Joshua, first, thanks for writing this code. Second: I've committed it to our subversion repository, with a few minor edits and split into a .cc and a .h file. Third, I've already fixed a couple of assertions you hit in your code that had to do with interfacing the FE_Nothing with FE_Q.
Now for the rest: I've modified your little test program into a program for our testsuite. Can I ask you to read through the "Regression tests" section of this page: http://www.dealii.org/developer/development/testsuite.html If you check out the testsuite from svn (along with the current version of the library proper) you'll find your program as hp/fe_nothing_01.cc. It still fails, but at a later point from where it stopped for you -- in FiniteElement::compare_for_face_domination, a function that you also need to overload in FE_Nothing. Face domination is a concept that is explained in a bit detail in http://www.math.tamu.edu/~bangerth/publications/2006-hp.pdf In essence, in the implementation of this function you'll have to say whether the FE_Q or the FE_Nothing is the master -- a decision the implementations in the two classes have to agree upon as they both could be asked this question. In FE_Q I've picked the FE_Q as the master, so you'll have to do the same in FE_Nothing. See how far you'll get after implementing this function and let me know when you don't know any further again! Again, thanks for working on this! Wolfgang PS: When you have time, would you mind copying (and if necessary augmenting) the comments for all the functions in FE_Nothing from FiniteElement so they would show up in the documentation pages for FE_Nothing? -- ------------------------------------------------------------------------- Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
