Changeset 2909

Timestamp:

05/12/07 11:38:46 (3 years ago)

Author:

lorenzen

Message:

Release preparations:

* Removal of the Texinfo manual. Instead, a PDF version of the Wiki manual

is included in the distribution.

* Updated the README and NEWS files.
* Small fix in the RPM generation concerning the documentation files.
* Updated the ChangeLog? (more to test svn2cl.sh, it will have to be done

once more anyway).

Location:

branches/2.1.x

Files:

• ChangeLog (modified) (1 diff)
• Makefile.am (modified) (1 diff)
• NEWS (modified) (1 diff)
• README (modified) (2 diffs)
• build/octopus.spec.in (modified) (5 diffs)
• build/var2html.pl (modified) (1 diff)
• configure.ac (modified) (1 diff)
• doc/Makefile.am (modified) (1 diff)
• doc/authors.texi (deleted)
• doc/copying.texi (deleted)
• doc/description.texi (deleted)
• doc/examples.texi (deleted)
• doc/external.texi (deleted)
• doc/index.texi (deleted)
• doc/installation.texi (deleted)
• doc/introduction.texi (deleted)
• doc/octopus.texi (deleted)
• doc/parser.texi (deleted)
• doc/pdf (deleted)
• doc/undocumented.texi (deleted)
• doc/update.el (deleted)

