################################################################
TODO: [Mainly fixes go here...]
=====

* replace `rest()' for `substraction()' or similar in FastMat2, since
  `rest()' is erroneously translated from spanish `resta'. 
* Add new print version feature to advdif and others. 
* Missing some PetscSynchronizedFlush() calls in the PETSCFEM_ASSERT
  macros in `fem.h' (see commit mstorti-v31-end-25-gb9761e4)
* surf2vol with `quad2hexa' is buggy when `use_exterior_normal=1'
* In `trcprtf.cpp' check if jacobian of element is null. Perhaps in
   the first time step? 
* ADVDIF is not completely invariant under Galilean transformations, 
  even with the addition of BCCONV+ALE. See test/alebecconv test. 
  2008-12-24 14:07:52
* With `print_vector_rota': if nfile is not set, then it does not save
  correctly. 
* Should print `nodedata' hash table in `readmesh' after reading. 
* Initialize members to NULL and void values in Gpdata::Gpdata()
* In FastMat2, static declarations should be removed, otherwise it is
   not reentrant. (I think this is done already, CHECK).
* using `shock_capturing' in `nsi_tet_les_fm2' gives SIGSEGV
* using nsi_tet_les_fm2 with fractional step produces  SEGV
* In test/sector/Makefile: change '-chmod +w sector.depl'
       by calls to make-depl
* Absorbing b.c.'s: 
    Make emphasis in the doc that the normal is the exterior normal to
    the fluid. 
* `$ make sw' crash if `DX' not installed
* In `gatherer.cpp' default value for `ndimel' should be `ndim', not
  `ndim-1'. In `set_pg_values' we should pass `Jaco' always, and `n'
  (the normal) should be passed as a public member in the class. 
* Running large problem FS with (procsel -l) (0 weight in server) leads
  to dead-lock apparently in `iisdcr.cpp'   
  block: if  (iisdmat_print_statistics) {...}
* In advdif, make `CHECK_JAC' to be an option to be read from the data
    file at run time. 
* When using `ndim=2' in global options and 'ndim=3' in an elemset
   some things go wrong. (reported by Beto).
* Set `node=nel-2' in nonlres.cpp:250?
* Read  `sqcav', `cubcav', `pfmat2' to test/Makefile
* USE_DX should be tied to USE_PTHREADS. Also it gives an error with
    `make depend' if DXARCH is not set and USE_SSL is set (Or
    something like that). 
* make `epsilon' relative in `srfgath' 
* `advdif/id' elemset does some `proj' task or something. We should write
    an `ask' function for it. 
* Add `start_step' feature. 
* List grants in doc
* Add `dataflow.obj' to doc.
* Finish completing_read with petscfem_options
* There is a conflict between the computation of residual jacobians 
	  in advdif and the use of masks/profiles for element
	  uploading. 
* Check the `advdif/bcconv' elemset in 3D. Seems that there is a sign
     problem. 
  Mon, 08 Dec 2003 08:56:00 -0300: See discussion and patch in
     `notes.txt'. Patch is not applied yet, though. 
* Document the element locker feature
* Using DX cache. dx/epimport.cpp deletes too many objects and DX
    complains. Also is hangs frequently. 
* Print a banner saying if a version is stable/unstable. 
* Let PETSc-FEM have knowledge of the name of the nodes 
* verify `nsid.cpp' (ns_id_fac*ns_id_cn1) and `lumped_vc'
* replace epart2 -> epart in readmesh.cpp and other parts of elemset,
     etc... 
* Pass fix to `ns.cpp' (define MPI_RANK *before* `print_copyright') to
      stable 3.22
* Make user can call PETSc-FEM assert call abort() or exit(). 
* Check caching DX objects for multiple elemsets
* Code and test changing coords and elemset conectivities 
  caching by the user/ appl. writer. 
* `src/secant.cpp' is compiled differently. why?
* Make a test for Newton convergence with ns_id and lag_mul mixed. 
* Correct DOC++ doc in `lagmul.h'
* why `make finaldoc' in ./ doesn't launch latex2html?
