3.0.4 (December 9, 2018): allow general groups in Jacobian definitions; fixed
string parser regression.

3.0.3 (October 18, 2018): new AtGaussPoint PostOperation option; bug fixes.

3.0.2 (September 10, 2018): small compilation fixes.

3.0.1 (September 7, 2018): small bug fixes.

3.0.0 (August 22, 2018): new extrapolation (see SetExtrapolationOrder) in
time-domain resolutions; new string macros; added support for Gmsh MSH4 file
format; new file handling operations and ElementTable format in PostOperation;
added support for curved (2nd order) simplices; enhanced communication of
post-processing data with ONELAB; many new functions (Atanh, JnSph, YnSph,
ValueFromTable, ListFromServer, GroupExists, ...); various small bug fixes.

2.11.3 (November 5, 2017): new 'Eig' operator for general eigenvalue problems
(polynomial, rational); small improvements and bug fixes.

2.11.2 (June 23, 2017): minor build system changes.

2.11.1 (May 13, 2017): small bug fixes and improvements.

2.11.0 (January 3, 2017): small improvements (complex math functions, mutual
terms, one side of, get/save runtime variables) and bug fixes.

2.10.0 (October 9, 2016): ONELAB 1.3 with usability and performance
improvements.

2.9.2 (August 21, 2016): small bug fixes.

2.9.1 (August 18, 2016): small improvements (CopySolution[], -cpu) and bug
fixes.

2.9.0 (July 11, 2016): new ONELAB 1.2 protocol with native support for lists;
simple C++ and Python API for exchanging (lists of) numbers and strings;
extended .pro language for the construction of extensible problem definitions
("Append"); new VolCylShell transformation; new functions (Min, Max, SetDTime,
...); small fixes.

2.8.0 (March 5, 2016): new Parse[], {Set,Get}{Number,String}[] and
OnPositiveSideOf commands; added support for lists of strings; various
improvements and bug fixes for better interactive use with ONELAB.

2.7.0 (November 7, 2015): new Else/ElseIf commands; new timing and memory
reporting functions.

2.6.1 (July 30, 2015): enhanced Print[] command; minor fixes.

2.6.0 (July 21, 2015): new ability to define and use Macros in .pro files; new
run-time variables (act as registers, but with user-defined names starting with
'$') and run-time ONELAB Get/Set functions; new Append*ToFileName PostOperation
options; new GetResdidual and associated operations; fixes and extended format
support in MSH file reader; fixed UpdateConstraint for complex-simulated-real
and multi-harmonic calculations.

2.5.1 (April 18, 2015): enhanced Python[] and DefineFunction[].

2.5.0 (March 12, 2015): added option to embed Octave and Python interpreters;
extended "Field" functions with gradient; extended string and list handling
functions; new resolution and postprocessing functions (RenameFile, While, ...);
extended EigenSolve with eigenvalue filter and high order polynomial EV
problems; small bug fixes.

2.4.4 (July 9, 2014): better stability, updated onelab API version and inline
parameter definitions, fixed UpdateConstraint in harmonic case, improved
performance of multi-harmonic assembly, fixed memory leak in parallel MPI
version, improved EigenSolve (quadratic EVP with SLEPC, EVP on real matrices),
new CosineTransform, MPI_Printf, SendMergeFileRequest parser commands, small
improvements and bug fixes.

2.4.3 (February 7, 2014): new mandatory 'Name' attribute to define onelab
variables in DefineConstant[] & co; minor bug fixes.

2.4.2 (Septembre 27, 2013): fixed function arguments in nested expressions;
minor improvements.

2.4.1 (July 16, 2013): minor improvements and bug fixes.

2.4.0 (July 9, 2013): new two-step Init constraints; faster network computation
(with new -cache); improved Update operation; better cpu/memory reporting; new
-setnumber, -setstring and -gmshread command line options; accept unicode file
paths on Windows; small bug fixes.

2.3.1 (May 11, 2013): updated onelab; small bug fixes.

2.3.0 (March 9, 2013): moved build system from autoconf to cmake; new family of
Field functions to use data imported from Gmsh; improved list handling; general
code cleanup.

2.2.1 (July 15, 2012): cleaned up nonlinear convergence tests and integrated
experimental adaptive time loop code; small bug fixes.

2.2.0 (June 19, 2012): new solver interface based on ONELAB; parallel SLEPC
eigensolvers; cleaned up syntax for groups, moving band and global basis
functions; new Field[] functions to interpolate post-processing datasets from
Gmsh; fixed bug in Sur/Lin transformation of 2 forms; fixed bug for periodic
constraints on high-order edge elements.

2.1.1 (April 12, 2011): default direct solver using MUMPS.

