###############################################################################
# #
# Trilinos Release 10.10 Changelog #
# #
###############################################################################
Release update 10.10.3 July 27, 2012
1. Epetra - Performance improvements for FEVector.
2. Epetra - Fix a bug when using Insert mode for GlobalAssemble.
3. Kokkos - Fix a bug with the default node selection.
-------------------------------------------------------------------------------
Release update 10.10.2 April 24, 2012
1. Epetra - Fix performance hot spot in Epetra_FEVector.
2. Epetra - Fix a checked stl bug in Epetra_FECrsMatrix.
3. EpetraExt - Use correct Epetra_MpiComm object for HDF5 class.
4. EpetraExt - Fix missing return in restrict_comm.
5. Multiple Packages - Fix Clang/llvm build errors in several packages. Trilinos
now builds with Clang/llvm.
6. Zoltan - Address a few compiler issues.
7. Zoltan - Fix the eigen computation for inertial matrices in
rcb/rib/hsfc.
-------------------------------------------------------------------------------
Release update 10.10.1 February 22, 2012
1. Epetra - Removed references to Epetra_MpiSmpComm.h from
EpetraExt_HDF5.cpp. This file is not distributed and was
causing build failures.
2. Epetra - Fixed two assignements to NULL that were supposed to be
comparisons against NULL in Epetra_FEVector.
3. Epetra - Add missing over-ride of InsertGlobalValues to FECrsMatrix.
4. Epetra - Improve performance of non-local data insertion in
FECrsMatrix.
4. Framework Level - Fixed an issue where the installed cmake export files
referenced the wrong directory for include files when
Trilinos_INSTALL_INCLUDE_DIR was specified manually. The
issue stemmed from a change to support installers that can
change the install directory at install time instead of
relying on the configure time settings. The fix allows both
an installer to change the installation path and manually
setting the variable Trilinos_INSTALL_INCLUDE_DIR provided
that it is relative to CMAKE_INSTALL_PREFIX. This relative
restriction has actually been silently assumed in the
makefile export files for some time it is now a stated
restriction to the entire export system.