branches/2.1.x/ChangeLog

@@ -1 @@
-2006-02-16 23:22  appel
-
-        * AUTHORS: Adding current e-mail addresses.
-
-2006-02-16 17:23  marques
-
-        * AUTHORS: updated list
-
-2006-02-16 16:50  marques
-
-        * external_libs/: expokit/COPYING, expokit/Copyright,
-          metis-4.0/COPYING, metis-4.0/Copyright: Changed filenames from
-          Copyright COPYING
-
-2006-02-16 16:49  marques
-
-        * configure.ac, liboct/Makefile.am, liboct/oct_f.c,
-          build/acx_metis.m4, src/Makefile.am,
-          external_libs/metis-4.0/Makefile.am: *) Metis is now only
-          compiled when needed *) changed version number to 2.0 *) octopus
-          fails a couple of tests in my machine. I will take a look before
-          committing
-
-2006-02-16 16:02  appel
-
-        * liboct/: Makefile.am, oct_f.c: The header file metis.h was still
-          included as <metis/metis.h>. Now we use the one that we ship in
-          external_libs.
-
-2006-02-16 12:37  appel
-
-        * debian/changelog: Switching from 1.99devel to 2.0 for the
-          upcoming release.
-
-2006-02-16 12:36  appel
-
-        * configure.ac, build/acx_metis.m4: I forgot to treat the
-          HAVE_METIS properly, when I added the library to external_libs.
-          Now HAVE_METIS is enabled by default, but can be disabled with
-          --disable-metis at configure time.
-
-2006-02-13 16:58  acastro
-
-        * src/spectrum.F90: Bug fixed.  The units were not properly handled
-          in the cross_section_tensor subroutine.
-
-2006-02-12 23:13  acastro
-
-        * testsuite/rc_bugs/12-spin: Bug added.  Sorry, but I had to add
-          another bug, I think that it is important.  See the explanation
-          in the file. I will try to think about it the next days, to solve
-          it asap.
-
-2006-02-09 23:38  appel
-
-        * src/liboct_parser.F90: o Adding the full filename to the
-          variables error message.  o Removing stdin stuff completely.
-
-2006-02-09 23:32  appel
-
-        * Makefile.am: There was a line continuation \ missing. During make
-          install the share and testsuite directories have not been taken
-          into account.
-
-2006-02-09 22:49  appel
-
-        * src/out.F90: The compilation failed when using --with-netcdf. The
-          netcdf module doesn't have _m. Too much Perl :) ...
-
-2006-02-09 21:43  appel
-
-        * COPYING: Updating the top-level license file to account for
-          Expokit and Metis as 'external libraries'. I explicitely state in
-          the text, that both are _not_ GPL.
-
-          Please have a look if you agree with the style.
-
-2006-02-09 21:39  appel
-
-        * Makefile.am, configure.ac, external_libs/Makefile.am,
-          external_libs/expokit/Copyright,
-          external_libs/expokit/Makefile.am,
-          external_libs/expokit/expokit.f90, src/Makefile.am,
-          external_libs/metis-4.0/Copyright,
-          external_libs/metis-4.0/Makefile.am,
-          external_libs/metis-4.0/balance.c,
-          external_libs/metis-4.0/bucketsort.c,
-          external_libs/metis-4.0/ccgraph.c,
-          external_libs/metis-4.0/coarsen.c,
-          external_libs/metis-4.0/compress.c,
-          external_libs/metis-4.0/debug.c, external_libs/metis-4.0/defs.h,
-          external_libs/metis-4.0/estmem.c, external_libs/metis-4.0/fm.c,
-          external_libs/metis-4.0/fortran.c,
-          external_libs/metis-4.0/frename.c,
-          external_libs/metis-4.0/graph.c,
-          external_libs/metis-4.0/initpart.c,
-          external_libs/metis-4.0/kmetis.c,
-          external_libs/metis-4.0/kvmetis.c,
-          external_libs/metis-4.0/kwayfm.c,
-          external_libs/metis-4.0/kwayrefine.c,
-          external_libs/metis-4.0/kwayvolfm.c,
-          external_libs/metis-4.0/kwayvolrefine.c,
-          external_libs/metis-4.0/macros.h,
-          external_libs/metis-4.0/match.c,
-          external_libs/metis-4.0/mbalance.c,
-          external_libs/metis-4.0/mbalance2.c,
-          external_libs/metis-4.0/mcoarsen.c,
-          external_libs/metis-4.0/memory.c, external_libs/metis-4.0/mesh.c,
-          external_libs/metis-4.0/meshpart.c,
-          external_libs/metis-4.0/metis.h, external_libs/metis-4.0/mfm.c,
-          external_libs/metis-4.0/mfm2.c,
-          external_libs/metis-4.0/mincover.c,
-          external_libs/metis-4.0/minitpart.c,
-          external_libs/metis-4.0/minitpart2.c,
-          external_libs/metis-4.0/mkmetis.c,
-          external_libs/metis-4.0/mkwayfmh.c,
-          external_libs/metis-4.0/mkwayrefine.c,
-          external_libs/metis-4.0/mmatch.c, external_libs/metis-4.0/mmd.c,
-          external_libs/metis-4.0/mpmetis.c,
-          external_libs/metis-4.0/mrefine.c,
-          external_libs/metis-4.0/mrefine2.c,
-          external_libs/metis-4.0/mutil.c,
-          external_libs/metis-4.0/myqsort.c,
-          external_libs/metis-4.0/ometis.c,
-          external_libs/metis-4.0/parmetis.c,
-          external_libs/metis-4.0/pmetis.c,
-          external_libs/metis-4.0/pqueue.c,
-          external_libs/metis-4.0/proto.h,
-          external_libs/metis-4.0/refine.c,
-          external_libs/metis-4.0/rename.h,
-          external_libs/metis-4.0/separator.c,
-          external_libs/metis-4.0/sfm.c, external_libs/metis-4.0/srefine.c,
-          external_libs/metis-4.0/stat.c, external_libs/metis-4.0/struct.h,
-          external_libs/metis-4.0/subdomains.c,
-          external_libs/metis-4.0/timing.c, external_libs/metis-4.0/util.c:
-          Both, Expokit and the Metis library are now declared as external
-          libraries. The build system has been adapted accordingly to
-          reflect this change.
-
-          The inclusion of Metis should now finally solve the build
-          problems, that people were encountering frequently with the MPI
-          executable.
-
-2006-02-09 21:34  acastro
-
-        * src/: par_vec.F90, simul_box.F90: Bug fixed.  Just an absolute
-          value, but it created some awful problems.  I also put an ASSERT
-          in par_vec at a point where, if it had been before, it would have
-          saved me quite a few hours to debug. Maybe it is useful in the
-          future.  This bug was pointed out by some Carlo Camilloni.
-
-2006-02-09 21:08  appel
-
-        * src/: atomic.F90, blas.F90, broad.F90, casida.F90,
-          centergeom.F90, cf.F90, cf_inc.F90, cross_section.F90,
-          crystal.F90, curv_briggs.F90, curv_gygi.F90, curv_modine.F90,
-          curvlinear.F90, datasets.F90, derivatives.F90,
-          derivatives_inc.F90, eigen.F90, eigen_arpack.F90, eigen_cg.F90,
-          eigen_evolution.F90, eigen_plan.F90, eigen_trlan.F90, epot.F90,
-          epot_inc.F90, f.F90, f_inc.F90, fftw2.F90, fftw3.F90,
-          geom_opt.F90, geometry.F90, global.F90, grid.F90, gs.F90, h.F90,
-          h_inc.F90, h_so.F90, hgh.F90, hs.F90, io.F90, lasers.F90,
-          lcao.F90, liboct.F90, liboct_parser.F90, libxc.F90,
-          linalg_adv.F90, linalg_basic.F90, linear_response.F90,
-          linear_response_inc.F90, logrid.F90, main.F90, math.F90,
-          mesh.F90, mesh_init.F90, mesh_lib.F90, messages.F90, mf.F90,
-          mf_inc.F90, mix.F90, mpi.F90, mpi_debug.F90, multicomm.F90,
-          multigrid.F90, nl_operator.F90, ode_solver.F90,
-          ode_solver_inc.F90, opt_control.F90, out.F90, out_inc.F90,
-          par_vec.F90, par_vec_inc.F90, pes.F90, pes_mask.F90, pes_rc.F90,
-          phonon_spectrum.F90, phonons.F90, poisson.F90, poisson1D.F90,
-          poisson2D.F90, poisson3D.F90, poisson_cg.F90,
-          poisson_corrections.F90, poisson_multigrid.F90, profiling.F90,
-          ps.F90, pulpo.F90, restart.F90, restart_inc.F90, root_solver.F90,
-          root_solver_inc.F90, rotational_strength.F90, run.F90,
-          scalar_mesh.F90, scf.F90, simul_box.F90, sparskit.F90,
-          sparskit_driver.F90, sparskit_inc.F90, specie.F90, spectrum.F90,
-          spline.F90, states.F90, states_inc.F90, states_kpoints.F90,
-          static_pol.F90, static_pol_lr.F90, stencil_cube.F90,
-          stencil_star.F90, stencil_starplus.F90, stencil_variational.F90,
-          string.F90, systm.F90, td.F90, td_calc.F90, td_exp.F90,
-          td_exp_split.F90, td_init.F90, td_rti.F90, td_write.F90, tm.F90,
-          units.F90, unocc.F90, v_ks.F90, v_ks_inc.F90, varinfo.F90,
-          wave_matching.F90, xc.F90, xc_KLI.F90, xc_OEP.F90,
-          xc_OEP_SIC.F90, xc_OEP_inc.F90, xc_OEP_x.F90, xc_axc.F90,
-          xc_functl.F90, xc_fxc.F90, xc_kxc.F90, xc_vxc.F90,
-          xyz_adjust.F90, xyz_file.F90, xyzanim.F90: The MVAPICH
-          implementation has serious I/O problems when one of the modules
-          in the code is called 'system'. Besides the problem with reading
-          from stdin, this was the main issue in Sahu's problem report. I
-          spent more than a day to figure this out. Also thanks again to
-          Sahu for his patience, while we were debugging this issue.
-
-          As there were some thoughts on this previously, I take the
-          opportunity to rename all modules according to
-
-           sample/sample_mod  ->  sample_m
-
-          Similarly all types have been converted according to
-
-           sample_type  ->  sample_t
-
-          This is actually Linux kernel convention :).
-
-2006-02-09 19:13  appel
-
-        * src/liboct_parser.F90: Some MPI implementations, e.g. the MVAPICH
-          implementation for infiniband interconnects, do not forward stdin
-          to all compute nodes. This problem was initially reported by Sahu
-          on the octopus-users mailing list and recently showed up in a
-          second posting.
-
-          Reading the input file from stdin via e.g.
-
-           octopus < Na2
-
-          is now disabled. This caused too many confusions, even in the
-          serial case.  Often people already had a file with name 'inp' in
-          the working directory while trying to read the input via '<'. In
-          that case they got tricked by the precedence.
-
-          Octopus is now printing an error message, mentioning that reading
-          input from stdin is deprecated.
-
-2006-02-06 17:18  acastro
-
-        * src/messages.F90: The write_warning and write_fatal were writing
-          an awkard "WARNING: From Node x", with no information afterwards,
-          from all the nodes.  Only node zero was really printing the
-          information.
-
-          Maybe in the future we should allow for a way for all non-head
-          nodes to write warning and fatal errors.
-
-2006-02-05 12:33  acastro
-
-        * src/: atomic.F90, hgh.F90, tm.F90: Not really a bug fix.  The
-          problem with the Gadolinium bug report ("rc_bugs/07-Atomic_Gd")
-          is deep in the "egofv" routine, and I don't quite understand it.
-          What I have done is:  - for HGH pseudopotentials, the atomic
-          wavefunctions are not strictly neccesary, so if they are not
-          found, they are just not used. "egofv" does no longer kill the
-          program, but passes back an error flag and the program emits a
-          warning. The program can no longer do LCAO, but hopefully it
-          could converge. I could not try it in the case of Gadolinium,
-          because in any case the memory needed to get all the projectors
-          is so large that it does not fit in my laptop.   - for TM
-          pseudopotentials, the wavefunctions are needed to build the
-          projectors. So if the error is produced, the program stops. But
-          the error message emitted now is more friendly than before.
-
-          In any case, the proper solution to these issues will have to be
-          done for next release.
-
-2006-02-04 20:22  acastro
-
-        * src/mesh_init.F90, testsuite/rc_bugs/08-TDELF_C2H4: Bug fixed.
-
-          This bug was the one reported in 08-TDELF_C2H4, which I remove.
-          The input of the file was not consistent any more, in any case.
-
-          The bug showed up when trying to write a dx file. There was a
-          seg-fault whent trying to pass the mesh to a cube, and I traced
-          it back to be some kind of round-off error in the point that I
-          changed in mesh_init. Probably it is a quick and dirty trick, but
-          we should leave a deeper clarification of the mesh module for
-          after the release.
-
-2006-02-04 15:18  acastro
-
-        * src/poisson.F90: Bug fixed.  Some compilers (intel) seem to have
-          problems manipulating big 3D arrays.
-
-2006-02-04 02:45  acastro
-
-        * src/out_inc.F90: Bug fixed.  Under some circumstances, this bug
-          stopped some parallel runs in marenostrum. I don't know why it
-          did not show up always.
-
-2006-02-03 19:22  acastro
-
-        * src/systm.F90: The argument parallel_mask of system_init was
-          flagged as optional.  Then the internal argument parallel_mask_
-          was set to be equal to parallel_mask when it was present, as
-          usual... except that then it was not used.  Instead of fixing
-          this, I noticed that system_init is only called once, and
-          parallel_mask was passed. So no need for an optional argument
-          there.
-
-2006-02-03 18:27  acastro
-
-        * src/run.F90: (Possible) bug fixed: The variable mode (now
-          parallel_mask), used in the routine get_mode_parallelized (now
-          renamed to get_parallel_mask), was not initialized to 0. Some
-          compilers do not initialize variables to 0, so this could cause
-          trouble.
-
-2006-02-02 19:20  acastro
-
-        * liboct_parser/: parse.c, parser_f.c: Bug fixed.  *
-          parse_block_string function (file parse.c): I think that it
-          should return an error code whatever is read is not a string. For
-          other case it is OK to convert, but not in this case.  *
-          parse_block_error (file parser_f.c): I am not sure about this
-          one.  I uncommented the exit function. I believe that the code
-          should stop if an error is encountered. But whoever commented
-          that code probably had reasons for it, so take a look, maybe I am
-          wrong.
-
-          Now the code seg-faulted if a non-string was found where a string
-          should be found.
-
-          For next release we have to improve the input error handling.
-
-2006-01-31 12:11  acastro
-
-        * libxc/gga_lb.c, testsuite/rc_bugs/05-Na2_LB94_Casida: Bug fixed.
-          Removed the 05-Na2_LB94_Casida bug.  The LB94 is still rather
-          under suspiscion, in any case.
-
-          I don't know why the bug was there, in fact, it looked that the
-          code that I removed was there for a reason, but I could not trace
-          it down because it was there since the first revision...
-
-2006-01-30 16:55  acastro
-
-        * doc/Makefile.am: Bug fixed.  The compilation of octopus depended
-          on the existence of emacs. I think this should not be the case,
-          so I changed the Makefile.am in order that if emacs is not found,
-          octopus.info is not attempted to be rebuilt.  This way the
-          compilation proceeds.
-
-2006-01-26 13:25  appel
-
-        * liboct_parser/parser_f.c: Correcting cut'n paste error that was
-          found by Francesco Sottile.
-
-2006-01-22 11:15  appel
-
-        * src/: casida.F90, run.F90: Removing unused variables.
-
-2006-01-22 11:15  appel
-
-        * share/util/: Makefile.am, display_partitions.sh,
-          oct-display_partitions: Renaming display_partitions.sh to
-          oct-display_partitions and adding it to bin_SCRIPTS.
-
-2006-01-14 07:44  acastro
-
-        * src/epot.F90: The code should stop if a magnetic field is
-          required, but the executable is the one that only has real
+2007-05-07 06:59  lorenzen
+
+        * : Inserted forgotten word in variable description.
+
+2007-05-04 16:18  lorenzen
+
+        * : Documented the DerivativesOrder variable.
+
+2007-05-04 11:03  xavier
+
+        * : * Made fftw mandatory for 2.1, because the code doesn't compile
+          without it and it is required for a decent perfomance.
+
+2007-03-23 18:35  xavier
+
+        * : * Changed the files to build a Debian package in 2.1.x, now the
+          the
+          octopus deb can be built on a clean debian etch (or ubuntu)
+          system
+          using gfortran.
+          
+          * I had to disable the parallel binary, because the mpich
+          available in
+          Debian is still compiled with g77 and there are some problems
+          compiling octopus with openmpi.
+          
+          * I moved the old debian/rules file to
+          debian/rules.other_compilers .
+
+2007-03-23 18:17  xavier
+
+        * : More 2.1.x fixes:
+          
+          * Modified the README file to point user to the wiki.
+          * A variable was defined twice when compiling with fftw2.
+
