  Also there still remains work to do
on the PLMESHC, PLOT3DC, PLSHADE1, PLSHADES, PLSURF3D, and PLSTRIPC wrappers
to implement all the common API.

  plplot  bindings/f77/plstubs.h  1.29
  plplot  bindings/f77/scstubs.c  1.30


Commit from airwin (2003-12-23 07:12 CET)
------------------

* Replace file designed to be processed by m4 by result of m4 processing of
that file using double precision.  Fortran passes everything by reference,
and there is no actual numerical computation or constants in this fortran
source.  Thus, I believe the double precision is only technical in this case
and needs no further processing in the single-precision case.

* Handle common blocks by an include file, sfstubs.h.

  plplot  bindings/f77/sfstubs.fm4  1.8


Commit from airwin (2003-12-23 07:05 CET)
------------------

* Move from m4 processing to a simple cp.

* sfstubs.h added to file lists

  plplot  bindings/f77/Makefile.am  1.11


Commit from airwin (2003-12-23 07:03 CET)
------------------

Initial commit of common block definitions that are required (as an
include file) by routines in sfstubs.f(m4).

+ plplot  bindings/f77/sfstubs.h  1.1


Commit from airwin (2003-12-23 07:01 CET)
------------------

Initial commit of documentation about how Fortran API is
controlled/generated by the various files in this directory.

+ plplot  bindings/f77/README.f77API  1.1


Commit from airwin (2003-12-23 06:46 CET)
------------------

* plgriddata: sg parameter changed from input to output.

* tweak wording: "in in" ==> "in"

  plplot  doc/docbook/src/api.xml  1.58


Commit from airwin (2003-12-23 06:42 CET)
------------------

Style Changes:

* improve alphabetic ordering to help in maintenance

* plshades1 ==> c_plshades1 to be consistent with other common API functions.

  plplot  include/plplot.h  1.145


Commit from airwin (2003-12-23 01:43 CET)
------------------

x19f excluded from these Makefiles because this example invokes plmap and
plmeridians, and these functions are not implemented as part of the common
API (and probably won't be for a long time because of having to work around
the mapform type of argument for these functions).

  plplot  examples/f77/Makefile.am           1.13
  plplot  examples/f77/Makefile.examples.in  1.9


Commit from airwin (2003-12-22 20:21 CET)
------------------

Correct from "input" to "output" designation of certain plcalc_world arguments.

  plplot  doc/docbook/src/api.xml  1.57


Commit from airwin (2003-12-22 20:11 CET)
------------------

Adjust to new names of generated files

  plplot  pkgcfg/.cvsignore  1.2


Commit from airwin (2003-12-22 04:58 CET)
------------------

Change from int to PLINT arguments for plgriddata to conform to the pattern
for all other functions in our public API.

  plplot  bindings/c++/plstream.cc  1.18
  plplot  include/plplot.h          1.144
  plplot  src/plgridd.c             1.7
  plplot  examples/c/x21c.c         1.10
  plplot  examples/c++/x21.cc       1.3


Commit from rlaboiss (2003-12-21 12:59 CET)
--------------------

Debian release 5.2.1.cvs.20031030-2

From the changelog:

  * debian/rules: Call configure with --enable-static, such that the
    lib*.a files will be present in the *-dev packages (closes: #224550).
  * debian/control: Changed architecture of packages libplplot-dev and
    plplot-tcl-dev from "all" to "any", because of the *.a files included.

  plplot  debian/changelog            1.54
  plplot  debian/control              1.39
  plplot  debian/libplplot-dev.files  1.8
  plplot  debian/rules                1.41


Commit from rlaboiss (2003-12-16 18:08 CET)
--------------------

Made the Math::Trig module export only the symbol `pi', to avoid conflict
with PDL definitions.  Also, some small formatting cleanups were done.

  plplot  examples/perl/x01.pl  1.3


Commit from rlaboiss (2003-12-16 18:02 CET)
--------------------

Added exclusions file.