2.1.0 (October 24, 2010): parallel resolution using PETSc solvers; new Gmsh2
output format; new experimental SLEPc-based eigensolvers; various bug and
performance fixes (missing face basis functions, slow PETSc assembly with global
quantities, ...)

2.0.0 (March 16, 2010): general code cleanup (separated interface from kernel
code; removed various undocumented, unstable and otherwise experimental
features; moved to C++); updated input file formats; default solvers are now
based on PETSc; small bug fixes (binary .res read, Newmark -restart).

1.2.1 (March 18, 2006): Small fixes.

1.2.0 (March 10, 2006): Windows versions do not depend on Cygwin anymore; major
parser cleanup (loops & co).

1.1.2 (September 3, 2005): Small fixes.

1.1.0 (August 21, 2005): New eigensolver based on Arpack (EigenSolve);
generalized old Lanczos solver to work with GSL+lapack; reworked PETSc
interface, which now requires PETSc 2.3; documented many previously undocumented
features (loops, conditionals, strings, link constraints, etc.); various
improvements and bug fixes.

1.0.1 (February 6, 2005): Small fixes.

1.0.0 (April 24, 2004): New license (GNU GPL); added support for latest Gmsh
mesh file format; more code cleanups.

0.91: Merged moving band and multi-harmonic code; new loops and conditionals in
the parser; removed old readline code (just use GNU readline if available);
upgraded to latest Gmsh post-processing format; various small enhancements and
bug fixes.

0.89 (March 26, 2003): Code cleanup.

0.88: Integrated FMM code.

0.87: Fixed major performance problem on Windows (matrix assembly and
post-processing can be up to 3-4 times faster with 0.87 compared to 0.86,
bringing performance much closer to Unix versions); fixed stack overflow on Mac
OS X; Re-introduced face basis functions mistakenly removed in 0.86; fixed
post-processing bug with pyramidal basis functions; new build system based on
autoconf.

0.86 (January 25, 2003): Updated Gmsh output format; many small bug fixes.

0.85 (January 21, 2002): Upgraded communication interface with Gmsh; new
ChangeOfValues option in PostOperation; many internal changes.

0.84 (September 6, 2001): New ChangeOfCoordinate option in PostOperation; fixed
crash in InterpolationAkima; improved interactive postprocessing (-ipos);
changed syntax of parametric OnGrid ($S, $T -> $A, $B, $C); corrected Skin for
non simplicial meshes; fixed floating point exception in diagonal matrix
scaling; many other small fixes and cleanups.

0.83: Fixed bugs in SaveSolutions[] and InitSolution[]; fixed corrupted binary
post-processing files in the harmonic case for the Gmsh format; output files are
now created relatively to the input file directory; made solver options
available on the command line; added optional matrix scaling and changed default
parameter file name to 'solver.par' (Warning: please check the scaling
definition in your old SOLVER.PAR files); generalized syntax for lists
(start:[incr]end -> start:end:incr); updated reference guide; added a new short
presentation on the web site; OnCut -> OnSection; new functional syntax for
resolution operations (e.g. Generate X -> Generate[X]); many other small fixes
and cleanups.

0.82: Added communication socket for interactive use with Gmsh; corrected
(again) memory problem (leak + seg. fault) in time stepping schemes; corrected
bug in Update[].

0.81: Generalization of transformation jacobians (spherical and rectangular,
with optional parameters); changed handling of missing command line arguments;
enhanced Print OnCut; fixed memory leak for time domain analysis of coupled
problems; -name option; fixed seg. fault in ILUK.

0.80: Fixed computation of time derivatives on first time step (in
post-processing); added tolerance in transformation jacobians; fixed parsing of
DOS files (carriage return problems); automatic memory reallocation in
ILUD/ILUK.

0.79: Various bug fixes (mainly for the post-processing of intergal quantities);
automatic treatment of degenerated cases in axisymmetrical problems.

0.78: Various bug fixes.

0.77: Changed syntax for PostOperations (Plot suppressed in favour of Print;
Plot OnRegion becomes Print OnElementsOf); changed table oriented
post-processing formats; new binary formats; new error diagnostics.

0.76: Reorganized high order shape functions; optimization of the
post-processing (faster and less bloated); lots of internal cleanups.

0.74: High order shape functions; lots of small bug fixes.

0.73: Eigen value problems (Lanczos); minor corrections.

0.7: constraint syntax; fourier transform; unary minus correction; complex
integral quantity correction; separate iteration matrix generation.

0.6: Second order time derivatives; Newton nonlinear scheme; Newmark time
stepping scheme; global quantity syntax; interactive post-processing; tensors;
integral quantities; post-processing facilities.

0.3: First distributed version.