+2007-03-23 15:48  xavier
+
+        * : * Backport of the NP/NP_PART fix to 2.1.x
+
+2007-03-22 18:26  lorenzen
+
+        * : Revision of the RPM packaging infrastructure. The RPM spec file
+          now takes
+          care of building both serial and MPI executables.
+          
+          The rpm target of the main Makefile.am has been polished a bit to
+          respect local build
+          settings, so being root is not necessarily required any more.
+          
+          As soon as the 2.1 release will be done, I can take care of the
+          RPMs for 32 Bit and also 64 Bit.
+
+2007-03-12 12:35  xavier
+
+        * : * Bugfix: The declaration of an argument was missing, as
+          pointed out
+          in ticket 77. This was fixed in trunk.
+          
+          The other bug is already fixed in trunk and in the 2.1.x branch.
+          
+          This fixes #77.
+
+2007-02-25 21:27  acastro
+
+        * : Backport of the bug corrected in trunk. However, even if arpack
+          does
+          not crash, it does not produce very good numbers in 2.1.x (in
+          trunk,
+          however, it seems to work much better...)
+
+2007-02-07 15:12  lorenzen
+
+        * : This is a backport of revisions r2647 and r2659 to
+          branches/2.1.x.
+          Those changes renamed the following macros of global.h:
+          
+          DOUBLE -> REAL_DOUBLE
+          SINGLE -> REAL_SINGLE
+          PRECISION -> REAL_PRECISION
+          
+          Again, the rationale was that -- unfortunately -- DOUBLE
+          PRECISION is still
+          a Fortran 90 reserved word and it is a bad idea to shadow any of
+          those by
+          a macro. This caused a compilation error with the mpif.h header
+          file. With
+          this commit I also revert the last change of Xavier in the 2.1.x
+          branch, which
+          circumvented the problem by undefining the macros but became
+          superfluous by
+          this more systematic renaming.
+
+2007-02-06 19:06  xavier
+
+        * : * Compilation fix: DOUBLE and PRECISION have to be undefined
+          when
+          including mpif.h.
+
+2007-01-19 13:21  xavier
+
+        * : Some fixes before 2.1.0rc2 release:
+          
+          * Fixed address of the pseudopotential generator in doc of the
+          species
+          block.
+          * The file variables.local wasn't being added to the tarball.
+          * Two tests used old variable names for pseudopotential species.
+
+2006-12-13 06:48  acastro
+
+        * : Backport of trunk bugfix 2623, to 2.1.x.
+
+2006-12-09 12:18  acastro
+
+        * : 2.1.x now also compiles if FFTW was not found at configure
+          time.
+
+2006-11-03 15:25  xavier
+
+        * : * Backport of the LCAO fixes.
+
+2006-10-27 18:23  acastro
+
+        * : Backport of some bugs found in branch but also present in 2.1.x
+
+2006-10-06 09:52  xavier
+
+        * : Backport of alberto's bugfixes (changesets 2454 and 2403).
+
+2006-10-06 09:15  athimm
+
+        * : Missing netcdf detection fragement.
+
+2006-10-06 09:08  athimm
+
+        * : Remove the gui from 2.1.x.
+          
+          There are changes required that should change the variables'
+          interface
+          in octopus. Furthermore 2.1.x is release-ready other than the
+          gui.
+
+2006-10-06 09:06  athimm
+
+        * : Backports the netcdf detection fixes from trunk, fixes #39.
+
+2006-10-02 20:33  xavier
+
+        * : This error check can be made simpler.
+
+2006-10-02 16:37  xavier
+
+        * : This is the bug that appeared during the Benasque school. An
+          error
+          condition was not checked.
+
+2006-10-02 16:02  xavier
+
+        * : A newline at the end of this file was missing and this caused
+          problems
+          with autoconf.
+
+2006-09-13 13:58  acastro
+
+        * : [Porting one bug already fixed in trunk]
+          
+          Bug fixed.
+          
+          The Portland compiler did not like very long lines even if they
+          are
+          just comments, and did not like to mix, in an interface block,
+          the
+          function or subroutine interface definitions, with the module
+          procedure statements. I have looked up the standard, and I
+          believe
+          both things are legal, but in any case this way that compiler is
+          happy
+          and it does not harm the others compilers.
+
+2006-08-24 20:25  appel
+
+        * : A few more places where check_inp was still missing.
+
+2006-08-24 15:07  acastro
+
+        * : Bug fixed.
+          
+          The problem with the 11-sic.test (never converging) was the use
+          of
+          density mixing, instead of potential mixing. The potential mixing
+          is
+          necessary when using orbital dependent xc functionals.
+
+2006-08-24 14:17  athimm
+
+        * : Default to not use python.
+          
+          It is currently of little value and only adds to the release
+          process
+          (like forcing the release manager to perform gettext/varinfo.xml
+          steps).
+
+2006-08-24 14:14  athimm
+
+        * : Make branch point to the next release (candiate).
+
+2006-08-23 20:39  appel
+
+        * : Creating branch for 2.1.x development series.
+
+2006-08-23 20:36  appel
+
+        * : Preparing the first release candidate of octopus 2.1.0. Setting
+          version strings to 2.1.0rc1.
+
+2006-08-23 09:39  appel
+
+        * : Make the various output options public, so that they can be
+          used in other parts of the code for calls to X(output_function).
+
+2006-08-23 09:34  appel
+
+        * : This push_sub had the wrong routine name.
+
+2006-08-23 09:18  appel
+
+        * : When using UserDefinedStates as initial states for a time
+          propagation, octopus was always trying to read ground state
+          files from the tmp directory, even when all initial orbitals
+          were given as formula strings (I didn't think of this case,
+          when I was initially adding the UserDefinedStates feature).
+          Adding therefore a logical variable OnlyUserDefinedInitialStates
+          (with default "no"). If true, then one can use arbitrary
+          initial states for a time propagation (via UserDefinedStates)
+          and octopus is not attempting to read ground state orbitals.
+          
+          This is e.g. useful to propagate coherent states, or gaussian
+          wave packets. Might also be useful for excerices on time
+          propagations in the summer school.
+
+2006-08-23 08:58  appel
+
+        * : Avoid division by zero, when scalar meshes with "zero"
+          points are created.
+
+2006-08-23 08:56  appel
+
+        * : A !%Variable header was missing for one of the variables.
+
+2006-08-23 08:55  appel
+
+        * : A check_inp and a pop_sub were missing.
+
+2006-08-22 10:16  xavier
+
+        * : Added a new variable, LinearSolverMaxTol, that is the maximum
+          tolerance that will be applied to the Linear Solver.
+
+2006-08-18 12:28  xavier
+
+        * : I forgot to change the filename in this test.
+
+2006-08-17 13:36  xavier
+
+        * : * Reverted from allocatables to pointers, because the code was
+          not F90
+          standard.
+
+2006-08-17 12:20  xavier
+
+        * : * Iterations are counted once, even if the apply the
+          hamiltonian twice
+          per iteration.
+
+2006-08-17 12:17  xavier
+
+        * : * Added GPL header.
+
+2006-08-17 11:52  xavier
+
+        * : * Renamed the output file of dynamic polarizability to
+          'linear/alpha'.
+
+2006-08-17 11:38  xavier
+
+        * : * In the nl_operator class, the arrays were defined as
+          pointers, i
+          changed the definition to allocatables, this gives the compiler
+          more
+          room for optimization.
+          * I had to change some code that relied on pointers to this
+          arrays.
+          Instead i used a pointer to the operators.
+          
+          It would be nice if someone can review this changes, specially
+          for
+          mpi, i don't know if this code is checked by the testsuite.
+
+2006-08-17 10:17  appel
+
+        * : When adding multi-dataset lables to the input files yesterday I
+          forgot to change the directories in the *.test files that are
+          used for the matches. This fixes the failed periodic tests of
+          today.
+
+2006-08-17 09:50  acastro
+
+        * : Two *_inc.F90 files were missing from the noinst_HEADERS,
+          so that the package constructed by make dist was incomplete.
+
+2006-08-16 23:53  acastro
+
+        * : The regeneration of the info file, when typing "make dist" was
+          broken,
+          due to some repetition of node names which apparently is not
+          allowed.
+          Maybe we are going to get rid of this file anyways, but anyways
+          I fix this.
+          
+          The "make dist" still does not work, it breaks now in the python
+          directory, complaining about config.rpath missing.
+
+2006-08-16 12:57  appel
+
+        * : Add routines which take care of resizing the simulation
+          box for scattering calculations (by multiples of the asymptotic
+          unit cell). The resizing is done such that the spacing
+          of the "central region" matches in the end the spacing
+          that is used for a periodic calculation of the asymptotic
+          region.
+
+2006-08-16 09:57  appel
+
+        * : Only index 0 is available for spin unpolarized
+          calculations. This will fix the failed periodic
+          tests of today (that will arrive later).
+
+2006-08-15 21:05  appel
+
+        * : For spin-polarized calculations also output spin resolved bands
+          and DOS.
+
+2006-08-15 21:05  appel
+
+        * : Add unscaled k-vectors to the output of the Wigner-Seitz cell.
+
+2006-08-15 21:04  appel
+
+        * : Add labels for multi dataset mode.
+
+2006-08-15 21:04  appel
+
+        * : Uncomment this for the moment, since Input/OutputDir is not
+          (yet)
+          used and we would create non-necessary directories in multi-
+          dataset mode otherwise.
+
+2006-08-13 21:47  appel
+
+        * : Also include unscaled k-vectors in the output of the
+          fermi energy.
+
+2006-08-13 21:16  appel
+
+        * : The scaling of kpoints in crystal_init was causing all my
+          confusion about factors of two. Now I made my Wigner-Seitz
+          routine compatible to crystal_init.
+
+2006-08-13 18:17  appel
+
+        * : For speed reasons Intel 9.1 is now also the default
+          compiler for the i386 deb.
+
+2006-08-13 10:10  appel
+
+        * : When using a large number of states the record length was
+          exceeded during output. Revert again to direct io.
+
+2006-08-13 10:08  appel
+
+        * : Reciprocal cell sizes were still not correct.
+
+2006-08-12 20:24  appel
+
+        * : o Allow to use FullWignerSeitzCell also for 1D periodic
+          systems.
+          
+          o Correct some factors of two.
+
+2006-08-12 20:24  appel
+
+        * : o Add routine to compute the degeneracy matrix for scattering
+          states.
+          
+          o Add option to choose between scaled and unscaled k-vectors
+          for the output of bands.
+          
+          o Fix a couple of cases which were not MPI clean.
+
+2006-08-12 20:06  appel
+
+        * : Add routine to write mesh info to disk (similar to
+          simul_box_dump).
+
+2006-08-12 20:05  appel
+
+        * : Also write total Kohn-Sham potential.
+
+2006-08-12 17:39  acastro
+
+        * : Bug fixed.
+          
+          The call to write_momentum should be done by all nodes,
+          or else it stops when running in parallels.
+          
+          I removed the last digit of the test; I think that it is asking
+          for too much precision, and that last digit is not obtained
+          when running with different compilers or with different number
+          of processors.
+
+2006-08-12 16:20  acastro
+
+        * : I disconnect the last match of this test, since the code
+          produces different numbers depending on the compiler flags.
+          Obvioiusly something weird is inside the algorithm.
+
+2006-08-12 08:17  xavier
+
+        * : * Renamed the "pol" calculation mode to "pol_fd" and the output
+          file
+          to "linear/polarizability_fd".
+
+2006-08-12 02:10  acastro
+
+        * : Bug fixed.
+          
+          The atomic wave functions of some atoms (like Ag with semicore)
+          could
+          not be calculated by the Schrodinger solver in the hgh module.
+          This
+          little change, for some reason, seems to solve this issue.
+
+2006-08-12 01:24  acastro
+
+        * : Bug fixed.
+          The Ag_sc pseudopotential file had an extra line.
+
+2006-08-10 18:45  appel
+
+        * : Adding routine to calculate total and band resolved density of
+          states. When running with periodic boundary conditions octopus
+          now also writes the density of states (including band resolved
+          DOS) and the fermi energy to 'static'.
+
+2006-08-10 18:44  appel
+
+        * : Adding to X(Hpsi) the possibility to compute (H-E)|psi>.
+          This will be needed for scattering states.
+
+2006-08-10 18:44  appel
+
+        * : Forgot to update header.
+
+2006-08-09 19:38  acastro
+
+        * : After some update of the code (I could not find which one), the
+          last
+          number of this test stopped matching the previous value. However,
+          I
+          find it more likely that the number is correct now, rather than
+          before. So I changed the value of the test.
+          
+          Optimal control should still be tagged as an "experimental
+          feature,
+          anyway.
+
+2006-08-09 17:47  acastro
+
+        * : I deactivate one of the tests that is failing, since it will
+          take some time to figure out why, and it is not relevant at
+          this point.
+
+2006-08-09 12:38  xavier
+
+        * : * Added a very basic printing of frequency dependent
+          hyperpolarizability.
+
+2006-08-09 09:34  appel
+
+        * : Bug fixed.
+          
+          When running parallel in domains octopus was crashing for
+          periodic boundary conditions. This happened because the
+          mesh_index
+          function was assigning an index 0 to several entries of opi%i
+          in nl_operator.F90. Not only the "removed point" was missing
+          in the interior mesh (as it should), but also the mirror point of
+          the "removed point". All of that occured because simul_box_in_box
+          was returning the wrong answer for the "mirror points" in the
+          case
+          of periodic boundary conditions.
+
+2006-08-09 09:10  appel
+
+        * : Similar to the time measurements of td steps, octopus now
+          also prints the elapsed (wall) time of a SCF step. The same
+          format as in the td case is used.
+          This turned out to be useful for estimating how much time
+          octopus spends in Hpsi operations during the scf cycle for
+          a chosen grid size.
+
+2006-08-08 19:53  acastro
+
+        * : Bugs fixed.
+          
+          Related to the failure of the 04-box_oct_long.test
+          (optimal-control).
+          It gave run-time errors and crashed. However, now the tests keeps
+          failing due to discrepancies in the numbers.
+          
+          I do not know now which numbers are correct, so I leave it as it
+          is for the moment.
+
+2006-08-08 16:11  acastro
+
+        * : Bug fixed.
+          
+          The Casida test was failing, when the run-time checks are
+          enabled, due
+          to this very stupid bug. The bug was harmless, when the checks
+          were
+          not done.
+
+2006-08-08 10:58  appel
+
+        * : Add possibility to choose the full Wigner-Seitz cell of the
+          reciprocal lattice for the sampling of the k-points. If the
+          variable FullWignerSeitzCell is set to true no symmetry is
+          taken into account for the choice of k-points (i.e. no
+          Monkhorst-Pack scheme will be used). Instead the k-point
+          sampling will range over the full Wigner-Seitz cell. Currently
+          cubes, octahedrons and dodecahedrons are implemented. The
+          shape of the Wigner-Seitz cell has to be selected manually by
+          setting the variable WignerSeitzCellType. A more sophisticated
+          automatic detection could be added later.
+
+2006-08-06 15:58  appel
+
+        * : Similar to angular momentum, octopus now also computes
+          the expectation values of the momentum operator for the
+          KS states at the end of a ground-state calculation.
+
+2006-08-06 14:30  appel
+
+        * : Octopus can now output also wavefunctions with index > 999,