+ plplot  examples/perl/.cvsignore  1.1


Commit from rlaboiss (2003-12-16 18:00 CET)
--------------------

Use module Math::Trig to get definition of number pi.

  plplot  examples/perl/x01.pl  1.2


Commit from rlaboiss (2003-12-15 19:07 CET)
--------------------

PDL demos for PLplot.

Added Perl examples directory, which work with the PDL::Graphics::PLplot
module (PDL version 2.4 or later).  For now, only example x01.pl is written.

The files README.perldemos and x01.pl will be unconditionally installed in
the $(DATA_DIR)/../examples/perl directory.  Of course, the scripts only
work if PDL is installed.

  plplot  configure.ac                    1.116
  plplot  examples/Makefile.am            1.4
+ plplot  examples/perl/Makefile.am       1.1
+ plplot  examples/perl/README.perldemos  1.1
+ plplot  examples/perl/x01.pl            1.1


Commit from airwin (2003-12-12 04:48 CET)
------------------

Convert all floating-point constants in fortran examples to standard
double-precision form.

e.g., .3 ==> 0.3d0.

Also, convert ugly dble (0.3), etc.  in argument lists to 0.3d0.

This was all done by a fairly well tested sed script that I have developed
for a different project.

Examples 3, 9, 11, and 13 gave slightly different (but still visually
acceptable) answers as a result.

  plplot  examples/f77/x01f.fm4  1.8
  plplot  examples/f77/x02f.fm4  1.6
  plplot  examples/f77/x03f.fm4  1.8
  plplot  examples/f77/x04f.fm4  1.7
  plplot  examples/f77/x05f.fm4  1.8
  plplot  examples/f77/x06f.fm4  1.7
  plplot  examples/f77/x07f.fm4  1.7
  plplot  examples/f77/x08f.fm4  1.7
  plplot  examples/f77/x09f.fm4  1.6
  plplot  examples/f77/x10f.fm4  1.6
  plplot  examples/f77/x11f.fm4  1.7
  plplot  examples/f77/x12f.fm4  1.7
  plplot  examples/f77/x13f.fm4  1.7
  plplot  examples/f77/x16f.fm4  1.7
  plplot  examples/f77/x17f.fm4  1.2
  plplot  examples/f77/x18f.fm4  1.2
  plplot  examples/f77/x19f.fm4  1.2


Commit from rlaboiss (2003-12-09 18:27 CET)
--------------------

I just noticed that the Octave function octave_config_info accepts an
argument.  This slightly simplifies the OCTAVE_CONFIG_INFO macro.  This
change was tested only with Octave 2.1.50.  Hopefully it will work with any
version of Octave supported by PLplot.

  plplot  sysloc.in  1.59


Commit from rlaboiss (2003-11-18 21:26 CET)
--------------------

Removed entry about --enable-strip-libraries option, sicne this is already
implemented by Automake.

  plplot  TODO.AM-LT  1.4


Commit from rlaboiss (2003-11-18 17:42 CET)
--------------------

Revamped ToDo list

Reorganization of this ToDo list, separating task that are already completed
or abandoned from those issues that are still open.  Also, added an item
about the implementation of the --enable-strip-libraries option of configure.

  plplot  TODO.AM-LT  1.3


Commit from jcard (2003-11-14 01:33 CET)
-----------------

-Make default visual the default, use "-drvopt defvis=0" to not use the default visual.
-Add "usepth" driver specific option; enable pthreads by default, use "-drvopt usepth=0"
 to not use pthreads to redisplay the plot.
-Fix subtle bug when using pthreads with two or more plot windows; when resizing one of them
 the resize/replot was directed to the plot window on the current stream.

  plplot  drivers/xwin.c  1.113


Commit from rlaboiss (2003-10-31 14:44 CET)
--------------------

Deleted obsolete files.

- plplot  debian/libplplot5.README.Debian  1.7
- plplot  debian/libplplot5.files          1.3


Commit from rlaboiss (2003-10-30 10:10 CET)
--------------------

Debian release 5.2.1.cvs.20031030-1.

  plplot  debian/.cvsignore  1.6
  plplot  debian/changelog   1.53

I found that on my debian sid machine that the detection
of the docbook html and print stylesheets failed.
(Not sure why, but it works OK on my Mandrake 9.2 machine.)

