v 2.24-1
  - correct periodic condition  in case of one edge 
  - correct doc file
v 2.24
  - bug with border and func
     this kind of code now works
       func abc=  a(6) + b(4) + c(4)  ;
       func def = d(4) + e(4) + f(6);
       func bbb= abc  + def;
      plot(bbb); mesh Th=buildmesh(bbb);
     whare : a,c,b,c,d,e,d are  border. 

  - correct bug in periodic Boundary Condition( find by Alexandre Masserey <masserey at ycoorsystems.com>)
       mistake in abscisse choose.
       sometime assert fail (loop).
     add option to the buildmesh that the boundary is fixe (fixeborder=true)
     see : examples++-tutorial/periodic4bis.edp   
  -  and  R1.hpp, R2.hpp, R3.hpp files for the 3d version.  
v 2.23-2
  - correct trap in adapdmesh (in mortar-DN-4.edp) in very rare case 
    du to  intialized   adj  of edges on geometry (very old bug) 
    in case of multi points.
  with vebosity=10; you get this message: 
 -- Begin of insertion process 
 bug  2
 Bug double points in 
....
 Fatal error in the meshgenerator 5
....

v 2.23-1
  - correct assert fail in check in case of no UMFPACK
v 2.23
  - add new edge finite element (P1 to P5) 
   in examples++-load/Element_PkEdge.cpp examples++-load/testFE-PkEdge.edp  
  - preconditionned  mortar example Neuman -> Dirichet in scalar and parallele
    see examples++-mpi/mortar-DN-4-mpi.edp examples++-tutorial/mortar-DN-4.edp
  - add vectorial  operator  d=  a ? b : c  where d,a is a array, and b,c can be array or scalar 
    usefull for boundary condition
  - correct probleme in mpi recive message
  - correct load dynamic on MacOs (bug in fortran find library) in configure.ac
v 2.22-2
  - small change in configure.ac of find libumfpack on debain systeme
  - correct checking code in probleme to same the mesh must be unique in problem of solve definition.
v 2.22-1
  - make change in Makefile to build universal binary on MacOS
  - change the README file for compilation on ubuntu Linux.
v 2.22   (december, 18, 2007)
  -  correct problem of string argument in function
  -  correct all elasticity problem (missing sqrt(2) )
  -  correct save of log file under  Windows XP, ..
v 2.21   (november, 18, 2007)
  -  correct documentation for new edition
  -  correct configure for no X11 and UMFPACK in sparsesuite
v 2.20-1 (october, 24, 2007)
  -  correct compilation problem without umfpack
  -  add  formal operator  for vectorial operator in variationnal form
        +, -  on  [  ]  array
        trace operator  matrix like [ [ ... ], .., [... ] ] 
        take element of an array [ ...][2] 
v 2.20  (october, 2, 2007)
  -  add in download automatic generation of f2c and fort77 to get a free  fortran compiler
	just  go in 
	(download/f2d;make install)
  -  add automatic find of libf2c in case of F77=fort77 (driver of f2c) in configure 
  -  add tools to change the sparse linear solver with load dynamic, see doc and SuperLu example
     in examles++-load
v 2.19   (august, 19  2007)
  - correct buildmesh from geometry file and add nbvx= named parameter
    to infore the maximal number of verter in a mesh.
  - add  Th[k].area to get the area of the k-th triangle of Th.
  - add  Th[k].region to get the region number (label)  of the k-th triangle of Th.
  - correct string delete in plot of border (mesh.edp)
  - correct mean and jump in  interpolation operator.
v 2.18-1 (august, 2007)
  - correct old bug under Windows: the std input stream  now works (cin) 
v 2.18  (july, 19, 2007)
  - add sort of real array or int array, syntaxe: 
      a.sort, and we have : a[i-1] <= a[i] for i =0 to a.n-1
  - add  v=a.quantile(p) ;  statisticial function  and commute v such than 
     #{ i / a[i] < v } = r*a.n   <=>    v = a[r*n]  when the array is sorted.   
  - correct P4nc, P4 finite element
v 2.17-2 (19 june 2007)
  - correct double   $(BLASLIB) in configure.ac (one more)
  - correct old bug in bamg, in case of bogus boundary (auto
    crossing)and   add code  to  handle  this error correctly with exception
    (see "test to catch bogus boundary"  of example++-tutorial/mesh.edp).
v 2.17-1
  - add the command line parameter [-v nn] in all freefem++ program to set 
    the level of verbosity to nn  before all, and if nn == 0 then theoriticaly no more
    spurious output.
  - correct  automatique choose of  color in plot of arrow. 
v 2.17
  - add bessel function   j0, j1, jn, y0, y1, yn -- bessel
        functions of first and second kind
  - add  erf, erfc -- error function operators
  - add   tgamma, lgamma  -- gamma and log of gamma
  - add   tool set and get   line,colunm and value of a sparce matrix
             [I,J,C]=A; // resizing  array I,J,C
             A=[I,J,C] ;
            where
             int[int] I(k),J(k); real[int] C(k);
  - add   in tools to get eigen vector of None FE problem
   like:
     int nev=5;  // number of computed eigen valeu close to sigma
     real[int] ev(nev); // to store nev eigein value
     real[int,int] eV(AA.n,nev);   // to store nev eigen vector
     int k=EigenValue(AA,BB,sym=true,11value=ev,rawvector=eV);