+          or 1024 respectively.
+          The old bit-shift code in liboct/oct_f.c was clearly more
+          compact than this but it was bound to work only for state
+          indices < 32*32 (increasing val in (1 << val) beyond 31
+          would cause a left shift count which is larger than the width
+          of the type).
+
+2006-08-06 12:01  appel
+
+        * : Bug fixed. The output of files in multi-dataset mode
+          was not always ending up in the right directories.
+
+2006-08-04 17:08  xavier
+
+        * : * Bugfix: Some indexes were wrong, so trying to calculate the
+          lr_elf resulted
+          in segmentation fault.
+          * Added push_sub/pop_sub to lr_calc_current.
+
+2006-08-03 20:25  appel
+
+        * : o Measure total runtime of the testsuite. This could be used
+          to compare how fast octopus is running on a given hardware
+          platform or for a given compiler.
+          
+          o Add option to the script to run testsuite without MPI runs.
+          
+          o Add datestamp to the mail subject of the mail-report.
+
+2006-08-03 12:06  xavier
+
+        * : * Initial version of dynamic hyperpolarizabilities. Now comes
+          testing.
+
+2006-08-03 09:33  xavier
+
+        * : I had disabled this test by mistake.
+
+2006-08-03 08:56  xavier
+
+        * : Updated the matches of long tests to take into account the use
+          of the
+          RR projection.
+
+2006-08-02 20:47  xavier
+
+        * : Nothing important, only a bit of i18n ;-)
+
+2006-08-02 15:28  appel
+
+        * : No output log was appended to the mail reports of the
+          daily testsuite. Let's see if this helps.
+
+2006-08-02 15:26  appel
+
+        * : The suffix was in the case of the MPI executables not really
+          a suffix, rather a "inbetweenfix". Change octopus-suffix_mpi
+          to octopus_mpi-suffix.
+
+2006-08-02 15:16  appel
+
+        * : o Adding Newton-Raphson method for matrices. Starting from a
+          sufficiently good initial guess, this routine can be used
+          to polish inverse matrices. In the case of Vandermonde matrices
+          e.g. the Parker-Traub algorithm can be used to compute an
+          initial guess.
+          Due to the required matrix-matrix multiplication Newton-Raphson
+          is however O(N^3).
+          
+          o Adding routine to estimate the quality of a inverse matrix.
+
+2006-08-02 14:51  appel
+
+        * : Adding an implementation of the Parker-Traub algorithm for the
+          inversion of Vandermonde matrices.
+          The algorithm inverts a Vandermonde matrix in O(N^2) operations
+          compared to O(N^3) for methods that don't take structure into
+          account.
+          
+          Refs.:
+          F. Parker, Inverses of Vandermonde matrices, Amer. Math.
+          Monthly, 71, 1964, p410-411
+          J. Traub, Associated polynomials and uniform methods for
+          the solution of linear problems
+          SIAM Review, 8, No. 3, 1966, p277-301
+          
+          This will be needed for the method that I developed to compute
+          scattering states.
+
+2006-08-01 10:10  appel
+
+        * : Bug fixed. The linear/excitations directory was not created
+          in multi dataset mode because loct_mkdir was used instead of
+          io_mkdir. However, loct_mkdir is not aware of multi datatsets.
+          The 10-casida.test was failing because of this.
+
+2006-08-01 08:54  xavier
+
+        * : Added two new methods: wfs_are_complex and wfs_are_real.
+          
+          Both functions receive an states_t object and return true when
+          wavefunctions are of the type specified in the name and false
+          otherwise.
+
+2006-08-01 08:50  appel
+
+        * : Forgot ${exec_suffix} inside mail_report().
+          
+          ... we are getting close to revitalizing the nightly builds and
+          regression tests.
+
+2006-08-01 08:32  appel
+
+        * : Also the variable "last" was used without beeing intialized
+          first.
+
+2006-08-01 08:30  appel
+
+        * : o Don't use compile flags as linking flags.
+          
+          o Remove duplicated preprocessing code. Don't delete prerequisite
+          instead.
+
+2006-07-31 19:03  xavier
+
+        * : I forgot to put some of the lines required to create
+          atomic_oct.f90.
+
+2006-07-31 18:07  xavier
+
+        * : * The FCDEBUGFLAGS was not used when linking.
+          
+          * The atomic.o file was rebuilt every time, changing the rule
+          from
+          "atomic.o: atomic_oct.f90" to "atomic.o: atomic.F90" solves the
+          problem.
+          
+          * Using FCDEBUGFLAGS is possible to compile octopus using
+          interfile
+          optimizations, for example -ipa with pathf90 or -ipo with ifort.
+          This
+          is because with this flags, the compiler doesn't produce binary
+          objects until the linking step, so the configure script doesn't
+          work
+          properly.
+
+2006-07-31 17:57  xavier
+
+        * : * More explicits bounds in vectorial operations.
+          
+          * The quadratic box test has undeterminisc behaviour. Leaving the
+          default mixing instead of Mixing=0.5 "solves" the problem, but I
+          am
+          not 100% sure that this is the right thing to do.
+
+2006-07-31 16:14  xavier
+
+        * : * Using the Rayleigh-Ritz subspace diagonalization, the
+          unstability of
+          the self consistency iteration disappears, so I am enabling it by
+          default. There is still the problem of non-determinisc behaviour,
+          but
+          now it takes 31 or 32 iterations for benzene and not 32, 33 or
+          127. I
+          still have to check what happens with other compilers.
+          
+          * The results of some test changed slightly with this. So i
+          updated
+          them.
+          
+          * I put back the NP_PART size in CG eigensolvers, but i couldn't
+          find
+          why it should be like that.
+
+2006-07-31 12:20  xavier
+
+        * : A bit of cleaning in the CG Eigensolver:
+          
+          * All arrays were defined with NP_PART size, altough it's not
+          needed.
+          I only kept NP_PART size in arrays were the hamiltonian is
+          applied.
+          * Put explicit limits in vectorial operations.
+
+2006-07-31 10:19  appel
+
+        * : Besides FCFLAGS there is now a variable FCDEBUGFLAGS. I had to
+          separate debug flags from regular flags since some files cannot
+          be compiled with debug flags. The prime example is atomic.F90.
+          If this file is compiled with a full set of debug flags all sorts
+          of runtime errors occur, causing a lot of tests to fail. I guess
+          nobody really bothers to fix the issues in the F77 spaghetti
+          code.
+          Therefore we exclude this file from runtime checking so that we
+          can enable the cheks in the nightly builds *for the rest of the
+          files*.
+          
+          To achive a conditional compilation with different FC flags I
+          had to introduce in src/Makefile.am pattern based rules instead
+          of suffix rules for the compilation
+
+2006-07-31 09:57  xavier
+
+        * : Put some explicit limits in vectorial operations. This is an
+          (failed) attempt
+          to solve bug 37.
+
+2006-07-30 20:55  appel
+
+        * : The variable first was being used without being defined.
+
+2006-07-30 18:57  appel
+
+        * : else clauses where missing for the case when no exec suffix
+          is given.
+
+2006-07-30 17:00  appel
+
+        * : Seems to be smarter to sort this out by symlinks.
+
+2006-07-30 16:38  appel
+
+        * : Also need to take care of manual installs.
+
+2006-07-30 15:55  appel
+
+        * : o Add also program suffix for executables build with NAG.
+          o Add missing LDFLAGS for MPI flavours.
+
+2006-07-30 15:02  appel
+
+        * : Add support for program suffixes like -debug. This allows to
+          run the testsuite for different flavours of executables (e.g.
+          with or without bound checking, etc).
+
+2006-07-30 14:58  appel
+
+        * : o On amd64 linking to the gdlib causes octopus to segfault
+          always when
+          statically linked executables are used (independent of the
+          compiler).
+          octopus doesn't even get to the point where it reads the input
+          file.
+          Add --disable-gdlib so that -Bstatic/-static can be used again.
+          With
+          commerical compilers the -static options are more important than
+          having
+          gdlib support, I think.
+          
+          Since the nightly builds were not running while I was away, this
+          was not
+          captured. In the end I had to do bisection in the revision
+          numbers 2156-2310
+          to figure out that this problem was introduced with Miguels
+          commit in
+          r2185/r2186.
+          
+          o Adding optimized and debug executables to the package. They are
+          build with
+          the Intel compiler. The executables carry the program suffix
+          -intel-9.1 and
+          -intel-9.1-debug respectively. This is mainly used for nightly
+          builds and
+          regression testing. It increases the size of the packages
+          considerably,
+          so that we will use it only for internal purposes.
+          
+          o Remove outdated metis configure options.
+
+2006-07-29 01:20  appel
+
+        * : o adding extra treatment for scalar meshes with only a single
+          point.
+          o removing variable initialization inside the type definition
+          of scalar_mesh_t. this is not conform with the f90 standard.
+
+2006-07-29 01:16  appel
+
+        * : o adding calling interface for transpose operators. this is
+          required
+          for solvers like bi-conjugate gradients or some gmres flavours.
+          o adding verbosity flag and a bit more relaxed warning criteria.
+
+2006-07-29 00:56  appel
+
+        * : o adding emacs encoding lines also to the testsuite files.
+          (I've chosen shell-script mode to get a reasonable syntax
+          highlighting)
+          o some more missing svn:keywords
+
+2006-07-29 00:50  appel
+
+        * : o add emacs encoding line
+          o release preparation: update year entries in
+          the header (200x -> 200x-2006)
+
+2006-07-28 20:27  xavier
+
+        * : * Now when the restart lr density is not found for a certain
+          frequency, then octopus uses the restart lr density of the
+          closest
+          frequency available.
+
+2006-07-28 20:24  appel
+
+        * : Removing the nl_operator_operate_cmplx routine. Complex weights
+          are now treated by nl_operator_operate directly if op%cmplx_op
+          is enabled.
+
+2006-07-28 20:14  appel
+
+        * : Bug fixed. This was a copy'n paste mistake. Should be
+          op%cmplx_op
+          instead of op%const_w.
+
+2006-07-28 20:12  appel
+
+        * : stencil_center is a new field in nl_operator_t. The variable
+          contains the index value of the central stencil point (i.e. the
+          position which matches the application point of the stencil).
+
+2006-07-28 20:09  appel
+
+        * : o Fixing svn:keywords for the testsuite
+          o Adding a bunch of missing $Id:$
+
+2006-07-28 20:07  xavier
+
+        * : As pointed by Heiko, nspin was not correctly initialized. In
+          this case
+          is always 1, because information is not separated by spin
+          components.
+
+2006-07-28 19:50  appel
+
+        * : Remove unused variables.
+          
+          Xavier: can you have a look at this warning of the NAG compiler
+          
+          Warning: em_resp.F90, line 578: Symbol NSPIN referenced but never
+          set
+          detected at DYNAMIC_OUTPUT_INIT@<end-of-statement>
+          
+          Thanks.
+
+2006-07-28 19:49  appel
+
+        * : o The NAG compiler was complaining about too long subroutine
+          names (> 31 characters)
+          
+          o Adding GPL header to _inc file
+
+2006-07-28 19:49  appel
+
+        * : o Remove unused variables.
+          o Use tabstops of 2 instead of 4.
+          o Cleanup formatting a bit to make this code a
+          bit more coherent with the rest.
+
+2006-07-28 18:40  appel
+
+        * : svn:keywords "Id" property was missing for these files.
+
+2006-07-28 18:37  appel
+
+        * : o Currently the generation of info files from svn sources is
+          broken. Uncomment dh_installinfo so that deb packages can be
+          build without info files (master source for documentation is
+          the wiki anyways).
+          
+          o Build executables for x86_64 dynamically (used to revitalize
+          the nightly builds and testuite runs).
+
+2006-07-28 18:34  appel
+
+        * : If available use svn info to extract the current revision
+          number.
+          In this way the find run is avoided. Gives a noticeable speedup
+          on slow disks (e.g. Laptops).
+
+2006-07-28 11:31  acastro
+
+        * : Bug fixed.
+          
+          One more on the problem of passing strings from Fortran to C.
+
+2006-07-27 23:37  xavier
+
+        * : Now restart information for linear response is stored for every
+          frequency. Only the delta density is written, this appears to be
+          a
+          good compromise between disk space and calculation time.
+
+2006-07-27 22:56  acastro
+
+        * : Bug fixed.
+          
+          One of the optimal control tests was failing because of this.
+
+2006-07-27 21:39  acastro
+
+        * : About thirty something tests were failing with the NAG compiler
+          on the
+          opterons. The reason for most of the bugs is removed with the
+          change
+          that I now commit, which has to do with the Fortran to C
+          interface --
+          especially the string handling --, and with the difference
+          between int
+          and unsigned long int in 64 bit machines (I think).
+          
+          *However*:
+          This change removes the bug, but I am very uncomfortable changing
+          these subtleties in C, so I ask any of the C experts to take a
+          look.
+          
+          *And*:
+          Still five tests fail, due to unrelated reasons.
+
+2006-07-26 13:39  xavier
+
+        * : Corrected indentation.
+
+2006-07-26 10:44  xavier
+
+        * : * Now static polarizability is calculated by the dynamic
+          polarizability code. The old code can be used by setting the
+          PolStatic
+          variable.
+          * Updated tests.
+
+2006-07-26 09:13  xavier
+
+        * : Bigfix: An array was deallocated even when not allocated.
+
+2006-07-25 13:26  acastro
+
+        * : The construction of excited states was not possible in
+          the one-electron case if the spin-unpolarized mode was being
+          used. This was annoying.
+
+2006-07-24 09:47  xavier
+
+        * : Bugfix: pop_sub() wasn't been called when using return to exit.
+
+2006-07-14 10:31  xavier
+
+        * : * The sign of the complex part of polarizabilty was uncorrect.
+          * Now a linear/cross_section_tensor file is created, from the
+          information of
+          linear response. The format is the same of the
+          cross_section_tensor file
+          generated by oct-cross-section, altough it is not separated by
+          spin.
+
+2006-07-14 10:17  xavier
+
+        * : Added the option to print the values of the coordinates over
+          the grid, this
+          is not very informative, but can be used to calculate new
+          functions from
+          other functions, for instance dipole density can be calculated as
+          the
+          density times r.
+
+2006-07-11 22:05  janwer
+
+        * : removed some debug statements, td local targets should work
+
+2006-07-11 21:33  acastro
+
+        * : Bugs fixed.
+          
+          1. The loct_parse_expression routines were not working properly,
+          because the conversion from "Fortran string" to "C string" was
+          not
+          done.
+          
+          2. In the laser_field routine, when the laser field was read from
+          file, the index "i" was used to loop over the various laser
+          fields,
+          but in this case there is only one (this should be eventually
+          extended). This caused an out of bounds.
+
+2006-07-11 13:29  janwer
+
+        * : bug fixed in laser.F90. Now we can read the optimized field
+          directly for a td calculation.(thanks to Alberto)
+
+2006-07-10 18:33  janwer
+
+        * : I have changed the routines that read a laser pulse from a
+          file, so that it is compatible with the output of the optimal
