This is the list of changes made after the release of ASPECT version 1.1 for version 1.2. All entries are signed with the names of the author.

1. The default for the number of zones in "depth average" got reduced from 100 to 10. A large number of zones causes erroneous output on coarse meshes.
   (Timo Heister, 2015/01/19)

2. New: ASPECT now uses meshes for the spherical shell where each node is placed on concentric shells. Previously, upon mesh refinement, new vertices were placed at an averaged location of the vertices of the mother cell, leading to a certain degree of distortion of cells. It also led to a situation where not all cells are equally shaped due to this distortion. The new mesh, in contrast, is completely symmetric. Additionally, a higher order mapping to represent curved faces is now also used for the interior cells using the manifold description of the shell/sphere.
   (Wolfgang Bangerth, Timo Heister, 2015/01/16)

3. New: GeometryModel::has_curved_elements() will allow for optimizations if all boundaries (and manifold description) are straight.
   (Timo Heister, 2015/01/16)

4. Fixed: the inner solvers for the Stokes block preconditioner now start with a zero initial guess. This should make the solvers more robust.
   (Timo Heister, 2015/01/16)

5. ASPECT now requires deal.II version 8.2 or later.
   (Timo Heister, 2015/01/13)

6. New: The parameters for the inner Algebraic Multigrid preconditioner used in the Stokes solve have been tuned to achieve a speedup of the solver phase by a factor between 1.2 and 2, depending on application.
   (Wolfgang Bangerth, 2014/12/05)

7. New: There is now the option to specify an origin when using the box goemetry model.
   (Ryan Grove, 2014/12/09)

8. Changed: The behavior when one changed the visualization output_interval during a checkpoint restart was previously undefined, and working in slightly unexpected ways like never writing output for the first timestep after the restart. This now works as one would expect, e.g. every timestep that ends more than output_interval after the last output time step will produce output. Old checkpoint files will continue to work, only with a possible short gap in visualization output right after restart.
   (Rene Gassmoeller, 2014/12/03)

9. New: There is now the possibility to interpolate the visualization output to a refined output mesh. This accounts for the fact that most visualization software only offers linear interpolation between grid points and therefore the output file is a very coarse representation of the actual solution field. The new output is still a linear interpolation, but at least on a finer grid. The visualization output quality and used disk space will increase when activating this option.
   (Rene Gassmoeller, 2014/09/14)

10. Fixed: There was a race condition in writing output files: we write to them in a temporary location and then move them to their final location. We used to do the second step before actually closing the file we just wrote to. Under some circumstances, this could lead to incomplete or empty output files. This is now fixed.
    (Wolfgang Bangerth, 2014/09/03)

11. Fixed: Running in release mode and with user-defined libraries loaded that were compiled against the debug version of deal.II, or the other way around, likely produces effects that are undesirable. ASPECT now tries to detect these situations and abort the program.
    (Wolfgang Bangerth, 2014/09/02)

12. New: There is now a new section in the manual that documents running a free surface computation with a crust as a stagnant lid overlying a convecting mantle.
    (William Durkin IV, Wolfgang Bangerth, 2014/09/01)

13. New: There is now a new section in the manual that documents running the benchmarks proposed by Davies et al.
    (William Durkin IV, Wolfgang Bangerth, 2014/08/29)

14. Fixed: The initial condition 'adiabatic' had a minor bug in the case of using a model without adiabatic heating and a prescribed bottom boundary layer. In this case the model should create a constant temperature profile with boundary layers to the boundary temperature. Due to a bug the bottom boundary layer amplitude was calculated against an adiabatic profile instead of a constant temperature. This is fixed now.
    (Rene Gassmoeller, 2014/08/19)

15. Changed: The GPlates velocity boundary plugin prescribed only the normal vector of the model plane of 2D models. Therefore it was very hard to rotate the model into its actual orientation, when comparing it with other datasets. Now the first of the user input points is always rotated to a known position (0,1,0), and the plugin outputs the according rotation angles to rotate the model into its proper orientation and the inverse angles to rotate other datasets into the model plane.
    (Rene Gassmoeller, 2014/08/15)