* FastMat2 doesn't detect errors like this
  `tmp5.prod(dshapex,tmp4,-1,2,-1,2);' i.e. the free indices are not
  in a sequence starting at 1. (see fstepfm.cpp, Fri Jul 26
  12:51:27 ART 2002). non contracted indices should be 1,2,.... 
  neither this: tmp11.prod(dshapex,tmp12,-1,1,-1);
* default values for npg
* DX: lets to call field generators from dynamically load C/C++
         modules. 
* DX: make `dxhook' to be launched through a flag of type `dx' not
          through the standard `hook-list' option line. 
* DX: check for memory leaks, both in `epimport.cpp' and in  petscfem
          itself. 
* Convert all `assert()' macros to a PETSCFEM_ASSERT or similar. 
* Why it takes so much time after "Start partitioning" when running
    large problems in one proc.
* Generate test for LES 
* `html' target in `doc' is duplicated and `make' complains. 
* Add a test for advdif with null weight processor
* Add edge weights to graph partitioning in `metisprt.cpp'
* Make `report_consumed_time' to report the time spent in computing
        residues, and assembly. 
* Document new options for IISDMat
* In `eperlini.pl' we have to check strings/number with `is_numeric()'
   defined in 'perl/utils.pl' but this requires to update eperl in
   spider. 
* There is a problem with the `fstack' close and destruction
  functions. It gives a SIGSEGV for the following version of rosi2/pfm
> ----------
> tag: v0.08.pl58 on Tue, 11 Jun 2002 21:05:52 -0300 by mstorti in spider
>      > Reads data from initialization files
* Add test for axisymmetric option in nsi_tet_les_fm2
* generate tests for gatherer
* generate a test for Levi-Civita (eps_LC())
* Make dof map to a `fixation_entry_list', that contains many entries
    of the form (double,Amplitude) so that we can have an Amplitude
    function that is a sum of several Amplitude functions. 
* Make advdif not to need the CHECK_JAC compilation flag, so that it
     doesn't need recompilation to check the jacobian. 
* Continuation lines behave extrangely when ending in a comment. For
	instance: 
  blah blah blah blah \
    blah blah blah \
	blah blah # comment
   See ROCKET `slosh.epl' example Thu Feb 28 09:11:50 ART 2002
* Fix `delete' and `new' not balanced in `NewAdvDifFF' derived
         classes. 
* add tests  `cubcav' 
* Fix documentation. New version of latex2html screwed all up. 
* write doc for Sparse::Mat, IISD with SuperLU local solver. 
* put `const A' in the argument to PFMat::duplicate
* interface preconditioning for IISD
* Put order in the headers. Now, a modification in some header launchs
    recompilation of almost the whole code. 
* make `da' in profiles to be dimensioned of size `neqp' (number of
   dofs in local processor, not of size `neq' (number of total dofs)
* write better error messages in FM2 (with threads??)
* new scheduling algorithm for load balancing in `elemset.cpp'
* adapt generic non-linear restriction element to general case
    (arbitrary argument list)
* fix some incompatibilities that exclude compilation with LES under old RH5.2
         of some STL vector usage.
* Dofmap may be eliminated of the arguments to the elemset 'assemble'
     routine (not from the global assemble)
* classify sections in doc as user / application developper / programmer.
* Review impact of '-funroll-loops' in FastMat2 efficiency. 
* Make ElementIterator a nested class of Elemset
* Convert flux functions (burgers, shallow water), to new ff's...
* In the advective elemsets we don't destroy the flux function objects
     created for instance in 
        `newadvdif_burgers() : NewAdvDif(new burgers_ff(this)) {};'
* make __INCLUDE__ can accept "" enclosed strings
* Throw an error if nothing is read (a void file).
* Check headers for functions (the __FUNC__ macro). 
* A faster (cached version) of mydet and mydetsur
* Fix makefile: runs bison and compiles getarray.cpp even if it's not
  necessary
* document coding of fastmat2. Document the eperl utilities.
* Should State have a TimeData instead of a Time
* Why not to do a single loop over node_list in the print_some()
       routines?  
