The inflow/outflow condition cause the NaN problem

[GoogleCodeExporter]

What steps will reproduce the problem?
1. I perform the experiment with the channel flow just according to the 
official code
2. And then I tried to modify the boundary condition(0 2 0) to (2 2 0),so that 
I can add the inflow and outflow condition.I did the modify because I found 
that if nclx==2(incompact3d.f90),then the inflow/outflow is added in the 
simulation. 
3. Oohs,sucks!  I tried to track the result with times going on, but the 
divergence U move to NaN.  Later on,I check the boundary conditions of the 
computational domain in the 《high order compact  schemes ...》:"The use of 
the Dirichlet conditions on the velocity(for no-slip or open conditions)need to 
be defined acoording to the time advancement procedure".
4. So,I vim the schemes.f90,but I don't know how to modify it,can someone give 
me some ideas?
-------------------------------------------------------
What is the expected output? What do you see instead?

   My expected output is the turbulent development process,rather than the Tollmien–Schlichting wave(http://en.wikipedia.org/wiki/Tollmien%E2%80%93Schlichting_wave) with the parallel flow assumption.
--------------------------------------------------------
What version of the product are you using? On what operating system?
  Incompact3d under ubuntu13.10 

Please provide any additional information below.

Original issue reported on code.google.com by [email protected] on 18 Apr 2014 at 4:32