From: Roy S. <ro...@st...> - 2008-08-14 19:29:28
|
On Thu, 14 Aug 2008, John Peterson wrote: > On Thu, Aug 14, 2008 at 1:32 PM, Roy Stogner <ro...@st...> wrote: >> >> One thing: we should tag (and tarball, etc.) 0.6.3 now, but we >> probably ought to write up a changelog (both for new features and API >> changes!) before we post an announcement about it. It's been a while >> since 0.6.2, and I'll have to go back through the SVN logs before I >> remember what's been new since then. But we'll definitely want to >> warn people about, e.g. the Mesh->MeshBase change and the prepending >> of libMesh_ to our Makefile variables. > > I think the changelog is a good idea. It has been quite a while since > the last release! Here's my bits: New features: Continuing experimental ParallelMesh development More internal parallelization Operator overloading on scalar/vector/tensor operations with mixed underlying types Exception throwing, tracefile output on errors More norm options for some library methods Better PETSc nonlinear solver integration Better line search in libMesh NewtonSolver Assorted bug fixes, compatibility fixes, efficiency improvements API changes: The macro error() has been replaced with libmesh_error() to avoid name conflicts, and internal assert() calls have been replaced by libmesh_assert() to enable exception-throwing and tracefile output on assertion failures. The libMesh::init()/close() methods have been replaced by a libMeshInit object's constructor and destructor, to simplify application code slightly and to provide better cleanup behavior when an uncaught exception is thrown. I think that's it? The Mesh->MeshBase changes already made it into 0.6.2, I suppose. --- Roy |