* Check for memory leaks in constructing and destructing filters.
* sqcav and burgers test
* Document sttfilter etc...
* Fix problems with b.c.'s when Dirichlet and periodic superimposed. 
* Making html doc, doesn't reuse the generated gifs, which is _very_
           time consuming.

################################################################

DONE:
=====

* Check that in some ALE elements the check for enough columns in the
  `H' fields is inverted, see commit mstorti-v36-start-6-g67321c4
* Tests sqcav/check.zwproc.tmp and sqcav/check.zwproc2.tmp give now a
   little error. See comments in notes.txt
   Mon Feb 23 15:37:21 2009 mstorti@galileo
   21:46:22 -- mstorti-v24-start-20-g1cf5a46
   2009-06-05 19:53:19: Seems that now it works OK
* Gives PETSc error when run with ncore=8 and np=2
  I THINK both these and below are the same and due to the 
  problem reported
  notes.txt/2009-05-25.20:21:49.mstorti-v25-end-10-g8b7bce3.
  FIXED!! there was a bug in PETSCFEM_ERROR and _ASSERT macros. 
* gives FPE w. ncore=8, see notes.txt:09:38:01 --
  fastmat2-omp-start-28-gd9d7fd1 see above. 
  FIXED!! there was a bug in PETSCFEM_ERROR and _ASSERT macros. 
* 2009-06-03 19:13:33
  FIXED: now report() is not called by default in the destructor. 
  See commit f3a697c9499c0d084e55e91d14b59dbf4b840396 fastmat2b
  do not open log. 
* 2009-05-25 17:23:59
  FIXED a BUG: in fem.h in the definition of macros PETSCFEM_ERROR and
  PETSCFEM_ERROR0, `templ' should go as an additional arg, and NOT 
  concatenated to the template, because in that case `%' characters
  are interpreted as a replacement pattern for `printf()'.   
* In FastMat2: check that FastMat's created in sucaches be linked to 
  current CacheContext. See commit b3f9a09f2e11b454 for how it was
  done for the `norm_p_all' and similar functions. 
  Only cache affected seems to be `detsur_cache'
  2009-02-20 12:38:44: DONE
* 2009-02-19 08:31:19 ELIMINATE use of static FastMat2 members in
  _all operations like norm_p_all()!!
  2009-02-20 12:26:20 DONE 
* In `ns.cpp' some warnings about uninitialized variables remain (see
notes.txt) FIXED 2007-05-05 11:43:19
* in advdif: setting ndim=3 but using 2d elements gives a FastMat2
  error. Should be detected earlier? CAN'T REPRODUCE 2007-02-23 21:12:58
* doesn't compile without `pthreads' FIXED:= in  `beta-3.49.73'
* mmmove2 doesn't work with FastMat2 cached -> FIX
  2007-02-23 14:23:35 FIXED:= `before_chunk()' must be called _before_
   the element loop
* read-bad-state-gives-error: When reading an incomplete state should
   give an error. (See notes.txt)
   FIXED: Added `check_initial_state_correct_size' option in version `beta-3.49.70'
* Let `upload_vector_fast' to have several masks with integer values,
    e.g. [1 1 1 0 0; 1 1 1 0 0; 1 1 1 2 2; 0 0 2 2 2; 0 0 2 2 2]
* Doesn't report correctly acesses to the last thash table, probably 
	  related to the weights flag. (Reported by
	  psanchez@intec.unl.edu.ar) Wed, 15 Oct 2003 09:12:34 -0300
    Mon, 08 Dec 2003 09:51:12 -0300: Can't reproduce
* Sends multiple messages (from all processors) when can't open file
    in 
    > Couldn't open file "spillwgc.bcconv-all.tmp"!!
    > Couldn't open file "spillwgc.bcconv-all.tmp"!!
    > Assertion failed: "ierr==0"
    > Couldn't find __END_ELEMSET__ tag
  DISCARDED: Sat, 06 Dec 2003 13:15:41 -0300: Can't reproduce
* Check that absorbing b.c.'a options are included in the doc. 