  - correct the quadrature formular automatic choose with the qforder= field
  by default the order is correct until  order 10 (so exact for P9 polynomals)
  to add new Quadrature formalur up two  order 26 add  dynamic load qf11to25
v 2.16-2
  correct configure.ac for windows ide compile
  correct meshsplit, trunc  function 
v 2.16 april 17 2007
  Change umfpack download version to the last one
  add string input from file or cin
  add Compressible Neo-Hookean Materials exemple
v 2.15-1
  Correct mistake in cas of splitmesh with internal boundary (assert)
  Add dump  of the mesh regulary (pesonnal work, FH)
v 2.15  (27 mars 2007)
  The blas.tgz file change on http://www.netlib.org/blas/blas.tgz
  correct the makefile  (remove BLAS dir) 
v 2.14-3 (24 mars 2007)
  Correct bug in vectorial operation of type  (b - 3.14*d) (Thanks to F. Dortu)
v 2.14-2 (22 mars 2007)
  correct integration problem on none classical mesh (mesh of curve of mortar technique)
  the mesh::Find methode is boggus of the kind on meshes, i make an optimisation to remove
  this call in case build matrix with varf int1d(Th)( u*v) , in case of Th, u or v is def.
  on same meshes.	
v 2.14-1 (0 mars 2007)
   correct missing speeling in  configure.ac  mpi++.h -> mpi++.h for mpich.1.2
   add computing linear form of 2 meshes  (ex. v on mesh Th1, and  compute int2d(Th1)(v) )
v 2.14  (7 mars 2007)
   make correction for solaris bluid
   add P3,P4 and new quadrature formular on triangle up degree 25 
   with dynamics load. See examples++-load/*P[43]*.edp
   correct mistake when we build matrix from varf with 3  meshes (FH).
v 2.13
   Add formal operator * and ' (to day just transpose) 
   on vector [ .. ]  and matrix [ [],...,  []]
   so we can write  
    macro grad(u) [ dx(u),dy(u)] // EOM
    int2d(Th)( grad(u)'*grad(v) )   
    add scalar term in block matrix as 1x1 matrix.
v 2.12-2
  correct problem in splitmesh and trunc
  trunc build always a part of a subdivision mesh
  splitmesh given more pretty mesh, before if we split 2 adjacent
  triangle resp in 3 and 4, before we put 2+3 points on the common edge
  now the put 3 points. 
v 2.12-1 (jan 15,2007)
  correct early delete of return pointeur in freefem++ function (trap in string.edp under windows )
v 2.12   ( jan 10,2007)
  correct mistake in none square block matrix 
v 2.11-2 (nov 29,2006)
  correct integration problem on none classical mesh (mesh of curve of mortar technique)
  the mesh::Find methode is boggus of the kind on meshes, i make an optimisation to remove 
  this call in case build matrix with varf int..(Th)( u*v) , in case of Th, u or v is def.
  on same meshes.  
v 2.11-1 (nov 28,2006)
  Correct bug in interpolation, we building fespace Wf(Th,[RT0,P0]); 
v 2.11 (nov 10,2006)
 Correction of  the problem in arpack (eigenvalue computation) under i383 processor.
 After 2 week of hard work, a found the bug.
 In new version of LAPACK second is a real function
 and in ARPACK second is a subroutine  (Thank the fortran).  
 So I change the Makefile to change the name of function second into secnd2. 
 How to find the bug, in this case the stack of the floatting register (info float under gdb) 
 are growing. So after any real function can return a NaN (one time) due to stack float register 
 overflow.  A big Thank to D. Bernardi of the help (Merci au fortran et Merci au i383).

v 2.10 (oct 26,  2006)
  correct problem with eigen value on window and some linux distrubition (Fedora)
  add a patch of two functions in lapack  see  arpack/arpack-patch-lapack.tar.gz

v 2.9 (sept 19, 2006 FH) 
  change Send and Recv in mpi  to Isend and Irecv to make async communication
  so now freefem++-mpi work with openmpi.

