Bugs fixed in 1.0pre10:
* Bug #11530: "cannot install..." [bison checking in configure]
* Bug #7709 : "Mesh Transformation code not finished" [code finished]
* Fixed Debian Bug#420922: FTBFS with GCC 4.3: missing #includes 
* The code is built even when cannot build documentation
* P2 and Q2 elements are now fully working (thanks to Edouard Oudet's reports)

New in 1.0pre10:
* Added function building simplifier [optimization]
* Removed deprecated functions from language syntax
* Save multiple functions and fields into one VTK file
* Many ff3d feature are now tests in the test suite
* Output improvements: save/read in vtk format for vectors, fields,...



Bugs fixed in 1.0pre9:
* Fixed advection operator on cartesian meshes
* Language: fixed "on M" for non-cartesian meshes boundary conditions
* Fixed Debian Bug#358053: FTBFS with G++ 4.1: extra qualification
* Fixed P1: first order and second order derivative right hand side
  for variational formulae
* Fixed #16041: related to bad error message

New in 1.0pre9:
* Completely rewrote the function algebra
* Added (at last!) a set of non-regression tests
* Added reader for 2D 'am_fmt' mesh format
* Added domain by function definition (e.g. domain(f<g))



Bugs fixed in 1.0pre8:
* Fixed natural boundary conditions for P1 (unstructured meshes)
  thanks to the report of Pierre Michaud.
* Replaced the 'opendx' keyword by 'raw': 'opendx' keyword was confusing
  since it is not available yet.
* Fixed #14841: "normal vector of surfaces"
* Fixed unknown number in language for vectorial boundary variational
  formulae
* Important bug fixed: non-symetric systems defined using variational
  formula were not correctly discetized (coeffcients were not stored
  correctly in the matrix)

New in 1.0pre8:
* Added P2 and Q2 finite elements [this is still under development]
* Added "normal" component functions: nx,ny and nz.
* Began non-regression testsuite using autotest
* For clarity, changed syntax for integrals:
	"int(set)(f)" to "int[set](f)"
  this also applies for boundary terms in variational formulae
* Cleaned-up output
* Improved exception handling
* Faces and edges list are built on-demand, this should lower memory
  use.
* Matrix-free computations should now give the same result as
  assembled-matrices computations up to the machine precision.
* Improved GUI and ported it to Qt4 and Vtk5 [this code is still
  experimental]


Bugs fixed in 1.0pre7:
* fixed Debian Bug#293018: g++-4.0 compilation issue
* closed Bugs (see http://savannah.nongnu.org/bugs/?group=ff3d)
  #11330: fixed compilation without autogen
  #11331: libtool.m4 file is now part of the distribution
* fixed mesh file checking. Bad vertices numbers could lead to
  segmentation fault when dealing with quadrilaterals, tetrahedra or
  hexahedra meshes
* improved mesh reading: uses fscanf instead of ifstream. Reading
  meshes files has been speed-up by a factor at least 10.

New in 1.0pre7:
* Added periodical Cartesian mesh support! The syntax makes periodic
  computations very easy: 'mesh M=periodic(givenMesh,0:1[,...]);'
  where '0:1' means that faces 0 and 1 define periodicity. Faces
  numbering is as usual 0=xmin, 1=xmax, 2=ymin, 3=ymax, 4=zmin,
  5=zmax. 
* Added support for reading gmsh version 1 meshes.
* Added 'finemesh' and 'coarsemesh' keywords. These keywords modify
  the behavior of ff3d. 'finemesh;' instruction tells ff3d to built
  precisely intersections of objects when building surface meshes.
  This was the previous behavior. 'coarsemesh;' is now the default
  behavior, it leads to non computing triangles intersections. This
  produces sufficient meshes for fictitious domain methods and is
  much stable and faster.
* ff3d now uses exception mechanism. This allows better error messages
  and is necessary to the creation of the future GUI.
* Multi-threading support really began in this release. This will as
  well serve the GUI management and performances enhancements.
* AutoOpts was dropped in this release, command line is now parsed by
  hand. There are two reasons for this. First concerns portability,
  end-users had to install autogen and libopts to access some
  functionalities. The second reason is more important: it seems that
  it was not easy to mix the use of both AutoOpts and Qt; I decided to
  let down the first. The problem is that info and manual pages are no
  more autogenerated.