  DONE: they are added and now we have the papers. Sun Nov 13 20:28:32 2005
* In `FastMat2::is()' if the index is greater than the dimension of 
  the matrix, that index is discarded in the mask in place of throwing
  an error. Now I think that it should throw an error. 
  DONE: seems that has been fixed in 1.20 of `fastmat2.cpp'
  on  2005/05/02 00:35:12. But the bug was reported on aug 2005. 
* Add to FastMat2 documentation that maskes `ir()', `is()' should not
    traverse `branch()', `choose()' boundaries. 
  DONE: added a section. Sun Nov 13 20:21:30 2005
* When in dx-hook mode, it should not crash if the state file is not
  found. 
  > dx_hook: reading state from file curvjet.dx-state.tmp, record 0
  > dx_hook: Reading from formatted file.
  > dxhook.cpp:326: Can't read line.
  DONE:= now sends 0 if file is not found. Sun Nov 13 20:21:30 2005
* make `TextHashTable::add_entry()' warn _optionally_ when the
  entry for an old key is modified. Now warns always. 
  DONE: Sun Nov  6 13:01:11 2005
  Implemented, not very well tested though. 
* Lacks of preco_type=jacobi for FS in ns.cpp?
  Apparently fixed with the addition of `fractional_step_solver_combo'
  option. 
* when `fixa_amplitude' has no function name it givs segmentation
  violation. Ver `petyscfem-cases/gfabso-examples/gfabso.epl' 
  Tue Aug 23 11:45:09 2005. 
  FIXED: Sun Nov  6 11:52:25 2005
  Added controls to token, after calling strtok().
* Escape ! in options doc. For instance in line
  TGETOPTDEF(go,int,dx_state_all_fields,!dx_split_state_flag);
  in `dxhook.cpp'. It comes without the `d' in the PDF. 
  DONE: added escaping for `!'. Not well tested because all options
  in `odocall.tex' are not processed in the main documentation file
  `petscfem.pdf'. 
* Make `advdif' process hooks _before_ assembling profile and reading
  initial vector
  DONE: Sun Nov  6 10:09:37 2005. Simply moved code as for NS. Seems
  to work well. 
* When not using __INCLUDE__ in the constraint section (for instance)
      doesn't complain. (see version `beta-3.43.21'). 
> constraint
> gfshock.peri.tmp
> gfshock.v-peri.tmp
> gfshock.slip.tmp
> __END_CONSTRAINT__
  FIXED: Sun May  1 15:37:50 2005
	 Now gives error if can't read a token. 
* Apparently crashes when in the elemset section finds a string that
    is not elemset, for instance
> elemset qharmm 8
> conductivity 1.
> C 0.
> Cp 0.
> geometry cartesian3d
> name fluid
> npg 8
> dx 1
> __END_HASH__
> 1 2 3 4 5 6 7 8
> 9 10 11 12 13 14 15 16
> __END_ELEMSET__
> 
> ##---:---<*>---:---<*>---:---<*>---:---<*>---:---<*>---:---<*>---: 
> name wall
> dx 0
> resistance 1.
> penalization_factor 10000.0
> __END_HASH__
> 5  9
> 6 10
> 7 11
> 8 12
> __END_ELEMSET__
>
FIXED: Sat Apr  9 18:39:58 2005 There was an error in the
    `BasicObject' factory that when the Factory 

* Can't open `weights.dat' gives a SIGSEGV and not in parallel. 
> Couldn't open file "weights.dat"!!
> --------------------------------------------------------------------------
> Petsc Version 2.1.6, Patch 4, Released Aug 14, 2003
>        The PETSc Team
> ...
> --------------- Stack Frames ---------------
> Couldn't open file "weights.dat"!!
> --------------------------------------------------------------------------
> Petsc Version 2.1.6, Patch 4, Released Aug 14, 2003
>        The PETSc Team
>     petsc-maint@mcs.anl.gov
> ...
* When can't open a file (included with __INCLUDE__) it reports 
> Assertion failed: "ierr==0"
> Can't find __END_FIXA__ tag
> ---------------
  which is confuse. 
  Mon Feb 21 16:11:32 2005: FIXED
* `gatherer.cpp' doesn't use `set_error' for handling detjaco<0
  Mon Feb 21 15:55:02 2005: FIXED
* Not reporting error in 
  parallel "Couldn't open file "../proy.skin.tmp"!!" while reading
  fixations. 
* Bad parallel error reporting when not known elemset type 
      in NS (and advdif??)
   > not known elemset type: "ns_id"
   > rm_l_8_1498:  p4_error: net_recv read:  probable EOF on socket: 1
   > not known elemset type: "ns_id"
   > not known elemset type: "ns_id"
   > not known elemset type: "ns_id"
   > not known elemset type: "ns_id"
   > rm_l_2_1523:  p4_error: net_recv read:  probable EOF on socket: 1
   > rm_l_4_1450:  p4_error: net_recv read:  probable EOF on socket: 1
   > not known elemset type: "ns_id"
   > not known elemset type: "ns_id"
  Sat, 06 Dec 2003 13:07:34 -0300: Hey!! can't reproduce
* Fix warnings that are currently disabled with '-Wno-deprecated'
       Mon Feb 21 15:00:28 2005: Fixed, compiles OK without
       '-Wno-deprecated'. 
* Add test for `surf_gatherer' in test/cubcav. Apparently it is
       almost already done. 
       Sun Feb 20 21:37:22 2005: Added test
* Breaks uncleanly when include directive is pasted to filename, e.g. 
            __INCLUDE__xxxxxx.file ???
* OK Test gasflow_abso in supersonic flow. 
     OK Test gasflow_abso in 3D. 
     Test gasflow_abso with transverse component. 
     Test on a 2D geometry (bump, nozzle, flat plate?).
* Eliminate the DRH60 flag. 
* Disable ROSI_COUPLING_MODULE
* Error in fixations is not parallel safe. Tue Sep 14 21:22:17 2004
    > Fixation 3169, imposed on an invalid node/field combination.
    > Hi user!          Today is Tue Sep 14 21:21:24 2004
    > Have fun and a nice run! :-)   [The PETSc-FEM team]
    > -------------------------------------------------
    > Hi user!          Today is Tue Sep 14 21:21:23 2004
    > Have fun and a nice run! :-)   [The PETSc-FEM team]
    > -------------------------------------------------
    > Hi user!          Today is Tue Sep 14 21:21:17 2004
    > Have fun and a nice run! :-)   [The PETSc-FEM team]
    > -------------------------------------------------
    > Hi user!          Today is Tue Sep 14 21:21:33 2004
    NO, OK: The error was in printing the `salute'. This as already
    fixed. 
* in dir `test/newff' eperl complains with 
   `not implemented test case 'dont_use_fastmat2_cache' at - line
   450.'
    FIXED: Sat, 22 May 2004 11:17:42 -0300
* Add `sw_abso' to tests
    DONE: Sat, 22 May 2004 11:17:19 -0300
* seems that `__SWITCH_FILE__' directives in `myexpect.pl' doesn't
	 give an error if the file doesn't exist. 
    NO, OK: the lacking file is an output file, so that it only gives
	 a `couldn't open' error. 
* `$ make sw' attempts to build `doc/odocall.tex'. 
    DONE: Mon, 08 Dec 2003 23:15:13 -0300
* Make use FastMat2 cache optional in advdife via an option
    DONE: Mon, 08 Dec 2003 20:33:59 -0300
* Add `sw_abso' to tests
    DONE: Mon, 08 Dec 2003 20:24:09 -0300
* `advdif/id' elemset does some `proj' task or something. We should write
    an `ask' function for it. 
    DONE: Mon, 08 Dec 2003 20:24:09 -0300
* Make a test for block_uploading. 
       Mon, 08 Dec 2003 08:58:53 -0300: Almost done in
       `cubcav'. Finish it. 
