Here, just check that a valid value is set for the finite difference formula. Later on,
we should then not ever get an invalid value any more, but if we do, we can just say
that whatever we got is not implemented.
Improve an error message by duplicating text from the input file.
This is motivated by a question on the mailing list. We only printed the internal
number of the cell, edge, or point, but should really be printing the number that
is used in the input file.
Timo Heister [Mon, 5 Sep 2016 19:40:34 +0000 (15:40 -0400)]
doxygen: format exception messages
This creates pretty formatting in the doxygen documentation for the
exception texts. We do this by including a javascript function that does
a couple of string replacements. It is difficult to do this
transformation before generating the documentation as detailed in #2571.
David Wells [Sat, 3 Sep 2016 16:31:49 +0000 (12:31 -0400)]
Set --options=none when indenting .inst.in files.
This works around a problem where, if one has global astyle options
turned on, they will be used to indent all of the .inst.in files when
one calls 'make indent'.
To assist debugging, where possible the error thrown internally when the
UCD grid is created is now displayed before the terminal error message
is given.
the assert was copy-pasted from p::d::Tria::locally_owned_subdomain()
to the based class p::Tria. Not sure why it was there for distributed
case, but for shared tria + metis, everything should work ok.
David Wells [Sun, 28 Aug 2016 18:38:55 +0000 (14:38 -0400)]
Fix get_intermediate_point in 1D.
Since geometrical entities with spacedim == 1 are line segments, we
should just do a linear combination to find the intermediate
point (there is no need for angle calculations).
David Wells [Sun, 28 Aug 2016 01:04:29 +0000 (21:04 -0400)]
Remove tests/mpi/coarse_grid_common.h.
The only functions in this header file that are actually used in this
directory are some extra LogStream::operator<< overloads: those
belong (logically) in the tests.h header file.
It was used in a function that has been used from namespace GridTools to
namespace GridGenerator, where the structure is now duplicated. It was
only used for MS Visual Studio in any case.
David Wells [Sun, 21 Aug 2016 17:54:14 +0000 (13:54 -0400)]
Remove duplicated tests.
This commit continues the work started in b9c64f0dac and removes tests
that are identical up to things the compiler sees (e.g., tests that are
the same aside from commenting or formatting). I did this by deleting
comments with the preprocessor and then running the result through
clang-format.
For the sake of posterity: I compiled this list by executing the python
script
# get rid of any line number indicators cpp may have left
subprocess.check_call(["sed", "-i", "/^# [0-9]/d", preprocessed_file_name],
stderr=sys.stderr)
# reformat source to canonical form
subprocess.check_call(["clang-format", "-i", preprocessed_file_name])
md5sum = hashlib.md5()
with open(preprocessed_file_name, "rb") as preprocessed_file_handle:
for block in iter(lambda: preprocessed_file_handle.read(65536), b""):
md5sum.update(block)
print(md5sum.hexdigest(), current_test_file_name)
What follows is a brief description of why each test was removed:
bits/block_sparse_matrix_iterator_03.cc
Same as bits/block_sparse_matrix_iterator_02.cc (up to a comment)
bits/sparse_matrix_01a.cc
Identical (up to a comment) to bits/sparse_matrix_01.cc
bits/step-16b.cc
Identical (up to a comment) to bits/step-16.cc
bits/vector_25.cc
Identical (up to a comment) to bits/vector_24.cc
The tests
dofs/dof_constraints_06_x.cc
dofs/dof_constraints_10_x.cc
dofs/sparsity_pattern_x.cc
dofs/sparsity_pattern_01_x.cc
were all duplicates (up to comments) of tests without '_x's in their
names. This was a result of deleting CompressedSparsityPattern et al.
fe/fe_support_points_q_qg0.cc
Identical (up to a blank line) to fe/fe_support_points_q_dg0.cc
fe/fe_values_view_10_single_04.cc
Identical (up to some code formatting and a comment) to
fe/fe_values_view_08.cc
The tests
full_matrix/complex_complex_full_matrix_05.cc
full_matrix/complex_complex_full_matrix_09.cc
were duplicates of the corresponding 'complex_real_full_matrix_*.cc'
test. Since the full matrices were real valued I kept the other ones.
grid/have_same_coarse_mesh_04.cc
Completely identical to grid/have_same_coarse_mesh_02.cc
matrix_free/matrix_vector_14b.cc
This test claims it uses more quadrature points than
matrix_free/matrix_vector_14.cc, but does not (and changing the number
of quadrature points is nontrivial for this test)
petsc/25.cc
Identical (up to a comment) to petsc/24.cc
trilinos/25.cc
Identical (up to a comment) to trilinos/24.cc
Get rid of the confusing ExcInvalidConstructorCall exception.
This exception did not print any useful error message, and it was used in a significant
number of places that just happened to check obscure conditions, without really
explaining in the error message what concretely they were doing. Replace all of these
with ExcMessage and provide explanations what exactly it is that's going wrong.