Changes between version 1.2 and version 1.3

This is the list of changes made after the release of ASPECT version 1.2 for version 1.3. All entries are signed with the names of the author.

1. Fixed: We now use the correct residual to calculate the stabilization for the advection of temperature and compositional fields. Both the latent heat term for the temperature and the reaction terms for the compositional fields were missing.
   (Juliane Dannberg, 2015/05/13)

2. New: There is now a mesh refinement criterion called Boundary based on refining near a user-specified list of boundary indicators.
   (Ian Rose, 2015/05/12)

3. Fixed: The advection solver had problems converging for constant compositional fields in timestep 1. This is fixed now.
   (Rene Gassmoeller, 2015/05/11)

4. New: There are now visualization postprocessors that can output the shear stress and full stress tensors.
   (Wolfgang Bangerth, 2015/05/02)

5. Changed: ASPECT accidentally used the type unsigned int in denoting boundary indicators in BoundaryComposition::Interface::composition() and BoundaryTemperature::Interface::temperature(), as well as derived classes' corresponding functions. It should have been the deal.II type types::boundary_id. This is now fixed. It may require fixing user plugins that overload these functions.
   (Wolfgang Bangerth, 2015/04/30)

6. New: ASPECT can now average material properties between the quadrature points of a cell. This greatly increases the stability of solutions in simulations with spatially varying coefficients, and also greatly accelerates the solution, at times up to a factor of ten.
   (Wolfgang Bangerth, 2015/04/25)

7. Fixed: Removal rigid body translations and rotations when the simulation has a nullspace was bugged. This feature has been fixed and extended, and a new section has been added to the manual describing its use.
   (Ian Rose, 2015/04/23)

8. New: vtu visualization output can now be grouped into an arbitrary number of files per time step by setting "Number of grouped files" to a number larger than 1.
   (Timo Heister, 2015/04/15)

9. Fixed: To make the right-hand side of the Stokes equation compatible to the matrix we need to apply a correction for imbalanced in-/outflow across the model boundaries. This correction was accidentally applied twice in the first iteration of the iterated IMPES solver. This does not change the results, because subsequent iterations will do it correctly, but it might prevent the model to converge or slow down convergence. This is fixed now by applying the correction directly after assembling the matrix and right-hand side.
   (Rene Gassmoeller, Timo Heister, Juliane Dannberg, 2015/04/14)

10. New: ASPECT can now read the list of input arguments from the default input stdin, instead of a file, by specifying "--" for the name of the input file.
    (Wolfgang Bangerth, 2015/03/20)

11. Changed: We use the current linearization point instead of the old solution in the assembly of the composition advection system now, and update the current linearization point with the solution after we have solved all of the fields. This allows for model parameters used in the composition advection (like the reaction term) to depend on solution variables and to be updated during nonlinear iterations.
    (Juliane Dannberg, 2015/02/20)

12. Changed: The default values for the latent heat release of melting in the 'latent heat melt' material model had the wrong sign. This likely resulted from a change in the latent heat terms in the assembly long ago. All tests were correctly set up, but the default values were forgotten. This is fixed now.
    (Rene Gassmoeller, 2015/02/18)

13. Changed: The unused parameter 'Activation enthalpies' was removed from the 'latent heat' material model. The active parameter for the same purpose is 'Thermal viscosity exponent'. If there are parameter files specifying the removed parameter it is safe to just remove the line.
    (Rene Gassmoeller, 2015/02/16)

14. Fixed: The compressibility in the 'latent heat' and 'latent heat melt' material models was used incorrectly to calculate the density. This did only affect compressible models with these material models and is fixed now.
    (Rene Gassmoeller, 2015/02/12)

15. Changed: We use a new method to calculate the initial residuals used for the iterated IMPES solver. Instead of using the first solution, we use the norm of the right-hand side, so that the residual is small for cases where the initial guess of a time step is already very good, and the number of iterations is decreased for these cases (e.g. when using a smaller time step size). Moreover, the screen output of the residuals is now in the same order as the output before (about which system is solved and how many iterations were needed).
    (Juliane Dannberg, 2015/02/09)

16. New: A new plugin called 'ascii data' is available for boundary and initial conditions that reads in text data from files provided by the user. The plugin works for box and shell geometries and allows for time-dependent or constant boundary conditions. The data files must provide data for the whole model domain and the plugin interpolates the data from a structured grid to ASPECT's mesh linearly.
    (Eva Bredow, Rene Gassmoeller, 2015/02/03)

17. New: There is now a mass flux statistics postprocessor that calculates the mass flux through every boundary interface. This is helpful for regional models with prescribed boundary velocities. In general one would want to set up models with as much in- as out- flux, but in some cases it is hard to know, if the boundary plugins actually ensure this. With this plugins the net flux can be monitored.
    (Rene Gassmoeller, 2015/02/02)

18. Fixed: There was a small bug in ASPECT 1.2 that only occurred in the adiabatic initial temperature plugin, when no temperature was prescribed at any boundary, the model was compressible and a bottom thermal boundary layer was included. This is fixed now.
    (Rene Gassmoeller, 2015/02/02)

19. Fixed: The residual of the iterated IMPES solver used to be nan when the initial residual of one of the solution variable was zero. This is fixed now, so that this variable is not considered for the overall residual anymore.
    (Juliane Dannberg, 2015/01/30)

20. Fixed: The Steinberger material model had a bug in case a material table was used that had not the same number of points in temperature and pressure direction. The table was truncated in pressure dimension to the number of lines of the temperature dimension. This is fixed now.
    (Rene Gassmoeller, 2015/01/29)