+          control run. (Very likely that it is still buggy...).
+          The user interface for td targets was also modified (still
+          buggy..).
+          Will fix both in the next days.
+          
+          jan
+
+2006-07-10 09:47  xavier
+
+        * : Fixed another memory leak.
+
+2006-07-08 10:06  xavier
+
+        * : Fixed a memory leak when \delta elf was calculated several
+          times.
+
+2006-07-07 15:49  xavier
+
+        * : * Now functions are printed even when calculating
+          polarizabilities for
+          several frequencies. The files go to linear/freq_w (with w the
+          frequency).
+          
+          * Modified the dynamical linear response tolerance code, so its
+          value is
+          never decreased with respect to the previous iteration.
+
+2006-07-07 14:54  acastro
+
+        * : Bug fixed.
+          The calculation of the determinant stopped emitting an error if
+          the
+          determinant was null. It should only stop if there is an attempt
+          to
+          calculate the inverse matrix.
+
+2006-07-07 12:36  acastro
+
+        * : Changed a bit the interface to the occupied_states subroutine.
+          Now it is used by the casida module.
+
+2006-07-06 19:26  acastro
+
+        * : I got tired of looking up the documentation for the possible
+          envelopes
+          of the laser field, so I added the corresponding !%Option in the
+          documentations for the TDLasers block.
+
+2006-07-05 16:12  xavier
+
+        * : * Now the LR SCF uses variable precision to calculate the
           wavefunctions.
-
-2006-01-13 16:41  xavier
-
-        * src/static_pol_lr.F90: Only a small modification, I was
-          multiplying and the diving by the same number.
-
-2006-01-12 18:58  acastro
-
-        * src/: epot.F90, global.F90, h_inc.F90: Bug fixed.
-
-          There was no definition of the gyromagnetic ratio; now it is
-          input by variable "GyromagneticRatio". Of course it is, in
-          principle, a physical constant, but this led to an error when
-          calculating the Zeeman term of quantum dots, which should use an
-          "effective" gyromagnetic ratio.
-
-          I still have some doubts about factors and signs in the Zeeman
-          term.
-
-2006-01-12 17:59  xavier
-
-        * src/static_pol_lr.F90: * The hyperpolarizability tensor now is
-          symmetrized by taking the average of all the corresponding
-          components.
-
-          * Now the output file only contains the components that are
-          different.
-
-2006-01-12 14:37  micael
-
-        * configure.ac, src/Makefile.am:  *) Bug fix: disable-debug mode
-          was not working.
-
-2006-01-12 10:02  micael
-
-        * src/v_ks_inc.F90, testsuite/finite_systems_3d/spinors.test:  *)
-          There was a bug in the calculation of the expectation value of
+          * Added the posibility to print the polarizability density.
+
+2006-07-04 17:41  xavier
+
+        * : Updated test to match previous commit.
+
+2006-07-04 17:40  xavier
+
+        * : * The imaginary eta term was been applied with the wrong symbol
+          to the
+          equation for negative frequency perturbation.
+
+2006-07-04 16:25  janwer
+
+        * : introduced polarization dependent filters and psiT.0xx is
+          produced during optimization,
+          to monitor the improvement during the iterations (thanks, Esa).
+
+2006-07-04 15:05  acastro
+
+        * : The calculation of the populations of the ground state and/or
+          excited
+          states is now done without stupid repetitions of calculations,
+          that
+          made it very slow. Nevertheless, for large systems it may still
+          be a
+          slow calculation.
+          
+          Also, in order to do these calculations, one has to use now the
+          keyword "populations", not "gs_proj", as before. [This keywork
+          had to be
+          passed to the "TDOutput" input variable]. Since probably nobody
+          except
+          me was using this (I think), I hope that nobody cares. The
+          generated
+          file also changed name, to "populations". These changes describe
+          more
+          accurately what the file contains: populations of the ground
+          state and
+          possibly also of excited states, no only the ground state as
+          before.
+
+2006-07-04 09:33  acastro
+
+        * : The calculation of the matrix of dot products between the
+          orbitals of
+          two Slater determinants is now an independent routine. This is to
+          avoid recomputations.
+          
+          Only one interface for the many-particle dot products:
+          X(states_mpdotp), valid for both dot products between single
+          Slater
+          determinants, and between single Slater determinants and linear
+          combinations.
+
+2006-07-03 22:01  xavier
+
+        * : Bugfix: Replaced FC_FUNC by FC_FUNC_.
+
+2006-07-03 18:25  acastro
+
+        * : Bug fixed.
+          
+          A couple of problems in the lcao part, that could something
+          to do with ticket #37. But I do not know if it solves it since
+          I did not attempt to reproduce it.
+          
+          The uninitialized variables problem can be usually tracked well
+          by passing the -fpe0 -ftrapuv options when using the Intel
+          compilers.
+
+2006-07-03 17:24  xavier
+
+        * : Some tests were calling the "lanczos" eigensolver while there
+          is no
+          eigensolver option called like that. The lanczos option belongs
+          to the
+          TDExponentialMethod variable. Numerically this value matched with
+          'plan'
+          option, preconditioned lanczos. To avoid confusions the test were
+          corrected
+          and another number was asigned to the 'plan' option.
+          
+          WARNING: Some existing input files can fail after this change, if
+          this is the case
+          just put 'plan' instead of 'lanczos' in the EigenSolver variable.
+
+2006-07-03 15:44  xavier
+
+        * : * I was going to implement the smoothing preconditioner of Saad
+          to use it for
+          linear response, when i discovered that it was already
+          implemented :-D . It
+          is used for the preconditioned Lanczos eigen solver. So I moved
+          it into a
+          new module called preconditioners_m, hopefully it'll have more
+          preconditioners in the future.
+          
+          * I also submit a fix for the docu of MainAxis, that had the
+          wrong format.
+
+2006-07-03 10:33  acastro
+
+        * : Bug fixed. Forgot to put one "skip_header" so that an
+          "excitations"
+          file could be properly read.
+
+2006-07-02 12:34  xavier
+
+        * : * The configure script was using AC_FUNC_MALLOC and
+          AC_FUNC_REALLOC, this
+          two macros search for a GNU compatible malloc and realloc. If
+          they don't
+          find it they define malloc as rpl_malloc, the problem is that
+          later in the
+          configure the gsl check tries to use malloc and fails. This bug
+          was seen in
+          OSF, using the Compaq C compiler. As we don't need a gnu
+          compatible malloc
+          (the difference is the way it handles zero size allocations) the
+          presence of
+          this functions is now checked by AC_CHECK_FUNCS.
+          
+          * Documented two variables, fromScratch and MainAxis.
+
+2006-07-01 19:31  xavier
+
+        * : Put explicit boundaries in a vectorial operation that gfortran
+          didn't like.
+
+2006-06-29 23:09  xavier
+
+        * : Moved the variable documention of MultigridLevels to Mesh
+          section.
+
+2006-06-29 18:29  acastro
+
+        * : * Bug fixed: Some division by zero could be done when printing
           the
-              exchange-correlation potential.
-
-2006-01-11 13:56  micael
-
-        * src/epot_inc.F90, src/h_so.F90, testsuite/rc_bugs/12-spin_orbit:
-          *) Spin-orbit bug fixed.
-
-2006-01-11 09:04  appel
-
-        * src/simul_box.F90: Bug fixed: The array lsize in simul_box_in_box
-          had a wrong index. When running with BoxShape = parallelepiped
-          the mesh was not symmetric around the origin (in z-direction) and
-          had the wrong size.
-
-2006-01-04 23:06  acastro
-
-        * src/td_rti.F90, testsuite/finite_systems_3d/restart.test,
-          testsuite/rc_bugs/10-Restart: Bug fixed.  Another one of the
-          NP/NP_PART that so much fun we are having with.  It affected the
-          "etrs" evolution method.
-
-          Now the restart.test is finally fully passed, so I enabled it
-          back, and removed the rc_bugs/10-Restart file.
-
-2006-01-04 19:40  acastro
-
-        * src/xc_vxc.F90, testsuite/finite_systems_2d/2dbox.test,
-          testsuite/rc_bugs/11-2dbox: Bug fixed.
-
-          This one was pretty serious,  since it affected all the
-          spin-polarized calculations,  at least  in  some  computers, such
-           as  in the  Berlin opteron  cluster.  It was  the  reason that
-          some  of  the tests  were disabled.
-
-          Now the 2dbox.test  and the first two matches  of the
-          restart.test are passed  OK  in the  opteron.  However, the  last
-           two  matches of  the restart.test  are  not  passed  in  the
-          opteron,  due  to  some  other different bug (so I did not enable
-          it again), and the 2dbox.test fails in my  PC. This is why
-          although  I enabled again 2dbox.test,  I do not remove its file
-          in the rc_bugs directory.
-
-2005-12-28 16:10  acastro
-
-        * src/poisson.F90: Just one warning.
-
-          It has to do with the bug reported by Francesco Sottile, about
-          divergent results with some 1.99 version and the rc1. In fact,
-          there is no such difference, but the fact that the cg_corrected
-          Poisson solver gives wrong results when using the "minimum" box
-          shape. The two runs reported by Francesco were done with
-          different solvers: fft and cg_corrected.
-
-          I have been trying to figure out if it is a bug, but I am
-          starting to think it is a fact: the method gives poor results
-          with odd box geometries. This is why I thought about putting a
-          warning/disclaimer, so that users don't get too mad, or screw
-          them if they do.
-
-          I always thought that cg and cg_corrected were essentialy
-          equivalent in the error that they provided, the difference being
-          that the latter is faster.  But it seems not to be true; the cg
-          method seems to behave better with odd geometries. That is why in
-          the warning I recommend using cg.
-
-          In any case I will try to investigate why and how this happens.
-
-2005-12-20 15:50  acastro
-
-        * src/scf.F90: The clean_stop was not working during the SCF
-          procedure since it actually was called twice -- one for checking
-          and then to really exit the SCF loop.  Since now the function
-          erases the "stop" file, the second call found no file at all.
-
-2005-12-20 13:48  xavier
-
-        * src/static_pol_lr.F90: Correction of an error message.
-
-2005-12-19 16:47  marques
-
-        * src/Makefile.am: Added integer.F90 to the Makefile so that it is
-          included in the .tar.gz
-
-2005-12-19 13:38  acastro
-
-        * src/: geometry.F90, xyz_file.F90: Corrected the bugs pointed out
-          by Carlo Camilloni.  I just had to apply the patches that he had
-          suggested, I think that they were just what was needed.
-
-2005-12-17 10:14  marques
-
-        * doc/: tutorial.tex, img/elf.eps, img/laser.eps, img/pi02.eps,
-          img/pi03.eps, img/pi04.eps, img/spacing.eps, img/spacing.ps,
-          pdf/Makefile.am, pdf/tutorial.pdf: The tutorial was moved to the
-          wiki. Therefore, I delete the corersponding files from the
-          distribution.
-
-2005-12-16 10:22  marques
-
-        * doc/pdf/Makefile.am: References to octopus-devel removed
-
-2005-12-16 10:21  marques
-
-        * doc/Makefile.am: References to octopus-devel removed.
-
-2005-12-16 10:20  marques
-
-        * doc/: octopus-devel.info, octopus-devel.texi, octopus.info,
-          octopus.texi, pdf/octopus-devel.pdf: The information contained in
-          the -devel.texi file was passed to the wiki. Therefore I am
-          deleting these files.
-
-2005-12-12 14:01  appel
-
-        * doc/html/Makefile.am: make was aborting, probably just a typo:
-          make[3]: *** No rule to make target `vars/*', needed by `all-am'.
-           Stop.
-
-          I replaced vars/* by vars*
-
-2005-12-12 12:59  marques
-
-        * doc/pdf/: octopus-devel.pdf, octopus.pdf, tutorial.pdf: The pdfs
-          are now updated
-
-2005-12-12 12:53  marques
-
-        * doc/Makefile.am, doc/html/Makefile.am, share/util/Makefile.am,
-          src/Makefile.am, build/octopus.spec.in, testsuite/Makefile.am:
-          There were several "bugs" in the Makefiles that made "make rpm"
-          not work.
-
-2005-12-12 12:52  marques
-
-        * configure.ac: Added one Makefile.am
-
-2005-12-12 09:34  marques
-
-        * configure.ac, doc/octopus.texi: As Heiko suggested, I removed the
-          "_" from the version number
-
-2005-12-12 09:33  marques
-
-        * src/: epot.F90, epot_inc.F90, h_inc.F90, states_inc.F90,
-          v_ks_inc.F90, xc_functl.F90: some cleaning... nothing really
-          important, except that I changed index->iatom in the
-          projector_type.
-
-2005-12-12 09:31  marques
-
-        * configure.ac, doc/octopus.texi: Retagged 1.88->2.0_rc1
-
-2005-12-11 16:45  marques
-
-        * src/spectrum.F90: Bug fixed. BTW, the _info routines in this file
-          are all terribly similar. Can we somehow merge them into one?
-
-2005-12-08 18:07  micael
-
-        * testsuite/rc_bugs/12-spin_orbit:  *) Adding a bug file:
-          spin-orbit is sometimes causing a segfault.
-
-2005-12-08 09:55  marques
-
-        * src/: mix.F90, td_write.F90: These continuation lines were
-          skrewing my octopus analyzer tool. Please try not tu use "&"
-          inside a type definition.
-
-2005-12-07 23:26  appel
-
-        * testsuite/oct-run_regression_test.pl: Forgot $ in front of opt_m.
-
-2005-12-07 22:51  appel
-
-        * testsuite/rc_bugs/02-Na_Chain: This input file is by now running
-          properly and can be removed.
-
-2005-12-07 22:50  appel
-
-        * testsuite/oct-run_regression_test.pl: Added option -m which
-          allows to perform only the matches of a given test without
-          running the code. This turned out to be useful for checking the
-          numbers in an existing working directory.
-
-2005-12-07 22:49  appel
-
-        * src/: gs.F90, lcao.F90, scf.F90, unocc.F90: In my opinion this is
-          definitely a compiler bug of the NAG compiler. If the arguments
-          of the routine lcao_wf are not sorted according to their intent
-          flavour, the compiler creates an executable that produces
-          non-deterministic! output, i.e.  you can run the very same
-          executable with the very same input several times from scratch
-          and you get different numbers (probably some garbage numbers from
-          memory were collected).  This was very akward to debug.
-          Especially since I had to run the code xillion times, as not
-          every run was giving the wrong numbers ...  Finally I wasted a
-          full day on this :(, but at least the 3D/1D periodic stuff is now
-          running properly on 64Bit and one of the tests in rc_bugs can be
-          removed.
-
-          Probably it is a good idea to sort all other arguments that are
-          passed to the routines in the code in a similar fashion.
-
-2005-12-07 22:47  appel
-
-        * testsuite/periodic_systems_1d/sodium_chain.test: Adding a test
-          for a chain of sodium atoms. Finally we have the first realistic
-          test for 3D/1D periodic systems. To speed up the runtime of the
-          test I had to stop the SCF cycle after one step, i.e. be careful,
-          the results are not converged.
-
-2005-12-07 17:17  marques
-
-        * src/multicomm.F90: Thre was a bug in the distribution of the
-          nodes that made the code crash with some combination of number of
-          nodes.
-
-2005-12-07 09:19  appel
-
-        * debian/info: Remove octopus-faq.info from the build system.
-
-2005-12-06 16:09  acastro
-
-        * src/casida.F90: Bug fixed.  The casida module was breaking when
-          using non-interacting electrons.  It is a bit absurd to run the
-          casida module with non-interacting electrons, but it is nice for
-          testing, and the program should not break like that in any case.
-