       DONE: Mon, 08 Dec 2003 09:34:23 -0300
* Write a FastMat2::trace() (tensorial sense)
	DONE: Sun, 07 Dec 2003 18:22:35 -0300
* Finish with the GenericError and CHECK_PARR_ macros in readmesh... 
  Doesn't crash correctly when	
  > 	  Error reading coordinates in line:
  > 	  " 186.479250051526 0 65.8289085858449 5 0 100526.561473902"
  > 	  Not enough values in line!!
	  FIXED: Sat, 06 Dec 2003 12:12:37 -0300
	  added GenericError(char *,...) and PETSCFEM_ASSERT_GE()
* Write test for `compute_fd_adv_jacobian'. 
	DONE: Wed, 26 Nov 2003 21:03:25 -0300
* Advdif crashes when no filename is pased via the -case option
	 FIXED: Wed, 26 Nov 2003 21:01:44 -0300
* Make killpar print username when other user
       DONE: Wed, 26 Nov 2003 21:01:50 -0300
* Printing "For elemset ..." element partitioning should
	   give the elemset name and pointer. DONE.
* When .ini file is not found advdif (don't know NS) 
       aborts only in server. FIXED: 
* Put `COMPUTE_FD_ADV_JACOBIAN' in `advdife' as a elemset flag. 
* Gather several instances of the same keyword in the same
     options.info page. 
* Mention in the doc that shell-script-mode is appropriate for fontifying
      PETSc-FEM data files. 
* Document `shell-hooks'. 
* Add doc for block_uploading
* Add doc for new features in `myexpect.pl' (__SWITCH_FILE__, 
      variable match, working directory)
* Fix tests in /burgers (they started to fail).
* Put version policy in a separate file. 
* in `./Makefile' local_sw: should FIRST do `chmod 755'
* a dynamically loaded Gauss Point adaptor elemset. 
* revert changes for locker in adaptor, etc...
* Crashes with SEGV if input file is void (should complain and abort). 
* FIXED: Grows indefinitely in memory if a `fixa' section is not closed by
    the corresponding __END_FIXA__
* Make VERSION to have a newline at the end of line so as to be a
           regular Unix text file.
* Generate a summary of `odoc' options in the doc with `odoc.pl'. 
* Generate a test for exception handling in `advdif'. 
* Convert negative jacobian errors in advdif to set_error. 
* DX: make epimport.cpp to print shapes in one line. 
* Seems that can't correctly read elements with 1 node. STATUS:
	 fixed. 
* put dependency of `compdate.cpp' on `compdate.cppin' in the
      Makefiles. STATUS: discarded. 
* Can't read element lines with only one node with one digit
* Generate a test for the disconnected graph bug. 
* Add `wiki' syntax to //o options (#tt#, *bf*, _em_ ...)
* Make PETSCFEM_ASSERT not to print a dot in the message i.e. "condition." 
* document wiki syntax for `odoc.pl'
* print the date of running on the start
* OK Write a `system' hook. (executes a shell command). 
* DX: let the user to dynamically split fields in scalar, vectors, etc...
* if it can't open the connectivity file, it hangs with a SIGSEGV
  write test. FIXED
* DX: Make a hook class for generating fields directly from
          PETSc-FEM. 
* change 2002 -> 2003 everywhere
* DX: make `dxhook' to auto-recognize geometries from the "geometry"
          option and dimensionality. 
* DX: test splitting with triangles
* DX: Code asynchronous visualization with threads. 
* DX: write documentation for `dx_hook' and `ExtProgImport'. 
* DX: check runs auto with sequencer
* DX: check runs OK with job in background
* DX: check if runs OK in parallel. 
* DX: make protocol more robust by sending acknowledgments from the client.
* DX: Make auto naming of elemsets od the form <elemset_type>_xxx
* DX: make two outputs to ExtProgImport: one with all the fields and
           other with all the arrays
* DX: Delete comments in `epimport.cpp'
* Generate test for SuperLU implementation. 
* Use lapack lib for `daxpy' in `test/Makefile:testfm2d'
* recover test `thash2.cpp'
* Changing 'distmap.h' launches recompilation of almost all
   PETSc-FEM. 
