order 7->9 seem to be wrong : the way shape functions are
computed is maybe numerically unstable
 

No commit message

No commit message

all contrib/ stuff should be optional -> defined HAVE_BFGS ; the Lloyd code must still be #ifdef'ed accordingly

No commit message

* the old code has been generalized to deal with these cases
* the new "in entity" boundary layer method(s) will not create a topology of
  "boundary layer entities" anymore, so this is superfluous

Added MElement::integrate, MElement::integrateCirc,
MElement::integrateFlux in similar fashion as in deprecated
Post/shapeFunctions.h

Integration of p-covector fields to riemann.

streamwise edge length is ok
functionIntegrator interface ok
functionevaluator with dgdof solution

linearSystemPETSc : move implementation to linearSystemPETSc.hpp included only in linearSystemPETSc.cpp. I hope it does not break petsc when compiled for complex. If it does, check the first lines of linearSystemPETSc.cpp (and send me an email :p --jon)

removed so benchmarks has to update)
Add function on fullMatrix to assemble matrix by perturbations

Integration on bulk and virtual interface terms are performed by a
matrix vector product. Add 2 functions in fullMatrix.h and cpp to
realize this operation