-2005-12-06 14:58  acastro
-
-        * src/scf.F90: I guess that line was there just for debugging...
-
-2005-12-06 10:52  marques
-
-        * doc/: Makefile.am, installation.texi, octopus-faq.info,
-          octopus-faq.texi, octopus.info, octopus.texi: Removed octopus-faq
-          and some info from octopus.texi that are now in the wiki.
-
-2005-12-04 22:17  marques
-
-        * src/: datasets.F90, syslabels.F90: This file was rename from
-          src/syslabels.F90 to src/datasets.F90 as explained in the
-          previous commit.
-
-2005-12-04 22:15  marques
-
-        * src/Makefile.am, src/broad.F90, src/casida.F90,
-          src/centergeom.F90, src/cross_section.F90, src/curv_briggs.F90,
-          src/curv_gygi.F90, src/curv_modine.F90, src/curvlinear.F90,
-          src/derivatives.F90, src/eigen.F90, src/fftw2.F90, src/fftw3.F90,
-          src/geom_opt.F90, src/geometry.F90, src/gs.F90, src/h.F90,
-          src/hs.F90, src/io.F90, src/lasers.F90, src/linear_response.F90,
-          src/main.F90, src/math.F90, src/messages.F90, src/mix.F90,
-          src/mpi_debug.F90, src/multicomm.F90, src/ode_solver.F90,
-          src/opt_control.F90, src/out.F90, src/out_inc.F90,
-          src/phonon_spectrum.F90, src/phonons.F90,
-          src/poisson_multigrid.F90, src/profiling.F90, src/restart.F90,
-          src/restart_inc.F90, src/root_solver.F90,
-          src/rotational_strength.F90, src/run.F90, src/scalar_mesh.F90,
-          src/scf.F90, src/simul_box.F90, src/sparskit.F90, src/specie.F90,
-          src/spectrum.F90, src/states.F90, src/static_pol.F90,
-          src/static_pol_lr.F90, src/systm.F90, src/td.F90,
-          src/td_write.F90, src/units.F90, src/unocc.F90, src/v_ks.F90,
-          src/wave_matching.F90, src/xc.F90, src/xc_functl.F90,
-          src/xyz_adjust.F90, src/xyz_file.F90, src/xyzanim.F90,
-          testsuite/periodic_systems_1d/cosine_potential.test,
-          testsuite/periodic_systems_1d/free_electrons.test: *) I finally
-          cleaned the run.F90 file. The reason is that, with the new
-          changes    allowing for multi-parallelization, the ability to,
-          e.g., run the gs if the    td calculation did not find the
-          restart wave-functions had to be removed,    as there could be
-          inconsistencies in the parallelization method. Now, if you
-          want to run both the gs and td in the same file, use the
-          %CalculationMode    block.  *) I changed slightly the
-          initialization of the old syslabels stuff, allowing    for
-          sensible defaults for the variables. Now the only thing you
-          really have    to say are the CalculationModes for each run. This
-          implied changing the order    of the lines in the block (now
-          modes, labels, and the order). You may have    to change you
-          input files to reflect this change.  *) To make this multilpe
-          runs useful, they had (sometimes) to share the same    tmp
-          directory (otehrwise, one couldn't start a td calculation after a
-          gs).     Until now they all had different tmps. Now by default
-          they have all the same     "tmp", but this can be changed by the
-          variable TmpDir.  *) I rename "syslabels" to "datasets". The
-          reason is that ABINIT has this    multiple datasets for years,
-          and they work more our less like ours. So I    thought it was a
-          good idea to make the names more compatible.
-
-2005-12-01 17:51  xavier
-
-        * src/static_pol_lr.F90: Fixes in dynamic polarizability, now it
-          gives the correct result for w = 0.
-
-2005-11-30 16:28  micael
-
-        * share/util/: vf.cfg, vf.net:  *) This is a DX program I wrote
-          some time ago to plot vector fields (very usefull to
-              plot currents ;) Probably it could be greatly improved, or
-          even merged into the
-              existing mf.net program, but for now I'm just commiting what
-          I have.
-
-2005-11-30 16:11  micael
-
-        * src/spectrum.F90:  *) Added the pre-factors that were missing
-          from the complex rotatory strength function.
-              I also added two sum rules.
-
-2005-11-30 15:17  marques
-
-        * src/: epot.F90, epot_inc.F90: I believe that this solves the
-          problem of Carlo (please check). There was a R_CONJ missing in
-          the non-local part of the pseudopotential.
-
-2005-11-30 14:18  rozzi
-
-        * src/static_pol_lr.F90: Intel compiler was confused by unmatched
-          single quote
-
-2005-11-30 11:09  xavier
-
-        * src/static_pol_lr.F90: * Dynamic polarizability is implemented
-          but not tested yet.  * Now an array of lr objects is used instead
-          of only one, indexes are direction, sign and frequency.
-
-2005-11-30 11:05  micael
-
-        * src/scf.F90:  *) Only the root node should output the local
-          magnetic moments.
-
-2005-11-30 10:51  marques
-
-        * src/: cross_section.F90, hs.F90, spectrum.F90, td_exp.F90,
-          td_init.F90, td_rti.F90: Rearranged a little the sections to
-          which the variables belong to.
-
-2005-11-30 10:14  marques
-
-        * liboct_parser/parser_f.c, share/variables.local, src/epot.F90,
-          src/xc.F90, src/xc_OEP.F90, src/xyz_adjust.F90: *) Documented a
-          couple more variables, and made the corresponding changes in
-          variables.local *) Fixed a bug in parse_expression. There is a
-          variable in symrec name def, that    defines if a variable has
-          been defined, and that we should not forget to set.
-
-2005-11-30 10:10  marques
-
-        * testsuite/finite_systems_1d/helium.test: I don't see any reason
-          why the helium test should not be done with th real executable.
-
-2005-11-30 10:07  marques
-
-        * testsuite/periodic_systems_1d/: cosine_potential.test,
-          free_electrons.test: The Units should be "a.u" and not a.u. As
-          this is the default value, I took away those lines.
-
-2005-11-30 00:27  appel
-
-        * src/Makefile.am: Forgot oct-make-st in EXTRA_PROGRAMS.
-
-2005-11-29 23:51  appel
-
-        * src/Makefile.am: make_st can also be removed from Makefile.am.
-
-2005-11-29 23:33  appel
-
-        * liboct_parser/parser_f.c, liboct_parser/symbols.c,
-          src/liboct_parser.F90, src/make_st.F90, src/specie.F90,
-          src/states.F90, src/states_inc.F90, src/td.F90: I know that we
-          are actually in feature freeze. But I believe that this is a must
-          have feature for every td code: there is now the possibility to
-          specify arbitrary user defined orbitals as initial states for the
-          time propagation.
-
-          Similar to user defined species this is done with formula strings
-          in the input file:
-
-          %UserDefinedStates    1 | 1 | 1 | "exp(-r^2)*exp(-i*0.2*x)" %
-
-          The first three columns specify dim, nst and nik value of the
-          respective orbital. The last column specifies the formula for the
-          orbital. Normalization constants are not required as the code
-          takes care of that after reading the formula. Also note, that the
-          parser now accepts "i" for the imaginary unit.
-
-          Currently octopus reads all orbitals from the restart_gs
-          directory and only the ones specified in the above block will be
-          replaced in the restart_gs determinant. Combined with ExtraStates
-          this can be used e.g. to setup scattering states.  The
-          modifications also obsolete the make_st program.
-
-2005-11-29 23:17  appel
-
-        * src/restart.F90: Only the root node should delete the stop file.
-
-2005-11-29 23:16  appel
-
-        * src/: out.F90, out_inc.F90: Using 'e' in the default Fortran
-          output format causes for very small numbers output like
-          9.43732460066152-133 instead of 9.43732460066152E-133. Gnuplot
-          cannot cope with the missing E.  Therefore I switched to 'f' in
-          gnuplot_mode.
-
-2005-11-29 23:15  appel
-
-        * src/messages.F90: The debug mode was extremely slow for the MPI
-          executable. Now the debug_trace files in the debug directory are
-          written only if DebugLevel >= 100. For smaller values only the
-          root node writes to stderr.
-
-2005-11-29 23:14  appel
-
-        * src/string.F90: Added routine to convert Fortran strings to C
-          strings.
-
-2005-11-29 23:14  appel
-
-        * src/lcao.F90: To deallocate only X(psi) a full fledged
-          states_end() was called in lcao_end. However, since there is no
-          states_init() in lcao_init() there were some arrays not
-          available. I got segfaults due to this (although this shouldn't
-          happen since states_end() checks if the pointers are associated.
-          Perhaps a compiler bug?).
-
-2005-11-29 23:13  appel
-
-        * src/io.F90: Now also flush_messages can be enabled/disabled on
-          the fly. Simply touch the files
-          enable_flush_messages/disable_flush_messages.
-
-2005-11-29 15:59  marques
-
-        * src/: epot.F90, fftw3.F90, mesh.F90, mesh_init.F90: A phase was
-          wrong in the code. The most important thing was a minus sign
-          missing. I believe that the potential is OK now.
-
-2005-11-28 17:32  xavier
-
-        * src/spectrum.F90: There was a problem reading the multipoles file
-          in spectrum_hs_from_mult.
-
-2005-11-28 15:50  marques
-
-        * src/spectrum.F90: There were a couple of '#' missing, which made
-          oct-cross-section crash while readin the _vector files. I also
-          took the opportunity of removing several lines that gave warning
-          with ifc7.
-
-2005-11-28 14:45  marques
-
-        * src/td_rti.F90: removed some unused lines of code.
-
-2005-11-28 14:41  marques
-
-        * liboct_parser/grammar.y, liboct_parser/parse.c,
-          liboct_parser/parse_exp.c, liboct_parser/symbols.c,
-          liboct_parser/symbols.h, src/poisson_multigrid.F90,
-          src/td_exp.F90, src/td_rti.F90: *) Added a very tiny feature to
-          the parser. Now it will stop with a input    error if one
-          attempts to use a variable that has not been defined. This
-          avoids misspells in the input file to be acepted as valid inputs.
-          (most of the diff is whitespace, as the files were indented
-          with tabs,     when they should be indent with spaces)  *) This
-          allowed to find a couple of variables that were using "-". This
-          is    forbidden, as he "-" is going to be confused with the minus
-          sign. For    example, suzuki-troter is read as "suzuki" -
-          "troter", and as these two    variables have not been defined
-          (i.e. are 0), the result will be 0, and    not the expected
-          value.
-
-2005-11-28 12:03  marques
-
-        * libxc/: lda.c, lda_c_vwn.c, lda_x.c, test.c: *) very tiny bug
-          corrected in VWN (one of the constants was wrong in the 5th
-          digit) *) fxc is now calculated numerically a bit better in case
-          of the ferromagnetic gas. Actually I don't really know how to
-          handle this case properly, as fxc should diverge (at least fx).
-
-2005-11-27 10:32  appel
-
-        * testsuite/oct-run_regression_test.pl: There was a \ missing.
-          Without the \ the -c option was creating buggy testfiles.
-
-2005-11-27 10:30  appel
-
-        * testsuite/: finite_systems_2d/2dbox.test,
-          finite_systems_3d/restart.test, rc_bugs/10-Restart,
-          rc_bugs/11-2dbox: There's something fishy with the precision on
-          64Bit. The last input in the 2dbox test takes currently 114 scf
-          cycles to converge on x86_64 but requires only 14 on i386. Also
-          the eigenvalues and total energy come out differently.  Similar
-          things apply to the last input file in restart.test.
-
-          To avoid spamming of octopus-devel I've disabled both tests
-          (Enabled: No) and have copied them to the rc_bugs directory.
-          Before we release this eventually has to be sorted.
-
-2005-11-26 18:58  acastro
-
-        * src/states.F90: The output of the information of which node
-          manages which KS states was not good for the case where there is
-          a combined parallelization scheme. I have redone it; now it is
-          also better since only root node is doing the output, as it
-          should be.  [I noticed this when doing a run with 128 processors:
-          32 domains and four states groups. OK, I could only do a 10
-          minutes run in the debug queue, but still, I think it is a
-          record. The speedup, by the way, was of about 80]
-
-2005-11-26 17:43  acastro
-
-        * src/: profiling.F90, restart.F90, td.F90: (*) Added the TIME_STEP
-          tag to the profiling module.  (*) A little bug fixed also in that
-          module.  (*) The clean_stop routine deletes the stop file, as
-          Heiko suggested. I     also found it annoying, never remembered
-          to delete it.
-
-2005-11-25 22:31  appel
-
-        * src/: io.F90, out_inc.F90, scf.F90, td.F90: I found this quite
-          useful for chasing bugs which were not showing up immediately:
-          There is now the possibility to enable/disable the debug mode on
-          the fly. Imagine a bug that shows up after 100 SCF cycles. One
-          definitely does not want to run the code all the way in debug
-          mode. The present modifications allow to run the first 99 cycles
-          in normal mode and to switch only then to debug mode.
-
-          To enable the debug mode on the fly you have to touch a file with
-          name enable_debug_mode in the working directory, similarly to
-          disable the debug mode simply touch disable_debug_mode. Contrary
-          to the stop file the code takes care of the cleanup of the
-          touched files in this case.  (should the code also take care of
-          the cleanup of the stop file?) What a pity that Fortran does not
-          support signal handling ...
-
-          If the code runs inside some queueing system which buffers output
-          this could also be used to quickly check if octopus is still
-          doing reasonable things or got e.g. stuck in an MPI deadlock.
-
-2005-11-25 20:27  appel
-
-        * src/states.F90, testsuite/finite_systems_2d/fock_darwin.test,
-          testsuite/finite_systems_3d/restart.test: Adjusting the failing
-          tests to the new format.
-
-2005-11-24 21:55  appel
-
-        * src/states.F90: I just noticed that I unintentionally modified
-          the format for the output of the eigenvalues in my last commit.
-          The altered format causes some tests to fail (the matches don't
-          work as intended).  I've now again changed the output of the
-          eigenvalues, since in the spin-polarized case more than 80
-          columns were used. Please have a look if you agree with the new
-          format (basically I use "up", "dn", "--" for spin up, spin down
-          and spin unpolarized respectively).  If you agree with the
-          modification I volunteer to update the test files accordingly.
-
-2005-11-24 19:42  appel
-
-        * src/par_vec.F90: Forgot to remove debug output.
-
-2005-11-24 19:37  appel
-
-        * src/: io.F90, mesh.F90, mesh_init.F90, messages.F90, par_vec.F90,
-          simul_box.F90, states.F90, xc.F90, xc_functl.F90: Support for
-          FlushMessages was broken. When running in parallel filehandles
-          were not treated properly.  Instead of writing directly to some
-          unit I've changed several places such that they use the write_*
-          functions.
-
-2005-11-24 18:36  appel
-
-        * src/mesh_init.F90: Array index was missing in the last commit.
-
-2005-11-24 17:08  rozzi
-
-        * src/: mesh.F90, mesh_init.F90: Bug fixed in the building of the
-          potential for periodic calculations: the box size was not correct
-
-2005-11-24 13:55  acastro
-
-        * src/mf_inc.F90: Just some corrections of the push_sub argument,
-          which seem to have no relation to the actual subroutine name in
-          many cases.
-
-2005-11-24 13:39  marques
-
-        * src/td_rti.F90: another NP NP_PART issue.
-
-2005-11-24 13:09  marques
-
-        * src/nl_operator.F90: Some cleaniong in nl_operator.F90. I think
-          the routines _operate are more readable now.
-
-2005-11-24 12:35  marques
-
-        * src/: derivatives.F90, eigen_trlan.F90, mesh_init.F90,
-          mf_inc.F90, multicomm.F90, nl_operator.F90, out_inc.F90,
-          poisson.F90, run.F90, systm.F90: *) Cleaned the #if
-          defined(HAVE_METIS) directives in the code. Now they only
-          appear in 2 or 3 places (where they are really needed) *) The
-          code now stops if you want to run in parallel in domains with a
