Targets ctest and ctest-large now run 3 parallel processes.

Lars Bilke authored 7 years ago and Dmitri Naumov committed 7 years ago

- Do not run in parallel with unit tests
- ctest-large target run tests sequentially

Fixing HM specific error.

Dmitri Naumov authored 7 years ago and Dmitri Naumov committed 7 years ago

This is important for the dN_u/dx * N_p products
for displacement and pressure shape functions.

Testing additionaly on Quad9, Tri6, and Hex20 meshes.
The latter with plain strain.

 - Fix error in the derivatives.
 - Rewrite the expressions using more simple and explicit formulation.

This is failing with adaptive time stepping scheme using LIS library and enters an infinite loop.
This is to be fixed by Wenqing.

Improved the convergence control of the coupling iterations

[PCS/LIE/HM] add output of nodal w and aperture

Small fixes

Patches for adaptive time stepping

added a warning message about the times of repeated rejected time steps.

From the standard: If the initializer list has no elements and T is
a class type with a default ctor, the object is value-initialized.

The warning appearing in gcc-4.9 is due to a c++11 defect
fixed in
http://open-std.org/JTC1/SC22/WG21/docs/cwg_defects.html#1324
which points to
http://open-std.org/JTC1/SC22/WG21/docs/cwg_defects.html#1301.

The warning is fixed in later versions of gcc:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55805

Use static_cast to unsigned integer because count is non-negative.
Check if iterator is not pointint to the list's end before access.

For others use more specialized includes like fstream.

It affected the initial guess for linear solver in the Picard linearization.

Restructure SmallDeformation process

The DisplacementDim is actually equal to the GlobalDim.
We never have used non-equal mesh dimensions and process
dimensions (e.g. 2d deformations in 3d space)