* int FastMat2: `mydetsur' with masks can't be cached (loses cached
      dims and claims dimensioning errors. See version "tag:
      beta-3.02.pl11 on Wed, 7 Aug 2002 16:31:37 -0300 by mstorti in
      minerva.arcride.edu.ar"
* A line like this: ' __INCLUDE__ cylinder.bott_no_slip.tmp' causes a
  SIGSEGV (note the line starts with a space). 
  Fix. See ROCKET/ident -D 'Sun, 14 Apr 2002 16:37:51 -0300'
  NOW: gives error. 
* the name for the elemsets doesn't accept the quotes "name" feature
   (see readmesh.cpp:191). NOW: read with `get_string'.
* latex2html doesn't work with `algorithmic' environment. 
	FIXED:      Latex2html doesn't like the `algorithm' (it is like a figure
     environment) environment, but DO support the `algorithmic'
     environment. Eliminated the `algorithm' environment in `doc/nsdoc.tex'. 
* latex2html should put the section number in the title (eg. section
        5.3)... 
* FIx in documentation, can't call environment `algorightm'. 
* Write a script that retrieves date corresponding to a light tag
          Sol: Use date in tag replacing `ART' or `ARST' by `-0300'
* What happens to the `G_body' item in `nsi_tet_les_fm2' ??
  It works... Added a test Wed Dec 19 21:33:23 ARST 2001
* Write tests for IISD with subpartitioning.
* Check if the `-O1 measperf.cpp' problem is fixed.
* add tests with `sqcav' and several processors
* merge turbulence enhancements.
* merge adaptors developed in Chile
* reenable the `cache_grad_div_u' feature. It works, but there is not
   very much gain: At most a 10%. 
* Write tests for `Graph' class
* In `ns.cpp' it seems that there is no convergence check on the
  internal Newton loop. 
* Use IISD `inside processors'. 
* In class Graph `ngbrs_v' should be a `set<int>' not `vector<int>'
* add in  `Elemset::weight()' to get weight from an option
* change arcride by ceride anywhere
* add reference to parallel computations in README
* make better scheduling algorithm for distmap (as described in the
   doc). 
* document distributed map
* IISD: Code for element contributions to `local-local' matrices in other
   processors. 
* Control no memory losses in IISDMat
* Make readmesh check dof numbers not being greater than... 
* while reading constraints complain if dofs out of range
* Write a norm_p function (like sum_square_all...)
* write a generic non-linear restriction element
* escribir una funcion "double pow_int(double base,int exp)"
* Cached version of detsur
* add wall laws to turbulent shallow water
* move 'awatch' from test/
* test for burgers
* generate doc in PDF
* add the genload elemset to advdif.
* cambiar 'ceride.gov' por 'arcride.edu'
* Exception: number of nodes in connec. table doesnt match
	     that entered in elemset header
* Add weak_form support to the new flux functions
* Write a test for linear advection diffusion
* Include an S_body for ffadvfm2
* write diag-like fastmat2 functions. 
* add doc for enthalpy jacobians
* fix the __VA_ARG__ incompatibility with older cpp (???)
* Add a "mass" to the advective dffusive eq.??
* Add enthalpy contrib to the P_supg term
* Add doc for new flux functions
* replace all  VOID_IT by clear(): Partially done. VOID_IT calls clear
* the link to DOC++ in index.tex is broken
* Cambiar la licencia en todos los archivos
* How to choose the correct jacobian when different combinations (eg:
    ndim*ndim, ndof, etc...) coincide...
* corregir normales en mydetsur y otros ... 
* Add options for turbulent shallow water
* Fix error while opening data file, when no case gives
           "Couldn't open file "^&*&^*&^ <garbage> *^%&^()#*$"!!
* Add targets to the Makefile like `distrib' `almost_clean' etc...
* Find a better hash function for the text hashes. 
* Put `const' keyword in `get_entry' and friends.
* ffswfm2t test

; Local Variables: *
; mode: text *
; mode: abbrev *
; End: *
