GitLab

Do not package tutorials into source package.

The MR just disables the tutorials repository from being included in the Nektar++ source packages if the submodules are cloned in the working directory.

See merge request !625

Fix/single mode

This MR fixes the linearised advection operators to work correctly with stability analysis using a full complex single mode.

See merge request !597

Fix for Debian shlibdeps not being called.

This MR fixes a regression in the Debian packaging which results in the package dependencies not being calculated with dpkg-shlibdeps.

See merge request !616

Extend some meshing modules to 2D

This MR extends the functionality of NekMesh in 2D:
- The cyl module now works in 2D. The cylinder center can now be passed as a parameter, instead of always being at the origin.
- The boundary layer module is extended to Quads.
- A new module was created that generates curved edge information in 2D based on a prescribed function y=f(x).


See merge request !583

Removed incorrect constraint requiring base flow to be 2D2C for SM.

Fix 3DH1D copy constructor for boundary extraction

This MR introduces a new constructor for ContField3DHomogeneous1D, providing a proper fix for the FieldConvert extract module in this case (this only worked by avoiding the copy constructor). We still need to forced Field.hpp to use a valid variable name, since using DefaultVar would lead to a field with no boundary expansions.

See merge request !590

Fix mpi problem when terminating after finding NaN.

This fixes a bug causing the simulation to hang when only some of the mpi processes encounter a NaN.

See merge request !605

Move executable name override to TestData class.

This MR moves the parsing of a command-line override of the executable to use for a test into the TestData class so we do not expect an tag in the .tst file.

See merge request !595

Fix problem in QuadExp with different number of points per direction

This fixes a few bugs in `QuadExp`, which prevent simulations using quads from working when the number of points is not the same in both directions.

See merge request !607

Fix/compressed io seg curved

Provide a fix to issue #24 where extracting a 1D manifold from a 2D mesh did not contain curvature when the file was in compressedIO format. This was due to an incorrect reference to a face map rather than an edge map. 

See merge request !619

fix lapack call

This fixes a small bug in the lapack call for eigensolve of full nektar matrices. Previously it was not assigning enough room in memory for lapack

See merge request !610

Fix FieldConvert processing of partitions in serial

This MR makes two changes related to using FieldConvert on individual partitions in serial:
- Fix the call to PartitionMesh inside OutputInfo. This was previously using `PartitionMesh(false)`, when the first parameter should be the number of partitions.
- Force FieldConvert to use the `forceoutput` option when using `procid`. If this was not done, whenever `procid != 0` the output would be automatically skipped. 

See merge request !612

Fix MultiLevelStaticCond with periodic BCs

This should fix issue #21, regarding the use of MultiLevelStaticCond with periodic boundary conditions in parallel. Basically, this involves forcing all periodic boundary conditions to be on the last level of the static cond. I expected doing this only for the dofs connected to other partitions would be enough, but apparently it is not since it leads to a segfault in one of the cases I tested.

See merge request !614