V 2.8-3
   correct freefem++-cs trap under windows (install the static version)
   add tools to format ostream  ( file or cout)  

V 2.8-2 
   (july)
   use f2c as a fortran compiler on  MacIntel architecture,
    correct mistake in mpi. 
   (june 2006)
   add nint, and correct mistake in ceil function
   add configure flag to remove cadna --without-cadna
V2.8
   Correct mistake in OtherMacOsLib.tgz store file and not link (after 1 june)
 * Correct bug in vectorial problem with same approximation, some time
   we do a renumbering of the previous  approximation space (a pointer mistake).
V2.7-1
  * change size of data in 64 bits architecture (in file AnyType.hpp)
V2.7
  * correct mpi Makefile
  * correct bug in construction of recursive composite finite element (see HISTORY)
    remark: the finite elemnt struct, so all dynamic library must be recompile. 
V2.6-1
 * add the examples-bamg  directory  in freefem++ and  correct bamg graphic
V2.6
 * Huge work (more than a full week) to improve the windows version 
   change all the .dll construction
   remove lots of trap. All example run without graphics
V2.5.1
 *  change nbve= to nev= in EigenValue function  (to be compatible with the doc)
 *  newtow to newton    (to correct misspelling)
 *  change in AnyType to remove da randow bug on windows.
 *  vim color syntax from Richard MICHEL (vim_highlighting_for_FF++.tar.gz)

V2.5-0 
  *  add C++ string operator :  int i; string s = "....";
       i=s.length ; i=s.size; 
       i=s.find("qsdqs");i=s.rfind("sqs"); i=s.find("qsq",2); 
       s(3:4)="sdfffsf";  // replace from char 3 to 4 by 
       // s[3] not alloued to day (char type is not well type in freefem++).
       getline(cin,s); 
  *  Correct bug when passing parameter string in function
  *  add freefem++  mode for mi editor on macos http://www.mimikaki.net/en/
    store in  mode-mi-edp.zip archive (unzip and put in the Folder  openned with the 
    mi "Option->Open Mode Folder" menu and set "mi" is the Default application 
    for all the .edp file). 
    
V2.4-2
  * add lot missing vectorial operation with sub array 
  * if u is a complex array (i.e. complex[int u] ..;) 
     then  u.re (resp u.im) are  the real array of the real (resp. imag ) 
     part of the vector u
  * in block matrix add utilisation of array to build the block matrix
    see new Laplace-lagrange-mult.edp file to have an example
V2.3-3
 * add outer  product to set matrix or full real or complex 2D array
    A = 2*a*b'; // where A is a "real[int,int]", b and c a "real[int]".
    A += 2*a*b'; or  A = 2*a*b';
   add matrix and array renumbering (see sec. 4.8 of freefem++.doc for all details)
    
V2.3-0
 * add three key word for future use try,catch throw 
   a exception handling (no memory management.) 
   try {   code ; }
   catch (...) { code ; }
 * use tolpivot= to set the pivottol in LU, crout, cholesky factorization 
V2.2-1
 * Add init file to set verbosity, includepath  , loadpath and 
   add preload (dynamic link file)
   the file are under unix and MacOs
	 /etc/freefem++.pref
         /Users/hecht/.freefem++.pref
         freefem++.pref
   under windows
         freefem++.pref
   and the syntaxe of the files is

     verbosity= 5
     loadpath += "/Library/FreeFem++/lib"
     loadpath += "/Users/hecht/Library/FreeFem++/lib"
     includepath  += "/Library/FreeFem++/edp"
     includepath  += "/Users/hecht/Library/FreeFem++/edp"
     #  comment 
     load += "funcTemplate" 
     load += "myfunction" 

 * Add search directory for include, and load  under Unix and Windows
   Just on the shell sh  set environment variable
    export FF_VERBOSITY=50 
    export FF_INCLUDEPATH="dir;;dir2"
    export FF_LOADPATH="dir;;dir3""
   remark the separtor of directory  is ";" and not ":" because ":" is use under Windows.
 * Examples of Add new Finite element with dynamic link/load,  Morley and Bernardi Raugel   
   see examples++-load 
V2.1-2
 * add more example of dynamic load example (thank to F. Dortu)
 * correct mistake in simple integral computation (not in variationnal 
   form)  before we alway use a quadrature formula of order 3, 
   now we can change the quadrature formula and the defaut is of order 5. 
V2.0-4
 * change in mshptg the size of the small possible edge from
   1/32000 to 1/1 000 000 000, this mesher is use in trunc, 
  splitmesh, and triangulate functions.

V2.0-3
 * correct operator x =/,  it does  nothing before. (v2.0-3)
   see end  examples++-tutorial/array.edp for all array operator
   =  +  -  *  /  .*  ./ += -= /= *=  : .l1 .l2 .linfty .sum .max .min ' 
   
 * correct bug when solving complex linear system with UMFPACK
   before we solve the system when the conjugate matrix


Major Change previous version
-----------------------------
 * add ternary C expression
     Symbol       Example      Explanation
       ?:          a?b:c     ternary operation
 
 * add 
   sparse matrix computation,
   block matrix construction 
   interpolation construction
   see 
     examples++-tutorial/sparse-cmatrix.edp  
     examples++-tutorial/sparse-matrix.edp
     examples++-tutorial/mat_interpol.edp
   not so well tested, but could be worse. 
   
 * add example to solve vartionnal inequation. VI.edp

 * add  possibility to build matrix where the 3 meshes 
   the integrale mesh, unkwon FE mesh, test FE function 
   can be different.  You must use varf to build the matrix 
   is not a standard porblem.
  
 * Improvement of  dynamic loading facility under the 3 systems
   add a fast fourier transform with dynamic loading
   see examples++-load/dfft.edp