Add workaround to be able to use SLEPc >= 3.5.0 with deal.II. This requires disabling support for STFOLD spectrum transformation type since it was removed in SLEPc 3.5.0 and newer.
agrayver [Wed, 13 Aug 2014 12:40:48 +0000 (14:40 +0200)]
Fix bug in project_boundary_values_curl_conforming related to the wrong indexing of high-order DoFs in internals::compute_face_projection_curl_conforming.
This commit removes the obsolete support for a 'Nightly' track and adds a
new 'Continuous' track. The idea is that in future 'tester' uses
exclusively the 'Regression Test' track and simserv04 uses 'Continuous'.
Both testing sites have different strategies: tester goes through all
commits with one configuration (for 32 and 64 bit indices), simserv04 just
checks the latest merge available but for a number of compilers and
configuration. The current mix of test results is just confusing and very
horrible to read :-/
Matthias Maier [Fri, 8 Aug 2014 23:22:38 +0000 (01:22 +0200)]
Silence a warning
A logical construct of the form "EXPR && EXPR || EXPR" in manifold_lib.cc
triggered a bunch of warning. Silence them by explicitly enclosing the
intended precedence: (EXPR && EXPR) || EXPR.
Timo Heister [Fri, 8 Aug 2014 21:31:21 +0000 (17:31 -0400)]
Merge pull request #74 from guidokanschat/doxygen_steps_missing
Fix a bug in doxygen causing missing tutorials
The tutorial steps were missing in the online documentation.
This bug fix makes them reappear by creating the string of input directories and files only after the list has been competed.
Timo Heister [Thu, 7 Aug 2014 13:31:54 +0000 (09:31 -0400)]
Merge pull request #65 from bangerth/fix-muparser-if
Change the parsing of functions in muparser.
muparser requires that there be no space between the name of a function and its argument
list. This is incompatible to the behavior we had with fparser that allowed for spaces
in this position. To allow for backward compatible behavior, this patch changes the
parsing in such a way that it eats spaces after the function name for all built-in
and deal.II-defined functions before passing the expression to muparser.
This addresses bug #218 at https://code.google.com/p/dealii/issues/detail?id=218 .
Luca Heltai [Thu, 7 Aug 2014 13:04:09 +0000 (15:04 +0200)]
Merge pull request #63 from tamiko/fix_manifold_instantiations
This commit reverts the quick fix in commit 982b0683.
The issue why we encountered unresolved references to
dealii::FlatManifold::FlatManifold(dealii::Point, double)
etc. was due to boilerplate instantiations of Triangulation::get_manifold
for cases with spacedim < dim (in order to make TriaAccessor happy).
This commit fixes the issue by using a nullptr-dereference in order to
return an invalid object - the specific code throws an error message before
that and should actually not be reachable at all.
Matthias Maier [Thu, 7 Aug 2014 11:07:45 +0000 (13:07 +0200)]
Fix instantiation issue with FlatManifold properly
This commit reverts the quick fix in commit 982b0683.
This commit closes #56
The issue why we encountered unresolved references to
dealii::FlatManifold<3, 2>::FlatManifold(dealii::Point<2, double>, double)
etc. was due to boilerplate instantiations of Triangulation::get_manifold
for cases with spacedim < dim (in order to make TriaAccessor happy).
This commit fixes the issue by using a nullptr-dereference in order to
return an invalid object - the specific code throws an error message before
that and should actually not be reachable at all.
This way, the filter script can replace them automatically with references to
the corresponding pages for each of the tutorial programs. This does not
currently work on http://dealii.org/developer/doxygen/deal.II/Tutorial.html .
Luca Heltai [Wed, 6 Aug 2014 12:33:08 +0000 (14:33 +0200)]
Merge pull request #50 from luca-heltai/manifold_id_step0
This is the first backward compatible version of the manifold id branch. Four methods were added to a class named Manifold, which contains a subset of the Boundary methods. Boundary is now derived from FlatManifold. This fixes issue #27.
Added manifold, copying over the old concept, but retaining dim. Made Boundary<dim,spacedim> derived from Manifold<dim,spacedim>.
Added specific instantiations for invalid accessors. Moved get_default_quadrature into an anonymous namespace.
Made Boundary derived from FlatManifold, and made sure that nothing broke. Added get_new_point_on_cell/hex. Made get_new_point_on_quad not throw an exception in 2d.
Created Manifolds namespace, and made specialization of get_new_point_on_hex only in 3d, instead of multiple specializations for when dim not equal 3.
Added test flat_manifold_01
Fixed 2d problems.
Fixed 3d. Now all tests work, except from machine precision errors.
Added a new test for flat manifold checks on faces.
Completed documentation of manifold and tria_boundary.
Removed commented out functions from Boundary<dim,spacedim>, whose default has been moved to Manifold<dim,spacedim>
Fixed periodicity of FlatManifold, and added a test for it.
Timo Heister [Wed, 6 Aug 2014 03:57:54 +0000 (23:57 -0400)]
Merge pull request #45 from guidokanschat/doxygen_cleanup
Some cleanup of doxygen issues:
simplify doxygen input path
reduce warnings from thousands to hundreds as long as major code is undocumented
fix displayed include paths in doxygen output
fix some of the warnings
Timo Heister [Wed, 6 Aug 2014 03:54:03 +0000 (23:54 -0400)]
Merge pull request #53 from bangerth/fix-nedelec-interpolation-matrix
Patch by Alexander Grayver: Fix face interpolation between FE_Nedelec elements.
This patch fixes a bug in the FE_Nedelec::get_face_interpolation_matrix related to the wrong indexing of the DoFs located on element faces. This patch ensures that the code produces an interpolation matrix with full column rank meaning that all DoFs of the FE with lower order are constrained.
Patch by Alexander Grayver: Fix face interpolation between FE_Nedelec elements.
This patch fixes a bug in the FE_Nedelec::get_face_interpolation_matrix related to the wrong indexing of the DoFs located on element faces. This patch ensures that the code produces an interpolation matrix with full column rank meaning that all DoFs of the FE with lower order are constrained.
Guido Kanschat [Fri, 1 Aug 2014 17:12:29 +0000 (19:12 +0200)]
simplify doxygen input path
reduce warnings from thousands to hundreds as long as major code is undocumented
fix displayed include paths in doxygen output
fix some of the warnings
Implement new initialize method for Trilinos AMG preconditioner: It can be based on an Epetra_RowMatrix object. We can use this code path for the standard initialize method with TrilinosWrappers::SparseMatrix as well.
Timo Heister [Mon, 4 Aug 2014 19:23:30 +0000 (15:23 -0400)]
fix wrong assert in notify_ready_to_unpack
The offset will always be >=4 because we store the CellStatus in the
data sent to p4est in front of the user data. This means the smallest
offset will 4, no matter how much data is attached.