16. New: There are numerous places in the input file where one can or has to input boundary indicators. These indicators identify individual parts of the boundary of the domain, such as the inner or outer surfaces of a spherical shell. However, remembering these numerical indicators when writing or reading input files is a hassle and error prone.
    Given this problem, there is now functionality in ASPECT so that geometry models can (and, for the existing models, do) provide symbolic names for each of the boundary parts. These symbolic names can then be used in the input files wherever it was previously necessary to use numbers.
    (Wolfgang Bangerth, 2014/07/21)

17. New: The spherical hexagonal perturbation initial temperature model has been generalized to allow other modes than just the hexagonal one.
    (Joey Durkin, 2014/08/04)

18. New: The spherical shell geometry model in 2d now takes additional parameters indicating the number of circumferential cells.
    (Joey Durkin, 2014/08/04)

19. Changed: Updated maximum refinement function with the same bugfix as the minimum refinement function. It now always declares dim number of variables. When using the 'depth' coordinate system, only the first variable is used. The others are set to zero.
    (Jonathan Perry-Houts, 2014/08/04)

20. Fixed: The viscosity mesh refinement criterion did not ask the material model for the viscosity, therefore it did not work with many material models. Some material models calculate the viscosity anyway, so it worked in that cases. Now the criterion asks correctly for the viscosity.
    (Rene Gassmoeller, Juliane Dannberg 2014/08/04)

21. Changed: The minimum refinement function now always declares three variables. In the 'depth' option only the first one is set to the depth, the other variables are set to zero.
    (Rene Gassmoeller, 2014/08/04)

22. New: There are now function parser plugins for the gravity model and boundary temperature model. Gravity can be prescribed in dependence of position and boundary temperature in dependence of position and time.
    (Sanja Panovska, Rene Gassmoeller, 2014/08/01)

23. New: There is now a postprocessor that outputs multiple material properties with just a single call to the material model. This is more efficient, but only matters for complex material models.
    (Rene Gassmoeller, 2014/08/01)

24. New: There is now a section in the manual detailing the "Burstedde" benchmark and its verification.
    (Iris van Zelst, Wolfgang Bangerth, 2014/07/07)

25. New: Added "maximum refinement function" plugin to limit the depth of the mesh octree.
    (Jonathan Perry-Houts, 2014/07/03)

26. Changed: The viscosity in the Steinberger material model can now be calculated by taking either the laterally averaged temperature or the adiabatic temperature at this depth as reference. (Rene Gassmoeller, 2014/06/20)

27. Changed: The minimum refinement function is now evaluated at every cell corner instead of the cell midpoint. This especially helps in refining thin boundary layers.
    (Rene Gassmoeller, 2014/06/20)

28. Changed: The internal heating statistics postprocessor now calculates the integral and averaged internal heating by mass averaging instead of volume averaging.
    (Rene Gassmoeller, 2014/06/20)

29. New: Added a visualization processor to output the local gravity vector.
    (Timo Heister, 2014/06/17)

30. Fixed: The direct solver now works with nonlinear schemes like iterated IMPES.
    (Timo Heister, 2014/06/11)

31. Changed: The maximum number of solver iterations for all linear solvers (including the inner solvers) is now set to 1000.
    (Timo Heister, 2014/06/08)

32. New: Initial temperature condition "S40RTS perturbation" has been added, that provides a perturbation for a shell geometry based on the S20RTS or S40RTS global shear wave velocity model data.
    (Jacky Austermann, 2014/06/06)

33. Fixed: When a linear solver fails to converge in a parallel program, every processor would output the same error message – leading to incredible amounts of entangled error messages. This has now been resolved: every processor still fails, but only processor 0 reports the error.
    (Wolfgang Bangerth, 2014/06/06)

34. Fixed: When setting "Use years in output instead of seconds" the velocity solution is now exported in m/year instead of m/s in visualization files.
    (Timo Heister, 2014/06/02)