Bugs fixed in 1.0pre6:
* closed Bugs (see http://savannah.nongnu.org/bugs/?group=ff3d)
  #8650: "natural convection example crashes".
          This particular crash is fixed. This does not mean the
          surface mesh generator is bug free...
* fixed log function name thanks to the report of Matthieu Guillon.
* fixed positive vectorial operators thanks to the report of Luigi
  Bardelli.
* Surface generator seems to be more stable (see below).
* Fixed all examples. It was quite a mess: some none symetric linear
  systems were "solved" using conjugate gradient, for instance :-(

New in 1.0pre6:
* Surface mesh generator now uses only tetrahedra mesh. It seems to be
  more stable :-) but even slower :-( This comes from some
  connectivity constructions that are performed only once.
* As suggested by Lawrence R. Frank, a 'plot(u,M)' instruction was
  added to allow plotting of solution using VTKDriver. This allows VTK
  hackers to proceed their own visualization editing only the
  'graphics/VTKDriver.cpp' file. The implemented code is sill just a
  toy. 
* Added a new tetrahedrization tool. This uses a domain and a
  background mesh, an extracts the relevant cells, ie: the one that
  carry degrees of freedom. This is intended to be used for
  visualization only, but is very important in the fictitious domain
  context, since visualizing the solution on the full mesh
  (background) is very difficult: null values will be plotted for none
  degrees of freedom vertices, so for instance, if the solution lives
  between 100 and 101 the visualization scale will be [0-101]. You
  will think that the solution is neer to a constant.
  Using the new visualization mesh, the visualization will not be
  perturbed by unrelevant 0 values. Here comes an example of use
  (let us assume that "m" is a background mesh, "D" a domain, and
  "u" the solution of some "solve (u) in D by M" statement):
	mesh V = tetrahedrize(D,M); // <- new
	save(medit,"u",u,V);
	save(medit,"u",V);
  and that's all! You can also use the new
	plot(u,V);
  instruction just to get a quick idea of the solution.


Bugs fixed in 1.0pre5:
* fixed Debian #258840 (g++-3.4 compilation) thanks to the report of
  Andreas Jochens.
* surface mesh referencing (thanks to Arun Viswanathan's report). This
  bug was introduced in 1.0pre5.
* reading of functions saved in Medit format
* fixed documentation Postscript file generation (thanks to Karl
  Hauschild remark)
* fixed compilation on Alpha architecture, renaming __alpha variables
  to __Alpha (Christophe Prud'homme report)

^L
Bugs fixed in 1.0pre4:
* Surface mesh generation fixed thanks to the report of Tobis
  Naehring: transformations were not taken into account for composed
  objects.
* Fixed boundary and surface mesh association for variationnal
  formulae.

New in 1.0pre4:
* Tarballs will now be produced following each (pre-)release (this
  simplifies the compilation task).
* doc/paramters.tex is now generated automagically
* --enable-opt became --enable-optimize in configure script
* Plugged P1 FEM functions derivatives evaluation


New in 1.0pre3:
* Added examples/ to tarball distribution (at last)
* Improved documentation generation (pdf and ps)
* Improved mesh reader (no more crash with Ridges, Edges and Corners)
* convection operator is now implemented for Tetrahedral meshes.


Bugs fixed in 1.0pre2:
* closed Bugs (see http://savannah.nongnu.org/bugs/?group=ff3d)
  #6932: "Not enough precise evaluation of domain intersection"
  #6933: "File presence not tested"
  #7461: "add 'inverse' keyword"
  #7482: "Sparse matrix initialization"
  #7545: "Not implemented output for boolean functions"
  #8139: "parse.ff.yy failes to compile with -Ox x>=1"
  #8140: "default compilation flags are not good"
* fixed center of mass for reference tetrahedron: (0.5,0.5,0.5) was
  used.

New in 1.0pre2:
* ff3d is distcheck clean now (prudhomm@debian.org)

Local variables:
mode: outline
paragraph-separate: "[ 	]*$"
end:
