NEWS

Please send CitcomCU bug reports to cig-mc@geodynamics.org.

Version 1.0.3

* Disable a system call, which was used to copy the input file, since
  some mpi implementation (e.g. on TACC Lonestar) does not support
  system call.

* Added velocity output as *.velo.* files.

* Allowing run on more than 100 processors and fixed a use of
  uninitialized memory problem in the marker routine.

* Fixed a bug in conjugate gradient solver, which was not functioning.

* Fixed some memory problems in the Stokes solver. Some arrays were not
  allcated with enough elements (off by 1). Writing to the end of those
  arrays can potentially overwrite other parts of the memory. In
  practice, it seems that this bug didn't affect the solver at all, but
  this is a subtle bug to detect.

* Fixed an index-overrun bug in advection-diffusion solver with heatflux
  boundary conditions.

* Replaced Parsing.c with the CIG CitcomS version to ensure that defaults
  settings are actually used.  Before, the respective function always
  returned NULL to avoid problems with strtok on Linux.

* Modified layers() such that the full range of 1...4 layers gets assigned
  to E->mat[], otherwise not all four layers would be used in
  Viscosity_structures.c

* Added two new rheology options.

* Added a shell script to convert the output to parallel VTK files.


Version 1.0.2

* Fixed a bug in kinematic velocity boundary conditions in spherical geometry.
  If kinematic boundary condition is used in spherical geometry, CitcomCU
  may impose incorrect velocity as the boundary conditions and gives 
  incorrect results. The error is only significant near the poles though.
  See Issue 59 in the bug tracker (http://geodynamics.org/roundup) for more
  detail.

* Fixed an off-by-one bug. This bug doesn't affect the results of CitcomCU. 
  See Issue 58 in the bug tracker (http://geodynamics.org/roundup) for more
  detail.


Version 1.0.1

* A few minor fixes for the function and variable declarations.


Version 1.0.0

* Initial public release.