[Plplot-devel] Merge of AT has now been completed. A new configuration era starts!

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

The merge of AT to MAIN has been completed so welcome to the new Unix/Linux
configuration era for PLplot!  There are also consequences for Windows, etc.
users, see note at the end.

Thanks to a suggestion from Maurice I was able to work around the current
problems with state in the repository (although those problems still need to
be fixed). The merge was completed with no conflicts, the diff between the
old MAIN branch and the merged local tree looked good (with only the
expected changes) so I committed the result.  Two of the files (COPYING and
configure.in) could not be retrieved afterward using a fresh checkout, but
it turned out those had been deleted long ago (the commit message this time
mentioned they were being put in the Attic) so essentially those names are
unusable for the MAIN branch.

COPYING is not really needed since it is replaced anyway by one of the
procedures in bootstrap.sh.  configure.in is a much more serious manner,
but fortunately autoconf (inside bootstrap.sh) accepts a more modern name
for the file, configure.ac, so that is what I checked in.  Interestingly,
with the different name autoconf works differently with more extensive
warning messages.  But the built result is the same.

I checked out a fresh copy and used the following recipe to produce
identical postscript files to before:

(IMPORTANT NOTE:

I used autoconf-2.53, automake-1.5, and libtool-1.4.2.  Probably you should
stick with these versions or later.)

./bootstrap.sh

Note, after bootstrap, you should be able to just pack up your whole tree in
a tarball, take it to any Unix platform, and do the remaining steps on that
platform.  (I think that brute-force approach is preferable to using "make
dist" since that command depends on keeping track of all the extra
non-standard files that should be in the tarball.  See the various
EXTRA_DIST lines I have put in the Makefile.am files.  I maintained that for
a while, but it got to be a pain so I let it slide.)  Anyhow, I strongly
encourage all the developers who have access to alternative platforms to
give PLplot a try there once they get it working on Linux.

./configure --prefix=wherever --enable-dyndrivers

make
make install
cd wherever/lib/plplot5.1.0/examples/c
make -f Makefile.examples
cd ../c++
make -f Makefile.examples
cd ../f77
make -f Makefile.examples
cd ../tk
make -f Makefile.examples
cd ..
./plplot-test.sh

Also, the recipes in examples/tcl/README.tcldemos and
examples/tk/README.tkdemos should work as well.

All drivers seem to work as plug-ins (i.e. as dynamic drivers) except for
the tk driver problem I mentioned before:

./x10c -dev tk
TCL command "plclient_init" failed:
invalid command name "plclient_init"
TCL command "plclient_link_end" failed:
invalid command name "plclient_link_end"
Segmentation fault

This one has me stumped, and I leave it to the tcl experts to figure out
why the tk driver can no longer find plclient.tcl.

The next priority for me is to get octave under autotools control, then
do the static drivers.  For a while there I thought I would be stuck in
AT limbo forever so it is great to be working on the MAIN branch again!

Notes for Unix users.

* The basic configuration changes are in the various Makefile.am files.
  automake and libtool have a very simple syntax which doesn't take that
  long to learn so I strongly suggest you just dive in if you find any
  glitches.  info autconf, info automake, and info libtool are your friends.
  am_conditionals just  make life very easy compared to the autoconf-only
  system we had before.

* plplot/cf is not used at all any more, but we should keep it around for a
  while for reference.

* Most changes to source code consisted of a change from PLPLOT_VERSION to
  VERSION wherever that occurred, and a change of the #include style to drop
  the "plplot/".  The one really important change I can remember to source
  is in the plcore.c file where the old dlopen procedure was replaced with a
  cross-platform friendly libltdl approach.

* The largest practical change you will notice (Joao has already commented)
  is no additional "flat" plplot/tmp install area is supported at this time.
  After we get some experience with autotools we conceivably could decide to
  bring that back.  Alternatively, there may be a way within the automake
  and libtool paradigm to make it very easy to build and test particular
  results that you are developing.  I just don't know the answer to such
  "style-of-working" questions at the moment, but I am sure we will find out
  through experience and asking such groups as the KDE developers how they
  use autotools during their development cycles.

Note for non-Unix users.  You will be affected by the change in the #include
style.  Also, my understanding is that the configure scripts generated by
autotools might just work on some of the non-unix platforms.  So it will be
worth your while to look up your platform in a google search on the net for
autotools to see where you stand.

Alan

email: ir...@be...
phone: 250-727-2902	FAX: 250-721-7715
snail-mail:
Dr. Alan W. Irwin
Department of Physics and Astronomy,
University of Victoria, P.O. Box 3055,
Victoria, British Columbia, Canada, V8W 3P6
__________________________

Linux-powered astrophysics
__________________________