From configure 
./configure --enable-builddoc --with-sgml-catalogs=/etc/sgml/catalog

...
checking XML::DOM... yes checking XML::Parser... yes
checking for GNU make... make
checking for openjade... found
checking for jadetex... found
checking for pdfjadetex... found
checking for dvips... found
checking for onsgmls... found
checking for makeinfo... found
checking DSSSL Style Sheet DTD... found
checking DocBook HTML Stylesheet... no
checking DocBook Print Stylesheet... no
checking DocBook DTD... found
checking XML::Parser::PerlSAX... yes
checking XML::Writer... yes
...

The jadeout.log file contains:

Checking for DSSSL Style Sheet DTD

Checking for DocBook HTML Stylesheet
openjade:doc/docbook/conftest.dsl:3:25:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified
openjade:doc/docbook/conftest.dsl:8:27:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified
openjade:doc/docbook/conftest.dsl:8:44:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified

Checking for DocBook Print Stylesheet
openjade:doc/docbook/conftest.dsl:3:25:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified
openjade:doc/docbook/conftest.dsl:8:27:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified
openjade:doc/docbook/conftest.dsl:8:44:E: an attribute value specification 
must be an attribute value literal unless SHORTTAG YES is specified

Checking for DocBook DTD

---

By escaping the quotes in the sgml attributes used in CHECK_DTD
in configure.ac, detection worked properly.

A possible patch for configure.ac is:

--- plplot-5.2.1.cvs.20031226/configure.ac	2003-12-26 08:19:09.000000000 -0800
+++ plplot/configure.ac	2003-12-25 21:52:16.000000000 -0800
@@ -1037,8 +1051,8 @@
   [DocBook HTML Stylesheet],
   [html_ss],
   [[[<!ENTITY dbstyle PUBLIC \"$DB_SS_HTML_PUBID\" CDATA DSSSL>]]],
-  [use="docbook"],
-  [<external-specification id="docbook" document="dbstyle">],
+  [use=\"docbook\"],
+  [<external-specification id=\"docbook\" document=\"dbstyle\">],
   [[[<!ELEMENT book - O (#PCDATA)>]]],
   [sgml],
   [\"$DB_SS_HTML_PUBID\"],
@@ -1050,8 +1064,8 @@
   [DocBook Print Stylesheet],
   [print_ss],
   [[[<!ENTITY dbstyle PUBLIC \"$DB_SS_PRINT_PUBID\" CDATA DSSSL>]]],
-  [use="docbook"],
-  [<external-specification id="docbook" document="dbstyle">],
+  [use=\"docbook\"],
+  [<external-specification id=\"docbook\" document=\"dbstyle\">],
   [[[<!ELEMENT book - O (#PCDATA)>]]],
   [tex],
   [\"$DB_SS_PRINT_PUBID\"],









Brian

======================================================================
Brian D. Wright                                 Tel: (415)476-1007   
Dept. of Physiology, Box 0444		      	Fax: (415)476-4929    
Keck Center for Integrative Neuroscience      	bdwright@... 
University of California, San Francisco	      
513 Parnassus Avenue
San Francisco, CA 94143-0444
======================================================================

* Brian D. Wright <bdwright@...> [2003-12-26 11:45]:

> I found that on my debian sid machine that the detection
> of the docbook html and print stylesheets failed.
> (Not sure why, but it works OK on my Mandrake 9.2 machine.)

This is very strange, since I generate the CVS tarball in an up-to-date
Debian sid system and I get:

    checking DSSSL Style Sheet DTD... found
    checking DocBook HTML Stylesheet... found
    checking DocBook Print Stylesheet... found
    checking DocBook DTD... found

However,

> By escaping the quotes in the sgml attributes used in CHECK_DTD
> in configure.ac, detection worked properly.
> 
> A possible patch for configure.ac is:
> 
> [...]

I confirm that your patch works for me and, actually, does the Right Thing
(TM).  I am going to apply it to CVS.  Thanks for suggestion.

-- 
Rafael