Re: density plot

[Geoffrey Furnish]

Joao Cardoso writes:
 > "(Ted Harding)" wrote:
 > > First, I can't compile plplot. Despite having tk.h in /usr/X11R6/include,

That seems to me like a very strange place for Tk.  For example, on my 
Red Hat 6 system:

valinux[25] ls /usr/X11R6/include/
X11/  Xaw3d/
valinux[26] ls /usr/X11R6/include/X11/
Composite.h   CompositeP.h  ConstrainP.h  Constraint.h  Core.h
CoreP.h       DECkeysym.h   HPkeysym.h    ICE/          Intrinsic.h
IntrinsicP.h  Object.h      ObjectP.h     PEX5/         PM/
RectObj.h     RectObjP.h    SM/           Shell.h       ShellP.h
StringDefs.h  Sunkeysym.h   Vendor.h      VendorP.h     X.h
X10.h         XKBlib.h      XWDFile.h     Xalloca.h     Xatom.h
Xauth.h       Xaw/          Xaw3d@        Xcms.h        Xdmcp.h
Xfuncproto.h  Xfuncs.h      Xlib.h        Xlibint.h     Xlocale.h
Xmd.h         Xmu/          Xos.h         Xos_r.h       Xosdefs.h
Xpoll.h       Xproto.h      Xprotostr.h   Xresource.h   Xthreads.h
Xutil.h       ap_keysym.h   bitmaps/      cursorfont.h  extensions/
fonts/        keysym.h      keysymdef.h   xpm.h
valinux[27] 

there is no evidence of tk.h there.  

 > > which is a path that seems to be searched in the configure file, "make"
 > > nevertheless tells me that it can't be found. 

Whenever reporting on things like this, it is useful for us to be able 
to see actual verbatim clippings from the failed output, in order to
be 100% clear on exactly what is failing.

 > I also get "can't find
 > > arrayobject.h so python won't work (but I don't seem to have this
 > > file at all).

Configuring Python and PLplot for sane cooperation is quite a bit more
work than for plain Tcl or Itcl.  The reasons for this are deep and
cannot be explained in short order.  I personally feel there is a
serious shortcoming in the way Python/Tk configuration is done, and
have requested changes in this department in the past, to no avail.

At the present time, the best I can offer, is that there is a text
file in the tarball that explains how I recommend bootstrapping a
combined PLplot/Python/Tk system.  It's frankly pretty involved.  See
bindings/python/README.  A quick perusal now shows that there is some
stuff in there which is getting kind of dated, so you will in fact
have to "think on your feet".  Use it as an approximate statement of
what to do...

 > > Finally, "make" collapses with the message
 > > 
 > > make[1]: *** No rule to make target `stupidtk.c', needed by `stupidtk.o'.
 > >  Stop.

Mmmm.  I'm not really  sure what's wrong.  Does the same thing happen
if you first install Tcl/Tk into a custom prefix, and then configure
plplot against the same prefix?  I'm guessing this problem is related
to the prior problem with its inability to find the real site of Tcl/Tk.

 > > The plplot configure file is massive, and I can't make out what to do.

Well, /nobody/ can read configure scripts.  If you want to get into it 
at this level, you need to get autoconf so you can regenerate the
configure script from it's sources, and then you can go down into
those cf/* files, and insert your own debugging quips, and then
regenerate the configure script, and then reconfigure PLplot.

There is much to say about autoconf, on both sides of the pro/con list
:-)...  It's very good for building adaptive configurations, but
setting it up is a black art, of which I am no master.  Based on the
behavior of configure scripts I see from others, one might wonder if
there are any masters of this black art alive in the world today...

Mmm, well, the XEmacs folks have done a stupendous job with autoconf.
But I don't know of any other package that configures as completely or
as slickly as that one.

 > I configure plplot with all but the xwin and tk driver; I would
 > recommend to even discard the tk driver for start. *my* configure
 > options are:
 > 
 > ./configure --with-double --with-gcc --disable-dp --with-opt \
 > --with-shlib --disable-cxx --disable-f77 --disable-itcl \
 > --disable-linuxvga --disable-python --disable-mskermit
 > 
 > you might also try leaving out the --with-shlib, which often is platform
 > dependent.

A somewhat less drastic approach, would be to install your own Tcl/Tk
into a custom prefix, and then configure PLplot with the same prefix.
This is what I almost always do myself.

 > My tcl/tk is in /usr/local, and is founded. But sometimes the generated
 > makefile contains the incorrect version of the libraries,
 > libtcl7.6.a(so) instead of libtk8.0.a(so) etc.

Mmm, I suppose we should work on this at some point...

 > Perhaps we should maintain any further conversation private? unless
 > someone says no.

I'm personally in favor of using the list for things that appear to be 
problems with support of certain platforms, since you are unlikely to
actually be the /only/ one suffering.  Nobody is /forced/ to read
email :-). 

-- 
Geoffrey Furnish            Actel Corporation        address@hidden
Senior Staff Engineer      955 East Arques Ave       voice: 408-522-7528
Placement & Routing     Sunnyvale, CA   94086-4533   fax:   408-328-2303

"... because only those who write the code truly control the project."
                                                      -- Jamie Zawinski



---------------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.  To ensure
that development continues, see www.che.wisc.edu/octave/giftform.html
Instructions for unsubscribing: www.che.wisc.edu/octave/archive.html
---------------------------------------------------------------------