it comes from bug fixes and improvements that I
made to version lifev-0.4.1.tar.bz2 (not CVS)
cause I had not CVS access when I began.
so this version doesn't include all your changes made between
0.4.1 and the last one.
Anyway this includes support for reading netgen meshed and writing
netgen solutions, a test_netgen in testsuite and many bug fixes.
I hope it's all ok, becouse I simply copyed over a checkedout version
my files recursively. If you have problems compiling you could try
autoconf at the beginning. Bye

you pass __ns->yesFlux(1). The flag 1 is for the flux problem. For the other problems the flag is set equal to 0.

  velocity function (preferably the exact solution if available)
  if not called, the convective terms are linearized around the last iteration
  in the steady case or around the bdf extrapolation in the unsteady case
* do not fix one pressure node for matrix definiteness as this breaks
  convergence in the time dependent case
* set first pressure node to zero by post processing
* make sure _u and _p contain the initial values after peeking into bdf

component at a different timestep (u0x at t=0, u0y at t=-dt, etc.)

This might increase convergence behaviour in tests with exact solutions :-)

to ease the reading of the formulas

Christian what do u think ?
I don´ t think I have made a bouh-bouh

see http://www.boost.org/libs/numeric/ublas/doc/vector_expression.htm

for more details

 * get rid of one warning

Christian can u check ?

only one for now: pointer function versus boost::function
in fact we should have something like benchmark since it is not really a test

for the moment I commit this. Now it works well. The NSWithFlux_new.hpp with the possibility to impose 2 fluxes is under check.

too low

of mean values useful for fields defined up to a constant

  -> useful for linearization around exact solution
* rewrapped for readability
* translated non-english comments to english

PLEASE USE ENGLISH WORDS AND LANGUAGE IN THE ENTIRE CODE (variable names,
comments, output, etc.) Not all LifeV users understand italian!

* print number of iterations and condition number estimate, if available

* use penalize jumps of divergence and streamline gradient by default

An interface between 2 tubes is created.
Two automatic tests are run (1 tube or 2).

A serious clean-up is now necessary.
A clever interface should be devised.
A multitube (>2) test should be implemented.
A bifurcation should be done also.

  from the design POV
* use the setSourceTerm()

* timeAdvance() more powerful by accepting now functors
* NS provides a setSourceTerm() sourceTerm() getter/setter for the source
  term so that it can be stored, it is useful for NavierStokesWithFlux_new
  in the 2 fluxes case

now we can get rid of the loops for pressure copies