-          fft    poisson solver. This made me waste 30 minutes ;((
-
-          Incidentally, I now get *exactly* the same result running in
-          serial or in parallel (and not only approximately like before).
-          This means that it was either a np<->np_part issue, or a problem
-          with the HAVE_METIS (in fact there were a couple of
-          inconsistencies in the code)
-
-2005-11-24 10:12  appel
-
-        * share/PP/defaults: Updating the PP/defaults file for Cd since by
-          now all elements are included.
-
-2005-11-24 09:59  marques
-
-        * share/PP/TM2/Cd.ascii: Sorry, but this pseudo has problems, so I
-          removed it.
-
-2005-11-23 17:15  xavier
-
-        * src/: linear_response_inc.F90, static_pol_lr.F90: * Bugfix in
-          linear response: Because density was put to zero before time, the
-          fxc term was not considered for the polarizability. This affected
-          calculations of polarizability and hyperopolarizability.
-
-          * Added a check for DevelVersion to protect users from the
-          dynamic polarizability part which is not working.
-
-2005-11-23 15:54  micael
-
-        * src/h_inc.F90:  *) Moved the Zeeman term from subroutine vlpsi to
-          subroutine magnetic_terms
-              as it seems more apropriate.
-
-2005-11-23 15:05  micael
-
-        * src/xc_axc.F90:  *) Some more np/np_part issues fixed.
-
-2005-11-23 14:54  micael
-
-        * share/variables.local:  *) This two lines can now go away.
-
-2005-11-23 14:50  micael
-
-        * src/xc_functl.F90:  *) Documented the current functionals.
-
-2005-11-23 14:21  micael
-
-        * src/: h.F90, h_inc.F90, scf.F90, v_ks_inc.F90:  *) Removed the
-          current-current interaction from the CDFT hamiltonian, because it
-          is
-              time-consuming and in practice it should be neglectable.
-
-2005-11-23 12:24  appel
-
-        * src/states.F90: Bug fixed. The bands were not written anymore.
-
-2005-11-23 12:02  micael
-
-        * src/h_inc.F90:  *) Bug fixed: again a np/np_part issue. This time
-          in the CDFT part.
-           *) Corrected another 1/2 factor that was wrong in the magnetic
-          terms.
-
-2005-11-22 23:22  appel
-
-        * src/: Makefile.am, global.h, par_vec.F90, poisson.F90,
-          td_exp_split.F90: These modifications of the build system allow
-          for newline characters inside CPP macros. This overcomes the
-          problems with line truncation that we had with some compilers.
-          I've chosen \newline as newline sequence as \n was colliding with
-          things like eg. \nabla inside comments.  The changes partially
-          obsolete Alberto's m4 check for long lines. However, I kept it
-          since it is also used for conf%fcflags.
-
-2005-11-22 08:51  appel
-
-        * src/v_ks_inc.F90: Because of too long lines the code did not
-          compile with the NAG compiler.
-
-2005-11-21 17:16  acastro
-
-        * src/casida.F90: Bug fixed.  The casida module was broken. Yet
-          another np/np_part issue.
-
-2005-11-21 14:14  marques
-
-        * src/: eigen.F90, eigen_evolution.F90, eigen_plan.F90,
-          eigen_trlan.F90, epot.F90, epot_inc.F90, f_inc.F90, fftw3.F90,
-          geom_opt.F90, geometry.F90, grid.F90, gs.F90, h.F90, h_inc.F90,
-          h_so.F90, hgh.F90, lasers.F90, lcao.F90, linalg_adv_lapack.F90,
-          linalg_basic_int.F90, linear_response.F90,
-          linear_response_inc.F90, logrid.F90, make_st.F90, math.F90,
-          math_cg_inc.F90, math_inc.F90, mesh_init.F90, mf_inc.F90,
-          mix.F90, multicomm.F90, multigrid.F90, nl_operator.F90,
-          ode_solver_inc.F90, opt_control.F90, out_inc.F90, par_vec.F90,
-          par_vec_inc.F90, pes_mask.F90, pes_rc.F90, phonons.F90,
-          poisson.F90, poisson1D.F90, poisson2D.F90, poisson3D.F90,
-          poisson_cg.F90, poisson_corrections.F90, poisson_multigrid.F90,
-          ps.F90, restart_inc.F90, root_solver.F90, root_solver_inc.F90,
-          scalar_mesh.F90, scf.F90, sparskit_inc.F90, specie.F90,
-          spectrum.F90, spline.F90, states.F90, states_inc.F90,
-          states_kpoints.F90, static_pol.F90, static_pol_lr.F90,
-          stencil_star.F90, systm.F90, td_calc.F90, td_exp.F90,
-          td_exp_split.F90, td_rti.F90, td_write.F90, tm.F90, unocc.F90,
-          v_ks_inc.F90, xc_KLI.F90, xc_OEP.F90, xc_OEP_SIC.F90,
-          xc_OEP_inc.F90, xc_axc.F90, xc_fxc.F90, xc_kxc.F90, xc_vxc.F90,
-          xyz_file.F90: I changed the rest of the allocate -> ALLOCATE. Now
-          all the code has some rudimentary memory allocation error checks.
-          I am still not very happy about this solution. I have plans to
-          improve it (allowing also for statistics) but this will have to
-          wait for the next version.
-
-2005-11-20 21:57  xavier
-
-        * testsuite/rc_bugs/09-H2O_POL: Removed input file from bug that
-          was fixed in the previous commit.
-
-2005-11-20 21:51  xavier
-
-        * doc/octopus-devel.info, doc/octopus.info, doc/octopus.texi,
-          src/eigen_evolution.F90, src/geom_opt.F90, src/h_inc.F90,
-          src/linear_response.F90, src/linear_response_inc.F90,
-          src/make_st.F90, src/phonons.F90, src/static_pol.F90,
-          src/static_pol_lr.F90, src/td_write.F90: Bugfixes related to the
-          np/np_part transition: * Linear response.  * Other parts where
-          st%X(psi) was initialized with np size.  * Added an assert to
-          check input array size in h_inc.XHpsi.
-
-2005-11-20 19:21  acastro
-
-        * src/profiling.F90, src/states.F90, src/xc_KLI.F90,
-          src/xc_OEP_inc.F90, src/xc_OEP_x.F90,
-          testsuite/rc_bugs/04-Na2_Exx: Bug fixed.  This time, 04-Na2_Exx.
-          It had to do with the parallelization.
-
-          I do not put it as a test because I am not fully certain of the
-          validity of the result. I will try to find exact known results
-          for EXX and make a test for it. Also for the rest of the
-          orbital-dependent functionals.
-
-2005-11-18 17:02  acastro
-
-        * src/poisson_corrections.F90: When running in parallel in domains,
-          I got a crash there due to the np/np_part problem.
-
-2005-11-18 15:56  acastro
-
-        * src/h_inc.F90: Bug fixed.  Forgot one pop_sub
-
-2005-11-18 15:12  acastro
-
-        * src/profiling.F90: I don't think the profiling_in and
-          profiling_out should be monitored by the push_sub/pop_sub, the
-          output gets rather ugly.
-
-2005-11-18 14:52  acastro
-
-        * src/out_inc.F90: The ASSERT(0) led to an arithmetic if in Fortran
-          that is not accepted by modern compilers.
-
-2005-11-18 14:41  acastro
-
-        * src/: h.F90, h_inc.F90, poisson.F90, profiling.F90, v_ks.F90,
-          v_ks_inc.F90: Added some more profiling tags. I know that we
-          should be in regression mode, but maybe we should have some way
-          to properly benchmark the 2.0 code against all future
-          developments.
-
-2005-11-18 13:54  acastro
-
-        * src/: nl_operator.F90, profiling.F90: Bug fixed in the profiling.
-           The nl_operator tag had two entries and no exit.
-
-          I improved a little bit the output in the profiling.xxx files, so
-          that it is more readable.
-
-2005-11-18 11:38  acastro
-
-        * src/mesh_init.F90: The code was crashing in the marenostrum
-          machine at unexpected points near the call to mesh_partition,
-          when attempting to run systems with large meshes. The reason
-          seemed to be a memory problem, probably the code was legal but
-          the memory managemente was badly done by the compiler (mpxlf
-          compiler from ibm). With the changes that I did, it seems to work
-          fine now. I don't know, maybe in the way that it was done before,
-          some odd copies were being made, filling the cache, or the stack,
-          or whatever, maybe the ventilators, god knows.
-
-          In general, it is probably not a good practice to pass a
-          structure and components of that structure separately to a
-          subroutine. Also, it is probably better to use allocate than
-          using automatic arrays.
-
-2005-11-17 17:36  xavier
-
-        * src/static_pol_lr.F90: The off-diagonal terms of the beta tensor
-          were wrong.
-
-2005-11-17 15:16  acastro
-
-        * testsuite/finite_systems_2d/fock_darwin.test: Added a test for
-          the magnetic field. For the moment, only in 2D, and with a null
-          Zeeman term. It is the spectrum of Fock-Darwin (one particle in
-          2D confined by a Harmonic potential, and subject to a
-          perpendicular static constant magnetic field).
-
-2005-11-17 14:58  acastro
-
-        * src/unocc.F90: The unocc mode was broken due to the usual
-          np/np_part problem.
-
-2005-11-17 14:33  xavier
-
-        * src/geometry.F90: The atom_write_xyz function was printing the
-          file in all nodes except root, as a result oct-center-geom didn't
-          write the adjusted.xyz file. So I added a .not.
-
-2005-11-17 11:43  acastro
-
-        * liboct_parser/parser_f.c: Bug fixed.  In some cases, the code
-          stopped in xyz_file_read when attempting to see if the block was
-          defined. This was due to the assert statement in the
-          loct_parse_block C function, which may detect non-null
-          block_name.
-
-          I am not too sure about why this was happening, but in any case I
-          think it is much more robust this way.
-
-2005-11-17 08:30  xavier
-
-        * src/static_pol_lr.F90: Fixed compilation issues with complex
-          version.
-
-2005-11-16 23:37  acastro
-
-        * src/: epot.F90, h_inc.F90: I tested the static external magnetic
-          field. Apart from a 1/2 factor that shouldn't be there, it was
-          OK.
-
-          I allowed for the possibility of having the external magnetic
-          field and *not* doing CDFT, they are independent things.
-
-          The test that I did was to reproduce the Darwin-Fock spectrum in
-          2D: eigenvalues of a 2D single-particle system, confined by a
-          harmonic potential, and subject to a perpendicular magnetic
-          field. This spectrum is analytical.
-
-          However, I did not check the Zeeman term.
-
-2005-11-16 17:55  xavier
-
-        * src/static_pol_lr.F90, testsuite/rc_bugs/09-H2O_POL: Some
-          corrections for first static hyperpolarizability, and now it
-          seems to work fine (the spin polarized case still needs some
-          work).
-
-          Unfortunately, the linear response code is broken after the
-          np->np_part transition, but I haven't been able to locate where
-          the problem is. So I am commiting a example input to the rc_bugs
-          directory.
-
-2005-11-15 13:46  xavier
-
-        * src/linalg_basic_blas.F90: Removed spaces before the # in a
-          couple of preprocessor directives, Compaq cpp doesn't like them.
-
-2005-11-15 13:12  marques
-
-        * src/global.h: This should fix the compilation issues from
-          yesterday.
-
-2005-11-14 20:58  marques
-
-        * src/: atomic.F90, broad.F90, casida.F90, cf_inc.F90,
-          cross_section.F90, crystal.F90, curv_gygi.F90, derivatives.F90,
-          derivatives_inc.F90, eigen_arpack.F90, eigen_cg.F90, global.h,
-          messages.F90: I spent more than half an hour searching for a bug
-          that didn't exist - it had been an allocate error (out of
-          memory), but, as usual, it appeared as a segmentation fault.
-          Therefore, I decided to put a very simple check for the
-          allocation errors. It is basically a preprocessor macro, and
-          changing allocate -> ALLOCATE. I started changing all allocates
-          in the code, but I will finish when I come back from Louvain.
-
-2005-11-14 20:00  micael
-
-        * src/v_ks_inc.F90:  *) Sorry, I forgot to remove those lines.
-
-2005-11-14 17:45  appel
-
-        * src/: broad.F90, out.F90: Remove unused variables.
-
-2005-11-14 17:45  appel
-
-        * src/sparskit_inc.F90: Documented all Sparskit variables.
-
-2005-11-14 17:26  appel
-
-        * src/v_ks_inc.F90: The code did not compile. I commented some
-          lines in v_hartree() and added a FIXME at the corresponing
-          position. Micael, can you check if this part is still required?
-
-2005-11-14 17:06  micael
-
-        * src/restart_inc.F90:  *) Increased the number of digits of the
-          number of iterations in file "wfns".
-
-2005-11-14 16:58  micael
-
-        * src/: h.F90, h_inc.F90:  *) Removed the effective-mass model as
-          there were some precision issues when using it.
-              Also, I realized there is a better way of including this
-          model: I will make it a system
-              of units.
-
-2005-11-14 16:15  marques
-
-        * src/eigen_arpack.F90: Small bug corrected. h has to have intent
-          inout now.
-
-2005-11-14 16:14  marques
-
-        * share/variables.local: Forgot to commit this file.
-
-2005-11-14 16:05  micael
-
-        * testsuite/finite_systems_3d/spinors.test:  *) Changed the spinors
-          test. Now it uses the Lanczos eigensolver.
-
-2005-11-14 15:52  marques
-
-        * src/: curvlinear.F90, global.F90, io.F90, main.F90: *) Moved
-          reading of DebugLevel from io to main *) Added variable
-          DevelVersion. As discussed with Heiko, the objective is that
-          when conf%devel_version is .false. certain areas of the code are
-          not    accessible. This allows to "disconnect" some options, and
-          reconnect them    with DebugLevel, i.w., without having to
-          comment the test in the source    file.  *) conf%devel_version is
-          only used now in curvlinear.F90
-
-2005-11-14 15:03  marques
-
-        * src/: derivatives_inc.F90, eigen.F90, eigen_cg.F90,
-          eigen_evolution.F90, eigen_plan.F90, f_inc.F90, geometry.F90,
-          global.h, h_inc.F90, lcao.F90, poisson_cg.F90,
-          poisson_corrections.F90, poisson_multigrid.F90, scf.F90,
-          states.F90, states_inc.F90, td.F90, td_calc.F90, td_exp.F90,
-          td_write.F90, xc_vxc.F90: This change may cause some problems in
-          some cases. The problem is that, with the changes that Florian
-          and Heiko did, the wave-functions (or better, any function that
-          we want to calculate the derivative of) must be defined from
-          (1:m%np_part). If I remember well, this had to do with the
-          communication of the ghost points. What happened is that in
-          several places in the code, the functions were defined from
-          1:m%np. An example, in gs.F90, st%psi is defined from
-          1:m%np_part, while in td.F90 st%psi is defined from 1:m%np.
-
-          This is therefore an attemp to to clarify this situation. All
-          functions have to be passed to derivatives defined from
-          1:m%np_part. All tests work at the moment, but it is possible
-          that some parts of the code will fail (for example, I am not 100%
-          sure of td_exp::lanczos)
-
-          Furthermore, there was a problem and the ASSERT macro was never
-          define. Now it is OK. I also added some ASSERTs to the
-          calculation of the derivatives.
-
-2005-11-14 13:56  micael
-
-        * src/eigen_plan.F90:  *) Bug fixed: Lanczos eigensolver was not
-          working as there were some
-              inconsistencies between the size of the arrays and the calls
-          to the blas.
-
-2005-11-14 13:36  marques
-
-        * testsuite/periodic_systems_1d/: cosine_potential.test,
-          free_electrons.test: Following the recent changes of Alberto, the
-          tests were no longer working.  Now they do.
-
-2005-11-14 12:10  acastro
-
-        * src/geometry.F90, src/scf.F90, testsuite/rc_bugs/06-Spinors: I
-          did not know that geometry_min_distance existed. It can be used
-          in geometry_atoms_are_too_close, of course.
-
-          Also, I made sure that a meaningful default lmm_r is set in scf
-          when there is only one atom. It was wrong before due to the huge
-          value returned by geometry_min_distance. This way the 06-Spinors
-          bug file can also go away.
-
-2005-11-14 11:38  acastro
-
-        * share/samples/Makefile.am, src/geometry.F90, src/messages.F90,
-          testsuite/rc_bugs/01-Hartree_1D: Added a
-          geometry_atoms_are_too_close check to make sure that there are no
-          atoms sitting on top of each other in the input file. This way we
-          can get rid of one of the bug files, 01-Hartree_1D.
-
-          I forgot to change the Makefile.am when I changed the sample
-          input files.
-
-          Added a new line in write_fatal, otherwise the output looked
-          ugly.
-
-2005-11-14 11:06  acastro
-
-        * share/samples/Na2td: I added the Na2td sample input file, so that
-          one time-dependent run is also in the samples directory.
-
-2005-11-14 11:01  acastro
-
-        * share/samples/: Na2, bench-gs, benzene: After so many changes in
-          the input variables, the sample file had got wildly inaccuarte. I
-          removed most of the variables, leaving a more minimal set, I
-          think Miguel asked for that some time ago. I also renamed
-          bench-gs into benzene, and chaged it slightly. We can still use
-          it for benchmarking, anyways, but the name is more descriptive.
-          We should maybe add more examples before the commit, and describe
-          them better.
-
-2005-11-14 10:24  marques
-
-        * src/: h.F90, multicomm.F90, out.F90, scf.F90, states.F90,
-          states_inc.F90, td_write.F90, varinfo.F90: A small change to make
-          the code more usable. I transformed the OutputXXX and TDOutputXXX
-          variables into flags. In this way, the ~20 variables of this kind
-          are transformed in only 3: Output, TDOutput, and OutputHow
-
-2005-11-14 10:18  marques
-
-        * testsuite/oct-run_regression_test.pl: Small change to allow the
-          use of ; (written as '\;') in the precommand part of the match.
-          This is useful if we want to use awk scripts in the match.
-
-2005-11-14 08:03  acastro
-
-        * src/: global.F90, hgh.F90: This fixes, I think, the compilation
-          errors of yesterday night.
-
-2005-11-14 00:34  acastro
-
-        * src/mpi.F90: I believe that the #else there should be an #endif.
-          Otherwise some problems could show up when running with one
-          processor but with one mpi executable.
-
-2005-11-13 22:22  acastro
-
-        * src/curv_gygi.F90, testsuite/finite_systems_3d/curvilinear.test:
-          Added one first test for the curvlinear coordinates stuff.  It is
-          a bit slow, but the initialization time of the curvilinear runs
-          are that way.
-
-          Also, corrected a small error in the documentation of the
-          CurvGygiA variable.
-
-2005-11-13 21:09  acastro
-
-        * src/ps.F90: Bug fixed.  The code crashed when trying to run
-          Helium with the HGH pseudopotential.
-
-2005-11-13 11:04  acastro
-
-        * src/: Makefile.am, oct_test.F90: Removed the oct-test program,
-          which was there for internal use, and should not be there in the
-          release.
-
-2005-11-12 23:07  acastro
-
-        * src/xc_functl.F90, testsuite/finite_systems_2d/2dbox.test: The
-          default correlation should not be the same in 3D and 2D.
-
-          I added tests for the exchange and correlation potential in 2D
-          (the Attacalite one).
-
-2005-11-12 20:47  acastro
-
-        * libxc/lda_c_amgb.c: Bug fixed.  One wrong character kept me busy
-          the whole afternoon.
-
-2005-11-12 11:25  appel
-
-        * src/messages.F90: Adapting string lengths in
-          messages_print_stress. Octopus was printing on x86_64 garbage
-          like this
-
-          ^@^@^PS©^@^@^@^@^@ìËÿ)^A^@^@^@~@S©^@^@^@^@^@PS©^@^@^@^@^@ÐQ©^@^@^@^@^@
-
-          to static/info. This was causing grep to assume it is a binary
-          file.  Therefore the matches of the testsuite were failing.
-
-2005-11-11 13:00  marques
-
-        * src/: Makefile.am, casida.F90, epot.F90, epot_inc.F90,
-          geometry.F90, global.F90, grid.F90, gs.F90, h.F90, io.F90,
-          liboct_parser.F90, main.F90, mesh.F90, mesh_init.F90,
-          messages.F90, mf.F90, mpi.F90, mpi_debug.F90, mpi_debug_inc.F90,
-          multicomm.F90, out.F90, out_inc.F90, par_vec.F90, profiling.F90,
-          restart.F90, restart_inc.F90, run.F90, scf.F90, specie.F90,
-          states.F90, states_inc.F90, td.F90, td_calc.F90, td_write.F90,
-          v_ks.F90, xc_KLI.F90, xc_OEP.F90, xc_OEP_SIC.F90, xc_OEP_inc.F90,
-          xc_OEP_x.F90: Some rearrangements (cleaning) in the mpi stuff
-          *) The stuff that was in mpi.F90 is now in mpi_debug.F90   *) Now
-          mpi_mod has low level MPI stuff   *) There is now a mpi_grp that
-          handles MPI groups (including a mpi_grp_is_root      that, in my
-          opinion, is much more readable than (rank == 0)   *) The rest of
-          the code was changed to use this infrastructure
-
-2005-11-11 11:11  acastro
-
-        * src/poisson.F90: I commit a commented subroutine, poisson_test,
-          in the poisson module.  It implements a check for the Hartree
-          integrator, either in 2D or in 3D (it calculates the Hartree
-          potential of a Gaussian charge, and compares it to the exact
-          analytical result). This I didn't do it for 1D because for the
-          soft-Coulomb interaction actually I couldn't think about an exact
-          result.
-
-          I am not sure about wether or not to put it commented out or not.
-          After all, we are in regression mode. But I wanted it to be in
-          the code after all. It would be nice to have this kind of x_test
-          routines for some of the modules, so that their particular tasks
-          can be checked independently.
-
-2005-11-10 23:24  acastro
-
-        * testsuite/finite_systems_2d/2dbox.test: I checked the 2D Hartree
-          solver, so I added a test to the 2D file: a qd with 2 electrons
-          interacting with only the Hartree term.
-
-2005-11-09 16:45  acastro
-
-        * testsuite/finite_systems_2d/2dbox.test: Added a first test for 2D
-          calculations. This is just an infinite potential box example. I
-          will put more as a continue checking the 2D part of the code.
-
-2005-11-09 14:06  acastro
-
-        * src/: mesh.F90, specie.F90: Just a couple of stupid bugs that
-          showed up in my 2D calculations.
-
-2005-11-09 12:01  marques
-
-        * src/: eigen_cg.F90, geometry.F90, grid.F90, hgh.F90,
-          messages.F90, multicomm.F90, run.F90, scf.F90, td.F90, tm.F90,
-          v_ks.F90, varinfo.F90, xc.F90: Some more "cleaning" of the output
-          of octopus. Now the variable "stars" disapeared, replaced by the
-          function messages_print_stress, that was given some more options.
-
-2005-11-09 10:13  marques
-
-        * libxc/: lda.c, xc.h, xc_f.c: *) Alberto, I think that the "break"
-          in the fxc calculation should be a    "continue" (i.e., he should
-          skip the rest of the current cycle, but    should get out of the
-          cycle). Can you please check? *) Xavier, I guess that the
-          calculation of Kxc should have a similar to    the one Alberto
-          put in for fxc. Can you please check? *) I cleaned a little bit
-          the includes, whitespaces, and changed "(*p)." to    "p->" as it
-          is more C-like.
-
-2005-11-08 23:01  acastro
-
-        * libxc/lda.c: Bug fixed.
-
-          The numerical functional derivation was not working properly in
-          some spin-polarized cases (Perdew-Zunger parameterization,
-          Na4(+1), it was just giving lots of NaN).  Now, at least for
-          those cases, it seems to work. I still have to do some more
-          testing on the Casida module, though.
-
-          In any case, the PZ kernel should be obtained analytically. Maybe
-          in the future.
-
-2005-11-08 16:08  acastro
-
-        * src/mix.F90: Removed two FIXMEs.
-
-2005-11-08 16:03  xavier
-
-        * src/: multigrid.F90, poisson_multigrid.F90: All multigrid
-          variables are documented now.
-
-2005-11-08 14:51  acastro
-
-        * src/: broad.F90, casida.F90: The LinCasida, LinEigenvalues and
-          LinPetersilka variables are gone.  The oct-broad utility just
-          broadens whatever spectrum it finds.
-
-2005-11-08 08:24  appel
-
-        * src/eigen_arpack.F90: Fix compilation issues from last night. The
-          MPI version of the code was not compiling.
-
-2005-11-08 08:18  appel
-
-        * testsuite/rc_bugs/08-TDELF_C2H4: This bug will probably a bit
-          harder to trace, but I believe it's a real bug.  The code stops
-          after 200 SCF cycles with an ** Address Error **.
-
-2005-11-07 20:14  marques
-
-        * src/: curvlinear.F90, derivatives.F90, eigen.F90,
-          eigen_arpack.F90, epot.F90, f.F90, geom_opt.F90, h.F90, main.F90,
-          mesh_init.F90, messages.F90, mix.F90, multigrid.F90, pes_rc.F90,
-          poisson.F90, poisson_multigrid.F90, restart.F90, run.F90,
-          scf.F90, spectrum.F90, states.F90, states_kpoints.F90, systm.F90,
-          td_exp.F90, td_init.F90, td_rti.F90, v_ks.F90, varinfo.F90: I did
-          some cleaning in the way of handling the input errors (making the
-          use of input_error more consistent), and in the output of the
-          options (using messages_print_var_option throughout the code).
-
-          There is some need now of "beautifying" the output, by adding
-          some spaces and stars. In any case, from my experience with
-          Silvana, I would be happier if octopus would output more about
-          what it is doing...
-
-2005-11-07 17:40  xavier
-
-        * src/specie.F90: Updated documentation for Species block.
-
-2005-11-07 16:41  marques
-
-        * liboct/varinfo.c: Small bug corrected regarding newlines in the
-          varinfo file.
-
-2005-11-07 15:43  appel
-
-        * src/: gs.F90, lcao.F90, scf.F90, unocc.F90: The 1D periodic tests
-          were not passed anymore on x86_64. Fixing pointer deallocation.
-
-2005-11-07 15:14  appel
-
-        * src/epot_inc.F90: Remove unused variables.
-
-2005-11-07 15:12  appel
-
-        * src/global.F90: The new compilerflag feature can produce rather
-          long source lines (depending on the amount of compiler options
-          used).
-
-          The code didn't compile with NAG, as this compiler doesn't like
-          long lines
-
-           Error: global.F90, line 181: Line too long - truncated
-           Error: global.F90, line 181: Unterminated character literal
-
-2005-11-07 14:50  acastro
-
-        * src/curvlinear.F90: The code will now stop if the two non-working
-          adaptive coordinate schemes attempt to be used. They will not
-          make it before the release.
-
-2005-11-07 14:40  acastro
-
-        * src/: Makefile.am, global.F90, main.F90: I thought that it would
-          be cool if the code outputs what compiler and what compiler flags
-          were used to make the executable.
-
-2005-11-07 10:05  appel
-
-        * testsuite/rc_bugs/: 01-Hartree_1D, 02-Na_Chain, 03-Na2_MGGA,
-          04-Na2_Exx, 05-Na2_LB94_Casida, 06-Spinors, 07-Atomic_Gd: This is
-          a set of non-working input files. I've used the format of the
-          test files so that we can use them eventually for the testsuite
-          if they are working properly. The files contain a short bug
-          description and the output of octopus respectively.
-
-          I decided to add them to the cvs (besides the copy I put in the
-          wiki) so that everybody has easy file access and we have a
-          revision history on the bug tracking process (poor mans bugzilla
-          :).
-
-          Perhaps some of the features used in these files should be simply
-          disabled for the release of 2.0. In that case view this list just
-          as reminder that we have to stop the code with an appropriate
-          error message. Otherwise we have to look into the bugs ...
-
-          Some further things I noticed
-
-           - According to a comment in xc_OEP.F90:57 full OEP seems to be
-          only half
-             implemented. Is the comment outdated, or should we stop the
-          code? The
-             same applies to the Amaldi correction term. In both cases the
-          code seems
-             to run so that the user could assume by mistake that the
-          output is correct.
-
-           - There are 9 FIXME's in the sources:
-
-              appel@miraculix:..octopus/src> grep -i fixme *.F90 | wc -l
-              9
-
-           - ... and 31 warnings:
-
-              appel@miraculix:..octopus/src> grep -i warning *.F90 | grep
-          -v write | wc -l
-              31
-
-2005-11-07 10:00  appel
-
-        * share/util/: oct-atomic_occupations, oct-run_periodic_table:
-          Update the scripts. oct-run_periodic_table has now support for
-          ExtraStates and ElectronicTemperature and oct-atomic_occupations
-          has improved ordering for occupation numbers (4s states appear
-          before 3d states).
-
-2005-11-07 09:59  appel
-
-        * testsuite/: oct-run_regression_test.pl,
-          finite_systems_3d/multigrid.test, finite_systems_3d/restart.test,
-          finite_systems_3d/so.test: There was a / missing.
-
-2005-11-07 09:58  appel
-
-        * src/: io.F90, ode_solver_inc.F90, root_solver.F90,
-          root_solver_inc.F90: Documented a bunch of variables.
-
-2005-11-05 16:06  acastro
-
-        * src/: derivatives.F90, f.F90: The default stencil when using
-          curvilinear coordinates should be the "stencil_starplus" one, not
-          just the star. In fact, the star should *not* be used with curv
-          coordinates, so I made the code stop in that case.
-
-2005-11-04 18:26  acastro
-
-        * src/run.F90: This is to make sure that the code does not crash if
-          running in parallel but without metis.
-
-2005-11-04 18:12  marques
-
-        * src/: poisson_cg.F90, scf.F90: poisson_cg1 now works (also in
-          parallel). This is the PoissonSolver = cg
-
-2005-11-04 17:29  marques
-
-        * src/: casida.F90, epot.F90, epot_inc.F90, geometry.F90,
-          global.F90, grid.F90, io.F90, liboct_parser.F90, main.F90,
-          mesh.F90, mesh_init.F90, messages.F90, multicomm.F90,
-          par_vec.F90, profiling.F90, restart_inc.F90, scf.F90, specie.F90,
-          states.F90, states_inc.F90, systm.F90, td_write.F90, v_ks.F90,
-          xc.F90, xc_KLI.F90, xc_vxc.F90: *) Forces wre not working in
-          parallel. Now they seem to work.
-
-          *) Sorry, Alberto, I had also fixed the xc. I changed your
-          solution, for the    following resaon: I think it is much more
-          clear to use the mf_integrate    function than to call MPI
-          directly. The disadvantage is that this is    (maybe) slower and
-          more memory consuming (although temporarely). However,    I think
-          it is worth it. The same applies to the forces.
-
-          *) I rename the variables in mpiv in global. I think it is more
-          clear now.     I will probably create a module called mpi_group,
-          or something like that     to contain rank/size/comm to be used
-          consistently in the code.
-
-2005-11-04 17:06  acastro
-
-        * src/: xc.F90, xc_vxc.F90: The exchange and correlation energies
-          were badly calculated when using the parallelization in domains.
-
-2005-11-04 16:32  acastro
-
-        * src/states_inc.F90: Sorry about the last commit, I just
-          replicated Migues's bug fix.
-
-2005-11-04 15:23  acastro
-
-        * src/states_inc.F90: Bug fixed.  When running in parallel in
-          domains, the code crashed at that point because the state
-          communicator is null.
-
-2005-11-04 09:47  rozzi
-
-        * src/: epot.F90, out.F90, states.F90, states_kpoints.F90: A
-          contribution to the documentation of the variables.
-
-2005-11-04 08:59  marques
-
-        * src/states_inc.F90: Small bug corrected in the parallel version.
-
-          Alberto, the Na2 ground-states seems to work in my machine.
-          You'll have to use the cg_corrected Poisson solver (the cg
-          crashes), and decrease the PoissonSolverThreshold to 1e-7. Note
-          that using the fft solver does not stop octopus with an error,
-          but it gives non-sense!!!
-
-2005-11-04 08:35  appel
-
-        * src/td_rti.F90: Fix compilation problems. Same story as
-          yesterday: Don't put single quotes ' in the documentation. This
-          causes the compilation of the code to fail ...
-
-2005-11-03 18:43  marques
-
-        * build/var2html.pl, doc/octopus.info, doc/octopus.texi,
-          share/variables.local, src/curv_gygi.F90, src/geom_opt.F90,
-          src/h.F90, src/hs.F90, src/io.F90, src/out.F90, src/pes.F90,
-          src/pes_rc.F90, src/phonons.F90, src/poisson3D.F90, src/scf.F90,
-          src/spectrum.F90, src/static_pol.F90, src/systm.F90, src/td.F90,
-          src/td_exp.F90, src/td_init.F90, src/td_rti.F90,
-          src/td_write.F90, src/unocc.F90, src/v_ks.F90: I finished putting
-          the documentatio of the variables. I will now make a list of
-          undocummented variables.
-
-2005-11-03 15:37  appel
-
-        * share/util/: Makefile.am, oct-atomic_occupations,
-          oct-run_periodic_table: There's now a utility to compute the
-          atomic occupations and ExtraStates for all Pseudopotentials that
-          ship with octopus:
-
-&nb